CIOMS

Management of Safety Information from Clinical Trials

CIOMS publications may be obtained directly from CIOMS,
c/o World Health Organization, Avenue Appia, 1211 Geneva 27,
Switzerland or by e-mail to [email protected]
Management of
Both CIOMS and WHO publications are distributed by the
Safety Information
World Health Organization, Marketing and Dissemination,
Avenue Appia, 1211 Geneva 27, Switzerland and are available
from booksellers through the network of WHO sales agents.
from Clinical Trials
A list of these agents may be obtained from WHO by writing
to the above address.
Report of CIOMS Working Group VI

Price: CHF 40.– Geneva 2005

GROUP6_COVER DEF.indd 1 7.8.2007 12:17:50

 Management of
Safety Information
from Clinical Trials

Report of CIOMS Working Group VI

Geneva 2005

group6_PH.indd 1 7.8.2007 12:19:13

 Copyright © 2005 by the Council for International
Organizations of Medical Sciences (CIOMS)

ISBN 92 9036 079 8

group6_PH.indd 2 7.8.2007 12:19:15

 Acknowledgements

T he Council for International Organizations of Medical Sciences (CIOMS)

gratefully acknowledges the contributions of the members of CIOMS
Working Group VI on the Management of Safety Information from Clinical
Trials as well as the drug regulatory authorities, pharmaceutical companies
and other organizations and institutions which supported the work that
resulted in this publication. Hard work, drafting and redrafting of papers,
their reviews and a number of debates in the Working Group required
patience, motivation and active collaboration from all members.

CIOMS acknowledges especially the co-chairs, Drs. Wendy Stephenson

and Gottfried Kreutz, for their capable leadership, and Ms Linda Hostelley,
the secretary of the group. The editorial group, comprised of Drs. Gerald Dal
Pan, Arnold J. Gordon, Marianne Keisu, Siddika Mithani, and Wendy
Stephenson, merits special mention and thanks. CIOMS also wishes to
express special appreciation to Dr. Gordon, who as chief editor of the final
report assured the quality of the publication.

CIOMS and the Working Group are thankful for important input
received on several topics from many senior experts outside the Group who
reviewed the entire manuscript and made valuable suggestions: Drs. Ric Day
(University of South Wales, Australia), Frank Rockhold and Rita Patwardhan
(GlaxoSmithKline, US), and Patrick Waller (Consultant, UK). Members of the
US and EU pharmaceutical industry associations (PhRMA and EFPIA) also
provided very helpful detailed comments and suggestions; special thanks
are due to Drs. Barry Arnold (AstraZeneca, UK) and Brian Edwards (Barnett
Parexel, UK) and to Cindy Engle (GlaxoSmithKline, US) for their roles in syn-
thesizing these contributions. Dr. Susan Ellenberg (FDA, US) contributed her
expertise to material covering Data and Safety Monitoring Boards. Dr. Susan
Sacks (F. Hoffmann-La Roche Ltd.) provided key information on sources of

group6_PH.indd 3 7.8.2007 12:19:15

 epidemiological and related databases. Thanks are also due to a number of
colleagues from pharmaceutical companies who completed the CIOMS VI
questionnaire (see Appendix 3). Finally, we are grateful to Meghan McLaren
(Health Canada) for her administrative assistance in preparing the draft and
final documents.

Geneva, April 2005

Juhana E.Idänpään-Heikkilä, MD. PhD

Secretary-General, CIOMS

group6_PH.indd 4 7.8.2007 12:19:15

 Dedication

T his work is dedicated to the many thousands of patients and other

volunteers who generously participate in clinical research programs so
vital for the development and advancement of medicines.

group6_PH.indd 5 7.8.2007 12:19:16

group6_PH.indd 6 7.8.2007 12:19:16
 Table of Contents

Page
VISION ................................................................................................ 13

PREFACE ............................................................................................ 15

I INTRODUCTION AND OVERVIEW .............................................. 19

a. Rationale for the CIOMS VI Project .................................................... 21
b. Results of the CIOMS VI Survey on Company Practices .................... 25
c. Areas Covered by the CIOMS VI Project ............................................. 26
• Terminology and definitions ...................................................... 27
• Ethical aspects of clinical trials ................................................. 27
• Overall pharmacovigilance/risk management system ............... 27
• Collection and proper management of safety data .................... 28
• Evaluation of safety data ........................................................... 28
• Statistical analysis of safety data ............................................... 29
• Regulatory reporting and communication to others
of safety information during clinical trials ................................ 29
d. Limitations of Clinical Trials for Understanding Safety ................ 30
e. Scope of the Project ........................................................................ 32

II ETHICAL CONSIDERATIONS FOR CLINICAL TRIAL

SAFETY MANAGEMENT................................................................ 35
a. Background .................................................................................... 37
b. The Stakeholders ............................................................................ 39
• Patients ...................................................................................... 39
• Regulatory Authorities and the Public Health Community ....... 40
• Investigators .............................................................................. 41
• IECs and IRBs ........................................................................... 41
• Data and Safety Monitoring Boards .......................................... 43
• Pharmaceutical Companies and Their Representatives ............. 44

group6_PH.indd 7 7.8.2007 12:19:16

 c. Evolving Regulatory and Societal Demands .................................. 44
• Privacy and Confidentiality of Personal Data............................ 45
• Informed Consent ...................................................................... 46
• Transparency in Availability of Clinical Trial Results............... 47
• Other Issues ............................................................................... 51

III GOOD PHARMACOVIGILANCE AND RISK MANAGEMENT

PRACTICES: SYSTEMATIC APPROACH TO MANAGING
SAFETY DURING CLINICAL DEVELOPMENT ........................ 53
a. Introduction .................................................................................... 55
b. Principles of a Systematic Approach .............................................. 57
• Begin early ................................................................................ 57
• Establish a procedure ................................................................ 57
• Establish a Multidisciplinary Safety Management
Team (SMT) .............................................................................. 58
• Establish a project management function.................................. 59
• Determine background data ...................................................... 60
• Ensure accessibility of data ....................................................... 60
• Develop a proactive approach ................................................... 60
• Establish timeframes and milestones ........................................ 61
• Decision making ........................................................................ 62
• Advisory bodies ........................................................................ 62
c. Components of a Development Risk Management Plan (DRMP) . 63
d. Role of Epidemiology..................................................................... 67
e. Specific Issues that Should Always be Considered ........................ 70
• Cardiac electrophysiology ......................................................... 70
• Hepatotoxicity ........................................................................... 71
• Drug-Drug and Food-Drug interactions .................................... 71
• Immunogenicity......................................................................... 71
• Bone marrow toxicity ................................................................ 72
• Potential for reactive metabolite formation
and hypersensitivity ................................................................... 72
f. Conclusion ...................................................................................... 72

group6_PH.indd 8 7.8.2007 12:19:16

 IV COLLECTION AND MANAGEMENT OF SAFETY DATA
DURING CLINICAL TRIALS ......................................................... 75
a. Introduction .................................................................................... 77
b. Who? .............................................................................................. 79
c. What?.............................................................................................. 81
• General Principles ..................................................................... 81
• Causality Assessment ................................................................ 84
• Diagnoses vs Signs and Symptoms ........................................... 86
• Adverse Events of Special Interest ............................................ 88
• Laboratory Chemistry Measurements ....................................... 88
• Morbidity and Mortality as Efficacy Endpoints ........................ 89
• Special Situations ...................................................................... 90
d. How?............................................................................................... 91
• General Considerations ............................................................. 91
• Serious and Other Important Adverse Events ........................... 94
e. When? ............................................................................................. 95
f. Safety Data Management Considerations ...................................... 97
• Clinical Description of Adverse Events .................................... 98
• Coding Procedures .................................................................... 101
• Dealing with Unblinded Data .................................................... 103
• Data Processing Issues .............................................................. 104

V IDENTIFICATION AND EVALUATION OF RISK

FROM CLINICAL TRIAL DATA .................................................... 107
a. Introduction .................................................................................... 109
b. Expectations and Limitations in the Identification
and Evaluation of Safety Information from Clinical Trials ............ 111
c. Points to Consider During Analysis and Evaluation of
Safety Information.......................................................................... 114
• Patient Population Characteristics, Including Natural
History of Disease ..................................................................... 114
• Current Therapeutic Standards .................................................. 115
d. Timing of Safety Evaluation........................................................... 116

group6_PH.indd 9 7.8.2007 12:19:16

 e. Safety-Signal Detection and Evaluation ......................................... 117
f. Consistent Causality Assessment – Adverse Events vs Adverse
Drug Reactions ............................................................................... 119
g. Important Types of Analyses .......................................................... 120
h. Review of Individual Cases ............................................................ 121
i. Considerations for Periodic Review and Evaluation
of Case Reports in Aggregate ......................................................... 122
j. Pooling of Data ............................................................................... 123
k. Evaluation of Clinical Laboratory Data ......................................... 124
l. General Benefit-Risk Considerations ............................................. 127
m. Aggregate Analysis and the DCSI .................................................. 127

VI STATISTICAL ANALYSIS OF SAFETY DATA

IN CLINICAL TRIALS ..................................................................... 129
a. Introduction .................................................................................... 131
b. Uses of Statistics for Clinical Safety Data ..................................... 133
c. Principle of Intention to Treat......................................................... 136
d. Some Key Problems in Safety Analyses ......................................... 137
e. Useful Approaches to Statistical Analysis of Continuous
Measurements: Laboratory Chemistries......................................... 139
f. Statistical Treatment of Binary Data .............................................. 142
• Power Considerations ................................................................ 142
• One-Sided vs Two-Sided Testing ............................................... 143
• The Consequences of Multiplicity ............................................ 144
• General Measures Using Binary Data ....................................... 145
• Confidence Intervals .................................................................. 148
• Accounting for Time on or off Treatment ................................. 150
• Statistical Tests Using Time Since Start of Treatment............... 155
g. Combining Data from Several Trials: The Role
of Meta-analytical Techniques ........................................................ 157
h. Analysis of Rare Events ................................................................. 160
i. Measuring and Expressing Effects in Ways Relevant
to Public Health .............................................................................. 161

10

group6_PH.indd 10 7.8.2007 12:19:16

 j. Comments on ICH Guidelines E3 and E9: Discussion of
Statistical Aspects of Clinical Safety Data ..................................... 163

VII REGULATORY REPORTING AND OTHER COMMUNICATION

OF SAFETY INFORMATION FROM CLINICAL TRIALS ........ 165
a. Introduction .................................................................................... 167
b. Expedited Reporting from Clinical Trials ...................................... 169
• Expedited Reporting to Regulatory Authorities ........................ 169
• Expedited Reporting: Causality ................................................ 170
• Expedited Reporting: Expectedness .......................................... 171
• Expedited Reporting: Unblinding ............................................. 172
• Expedited Reporting: Comparators ........................................... 173
• Expedited Reporting: Spontaneous Reports ............................. 175
• Prompt Reporting Other than Case Reports .............................. 175
• Expedited Reporting: Investigators and IECs/IRBs .................. 177
c. Periodic Communication of Safety Information
from Clinical Trials......................................................................... 180
• Development Safety Update Report (DSUR) ........................... 180
• Investigator Brochure and DCSI Updates ................................. 181
• Other Periodic and Ad Hoc Communications
to Investigators and IECs/IRBs ................................................. 182
• Safety Management Process...................................................... 184
d. Other Reporting Considerations ..................................................... 184
e. Informed Consent ........................................................................... 185
f. Other Communication Considerations ........................................... 187
g. Conclusion ...................................................................................... 188

VIII SUMMARY OF CONCEPTS AND PROPOSALS.......................... 189

APPENDICES ..................................................................................... 217

1. Glossary and Abbreviations............................................................ 217
2. Membership and Process of CIOMS Working Group VI ............... 241
3. CIOMS VI Working Group Survey of Pharmaceutical
Company Safety-Management Practices During Clinical Trials .... 245

11

group6_PH.indd 11 7.8.2007 12:19:17

 4. World Medical Association Declaration of Helsinki ...................... 259
5. Data and Safety Monitoring Boards (DSMBs) .............................. 265
6. Data Elements that Should Be Considered for Individual
Adverse Event Reports ................................................................... 271
7. Causality Criteria and Threshold Considerations for Inclusion
of Safety Data in Development Core Safety Information (DCSI) . 275
8. Sample Serious AE Report Data Collection Form
for Investigators .............................................................................. 279
9. Databases for Epidemiology and Pharmacoepidemiology ............. 287

INDEX ................................................................................................. 289

12

group6_PH.indd 12 7.8.2007 12:19:17

 Vision

Patients and prescribers expect approved medicines to be “safe and

effective.” The goal of those who produce and regulate proprietary me-
dicinal products is to ensure that this expectation is met. This requires that
clinical trials be planned and performed to provide good evidence that
tested medicines are effective and that patients can be reassured that the
benefits outweigh the risks, both during the development process and in
general use.
This report has implications for all stakeholders in clinical medicinal
research:
1) patients and other volunteers
2) investigators and their site staff
3) ethics review committees
4) data and safety monitoring boards
5) drug regulatory authorities and the public health community
6) pharmaceutical companies and other clinical research sponsors
The vision of the CIOMS VI Working Group is that this report will
enhance awareness of the ethical and technical issues associated with safety
in clinical trials and point out the need for increased care and scrutiny in
the conduct of research. It is also hoped that this work will advance the me-
thodology for collecting, analysing, evaluating and reporting information
on product safety ascertained in clinical trials, and help to set standards in
these areas. Establishing and maintaining standards by all involved groups
will benefit all participants in trials and improve public health for those who
take medicines.
Pharmacovigilance has traditionally focused on detection and evalu-
ation of signals in the post-approval environment in order to secure early
detection of new adverse reactions or patient subgroups of exceptional sen-
sitivity, and to introduce measures to manage those risks.
However, we believe that there is a need not only to incorporate newer
approaches for managing safety information in the clinical trial setting, but
also to adapt the methods and tools used in post-approval pharmacovigi-
lance to the early and late stages of pre-approval development of medicinal
products. It is our vision that the practical approaches provided
in this report will aid these processes and will enable a more seamless

13

group6_PH.indd 13 7.8.2007 12:19:17

 transition in conducting high quality pharmacovigilance from the develop-
ment stage to the post-approval period. We also hope that this work will
stimulate research in several unresolved areas.
Finally, we recognize that new regulations have recently been enacted
in the EU and are pending elsewhere, such as in the US. It is hoped that
this book will stimulate the regulators to reconsider aspects of regulations
pertaining to our proposals; we believe that our suggestions can help to
improve the ability to generate and analyze useful safety data and to protect
trial participants.

14

group6_PH.indd 14 7.8.2007 12:19:17

 Preface

Since 1986, when they began a series of projects dedicated to im-

portant drug safety issues, the CIOMS Working Groups on drug safety
have been recognized for creating the theoretical platforms and pragmatic
suggestions to advance the debates leading to harmonization of interna-
tional pharmacovigilance practices. The initiatives over the years, iden-
tified as CIOMS Working Groups I, IA, II, III, IV and V, have resulted
in six major published reports.1 The nature of their membership, senior
drug safety officials from many major regulatory agencies and the regu-
lated pharmaceutical industry, and their modus operandi as a “think tank”
seeking practical solutions to important problems, have facilitated their
unique contributions. All members have served less as representatives of
any single organization or interest and more as motivated colleagues, with
day-to-day responsibility in the drug safety field. All shared a commit-
ment to think beyond their local practices even if such thinking were in
disagreement with current rules and regulations, in order to optimize drug
safety procedures, particularly in an international context. Although the
Working Groups did not – indeed could not – develop regulations, its work
has always been intended to inform and encourage those with rule-
making responsibilities. Gratifyingly, many of the recommendations have
been incorporated into regulations, not only in the countries of the partici-
pating regulators, but elsewhere as well.
The CIOMS I Working Group introduced definitions, criteria and a
standard form (CIOMS I Form) for international reporting of medically
important (“serious”) adverse drug reactions (ADRs) to marketed prod-
ucts. It also served as a model for the development of the International
Conference on Harmonization (ICH) Guideline E2A on expedited ADR
case reporting for clinical trials.
The result of the CIOMS II deliberations was a set of proposed
standards for the format, content and frequency of periodic safety update
reports (PSURs) which has been adopted by many regulatory authorities.

1
International Reporting of Adverse Drug Reactions (CIOMS I) (1990); International Reporting of Periodic
Drug-Safety Update Summaries (CIOMS II)(1992); Guidelines for Preparing Core Clinical-Safety Informa-
tion on Drugs, First Edition (1995) (CIOMS III) and Second Edition, Including New Proposals for Investiga-
tor’s Brochures (1999)(CIOMS III/V); Benefit-Risk Balance for Marketed Drugs: Evaluating Safety Signals
(CIOMS IV)(1998); Current Challenges in Pharmacovigilance: Pragmatic Approaches (CIOMS V)(2001).
All published by the Council for International Organizations of Medical Sciences, Geneva.

15

group6_PH.indd 15 7.8.2007 12:19:17

 It also formed the basis for the ICH Guideline on periodic reporting, E2C,2
adopted in 1996 and subsequently implemented internationally. A recently
adopted addendum to ICH E2C3 represents further refinement of the har-
monization concepts espoused by the CIOMS V Working Group.
Coincident with CIOMS II and in recognition of the need for more ef-
ficient, automated techniques to document and report ADRs to regulators,
the CIOMS IA subgroup worked on a proposal for a harmonized format for
electronic submissions. The CIOMS IA recommendations were not pub-
lished but formed the basis for the ICH Guideline E2B (Data Elements for
Transmission of Individual Case Safety Reports, 1997).
The CIOMS III Working Group concentrated on best practices for ap-
plying the concept of “company core safety information” (CCSI) intro-
duced in CIOMS II. The Working Group developed a set of what have con-
veniently been referred to as “good safety information/labeling practices”
for post-approval drug safety data, including practical guidance on deter-
mining when the threshold has been reached for adding an adverse reaction
to the CCSI. In the CIOMS III/V report, the second edition of the CIOMS
III report, the concepts were extended to the pre-approval environment by
recommending use of Development Core Safety Information (DCSI).
One of the most important aspects of post-marketing safety surveil-
lance is the identification and analysis of new, medically important find-
ings that might influence the use of a medicine. In recognizing that there
existed at the time no guidance on a systematic approach for handling the
emergence of a major safety issue, especially one that might lead to impor-
tant regulatory action, CIOMS IV developed its proposals for approaches
to comparative benefit-risk evaluation, analysis of options for action, and
good decision making practices.
As acknowledged in the reports by each of the Working Groups, un-
resolved and un-addressed issues remained. Thus was born the CIOMS V
Working Group which focused on several difficult aspects of day-to-day
pharmacovigilance work that affect the management and interpretation of
safety data. The proposals and their rationale are the subject of the CIOMS
V report which has become a regular source of guidance to industry profes-
sionals engaged in the day-to-day management of safety reporting.

2
ICH E2C Clinical Safety Data Management: Periodic Safety Update Reports for Marketed Drugs, Step 5 as
of November 1996. (See http://www.ich.org)
3
Addendum to ICH E2C: Periodic Safety Update Reports for Marketed Drugs, Step 5 as of February 2003.
(See http://www.ich.org)

16

group6_PH.indd 16 7.8.2007 12:19:18

 Proposals emanating from CIOMS Working Groups I through V
were principally focused on post-marketing surveillance regulations and
activities. These are described in greater detail in published papers.4,5 The
CIOMS Working Group reports themselves can be ordered by sending a
request to [email protected].
The current report, that of the CIOMS VI Working Group, repre-
sents a shift from the management of post-marketing safety information,
which relies heavily on spontaneous reports, to the management of clini-
cal trial information, starting from the earliest clinical trials and extending
to the post-marketing environment. The CIOMS VI Working Group also
represents an expansion in membership to include regulatory, industry and
academic representation with experience in the conduct of clinical trials
and to include representatives from less developed regions of the world.
This book introduces proposals for enhancing the collection, analysis, eval-
uation, reporting and overall management of safety information from clini-
cal trials. It also discusses the importance of sponsors’ having a systematic
approach to managing risk during development, taking into account non-
clinical as well as clinical data. CIOMS VI is not intended to be a reitera-
tion of other available guidances and guidelines on these subjects 6,7,8,9,10
or a rehash of closely related recent work.11,12 Rather, the proposals in this
document should be considered along with the principles established in
those authoritative documents.
The views and recommendations in this book are those of the
CIOMS VI Working Group as a whole, generally reached through a consen-
sus process, or in some cases by a majority vote. They do not necessarily
represent the views of the participants’ sponsoring organizations.

4
Castle, W. Overview of the CIOMS Pharmacovigilance Working Group. Regulatory Affairs Focus, April 2000
(Regulatory Affairs Professionals Society; see www.raps.org)
5
Tsintis, P. and LaMache, E. CIOMS and ICH Initiatives in Pharmacovigilance and Risk Management.
Overview and Implications. Drug Safety, 27 (8):509-517, 2004
6
ICH E6 Good Clinical Practice: Consolidated Guideline, Step 5 as of May 1996. http://www.ich.org
7
ICH E9 Statistical Principles for Clinical Trials, Step 5 as of February 1998. http://www.ich.org
8
ICH E3 Structure and Content of Clinical Study Reports, Step 5 as of November 1995. http://www.ich.org
9
ICH E2A Clinical Safety Data Management : Definitions and Standards for Expedited Reporting, Step 5 as
of October 1994. http://www.ich.org
10
Guideline for the Format and Content of the Clinical and Statistical Section of an Application, US FDA
Center for Drug Evaluation and Research, July 1988. http://www.fda.gov/cder/guidance
11
Guidance for Industry: Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment, and
Premarketing Risk Assessment. US Food and Drug Administration, March 2005. (See http://www.fda.gov/
cder/guidance/6357fnl.htm and http://www/fda.gov/cder/guidance/6359OCC.htm), respectively.
12
ICH Guideline E2E. Pharmacovigilance Planning (PvP), Step 4 as of November 2004. (See http://www.ich.
org)

17

group6_PH.indd 17 7.8.2007 12:19:18

 It is recognized that some of the proposals, in particular those found in
Chapter 7, may be in conflict with existing regulations in various countries
and in newly enacted legislation in Europe. However, it is the hope of the
CIOMS VI Working Group that its recommendations may stimulate regula-
tors to rethink some aspects of their regulations in terms of the practicality
of implementation and the usefulness of the safety information provided to
stakeholders. This book is primarily aimed at providing guidance to spon-
sors of clinical trials. The hope is that these proposals, once adopted by
regulatory authorities, will enhance our ability to protect patient well-being
and optimize the development and use of new medicines.

18

group6_PH.indd 18 7.8.2007 12:19:19

 I

Introduction
and Overview

group6_PH.indd 19 7.8.2007 12:19:19

group6_PH.indd 20 7.8.2007 12:19:19
 a. Rationale for the CIOMS VI Project
Medical research on human subjects is capable of providing signifi-
cant advances to benefit individuals and public health. The universally
accepted motif for such endeavors is that all such research be designed
to create important scientific knowledge, that it maximize the potential
benefits to the subjects, and, most importantly, it minimize their risks.
Thus, although this report focuses on the technical, medical and regula-
tory aspects of “drug safety” during clinical research, the foundation for
all these activities must always be the respect for the rights and welfare
of clinical trial participants. Some harm to individual patients or subjects
may be considered tolerable – for after all, it is the proper balance between
benefits and risks that drives not only the research process but the use of
any marketed medicine.1,2
Regulations and guidelines in most countries govern the conduct and
requirements of clinical trial sponsors and, increasingly, investigators and
their institutions. The collection, monitoring, and regulatory reporting of
clinical safety information on trial subjects feature prominently in such
regulations, usually in connection with Good Clinical Practice (GCP) re-
quirements. In many cases, regulations pertaining to clinical trials have
been based on or influenced by guidelines established under the ICH pro-
cess. Among the most important are: timing of non-clinical studies in
relation to clinical exposure (ICH M3), extent of population exposure to
assess safety in a development program (E1), expedited regulatory safety
reporting (E2A), presentation of safety data in clinical study reports (E3),
dose response information (E4), ethnic factors (E5), good clinical practice
(E6), studies in special populations (geriatrics, E7 and pediatrics, E11),
statistical principles (E9), and choice of active or placebo control group(s)
(E10).3

1
For purposes of this report, the term medicine or drug refers to prescription or over the counter products,
whether they are “drugs”, vaccines, or biotechnology products for prevention, prophylaxis or treatment of a
disease or medical condition, and possibly for use in diagnosis.
2
The benefit-risk relationship is commonly imprecisely summarized by referring to a product as “safe and
effective,” a description that may be misleading. The words “safety” and “safe” in common usage infer the
presence or absence of harm. The Working Group believes there is an erroneous perception, especially by the
public, that once a drug reaches the market it is, or should be, risk-free. In clinical trials or in general product
use, patients are monitored for the presence or absence of harm (not “safety”), and the data are assessed to
evaluate the probability of such harm, in other words the risk associated with the treatment. Patients and trial
subjects will have different “acceptable” levels of harm or risk, and in that sense risk and harm are relative
concepts to the individual.
3
For full details and access to specific guidelines, see: http://www.ich.org

21

group6_PH.indd 21 7.8.2007 12:19:19

 Historically, the CIOMS Working Groups on drug safety have concen-
trated on post-approval pharmacovigilance4 while recognizing that pharma-
covigilance as a discipline and a science should be regarded as a continuum
throughout the life of a product, beginning with exposure in humans during
the first Phase 1 trials.5 Nevertheless, safety functions for new drug devel-
opment and post-marketing are often separated, even to the point of main-
taining different departments and responsibilities within some companies
and most regulatory bodies. Although the separation of pre- and post-
marketing clinical safety departments and responsibilities has been a
standard model for drug regulation, it may introduce an unnecessary
complication into the discipline of pharmacovigilance. There has been a
trend in recent years toward more integration of the two organizational
divisions. Even with organizational separation, the conduct of pharmaco-
vigilance is best accomplished with close functional collaboration between
the groups. One goal of this CIOMS VI project is to help bridge the gap
between pre- and post-approval activities to understand and manage risk.
Although general responsibilities for managing drug safety issues are
usually covered in GCP regulations or guidances, the details and increas-
ing complexity of the field would benefit from the development of more
specific, internationally based Good Pharmacovigilance Practices (GPP),
something beyond the scope of this Working Group’s efforts.
More than 130 pharmaceutical products have been withdrawn from
various markets over the past 40 years because of actual or perceived safety
concerns. An estimated third were withdrawn within two years of launch
and half within 5 years.6 The most frequent problems are reported to be as-
sociated with hepatic, hematological and cardiovascular complications. It is
unclear whether such problems could have been foreseen during the drugs’
development and, if recognized early, managed sufficiently to preclude the
harm done to individual patients and to establish a favorable benefit-risk
profile for a specified target population. However, many lessons have been
learned from these past experiences, among them the need to have a more

4
The term “pharmacovigilance” is not always used consistently among its practitioners. For example, there is
some debate as to whether it should be used for pre-approval safety. We recommend that it should be. Detailed
discussion on terminology and definitions for this and other concepts are covered in Appendix 1.
5
For the most recent CIOMS report, see “Current Challenges in Pharmacovigilance: Pragmatic Approaches,”
Report of CIOMS Working Group V, CIOMS, Geneva, 2001.
6
For example, see Fung, M., Thornton, A., Mybeck, K., Wu, J. H., Hornbuckle, K. and Muniz, E., Evaluation
of the Characteristics of Safety Withdrawal of Prescription Drugs from Worldwide Pharmaceutical Markets
– 1960 to 1999. Drug Information Journal, 35: 293-317, 2001 and Friedman, M. A., Woodcock, J., Lumpkin,
M. M., Shuren, J. E., Haas, A. E. and Thompson, L. J., The Safety of Newly Approved Medicines. Do Recent
Market Removals Mean there is a Problem? J. Am. Med. Assoc., 281:1728-1734, 1999.

22

group6_PH.indd 22 7.8.2007 12:19:19

 systematic and comprehensive approach to safety issues during the devel-
opment process and beyond. Identifying and managing real or potential
safety problems as early as possible during drug development might there-
fore be expected to help avoid premature termination of a development
program and thereby prevent the loss of therapies whose benefits might
indeed outweigh the risks when used appropriately in the right patients.
Similarly, a systematic approach will also enable identification at the ear-
liest possible time of a product that does not meet an acceptable benefit-risk
profile, and therefore allow for cessation of the program with minimal risk
to trial subjects.
Recent concerns have been raised by clinical researchers on possible
inadequacies in adverse event reporting and on the interactions between par-
ties with clinical safety oversight responsibilities.7 Some of their concerns
and suggestions include: valid assessment of individual AE cases needs
more information, including on efficacy, than that contained in the case it-
self; the use and roles of data and safety monitoring boards (DSMBs) need
improvement, especially vis-à-vis their relationship with institutional eth-
ics committees/institutional review boards (IECs/IRBs) (e.g., they should
provide the ethics committees with a periodic summary on whether or not
the safety and other parameters established for the trial are as expected).
Many problems faced by IECs/IRBs were also identified, such as their in-
ability to evaluate the overwhelming number and types of expedited reports
in isolation from previous data; lack of access to data from sites other than
their own to help place their data in proper perspective; confusing regulato-
ry terminology; and often a focus on regulatory compliance at the expense
of ongoing benefit-risk assessment at the local site.
From a different perspective, a review of safety reporting and reasons
for patient withdrawal for toxic effects during randomized trials has found
variability and inadequacies across several medical areas, with a call for
improved standards.8
Although there are some important differences between pre-marketing
and post-marketing safety monitoring and management, there is a growing
realization that there should be a much stronger and closer relationship

7
Morse, M. A., Califf, R. M. and Sugerman, J., Monitoring and Ensuring Safety During Clinical Research,
J. Am. Med. Assoc., 285: 1201-1205, 2001 and a response to the article – Burman, W. J. and Schooley, R. The
Role of Local Institutional Review Boards in Protecting Human Research Subjects. ibid., 285: 2713, 2001.
8
Ioannidis, J. P. A. and Lau, J., Completeness of Safety Reporting in Randomized Trials. An Evaluation of
7 Medical Areas. J. Am. Med. Assoc., 285: 437-443, 2001, and commentary on the publication, Safety Reporting
in Clinical Trials, ibid., 285: 2076-2078, 2001.

23

group6_PH.indd 23 7.8.2007 12:19:20

 between them, for among other reasons, to learn from and use the tools
and methodologies applied in the two environments. This same observa-
tion has been forcefully made by Vandenbroucke, who asserts that those
who conduct randomized trials and pharmacoepidemiologists are needless-
ly “worlds apart” in their approaches to understanding drug benefits and
harms.9 This connection is especially important for development programs
on new uses or dosage forms of products already on the market. In an age
when the “risk management” of medicines has almost become a discipline
on its own, the development process must incorporate planning for manag-
ing risks early in a product’s life, and well before launch. Such planning
will allow the transition from development to authorized use to be based on
a systematic and comprehensive pharmacovigilance plan that involves not
only clinical safety specialists (including pharmacoepidemiologists), but
also toxicologists, clinical pharmacologists, statisticians, and clinicians.10
Although existing regulations form the framework for how clinical
trial safety data should be monitored and reported to authorities and other
involved parties, there is a considerable lack of consistency and complete-
ness in the treatment of these issues. This is partly due to cultural and his-
torical influences that have an impact on individual country requirements.
There also are many important topics not even covered within regulations
or supporting guidelines that perhaps should be, and it is our goal to stimu-
late discussion of these topics in the hope that consensus may be reached
for change.
It is important to remind the reader that this book presents proposals
and recommendations that may or may not be in agreement with current
regulations and guidance from health authorities. Continued adherence to
current requirements and guidances is obviously essential unless and until
our recommendations are officially recognized and implemented.
All the principles and practices recommended throughout this re-
port are summarized in Chapter 8 (Summary of Concepts and Proposals),
which the reader may wish to consult for a convenient overview of the
main points. We also believe that it would be very beneficial for the readers
to familiarize themselves with the Glossary (Appendix 1), especially the
introductory explanations.

9
Vandenbroucke, J. Benefits and harms of drug treatments. Observational studies and randomised trials should
learn from each other. British Medical Journal, 329:2-3, 2004.
10
For example, see Perfetto, E. M., Ellison, R., Ackerman, S., Shorr, M., and Zaugg, A. M., Evidence-Based
Risk Management: How Can We Succeed? Deliberations from a Risk Management Advisory Council. Drug
Information Journal, 37: 127-134, 2003.

24

group6_PH.indd 24 7.8.2007 12:19:21

 A listing of the Working Group members and their affiliations along
with a summary of its activities over the nearly four years spent to bring this
project to fruition are found in Appendix 2.

b. Results of the CIOMS VI Survey

on Company Practices
In order to ascertain prevailing practices in the industry for many of
the areas under consideration here, a survey was conducted via the Internet
during February and March 2003. A summary of key findings follows, but
for details a copy of the questionnaire and the complete results are found
in Appendix 3. Of 19 European, 35 US and 5 Japanese companies, subsid-
iaries or other industry organizations approached, there was a total of 21
respondents: 9 from Europe, 8 from the U.S. and 4 from Japan. The names
of the specific companies are given in Appendix 3; however, all results
are anonymized as to origin of answers. The topics covered in the survey
included broad organization and policy issues (regarding, e.g., risk mana-
gement, Investigator’s Brochure management) as well as case processing
and data management issues (e.g., causality assessment, study/case blind-
ing, use of AE terms and coding dictionaries, and much more).
Although there has been a tendency to create one safety/pharmaco-
vigilance department covering both pre- and post-approval periods, 8 of 21
respondents reported separate organizations within their companies. It is
clear that “risk management” as a discipline has taken hold, with 20 of 21
companies having incorporated this approach; 7 of 17 responding indicate
that a distinct group is responsible for this area – 4 headed by the safety
department, 3 by clinical development.
Most companies (14/21) regard the signing of informed consent as the
starting point for collecting adverse event information on subjects/patients;
4 other respondents indicated that it was protocol specific. Only 4 of 21
required investigators to record signs and symptoms of an AE/ADR along
with a suggested diagnosis on the study case report form; however, if signs
and symptoms were provided, 13 of 21 code and enter the information in
their database with the diagnosis.
The CIOMS Working Group has deliberated the appropriateness and
practicality of using a single, standard form that all companies would use
globally for investigators to record data on suspected serious adverse events/
reactions; 16 of 21 support this idea. More discussion on this point is found

25

group6_PH.indd 25 7.8.2007 12:19:21

 in Chapter 4. Another Working Group concept involves the introduction
of periodic summary reports of serious suspected ADRs to investigators
and ethics committees as a substitute for sending individual cases as they
arise; this idea was supported by 19 of the 21 respondents. See Chapter 7
for more details. Some regulatory authorities or study sites already request
some sort of periodic report along these lines (e.g., under the European
Clinical Trials Directive or other country-specific requests (UK, Portugal,
Spain), as reported by 5 companies).
Safety information is supplied to ethics committees directly by the
company (9 of 21 responses) and/or to the investigators who then forward
it to the ethics committees (12/21); however, this practice is country-
dependent (11/21) and may vary from study to study (2/21).
“Introspection” was the choice for 12 of 21 with regards to what
causality method is used by the company for assessing whether an AE is an
ADR; 2 use a home-grown algorithm, 3 a published routine (e.g., Karch-
Lasagna), and 4 indicated “no specific method.” Nearly all respondents
(19/21) take the investigator’s causality assessment into account in their
analyses and regulatory reporting of safety cases.
The CIOMS III/V recommendation on the Development Core Safe-
ty Information (DCSI) concept for the Investigator’s Brochure11 has been
(6/21) or will be adopted (7/21) by many of the respondent companies.
Perhaps surprisingly, not all companies conduct regular aggregation
and review of all AE data from ongoing trial results; 16 of 21 do and 5 of
21 reportedly do not.
The results of the survey were helpful to the CIOMS Working Group
in formulating its proposals.

c. Areas Covered by the CIOMS VI Project

The CIOMS VI Working Group has developed proposals based on
scientific principles for harmonizing many aspects of the collection,
monitoring, analysis, evaluation/interpretation, and communication to all
relevant parties of clinical trial safety information. In so doing, it has

11
See Guidelines for Preparing Core Clinical Safety Information on Drugs. Second Edition, Including
Proposals for Investigator’s Brochures. Report of CIOMS Working Groups III and V. Council for International
Organizations of Medical Sciences, Geneva, 1999.

26

group6_PH.indd 26 7.8.2007 12:19:21

 developed an approach to “good clinical trial safety practices” that
embraces an overall safety surveillance/risk management program that
links pre-marketing to post-marketing safety environments.12
Underlying this effort was the need to explore the following specific
areas, and others, which are not adequately addressed in regulations, yet are
the subject of considerable uncertainty and debate:
• Terminology and definitions: How relevant are conventional
safety terms, definitions and categories, largely developed for post-
marketing regulatory reporting purposes, to the analysis and under-
standing of clinical trial safety data (for example, pharmaco-
vigilance; serious, non-serious; adverse event, adverse drug
reaction)? What is meant by a “signal” and “adverse events of
special interest”? Are terms like effectiveness, risk, benefit-risk
relationship clearly delineated and understood? Whose terms and
definitions should be used among those issued by WHO, ICH,
prior CIOMS groups, various regulatory authorities, and others?
The Glossary in Appendix 1 as well as discussions within the
Chapters attempt to answer these questions. The reader is urged to
read the opening section of the Glossary for an important perspec-
tive on term usage, abbreviations and definitions. In addition to
terminology related to pharmacovigilance and drug safety, the
Glossary also covers important statistical terms as used in Chapter 6.
• Ethical aspects of clinical trials: Events of the recent past have led
to changes in various rules and regulations governing proper ethical
practices and behaviour. Revisions of the Declaration of Helsinki,
modified roles and responsibilities of investigators, sponsors, eth-
ics committees, and data and safety management boards (DSMBs),
new privacy and data confidentiality laws, and other considerations
affecting the rights and welfare of study subjects all require an ex-
panded view of the role of ethics concepts in human drug research.
Chapter 2 covers these matters.
• Overall pharmacovigilance/risk management system: Can gen-
eral principles and practical guidance be developed for an overall
product safety system as a basis for the identification, assessment,
and management of potential and real safety issues for the product?
How can it be applied to the transition between a development

12
The reader will be interested in ICH guideline E2E (Pharmacovigilance Planning), that outlines a comprehensive
approach to pharmacovigilance/risk management of newly introduced products. See http://www.ich.org.

27

group6_PH.indd 27 7.8.2007 12:19:22

 program and introduction of a medicine after its marketing authori-
zation? Chapter 3 provides guidance on a pharmacovigilance/risk-
management process that can form the basis for any needed general
or specific pharmacovigilance plans during drug development.
• Collection and proper management of safety data: Does tradi-
tional study protocol language satisfy the needs of safety manage-
ment? When should data collection begin? Can standards be devel-
oped on what should be collected and when? Are there special
issues with regard to the collection and documentation of labora-
tory data? What is the relationship between safety data and clinical
efficacy endpoints (especially involving mortality or increased mor-
bidity)? Who has the responsibility to ensure complete and timely
data collection at the study site? Who is responsible at the sponsor’s
location (especially when work is outsourced to contract organiza-
tions or there are licensing relationships)? Is it feasible and practi-
cal to adopt a global, standard form or set of data elements for use
by investigators to report serious adverse events to sponsors? What
impact should an investigator’s causality assessment for an adverse
event have? How long after a patient withdraws from a study for
safety reasons, or completes a study (or takes the last dose), should
he/she be followed for potential adverse drug reactions or to moni-
tor an existing ADR? Once data are retrieved, what is the appropri-
ate way to ensure the proper choice and coding of AE/ADR terms
to ensure accurate and informative analysis and evaluation? Chap-
ters 4 and 5 deal with these and other issues.
• Evaluation of safety data: Can standard approaches be recom-
mended for the detection, analysis and management of safety sig-
nals? What is the proper place of individual case report assessment
vis-à-vis aggregate data analysis? How should blinded studies be
managed with respect to safety monitoring, reporting and analysis?
Among the various stakeholders, who should have responsibility
for, or participate in, the ongoing analysis and interpretation of ag-
gregate safety data (investigator, sponsor, Ethics Review Commit-
tees (ERCs)13, Institutional Review Boards (IRBs), data and safety
monitoring boards/committees (DSMBs/DSMCs), regulatory au-
thorities)? What factors should determine how often and to what
depth safety data should be analyzed and evaluated during a

13
May also be called Independent Ethics Committees (IECs), or Research Ethics Committees (RECs). See
Chapter 2 for more discussion on the roles and responsibilities of ERCs and IRBs.

28

group6_PH.indd 28 7.8.2007 12:19:22

 development program? What options are available for action based
on the findings, particularly those relating to stopping rules, emer-
gencies, or changes to study protocols? What approaches are needed
to assess the safety experience of special or sub-populations (such
as the elderly, pediatrics,14 organ impaired, women of child bearing
potential)? How precise and relevant are the medical terminology
and definitions that are used to describe AEs/ADRs, including the
use of specific coding dictionaries? What influence do they have
on the information to be included in the Investigator’s Brochure
(IB) and the eventual authorized product information (data sheets)?
Chapter 5 covers these topics in some depth.
• Statistical analysis of safety data: What is the appropriate use for
inferential and descriptive statistics and when should they be used?
Should “intention-to-treat” analyses be applied to safety data?
What impact do statistical power, multiplicity (multiple analyses)
and time dependency have on analysis and interpretation of the data
from individual trials? Is one-sided or two-sided testing preferred?
What are the correct approaches to analysis of continuous data
(e.g., laboratory chemistries) vs binary data (e.g., present/absent)?
Are survival analysis techniques (accounting for time on drug and
discontinuations) important and if so, when? How can meta-
analytic approaches be used to pool data from multiple studies?
How can background data from various sources outside the trials be
used for comparison of results? What are the best ways to express
risk information for healthcare providers and patients? Chapter 6
provides details and guides to these and other statistical issues.
• Regulatory reporting and communication to others of safety
information during clinical trials: Is there sufficient consistency
between different countries’ regulations to allow for standard global
practices by industry? How do recent changes (e.g., the European
Directive on Clinical Trials15) affect the monitoring, handling and
reporting of safety data? What should be communicated to inves-
tigators, ERCs, IRBs, DSMBs, and ultimately study subjects, not

14
The EMEA issued a concept paper as evolving guidance in March 2003 on pharmacovigilance in children
(http://www.emea.eu.int/pdfs/human/phvwp/483802en.pdf). Although it pertains directly to post-marketing
conditions, it could serve as a useful reference for clinical trials as well.
15
See http://europa.eu.int/eur-lex/en/oj/index.html for the Directive and http://pharmacos.eudra.org/F2/
pharmacos/new.htm for the associated Guidances, two of which relate directly to pharmacovigilance during
clinical trials.

29

group6_PH.indd 29 7.8.2007 12:19:23

 only as a trial progresses or at its completion, but prior to its ini-
tiation? When should the information be communicated? Whose
responsibility is it to communicate such information to the various
affected/involved parties? Chapter 7 addresses these issues in detail
and makes some recommendations for new approaches.
It is important that we rely on a more comprehensive and transparent
approach to risk management than in the past; our understanding of a
product’s safety profile evolves throughout its study and use. This CIOMS
report covers all the issues discussed above (and more) in an attempt to
provide practical guidance for the design and execution of a rational drug
safety surveillance plan during any clinical research program. It is directed
not only to pharmacovigilance/clinical safety specialists, but to all those
involved in the planning, design, and execution of the clinical research
process for the development of new medicines, as well as new uses and
preparations of already available products.
Another aspect of product safety that deserves careful consideration is
the possibility of medication errors – mistakes made in the prescribing, dis-
pensing, administration, and use of medicines – which can lead to adverse
reactions, sometimes serious. Although not usually associated with clinical
trials, there have been instances of such errors during development programs.
In addition, it behooves a sponsor to try to anticipate what kinds of errors
might occur once the product reaches the general population and to take steps
to minimize their possibility (e.g., avoid possible name, appearance, and
packaging similarity/confusion with other products).16 Similarly, it would be
prudent for a sponsor to attempt to foresee what if any off-label (unapproved)
uses might be made of the medicine once it is in general use; different or
unusual safety considerations might pertain in such circumstances. For a dis-
cussion and some recommendations on these issues, see Chapter 3.

d. Limitations of Clinical Trials

for Understanding Safety
The development process for virtually all new medicines represents a
compromise between two extremes: (1) acquiring a minimum, basic data
set on a drug’s properties in animals and humans, and (2) the desire to
learn as much as possible about a product’s safety (and efficacy) prior to its
16
For a set of definitions and a taxonomy of medication errors, see www.nccmerp.org (the US National Coordi-
nating Council for Medication Error Reporting and Prevention (NCCMERP)).

30

group6_PH.indd 30 7.8.2007 12:19:23

 approval and general use. In order to enable the introduction of new medicines
within a reasonable time and at acceptable cost, the regulatory and scientific
requirements must be practically achievable. As a result, there are limitations
of typical clinical development programs which are familiar and include:
small numbers and homogeneity of study subjects relative to the much larger
and diverse population that may use the product; statistical aspects of study
designs focus on efficacy (power calculations, etc.) rather than on safety; a
controlled, experimental environment that may not reflect the “real world”
(concomitant treatments, number of treatment visits, extent of intervention
and measurements, concurrent conditions, etc.); uncertain generalizability of
the data;17 and a relatively short duration of treatment (e.g., latent effects may
not be observable). With certain categories of medicines, such as anti-HIV
drugs, there may be additional pressure to shorten development time to satisfy
urgent public health needs, in which case our knowledge and understanding of
the safety profile will be even less complete and will be confounded or made
more complicated, especially by polypharmacy and the use of fixed combina-
tion products. In principle, randomization during clinical trials will mitigate
such confounders, unlike the situation in most post-marketing trials and ob-
servational studies.
Compliance with ICH GCP Guideline (E6), the most widespread stan-
dard in use for the conduct of clinical trials, provides assurance that the rights,
safety and well being of trial subjects are protected, and that the trial data are
credible. General principles and guidance are given on the roles and respon-
sibilities of sponsors and investigators for collection and reporting of safety
information. Another ICH Guideline, General Considerations for Clinical Tri-
als (E8), summarizes key principles and practices that govern scientific excel-
lence and explains the connection between the various ICH clinical guide-
lines. However, the field has become increasingly complex and many aspects
require renewed attention in spite of such widely agreed standards and the
availability of published treatises on pharmaceutical clinical research.18 While
it is believed that companies strive to establish internal global standards for
the collection, monitoring, processing, analysis, assessment, presentation and
reporting of safety data, some new thinking and practices are deemed advis-
able by the Working Group.

17
For a recent analysis of clinical trial vs “real world” ADR profiles that highlights the problems with attempting
to generalize results from trials, see Dieppe, P., Bartlett, C., Davey, P., Doyal, L., and Shah, E. Balancing benefits
and harms: the example of non-steroidal anti-inflammatory drugs, British Medical Journal, 329:31-34, 2004.
18
For example, see Guide to Clinical Trials by B. Spilker. Lippincott Williams & Wilkins, Philadelphia, 1991 and
Handbook of Phase I and II Clinical Drug Trials, Edited by J. O’Grady and P. H. Joubert. CRC Press, LLC,
1997.

31

group6_PH.indd 31 7.8.2007 12:19:23

 e. Scope of the Project
The concepts and proposals developed here are applicable to prescrip-
tion drugs, biotechnology products, diagnostic agents and over-the-counter
(non-prescription) products, as well as prophylactic therapies. The focus of
this work is on new product development programs, conventionally Phase
I through III trials, but it also is relevant for Phase IV trials (generally
regarded as post-authorization therapeutic use studies).19 This work also
applies to programs involving the use of pharmacogenetics. An added com-
plication for biological and biotechnology-derived products is their greater
sensitivity to quality (manufacturing) issues; the presence of foreign anti-
gens, specific DNA content or DNA contamination, pyrogens, and/or viral
contaminants all can play a crucial role in establishing a safety profile.
Such technical details are beyond the scope of the present work.
Although gene-therapy research and programs involving genetically
modified organisms are still highly exploratory and may be somewhat contro-
versial, we believe that the guidance provided in this report can be applied
to these areas. However, evolving knowledge pertaining to the underlying
science and potential quality issues must be taken into account.
The material here is also applicable to prophylactic and therapeutic
vaccines, although each represents a somewhat special situation, with the
former involving potential major public health implications. The testing
of huge populations is often involved, increasingly with a requirement to
assess immunological markers. In many cases, most of the confirmatory
research will take place in a post-authorization environment. The fact that
most vaccine programs are directed at infants and children heightens the
sensitivity to ethical considerations and informed consent.
This report does not address the increasingly frequent attempt under
some regulatory jurisdictions to include cost effectiveness of products when
prescribed within defined clinical situations. This concept involves many
perspectives and controversies within the regulation of medicines and may
be seen as a sociopolitical or economic rather than (or in addition to) a
scientific issue. Approaches to benefit-risk-cost analyses and decisions are
still in their infancy.

19
ICH Guideline E8 (General Considerations for Clinical Trials) has proposed that studies be categorized ac-
cording to their objectives (human pharmacology, therapeutic exploratory, therapeutic confirmatory, and
therapeutic use), as distinct from the temporal phases of drug development (I through IV). For example, hu-
man pharmacology studies (traditionally referred to as Phase I) can be and often are conducted throughout a
product’s lifetime.

32

group6_PH.indd 32 7.8.2007 12:19:24

 Although this work deals with medicinal products, it is believed that
the principles and practices invoked can apply to medical devices as well,
although it is recognized that there are some special issues associated with
their study and use.
This effort is directed not only at pharmaceutical companies as clini-
cal trial sponsors along with their agents (contract research organizations,
CROs), but also to independent clinical researchers and others not involved
in commercially-based medicines development, since the pursuit of en-
hanced safety standards is principally concerned with the protection of pa-
tients. It is hoped that this report will be read and used by academic clinical
researchers who are as important as any other stakeholders in the conduct
of clinical trials.
Finally, it is important to acknowledge that the risks of a drug cannot
and should not be considered in isolation from the established benefits. In
any development program, the ultimate goal is to evaluate and provide a
measure of the benefit-risk relationship for the anticipated conditions of
use, something that is critical to its approval and use in the general popula-
tion. A rational assessment of the benefit-risk (or benefit-harm) relation-
ship is notoriously difficult, whether done by regulators for populations or
by patients and healthcare professionals for individuals. There are many
potential biases and influences that affect decisions surrounding the rela-
tionship, decisions that are usually based on what is often referred to as
“subjective expected utility theory”.20 This report only indirectly addresses
the benefit side of the relationship and does not deal in a major way with the
evolving methodologies for qualitative and quantitative aspects of benefit-
risk weighing.21,22 No matter what method is used to derive and describe the
benefit-risk relationship for a specific product, it must be recognized that

20
For a thoughtful treatment of this subject with suggestions on how to understand why different stakeholders
interpret the benefit-harm balance of medicines differently, and how to form a basis for strategies to counter
cognitive and other influences, see: Greenhalgh, T., Kostopoulou, O., and Harries, C. Making decisions about
benefits and harms of medicines, British Medical Journal, 329:47-50, 2004.
21
For more details on benefit-risk considerations, see: Spilker, B., Incorporating Benefit-to-Risk Determinations
in Medicine Development, Drug News and Perspective, 7 (1), February 1994, 53-59; Chuang-Stein, C. A.,
New Proposal for Benefit-less-Risk Analysis in Clinical Trials, Controlled Clinical Trials, 15: 30-43, 1994;
and “Benefit-Risk Balance for Marketed Drugs: Evaluating Safety Signals,” CIOMS, Geneva, 1998. Although
this last reference focuses on marketed drugs, the principles and process described are valid for drugs in de-
velopment.
22
Holden, W.L., Juhaeri, J. and Dai, W. Benefit-risk analysis: a proposal using quantitative methods, Pharmaco-
epidemiology and Drug Safety, 12:611-616, 2003, and idem, Benefit-risk analysis: examples using quantita-
tive methods, ibid., 12: 693-697, 2003. Also, see Eriksen, S. and Keller, L. R. A Multiattribute-utility-function
Approach to Weighing the Risks and Benefits of Pharmaceutical Agents, Medical Decision Making, 13:2,
April-June 1993, 118-125.

33

group6_PH.indd 33 7.8.2007 12:19:24

 over time, especially once a product is in general use, it can change for the
better or worse. As new and improved products are authorized for clinical
use, subsequently developed products in similar therapeutic classes will
need to meet increasingly stringent benefit-risk requirements, which can
have a significant impact on new development programs.

34

group6_PH.indd 34 7.8.2007 12:19:25

 II

Ethical Considerations
for Clinical Trial
Safety Management

group6_PH.indd 35 7.8.2007 12:19:25

group6_PH.indd 36 7.8.2007 12:19:25
 a. Background
Most countries or regions have rules and regulations on safety sur-
veillance during clinical trials that address the responsibilities of sponsors,
investigators and ethics committees but the details are continuously evolv-
ing. The ethical underpinnings of all the regulations are based on several
regional and international guides that set out principles of research on hu-
mans. The most widely known and applied is the Declaration of Helsinki,1
which is incorporated or referenced within most countries’ regulations.
However, there are several other valuable works that can be consulted and
that have had an impact on standards in this area. The CIOMS Interna-
tional Ethical Guidelines2 provide guidance on how the ethical principles
of the Declaration of Helsinki can be applied effectively. The topic has
also been receiving increased attention in specific parts of the world, such
as Latin America3 and in developing countries.4 It should also be noted
that the Council of Europe has been preparing a “Protocol for Biomedical
Research,” a comprehensive regulation intended to be legally binding and
that Member States must ratify.5 Independently, UNESCO plans to develop
a “Universal Instrument on Bioethics,” to include a section on biomedical/
clinical research.6
For anyone designing and conducting a clinical trial, the fundamen-
tal principle should be that any study that is not scientifically sound can
be considered unethical. The basic ethical principles universally accepted
for dealing with the potential risks and benefits for human subjects are:
autonomy of the individual (respect for persons and their dignity), benefi-
cence (do good), nonmaleficence (“do no harm”), and justice (benefits and
burdens of research distributed fairly among all groups and classes).7 These

1
For the latest edition, see Appendix 4 or www.wma.net/ethicsunit/DeclarationofHelsinki
2
International Ethical Guidelines for Biomedical Research Involving Human Subjects, CIOMS, Geneva, 2002
(a guide to the application of the Declaration of Helsinki, particularly for research in developing countries;
available in several languages, including Japanese, Spanish, Italian, German, etc.); also, see Ethical Consid-
erations in Clinical Trials, Proceedings of an EMEA Workshop, 26 November 2001 (www.emea.eu.int).
3
Cavazos, N., Forster, D., Orive, O., Kaltwasser, G. and Bowen, A. J. The Cultural Framework for the Ethical
Review of Clinical Research in Latin America, Drug Information Journal, 36:727-737, 2002.
4
Zumla, A. and Costello, A. Ethics of Healthcare Research in Developing Countries, J. Roy. Soc. Med., 95:275-
276, 2002.
5
See www.coe.int/T/E/legal-affairs/legal_co-operation/bioethics.
6
For an interim report, see www.unesco.org/ibc/en/actes/s10/index.
7
The National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research. The
Belmont Report. Ethical Principles and Guidelines for the Protection of Human Subjects of Research. US
Department of Health, Education and Welfare, Washington, DC, 1979 (see http://ohsr.od.nih.gov/mpa/belmont.
php3).

37

group6_PH.indd 37 7.8.2007 12:19:25

 principles apply to actions by researchers and investigators, which have a
direct impact on the potential risks and benefits experienced by patients and
normal volunteers. Neutel8 has examined in detail the complicated issues
surrounding the ability of investigators to apply the principles of benefi-
cence and nonmaleficence so as to effectively minimize risk and maximize
benefit; improvements in the informed consent process play a key role.
The purpose of this Chapter is to discuss current thinking and regula-
tions associated with ethical aspects of clinical medicinal research, with
particular focus on clinical safety. This is a complicated and culturally de-
pendent topic and examination of clinical research ethics is an active pro-
cess by different groups.9 Thus, the Working Group believes it is currently
beyond its scope to make any firm proposals on possibly controversial mat-
ters of ethics. However, in a few areas it does provide recommendations and
provides information and ideas it hopes will inform ongoing debates.
New laws or regulations and new perspectives have expanded the ap-
plication of ethical concepts in clinical trials beyond the usual themes of
informed consent and indemnification (insurance) against study subject in-
jury. There is growing importance and sensitivity not only for patient rights
generally, but for clinical trials in non-industrialized, developing countries,10
vulnerable and socially underprivileged patients,11 transparency (including
on payments to investigators and to trial subjects), and the availability of
results of all trials, including those with “negative” findings.
Conflicts of interest (professional as well as financial) also represent a
source of concern. Potential conflicts of interest with respect to clinical tri-
als may compromise the integrity of the research and human research par-
ticipant protection, and therefore must be considered carefully. They may
involve the institution, the investigator, and independent ethics committees
(IECs) and their individual members. In addition to the obvious conflicts
such as financial benefits that might accrue to individuals and/or institu-
tions, there are subtle influences, such as professional recognition and pro-
motion. For example, when ethics committees are constituted within the

8
Neutel, C. I. The Dilemma of Using Humans as Research Subjects: An Assessment of Risks and Benefits,
Drug Information Journal, 38:113-126, 2004.
9
For example, see Emanuel, E. J. et al. Oversight of Human Participant Research: Identifying Problems to
Evaluate Reform Proposals, Ann. Intern. Med., 141: 282, 2004.
10
Idanpaan-Heikkila, J. E. Ethical principles for the guidance of physicians in medical research – the Declara-
tion of Helsinki, Bulletin of the World Health Organization, 79(4):279, 2001.
11
Vrhovac, B. Chapter 2. Ethical Considerations in Basic Guidelines for Pharmacological Research in Humans,
IUPHAR, August 2004 (Brisbane, Australia).

38

group6_PH.indd 38 7.8.2007 12:19:26

 institution that is scheduled to conduct the research, they must be careful
not to let the often large amounts of funding the institution might receive
influence their review and approval of the protocol and the subject protec-
tion measures.12 There is increased scrutiny of the potential undue influence
of monetary compensation to investigators and patients, something that has
to be managed carefully.13
Transparency with all affected parties is paramount when faced with
any potential conflicts of interest. Each institution, investigator, IEC/IRB
and its members should have a conflict of interest policy that ensures inde-
pendence from undue external influence of any kind that could cast doubt
on their ability to make unbiased decisions and fulfill their mandate to
protect the rights, safety and welfare of human research participants. This
would entail the clear separation of the approval, audit and oversight func-
tions of the IEC/IRB from the operational functions of the institution and
investigator(s), i.e., those involved with funding, initiating and conducting
the research, and who might stand to gain from its positive outcome. The
IEC/IRB and its members should be unaffiliated with both the research
sponsor and the trial subjects.
It is useful to consider the broad subject of ethics in the current context
under two headings: stakeholder roles and responsibilities, and regulatory
considerations.

b. The Stakeholders
The monitoring and management of pre-authorization clinical safety
data involve many parties, with their own perspectives and expectations, as
well as roles and responsibilities.
• Patients. Their willingness to accept risk is based on their percep-
tion of safety (“Is this study safe?”) and is tempered with expecta-
tions of a favorable safety and, aside from normal volunteers, effica-
cy outcome. They should always be regarded as full partners in the
research and thereby be kept well informed so that they understand
their role and importance. In this way, patients/volunteers will be in

12
For a detailed discussion and suggestions for dealing with conflicts of interest, see Responsible Research: A
Systems Approach to Protecting Research Participants, Institute of Medicine (US), Washington, DC, October
2002 (http://www.iom.edu/report.asp?id=4459).
13
See Reiser, S.J. Research Compensation and the Monetarization of Medicine, J. Am. Med. Assoc., 293:613-
615, 2005.

39

group6_PH.indd 39 7.8.2007 12:19:26

 a better position to make decisions regarding their participation and
continuation in a study, and can enhance their willingness to adhere
to all protocol requirements. This makes it all the more important
that they be given as much information as possible in a way that
maximizes comprehension, through the informed consent process
and throughout a study.14 These issues represent a significant chal-
lenge to sponsors, investigators and ethics committees.15 There has
been evidence for some time that consent documents in use are too
long and too difficult to read by many patients.16,17 This aspect is
beyond the scope of the CIOMS VI project. Patient privacy and
the confidentiality of their data are also of considerable importance
(see part c. of this chapter for more details).
• Regulatory Authorities and the Public Health Community. Gov-
ernments through their regulatory bodies18 have a statutory respon-
sibility to authorize the use of medicines only if they can be dem-
onstrated to be “safe and effective” and when they have the required
manufacturing quality. They develop the regulatory framework
to ensure that their scientific evaluation is based on reliable data
obtained from well conceived and conducted clinical trials. The
regulatory and public health communities protect the public through
their ongoing monitoring of the safety of a wide range of both ex-
perimental and authorized medicines. During clinical development
programs, the authorities have several options available to protect
trial subjects and ensure the scientific quality and integrity of the re-
search – from routine monitoring and audits to “clinical holds” (tem-
porary cessation/suspension of one or more trials), mandatory proto-
col and/or informed consent adjustments, periodic safety assessment
reports from sponsors, complete discontinuation of the program, and
other mechanisms. Once the products are authorized and used in the

14
See Guideline 5, “Obtaining informed consent: essential information for prospective research subjects,” in In-
ternational Ethical Guidelines for Biomedical Research Involving Human Subjects, CIOMS, Geneva, 2002.
15
For example, see Hochhauser, M. Why You Can’t Write a Consent Form at a Sixth-grade Reading Level, DIA
Forum, October 2002, p. 22-25 and DeMilto, L. Working with Institutional Review Boards and Informed
Consent, ibid., July 2002, p. 16-20.
16
Sharp, S. M. The Problem of Readability of Informed Consent Documents for Clinical Trials of Investiga-
tional Drugs and Devices: United States Considerations, Drug Information Journal, 38:353-359, 2004.
17
Recently, the US Office for Human Research Protection (OHRP) has provided guidelines and sample docu-
ments for obtaining and documenting informed consent for non-English speaking subjects. Go to http://www.
hhs.gov/ohrp/policy/index.htm/#informed for links to the information.
18
For example, U.S. Food and Drug Administration (FDA), U.K. Medicines and Healthcare Products Regula-
tory Agency (MHRA), Japan Ministry of Health, Labor and Welfare (MHLW), European Medicines Agency
(EMA).

40

group6_PH.indd 40 7.8.2007 12:19:27

 general population, the authorities must continue to monitor product
use to ensure that an acceptable balance between benefits and risks is
sustained for each authorized indication and sub-population.
• Investigators. Whether they are independent researchers or are con-
ducting trials on behalf of a company or other sponsor, the investi-
gator and his/her staff play the most important role in ensuring the
rights and safety of subjects and in the collection of complete, accu-
rate, protocol-required data. They are also pivotal in establishing and
maintaining effective communications with ethics and data and safety
monitoring committees, sponsors, and when required, with the health
authorities.
• IECs and IRBs.19 Responsibilities and membership for these bodies
are generally articulated in national and international regulations for
clinical research.20 However, new issues have arisen that affect their
governance and their roles. It has become increasingly difficult for
clinical research ethics committees to cope with ethical and technical
complexities involving, for example, use of placebos, equivalence vs
non-inferiority trials, use of the appropriate comparative agent, proper
dosing, and therapeutic endpoints.21 Increased scrutiny of IECs/IRBs by
government and public groups has also emerged as a result of serious
injury or death to some trial patients over the past few years and efforts
are being made to strengthen subject safety, including by legislation.22

19
In common usage, the names IRB (Institutional Review Board), ERC (Ethics Review Committee) and In-
dependent Ethics Committee (IEC) generally represent the same or similar bodies which are expected to
have the expertise to maintain study oversight and protection of the subjects. However, in some cultures and
institutions, ethics committees and review boards may interpret their roles differently and perform different
functions. IRBs are usually limited to an institution as the name implies, but are expected to be capable of re-
viewing and approving study protocols. For example, some may include statisticians and trial methodologists
to ensure that studies will obtain data of value; however, IECs focus on assuring that patients are not exposed
to undue risk and may not have scientific expertise. Furthermore, an ethics committee may be responsible for
all ethical issues within an institution, including issues related to clinical trials, whereas a separate IRB might
be established within the same institution for a specific purpose, especially for oversight of clinical research.
Because of the use of “centralized” IRBs or IECs for multi-site studies, some institutions have redefined and
separated the local boards’ roles and responsibilities (e.g., data privacy, animal research review, ethical con-
siderations). See the Glossary (Appendix 1) for more discussion.
20
For example, see Article 6 in the EU Directive on good clinical practice and conduct of trials (http://europa.eu.int/
eur-lex/en/oj/index.html). Official Journal of the European Communities, 1 May 2001, L 121/34 – L 121/44.
21
Garattini, S., Bertele, V. and Li Bassi, L. How can research ethics committees protect patients better?,
British Medical Journal, 326:1199-1201, 2003.
22
For example, see (a) Responsible Research: A Systems Approach to Protecting Research Participants, Insti-
tute of Medicine (US), National Academies Press, Washington, DC, 2003 (see http://www.iom.edu/report.
asp?id=4459). (b) The Globalization of Clinical Trials: A Growing Challenge in Protecting Human Subjects,
US Department of Health and Human Services Office of Inspector General, September 2001 (Report Number
OEI-01-00-00190, available at http://oig.hhs.gov.oei) and (c) Bailey, V. J., New Directions for IRBs, Food and
Drug Law Institute Update, November/December 2001 (www.fdli.org).

41

group6_PH.indd 41 7.8.2007 12:19:27

 Steps have also been taken to introduce accreditation of IECs and
IRBs.23 In the US, a new accreditation program (Partnership for
Human Research Protection, Inc, PHRP) began in 2003.24 Another
new organization, the Association for the Accreditation of Human
Research Protection Programs (AAHRPP) in the US has also cre-
ated an accreditation system.25 Its mission is to provide a process
of voluntary peer review and education among institutions, IRBs,
and investigators concerned with research involving humans, in
order to promote preservation of the rights and welfare of subjects
in research, and compliance with relevant ethical and regulatory
standards.
Other developments include the use of centralized IRBs, whereby
local site IECs/IRBs would accept a review of a multi-center trial
from an authorized (preferably accredited) body in lieu of indi-
vidual reviews by each local group.26 Under the EU Clinical Trial
Directive, each Member State must establish a mechanism for a
single opinion on approving clinical trials within that country.
Guidelines have been developed for auditing of ethics committees
in Europe.27 A practical proposal has been published for an inves-
tigator’s checklist to ensure proper IEC/IRB review of the protocol
and subject protection mechanisms for each study.28 Finally, there

23
Preserving Public Trust: Accreditation and Human Research Participant Protection Programs, National
Academy of Sciences, Washington, DC, 2001 (see www.nap.edu). Also, see the EU Clinical Trial Directive
approach to accreditation: http://pharmacos.eudra.org/F2/pharmacos/docs/Doc2004
24
This is a collaborative effort by JCAHO (Joint Commission on Accreditation of Healthcare Organizations)
and the NCQA (National Committee for Quality Assurance). For details, see www.phrp.org.
25
The overall goal of accreditation is to improve protection of human research subjects by developing per-
formance standards that encourage programs to adopt “best practices” in this area, and by recognizing the
programs that meet those standards. See www.aahrpp.org/ for details.
26
A few examples of centralized and independent IRB models operating in the US include: MACRO (Multi-
center Academic Clinical Research Organization), a reciprocal IRB approval process for several academic
medical centers (www.ccs.wustl.edu); WIRB (Western Institutional Review Board), www.wirb.com, which
offers international ethics review services; CIRB (Consortium of Independent IRBs), a group associated with
the US National Cancer Institute (www.ncicirb.org); Midlands L.L.C. IRB, which has the ability to review
studies in all States of the US (see www.midlandsirb.com); Coast Independent Review Board (www.coastirb.
com); The Copernicus Group (www.copernicusgroup.com). For the UK, see Multicentre Research Ethics
Committee (MREC) and Local Research Ethics Committee (LREC) requirements (www.corec.org.uk and
www.eric-on-line.co.uk/index.php). Also, see www.irb-irc.net for information on Independent Review Con-
sulting, an organization that provides IRB services and ethics review consultation.
27
See “European Guidelines for Auditing Independent Ethics Committees,” European Federation for Good
Clinical Practice (EFGCP) at www.efgcp.org.
28
Spilker, B. Creating an IRB Checklist to Protect Human Subjects in Clinical Trials, Applied Clinical Trials,
September 2002, p. 34-36.

42

group6_PH.indd 42 7.8.2007 12:19:28

 are associations dedicated to clinical research ethics that publish
and hold conferences on the roles of IRBs and IECs.29
• Data and Safety Monitoring Boards.30 The use of data monitoring
boards for randomized clinical trials has increased in recent years
and different approaches have been taken as to their responsibili-
ties and interactions with other stakeholders. Such committees can
play a critical role in the new drug development process, and their
regulatory status is changing.31 There is some debate, from a scien-
tific as well as an ethical perspective,32 regarding whether or when
DSMBs should have access to unblinded efficacy and safety data,
and with whom such data should be shared. Many of the under-
lying principles of such committees, their roles and responsibilities,
and levels of access to blinded data have been established by some
regulatory bodies. The FDA has issued a Draft Guidance for spon-
sors on the operations of clinical trial monitoring committees.33
The EU Pharmacovigilance Guidelines for the Clinical Trials Di-
rective address such issues as well.34 The WHO through its Special
Program for Research and Training in Tropical Diseases has also
created a draft operational guideline for DSMBs.35 The underlying
challenge for such boards/committees is to seek the proper balance
between maximizing the scientific value and validity of trials, and
their obligation to protect participating and future patients.36 Under

29
PRIM&R (Public Responsibility in Medicine and Research, www.primr.org); ARENA (Applied Research
Ethics National Association, www.aamc.org/research/primr/arena); ACRP (Association of Clinical Research
Professionals), a very large membership of research professionals (www.acrpnet.org).
30
An independent body with oversight for the monitoring and assessment of data from clinical trials to protect
study participants and to protect the validity and credibility of the trial. They may be referred to variously as
data monitoring boards or committees (DMBs, DMCs), data and safety (monitoring) boards (DSMBs), and
other terms. See the Glossary (Appendix 1) under Independent Data-Monitoring Committee for more discus-
sion, and Appendix 5 for a detailed description.
31
Ellenberg, SS, Fleming, TR and DeMets, DL. Data Monitoring Committees in Clinical Trials: A Practical
Perspective, John Wiley (Chichester, England), 2002. Also, see Ellenberg, S.S. Independent Data Monitoring
Committees: Rationale, Operations and Controversies, Statistics in Medicine, 20: 2573-2583, 2001.
32
Fleming, T.R., Ellenberg, S., and DeMets, D.L. Monitoring Clinical Trials: Issues and Controversies Regard-
ing Confidentiality, Statistics in Medicine, 21: 2843-2851, 2002.
33
Guidance for Clinical Trial Sponsors On the Establishment and Operation of Clinical Trial Data Monitoring
Committees, November 2001 (www.fda.gov/cber/guidelines.htm).
34
See ENTR/6422/01 at http://pharmacos.eudra.org/F2/pharmacos/dir200120ec.htm
35
Operational Guidelines for the Establishment and Functioning of Data and Safety Monitoring Boards,
UNICEF/UNDP/World Bank/Who Special Program for Research and Training in Tropical Diseases (TDR),
31 March 2004 draft (WHO, Geneva). For details, write to Dr. Juntra Karbwang, the clinical coordinator of
TDR, at [email protected].
36
Slutsky, A. S. and Lavery, J. V. Data Safety and Monitoring Boards, New England Journal of Medicine,
350:1143-1147, 2004.

43

group6_PH.indd 43 7.8.2007 12:19:28

 the auspices of the DAMOCLES project37 a comprehensive review
has been published on the use of DSMBs with recommendations as
to best practices.
• Pharmaceutical Companies and Their Representatives. It is
incumbent on companies and their contractual partners (CROs,
laboratories, licensors/licensees) to work with investigators to en-
sure that all needed steps will be taken to ensure that trials will
be conducted under the best scientific and ethical conditions so as
to maximize the quality of the work and to minimize the risk to
subjects – all while adhering to local and international regulations.
Multinational sponsors should strive for implementation of global
safety standards for their clinical trial practices and operations, in-
cluding the assurance that all study protocols adequately address
safety surveillance and reporting.
In addition to the above stakeholders, other participants in the clinical
research process, namely, statisticians and epidemiologists,38 have devel-
oped their own ethical and related guidelines. Although not directly related
to the clinical trial process, journalists for the professional and lay media
also have an ethical obligation to provide accurate and balanced reports on
information available to them on clinical research results.

c. Evolving Regulatory and Societal Demands

The ethical, technical and administrative requirements in regulations
governing the conduct of clinical trials are many and complicated and may
depend on the country or region where trials are conducted. A detailed dis-
cussion is beyond the scope of this project, but access to information on
the continuously changing picture is available.39 One of the more important
requirements relates to the expedited reporting to regulators of medically
important (serious) adverse events during clinical trials. Periodic (status)

37
Data Monitoring Committees: Lessons, Ethics and Statistics (DAMOCLES). For details, see Sydes, M. R.
et al., Systematic qualitative review of the literature on data monitoring committees for randomized controlled
trials, Clinical Trials, 1:60-79, 2004.
38
For example: Ethical Guidelines for Statistical Practice, American Statistical Association (for information,
[email protected]), and Guidelines for Good Epidemiology Practices for Drug, Device, and
Vaccine Research in the United States, Pharmacoepidemiology and Drug Safety, 5:333-338, 1996.
39
A useful source of current regulations and guidelines for many countries throughout the world is found at:
www.regsource.com

44

group6_PH.indd 44 7.8.2007 12:19:29

 reports during development programs are also required in some countries,
such as under US IND Rules and the EU Clinical Trials Directive.40

(1) Privacy and Confidentiality of Personal Data

Over the past several years, there has been considerable attention paid to
confidentiality and the protection of personal data. New laws and regula-
tions introducing increased subject data rights and data safeguards have
been mandated in the EU and its Member States, the US, Canada, Aus-
tralia, Japan, Argentina, and several other countries, all of which have an
impact on the collection, access to, and handling of personal data from
clinical trials, as well as the ability to transfer such data outside the source
country. 41 The increasing use of pharmacogenetics and DNA typing of
tissue samples in clinical research programs represent an especially sen-
sitive area.42 Some analyses of these laws and their impact on clinical
research and pharmacovigilance have been published.43 Adherence by in-
vestigator sites to the new provisions may fall under the scrutiny of IRBs
and IECs, thereby increasing their responsibilities; however, this is an
evolving area and no international standards have been established.44

40
For US IND Rules, see 21CFR312 (www.fda.gov) and for the EU CT Directive, see Directive 2001/20/EC of
the European Parliament and of the Council (4 April 2001), Official Journal of the European Communities,
L121/34, 1 May 2001 (http://europa.eu.int/eur-lex/en/oj/index.html). Implementation guidelines for the CT
Directive are available on the European Commission website at http://pharmacos.eudra.org/F2/pharmacos/
docs.htm#news.
41
For example, see European Parliament and the Council of the European Union “Directive on the Protection of
Individuals with Regard to the Processing of Personal Data and on the Free Movement of Such Data,” (Direc-
tive 95/46/EC), Official Journal of the European Communities, No. L 281, 31-50 (November 23, 1995). Also
available on the Internet at: http://europa.eu.int/eur-lex/en/lif/dat/1995/en_395L0046.html. The Directive has
been transposed into local law within the Member States of the European Economic Area. In the US, the De-
partment of Health and Human Services (DHHS) released its final rule on Standards for Privacy of Individu-
ally Identifiable Health Information on 20 December 2000 for implementation in April 2003; see http://www.
hhs.gov/ocr/hipaa.html.
42
Anderson, D.C. et al. Elements of informed consent for pharmacogenetic research; perspective of the
pharmacogenetics working group, The Pharmacogenomics Journal, 2:284-292, 2002. Also, see Pharmaco-
genetics – Towards Improving Treatment with Medicines. Report of a CIOMS Working Group, CIOMS,
Geneva, 2005.
43
M. Barnes and J. Kulynych, HIPAA and Human Subject Research: A Question-and-Answer Reference Guide,
Barnett International, Media, PA, 2003 (for information, see www.barnettinternational.com/edu-pubs.cfm);
The Effect of the New Federal Medical-Privacy Rule on Research, N. Eng. J. of Med., 346:201-220, 2002; and
Knudsen, L. E., Theilade, M. D., Gordon, A., Mascaro, J. and Bruppacher, R. Will Data Privacy Impact Health
Research?, Drug Information Journal, 36:465-4809, 2002.
44
National standards have been defined in the UK and were of a statutory nature from 1 May 2004 under the
new UKECA (United Kingdom Ethics Committee Authority). UKECA will authorize, inspect and certify
standards for all research ethics committees in the UK. See “MRC Ethics Series: Human Tissue and Biologi-
cal Samples for Use in Research ” (April 2001) at http://www.mrc.ac.uk. In the US, the Office of Human
Research Protection (OHRP) has issued a Guidance on Research Involving Coded Private Information or
Biological Specimens (August 10, 2004) which deals with the anonymization of data (http://www.hhs.gov/
ohrp/humansubjects/guidance/cdebiol.pdf).

45

group6_PH.indd 45 7.8.2007 12:19:29

 (2) Informed Consent
Although gaining informed consent is the cornerstone of all human
subject clinical research, there are situations where it may not be
possible or appropriate. This raises a dilemma: is a trial unethical if
informed consent was not obtained in advance?45 There are settings
which do justify such exceptions, including use of anonymized tissue
samples, some types of epidemiological research, and certain kinds of
survey research (to avoid biased results).46 For example, observational
studies rarely require informed consent, and it would be highly im-
practical if not impossible if it were needed.
Conducting clinical trials on medicines for emergency-case patients
(in ambulance or hospital emergency room) represents a circumstance
in which the patient is rendered incapable of providing informed con-
sent, and a legally authorized patient-representative is often not avail-
able. Examples are many, such as acute MI, stroke, sepsis, grand mal
seizure, accident trauma, and alcohol and related intoxications by
poisons, resulting in the need for emergency treatment. The available
guidances for such trial situations call for prior approval of the proto-
col by an ethics committee as usual, and inclusion in the protocol of
the detailed reason(s) for the inability to secure informed consent, as
well as details on how consent to remain in the study will be obtained
as soon as possible from the trial subject, or if not possible, from a
family member or a legally authorized representative.47
A topic that must be considered carefully by all parties conducting
clinical research relates to the need for “re-consenting” of trial sub-
jects. Under what circumstances and how should new, important safety
information be conveyed to trial subjects who have already given their
informed consent to participate? Situations that must be considered
include subjects who are still in the trial, those who are in a post-
treatment follow-up period, and those who have completed the trial.
There are several factors that must be considered in deciding what the

45
Dogal, L. Informed Consent in medical research: journals should not publish research in which patients have
not given fully informed consent with three exceptions, British Medical Journal, 314:1107-1111, 1997.
46
Dawson, A. J. Commentary: Methodological reasons for not gaining prior informed consent are sometimes
justified, British Medical Journal, 329:87, 2004.
47
(1) Guidance 6 in International Ethical Guidelines for Biomedical Research Involving Human Subjects,
CIOMS, Geneva, 2002. (2) Article 26 of the Declaration of Helsinki (Appendix 4, this report). (3) Code of
Federal Regulations (US), Title 21, Part 50, Subpart B, Section 50.24. Exception from informed consent re-
quirements for emergency research. (4) See Chapter 14. Research and innovative treatment, in Medical Ethics
Today. The BMA’s handbook of ethics and law, Second Edition, British Medical Association, London, 2004.

46

group6_PH.indd 46 7.8.2007 12:19:30

 obligations are of the investigators, ethics committees and sponsor. For
example, for how long after a patient leaves a trial should new infor-
mation be provided? Does it depend on the nature of the information?
Although some aspects of re-consenting are covered in Chapter 7, the
details of this topic were considered beyond the scope of the Working
Group; however, this issue has become the subject of debate within
and between companies and must be addressed.

(3) Transparency in Availability of Clinical Trial Results

One of the more complicated and controversial issues facing the
biopharmaceutical industry and the biomedical research commu-
nity is whether results of all completed clinical trials should be
made available to interested parties, and if so what information and
how. ICH Guideline E6 (GCP) states that there should be a writ-
ten publication policy for trials, either as part of the protocol or as
a separate agreement. It calls for both positive and negative results
to be made available for other researchers so that lessons learned
in trial design can be shared. Similar guidance is provided in the
Declaration of Helsinki and in the CIOMS Ethics Guidelines. However,
critics have pointed to underreporting of clinical research by company
sponsors, which allegedly leads to bias, especially because “negative
results” (e.g., lack of, or poor, efficacy) are rarely published or other-
wise made publicly available.48 One of the problems in this area is the
reluctance of journals to publish reports of studies with negative find-
ings. The absence of complete data not only compromises the ability
of independent researchers to conduct proper meta-analyses, but also
it has been opined that it denies practitioners and possibly the public
from information needed to make good treatment decisions.
The medical publishing community has a cooperative group (CON-
SORT) that has made proposals for improving the quality of papers,
reducing publication bias, and making transparent any conflicts of inter-
est.49 Many published papers reporting clinical trials, however, do not
comply with their guidelines and apparently most are deficient in the

48
Chalmers, I. Drug companies should be forced to publish all the results of clinical trials. How else can we
know the truth about their products?, New Scientist, 6 March 2004, p. 19; Herxheimer, A. Open access to
industry’s clinically relevant data, British Medical Journal, 329:64-65, 2004.
49
The Consolidated Standards of Reporting Trials Group (CONSORT) has developed a checklist and flow
diagram for the reporting of randomized clinical trials (http://consortstatement.org). See Moher D, Schulz KF,
Altman D, et al. The CONSORT Statement: Revised. Recommendations for improving the quality of reports
of parallel-group randomized trials. Journal of the American Medical Association, 285:1987-1991, 2001.

47

group6_PH.indd 47 7.8.2007 12:19:30

 reporting of adverse events.50 There have been proposals from various
groups, journals included,51 that a special database be created to contain
the results of all studies. Contrary to claims made by many parties that
development and use of such a database is straightforward, we believe
that it is actually very complicated; the details must be carefully con-
sidered to preclude possible unintended consequences of making very
large amounts of unfiltered data available to inexperienced parties.
CIOMS Working Group VI is sensitive to and concerned about this sub-
ject and endorses the concept of transparency of results and outcomes for
all clinical research, especially safety data; however, it is not in a position
to make concrete proposals or recommendations on this continuously
evolving subject. Nevertheless, we make the following points with the
hope that they help contribute to a logical and rational solution.
(a) In a global research and development environment, care must
be taken to avoid unilateral legislation or other requirements for
the creation of a master database of clinical research results; a
harmonized effort would be highly desirable.
(b) It is important that all parties understand the distinction between
a database of results and a registry of ongoing trials that is
informational for prospective patients and their health care
providers. There is evidence of confusion on the difference by
both the public and many journalists.
(c) There are legitimate concerns regarding possible proprietary
information associated with study designs and methodologies;
premature disclosure in the absence of a public health need would
not be appropriate.
(d) It is vital to take into account the huge number of clinical medicinal
trials conducted by independent clinicians, academic institutions,
managed care organizations, and public agencies.52
(e) Deciding on the structure and contents of a database requires
consideration of many parameters: Should it cover both pre- and
post-approval studies? Include all protocols used for the trials?
Include full data sets for individual patients? When some trial
study reports can reach hundreds or even thousands of pages in

50
Ioannidis, J. P. A. and Lau, J. Completeness of safety reporting in randomized trials – an evaluation of
seven medical areas, Journal of the American Medical Association, 285:437-443, 2001.
51
See Clinical Trial Registration: A Statement from the International Committee of Medical Journal Editors,
New Eng. J. of Med., 312:12, 2004.
52
By one recent estimate, drug manufacturers sponsor only about one-third of drug trials in the US.

48

group6_PH.indd 48 7.8.2007 12:19:31

 length, of how much practical value will they be to practicing
physicians, let alone patients? Should results of only prospective
trials be included or should observational study results be covered?
What is the best form and focus for result summaries?
(f) Who will design, create and maintain the database and who will
pay for it?
(g) How should access by the public to such complicated data be
arranged? How can they, or even healthcare professionals, evaluate
the quality of the study and interpret the statistics provided?
(h) How much useful information can be gleaned from individual
study reports without placing them in the context of a full research
program?
(i) Drug regulatory authorities have access to all the clinical study
results (from companies) and use their expertise to judge their value
and application for product information, including official labeling.
When many agencies make available to the public their summary
reviews of marketing application data along with considerable
details, is it necessary to create new or different systems?53 Is there
a risk in bypassing or usurping the role of the regulators?
(j) If a database is required for all studies, how does this affect the
peer review process for journal publication? Does it prejudice the
ability to publish such disclosed results?
(k) Will availability of results of studies covering unapproved uses of
medicines lead to increased off-label use and serve as an implicit
(but unintended) form of “promotion” for such use?
(l) There is at least one group that maintains a comprehensive register
of well-reviewed trials that meet certain minimum standards of
quality (Cochrane Central Registry of Clinical Trials).54 Can
lessons be learned from its experience and methods?
(m) There may be liability issues for companies when there are
differences between the official product information (data sheets)
for marketed products and the full panoply of data found within

53
The US FDA posts summaries of the medical reviews for new drug approvals on the drugs@FDA web-
site: http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfm. For results of pediatric studies: www.
fda/gov/cder/pediatric. Other sources for results of trials conducted in the US are: clinicaltrials.gov (service
of the National Institutes of Health), cancer.gov (National Cancer Institute), centerwatch.com (industry and
government sponsored trials) and trialscentral.org (web site of Brown University’s Center for Clinical Trials
and Evidence-Based Medicine covering worldwide trials). In Japan, the MHLW posts the results of studies
in their summary basis of approval after a drug is approved; further details on individual studies can then
be accessed from the companies on request. In the UK, there are at least two repositories of trial results:
http://www.cancerhelp.org.uk/trials/trials/ and http://www.controlled-trials.com/mrct/
54
See www.cochrane.org

49

group6_PH.indd 49 7.8.2007 12:19:31

 multiple study reports, even though those differences may be
perfectly understandable.
Some individual pharmaceutical companies have taken steps to make
results of all their trials available, some through their own web sites,
others through public institutional sites, such as the NIH in the US. The
US pharmaceutical industry, through its professional association, has
established a principle for all its members to make study results avail-
able on a voluntary basis.55 It has also established a central database to
contain the results of hypothesis-testing trials completed since Octo-
ber 2002 by member companies, mainly Phase III and IV trials, whether
published or unpublished.56 A “Joint Industry Position on Clinical Trial
Information Disclosure” was issued by EFPIA, IFPMA, JPMA and
PhRMA that commits to making publicly available the results of all
clinical trials (other than exploratory) on a drug marketed in at least
one country, and completed after the date the position paper was pub-
lished (6 January 2005).57 The Association of the British Pharmaceuti-
cal Industry (ABPI) sponsors a clinical trial database.58 Also in the UK,
the British Medical Journal publishes Clinical Evidence, an interna-
tional source of the “best available evidence for effective health care”
to foster informed decision making by summarizing what’s known and
not known about the treatment and prevention of nearly 200 medical
conditions.59
Some leading medical journals have announced an initiative that would
require listing of a trial in a public registry of the results before a paper
would be accepted for publication.60
Much more debate and work will be required before a useful and vali-
dated system for widespread documentation of clinical trial results can
be achieved.

55
See Updated Principles For Conduct Of Clinical Trials And Communication Of Clinical Trial Results,
PhRMA, June 2004 (http://www.phrma.org/mediaroom/press/releases/30.06.2004.427.cfm).
56
This Internet database is publicly available free (www.clinicalstudyresults.org).
57
See www.ifpma.org
58
See http://www.cmrinteract.com/clintrial.
59
See http://www.clinicalevidence.com
60
British Medical Journal, 329:637-638, 2004.

50

group6_PH.indd 50 7.8.2007 12:19:32

 (4) Other Issues
Another sensitive issue relates to clinical trials in resource-poor and
developing countries. Development of new treatments or new uses
of old treatments for “neglected diseases” or conditions prevalent in
developing countries, such as HIV/AIDS, malaria and other tropical
diseases, requires that trials be conducted in those locations. Guidance
for the ethical aspects of studies in such locations is available.61 The
potential study populations are often vulnerable and socially under-
privileged, and the issue often arises as to whether study medication
should be provided to the subjects after trial completion; an important
clarification on this issue has been made to the Declaration of
Helsinki.62 Many companies have a process for deciding under what
circumstances, and how, such treatment continuation should be imple-
mented. Limited financial and infrastructure resources have an impact
on the choice of authorized medicines used in such areas.
Among other difficult ethical and scientific questions that do not have
easy answers, and for which no regulations or guidance is available, is
the following: If a new, significantly safer and/or more efficacious drug
is approved after beginning a development program that uses the previ-
ous standard therapy as a comparator, how should a sponsor proceed?
The strategy will likely depend on how far along the development pro-
gram is (Phase I, II, III) and on other factors. It would be prudent for
the sponsor to discuss the situation with appropriate regulators.

61
International Ethical Guidelines for Biomedical Research Involving Human Subjects, CIOMS, Geneva,
2002. See Guidelines 3, 10, 11, 20 and 21. Also, see http://europa.eu.int/comm./european_group_ethics/
docs/avis17_complet.pdf
62
The Declaration of Helsinki covers this issue under Paragraph 30. For the update (Tokyo 2004), see
www.wma.net.

51

group6_PH.indd 51 7.8.2007 12:19:32

group6_PH.indd 52 7.8.2007 12:19:32
 III

Good Pharmacovigilance and

Risk Management Practices:
Systematic Approach to
Managing Safety during
Clinical Development

group6_PH.indd 53 7.8.2007 12:19:32

group6_PH.indd 54 7.8.2007 12:19:33
 a. Introduction
Although most of this CIOMS report focuses on the technical aspects
of safety surveillance, analysis and reporting, the Working Group believes
it important to consider first an overall framework for a pharmacovigilance
process for any clinical program. It is hoped that such a perspective can
help sponsors elucidate a thorough, systematic and disciplined approach to
clinical trial patient safety. The purpose of this chapter is to suggest some
important aspects of such an approach that should be taken into account
perhaps even before initiating the first Phase I study but certainly through-
out the clinical program. It is also recommended that a formal risk man-
agement plan be created. While the tone of the chapter tends to address
large pharmaceutical company research and development, the principles
can be adapted to other environments. For example, in a smaller organiza-
tion one person may serve multiple roles otherwise served by several in a
larger company. In addition, certain functions (e.g., clinical data manage-
ment or clinical expertise) may be covered internally by a larger company
and externally by a smaller one. Likewise the same principles that apply
to pharmaceutical company sponsors should apply to all other sponsors of
clinical trials.
Regardless of the setting, it is important to ensure that a well-defined
and well-structured process is in place that will allow sponsors to
readily identify, evaluate and minimize potential safety risks relative
to potential benefits for study subjects in pre-approval trials. Such a
process should start before initiating the first Phase I study and con-
tinue through post-approval use of the drug or biologic in the general
population. In establishing the process, it is important to consider and
define, in advance, the roles and responsibilities of individuals within
the organization who are expected to participate.
Depending on the business processes and organizational structure of
the company a formal plan should probably be created, and modified as
needed during a clinical program. In the initial planning stages of a new
clinical development program, one goal is to gather the necessary knowl-
edge and information to adequately plan the optimum program from the
standpoint of safety. This would be a good time to create the team that will
be responsible for the process and if desired prepare the initial risk manage-
ment plan.
In contrast to the post-marketing phase, little guidance is available on
the pharmacovigilance process during development.

55

group6_PH.indd 55 7.8.2007 12:19:33

 Although the term “pharmacovigilance” has traditionally been asso-
ciated with post-marketing activities, the CIOMS VI Working Group
recommends that the term be applied to the pre-marketing process for
collecting, managing and assessing safety information during devel-
opment. Likewise, the concepts of risk assessment and risk minimiza-
tion, components of risk management, are terms that are as applicable
to the pre-marketing environment as they are to the post-marketing
environment. (See Appendix 1 for more details on terminology.)
Concurrently with CIOMS VI, guidelines are being developed that
address the planning of post-marketing pharmacovigilance and risk man-
agement activities for newly licensed/approved products. These include
a set of guidance documents by the FDA1 and ICH E2E (Pharmaco-
vigilance Planning)2. In both cases, a major focus is on the creation of a
document, to be submitted to health authorities prior to approval, that
describes the company’s plan for gathering additional information to fill
remaining gaps in knowledge or for interventions to minimize the known
risks in the patient population, once the product is approved/licensed and
marketed. These guidance documents complement each other, appropri-
ately reinforcing the growing recognition of the importance of maintaining
a proactive stance toward safety surveillance throughout a product’s life.
The CIOMS VI Working Group suggests that such written plans might be
a natural outgrowth of a process that starts at the earliest stages of develop-
ment. It is not recommending any particular format for a development risk
management plan since this will likely vary depending on the circumstances
and would evolve as development progresses. Early in development, docu-
mentation of risk considerations and planned steps to deal with them would
most logically be part of the overall Clinical Development Plan. As risks
are better understood, they would be included in the Investigators Brochure
as part of the Development Core Safety Information. In later stages of de-
velopment, documentation would eventually evolve into a stand-alone Risk
Management Plan. (See section c. below.)

1
The FDA issued three guidance documents for industry in March 2005: (1) Premarketing Risk Assessment
(http://www.fda.gov/cder/guidance/6357fnl.htm), (2) Development and Use of Risk Minimization Action
Plans (http://www.fda.gov/cder/guidance/6358fnl.htm), and (3) Good Pharmacovigilance Practices and
Pharmacoepidemiologic Assessment (http://www.fda.gov/cder/guidance/6359OCC.htm).
2
ICH Guideline E2E: Pharmacovigilance Planning, Step 4 as of November 2004 (http://www.ich.org).

56

group6_PH.indd 56 7.8.2007 12:19:33

 b. Principles of a Systematic Approach
(1) Begin early. Consideration of patient safety is the most important first
step in clinical development. As a matter of principle, the process for
managing risk during development should start no later than when a
decision is made to begin human trials. The sponsor’s decision to pro-
ceed with development of a medicinal product will certainly need to take
into account a broad range of factors including but not limited to safety.
However, a determination of whether and under what conditions it is safe
to proceed should always be made independent of other considerations
and the safety review team should be involved in that decision. For new
chemical entities, the decision regarding safety will be made based on
non-clinical safety data and information on closely related compounds,
and therefore requires careful assessment and advice from qualified
toxicology specialists as well as careful and deliberate planning for
safety monitoring during early clinical trials.

(2) Establish a procedure. The first step in establishing a systematic ap-

proach to identify and manage risk during development is the creation
of a procedure that defines how it will:
❏ ensure the regular and timely review and evaluation of all available
safety information in order to identify potential risks
❏ clearly define roles and responsibilities
❏ enable timely and effective decision-making to minimize risks to
study subjects
❏ assure consistent implementation of risk minimization actions
across protocols and study locations
❏ enable realization of the implications for the intended target population
after approval so that appropriate post-marketing pharmacovigilance
and risk minimization activities can be designed and implemented.
Sponsors should establish standard operating procedures that define
a framework for a process that can be applied consistently across all
development programs, but which allows enough flexibility to meet the
needs of what will inevitably be a diversity of products and a broad
range of safety issues associated with them. In some cases it may
be appropriate to supplement standard operating procedures with
product-specific procedures.

57

group6_PH.indd 57 7.8.2007 12:19:33

 (3) Establish a Multidisciplinary Safety Management Team (SMT). The
procedure should clearly define the makeup and charter of a multidisci-
plinary team that will be responsible for the timely review, assessment
and evaluation of incoming safety data. The core team should include
a representative from each of the medical functions that play a role in
the development and post-marketing monitoring of the product. Other
members should be available on a regular or ad hoc basis depending on
the issues, e.g., epidemiologist, clinical pharmacologist, toxicologist,
chemist, biostatistician, regulatory affairs expert. Roles and responsi-
bilities should be clearly defined, for the team as well as for each indi-
vidual on the team. Each member of the team must have responsibility
and accountability for raising issues, in particular those emanating from
their respective disciplines. The team should be empowered to make de-
cisions that will accomplish the goal of minimizing risk while maximiz-
ing benefits to subjects in clinical trials, as well as anticipating the use
of the product once marketed. Decisions should take into account the
need to update the IB, DCSI, CCSI and/or informed consent, modify
or add new monitoring procedures, implement protocol amendments
or initiate prompt communications to investigators, ethics committees
and regulators. When applicable, consideration should also be given to
when and how prescribers and patients should be informed for a pro-
duct already marketed in one or more countries.
The composition of such a team will vary depending on a number of
factors such as:
❏ Structure and size of the company
❏ Development stage of the compound
❏ Type of compound under development
– First in class
– Follow-up compound within the company
– Line extension
Although it will depend on the product and the size and complexity of
the program, the following is an example of how such a team can be
composed and function:
❏ Global project/product physician has the overall medical responsi-
bility for the project, including assessment of the benefit-risk pro-
file of the product
❏ Global safety physician has the responsibility for identifying and
evaluating the risks relating to the product
❏ Global regulatory affairs director/manager has the responsibility
for advising the team on regulatory policy

58

group6_PH.indd 58 7.8.2007 12:19:34

 ❏ Project manager, specifically assigned to track and manage the
team’s decisions, ensuring appropriate follow-up and completion of
assigned tasks
❏ Other disciplines on an as needed basis

The leader of the multidisciplinary SMT can be made accountable for the
creation and appropriateness of the plan and for ensuring that the plan
is implemented. Responsibility for drafting the plan should be shared
among all members of the team.
As necessary, the SMT would work with appropriate staff (e.g., epidemio-
logist and toxicologist) in the quantitative assessment of identified risks,
the characterization of the safety profile of the substance under develop-
ment, the identification of signals, and the determination of changes in
the safety profile.
The global regulatory affairs director or manager is responsible for ensur-
ing that a plan is included with regulatory submissions (e.g., New Drug
Application or Marketing Authorization Application) when required, or
when the global project team determines that it represents an essential
element of the application.
When internal resources are limited, e.g., for smaller companies or for
sponsors in developing countries, teams may be smaller and individuals
may play more than one role. In these situations, greater consideration
might be given to involving outside experts or establishing an external
DSMB with a role wider than for one specific study. (See Appendix 5 for
more detailed discussion on the role of the DSMB.)
When licensing partners are involved, a joint safety review process, in-
cluding clear roles and responsibilities of the respective companies,
should be defined in advance with timelines for exchange and joint re-
view of data. Ideally the terms should be part of the initial contract, but
at the very least should be incorporated into a follow-on agreement on
safety matters.

(4) Establish a project management function. Key to the success-

ful implementation of a consistent and systematic approach is the es-
tablishment of a mechanism for scheduling meetings, tracking issues
and timelines, and assuring completion of action items. The CIOMS
VI Working Group recommends establishing a project management
function to manage these tasks, document any decisions, and ensure
compliance with internal procedures.

59

group6_PH.indd 59 7.8.2007 12:19:34

 (5) Determine background data. Although there will generally not be an
abundance of data for the team to review early in clinical development, it
is at those stages, if not sooner, that the team should begin to formulate
an understanding of the target population. It is advantageous to involve
epidemiologists to help describe the natural history of the disease being
treated, to aid in defining endpoints, especially for Phase III studies, and
to anticipate important adverse events of interest (e.g., serious, severe,
frequent or otherwise of clinical importance) that might be observed as
part of the background. It is also a good time to consider the target benefit-
risk profile, taking into account the natural history and associated risks
of the disease as well as the benefits and risks of available alternative
therapies. For more details, see section d. Role of Epidemiology, below.

(6) Ensure accessibility of data. Also key to successful implementation is

the accessibility of all relevant data. It should be a top priority to make
safety and other pertinent data readily available to the safety team from
the clinical trial and safety databases as well as from other relevant sourc-
es, such as the pre-clinical toxicology department (e.g., carcinogenicity
and development and reproductive toxicology), in vitro mutagenicity
studies, and pharmacokinetic and drug-interaction studies. In doing so, it
would be important to identify who is responsible for accomplishing the
retrieval and presentation of data in a format that can be readily evalu-
ated by the core team. If data management is being handled by another
party, e.g., a contract research organization (CRO), it is important for the
sponsor to define in the contractual obligations the mechanism for timely
accessibility to accurate data.

(7) Develop a proactive approach. During early stages of development it is

also advisable to begin formulating components of risk assessment and
risk minimization plans. If there are adverse events of particular interest
or concern, for example based on knowledge of the therapeutic or phar-
maceutical class, on animal toxicology studies, or on the known mecha-
nism of action, then consideration should be given to special monitoring
procedures.3 If there are populations that are considered to be potentially
at higher risk, then plans should be made for addressing the risk through

3
The CIOMS VI Working Group considers the term “known risk” to refer to a risk that has been observed and
is reasonably established for the investigational product itself; the term “anticipated risk” to refer to a risk that
has not yet been observed or established for the product but is expected to occur based on knowledge of the
class of drugs; and the term “potential risk” to refer to a risk that has not yet been observed in humans for the
investigational product itself or for other drugs in the class but for which there is reason to suspect it might
occur, based on animal toxicology studies or the known pharmacologic properties. In other contexts (e.g., ICH
E2E), what we refer to as anticipated risks are usually placed in the potential risk category.

60

group6_PH.indd 60 7.8.2007 12:19:35

 exclusions or special studies. The size, components and nature of the de-
velopment program will in large part be driven by anticipated or potential
safety issues. It is therefore important to identify those issues as early as
possible to ensure the adequacy of the program. For example, if there is a
special population that is expected to be at high risk but that risk is not yet
well characterized, there may be a need to plan for special studies in that
population or to ensure their sufficient representation in the pivotal trials.
If drug metabolism studies suggest a propensity for drug interactions, it
would be important to understand the likelihood that the target population
would be using concomitant therapies that might be of concern, and to
plan accordingly. ICH Guideline E1 provides guidance on the size of the
safety database for drugs intended for chronic use in non-life threatening
conditions.4 The FDA Draft Guidance for Industry on Premarketing Risk
Assessment includes a discussion of other factors to consider, such as the
value of long-term controlled safety studies, the diversity of the clinical
trial population, and exploration of dose effects (see footnote 1). See also
Chapter 5 for a discussion of the safety review process.

(8) Establish timeframes and milestones. Monitoring of safety during

development should be viewed as an intensive continuous process, es-
pecially in Phase I and II when little may be known about the risks.
However, the procedure should establish regular timeframes for review
of safety data by the multidisciplinary SMT. The CIOMS VI Working
Group recommends quarterly review of safety data as a reasonable
standard. More frequent reviews might be necessary in some circum-
stances, in particular very early in development when little is known
about the risks or benefits, or when a specific issue has arisen. On the
other hand, less frequent reviews might be appropriate for continu-
ing development of an approved product with a fairly well established
safety profile, or when the pace of new data acquisition from trials
is very slow. Whatever the cycle of reviews it would be important to
coordinate the timing with that of pre-approval periodic reports such
as the annual IND report or the newly proposed Development Safety
Update Report (see Chapter 7 for detailed discussion). If the product is
approved, such reviews should also be coordinated with PSURs, where
applicable. It would also be important to coordinate the review with
milestones such as end-of-Phase II, completion of pivotal trials, or
writing of the integrated summary of safety.

4
ICH E1: The Extent of Population Exposure to Assess Clinical Safety for Drugs Intended for Long-Term Treat-
ment of Non-Life-Threatening Conditions, Step 5 as of October 1994 (http://www.ich.org).

61

group6_PH.indd 61 7.8.2007 12:19:35

 The CIOMS VI Working Group recommends that sponsors create a dedi-
cated Safety Management Team (SMT) to review all the available safety
information on a regular basis so that decisions on safety can be made in
a timely manner. It also recommends that these reviews take place usu-
ally at least quarterly pre-approval and be coordinated with pre-approval
and, if applicable, post-approval periodic reporting. Quarterly and ad hoc
safety reviews should consider the overall evolving safety profile of the
investigational product, make necessary changes to the IB/DCSI, and de-
termine if any changes to the conduct of the trials need to be considered.

(9) Decision making: The focus of safety reviews should be on the iden-
tification of issues, a determination of their implications, what actions
should be taken, and monitoring and assessing the results of those ac-
tions. For each safety review meeting, there should be a clear deter-
mination of whether or not there are any new issues that warrant close
attention or any new developments on issues identified earlier. For any
ongoing or newly identified potential risk, consideration should be
given to the implications for the DCSI, informed consent, communica-
tions to investigators, ethics committees or regulatory authorities, any
changes to monitoring procedures, amendments to protocols or the IB,
or to the overall development plan itself.

(10) Advisory bodies: The SMT should have advisors readily available to call
upon when issues of significant concern arise. These might be in the form
of an internal safety committee, an issue-specific external advisor or advi-
sory board, or an independent monitoring board for a trial or program.
There will be situations where the SMT might benefit from a higher level
of internal review, to ensure awareness of the issue by more senior man-
agement, to obtain support for decisions that may have a significant im-
pact on the overall clinical development program, and to ensure consis-
tency of the timeliness and content of communications on a global basis.
Hence, establishment of an internal senior safety committee of execu-
tives with expertise in managing safety issues from a scientific, medical
and regulatory perspective might be considered. This would generally be
a single committee that would review and respond to issues presented by
the product-specific SMT. This senior safety committee would provide
scientific advice, consider implications for the overall development pro-
gram, and expedite decisions and their resultant actions through liaison
with other parts of the organization, where timeliness and consistency
across regions is of utmost importance.

62

group6_PH.indd 62 7.8.2007 12:19:36

 If not available internally, it may be necessary to obtain the advice of
outside experts on an ad hoc basis when issues arise. Expert advisors
or advisory panels may provide advice regarding the significance of
findings, make suggestions regarding usefulness and interpretation of
diagnostic or screening tests, develop decision rules for discontinua-
tion of study drug, or provide input on other risk minimization actions.
Expert advisory panels are particularly useful if there is a need for
ongoing review of accumulating cases. The advisory panel may also
provide overall advice regarding implications for the viability of the
program based on the emerging product profile as it relates to standard
of care and other available therapies.
There may also be circumstances where the use of a DSMB is advis-
able. Although DSMBs are generally responsible for a particular trial,
it would be important to ensure that they have access to any and all
information external to the trial that might have a bearing on its role of
monitoring safety. In exceptional circumstances, consideration might
be given to establishing a DSMB to monitor safety across the entire
program rather than just one or more trials. For example, if a new class
of oncology drugs is being tested in multiple tumor sites across a num-
ber of protocols with survival as an endpoint, it would be reasonable to
establish a DSMB to monitor safety in all the trials.

c. Components of a Development
Risk Management Plan (DRMP)
The CIOMS VI Working Group fully recognizes the demanding work-
load and pressures on companies, and the many committees, working groups
and reports that are already involved in drug development. Thus, it does not
take lightly the recommendation to create another plan and process. How-
ever, high quality pharmacovigilance is an essential component of any clinical
program. A natural outgrowth of the systematic approach described above is
the creation of a formal Development Risk Management Plan. Such a plan
would need to be compound-specific and perhaps form a section of the overall
Clinical Development Plan. It should include early documentation of known,
anticipated or potential risks along with plans for addressing them during de-
velopment and, where appropriate, the DRMP would eventually evolve into
a post-marketing risk management plan that will accompany the registration
application.

63

group6_PH.indd 63 7.8.2007 12:19:36

 The DRMP is not intended to be a legal or regulatory document, but
rather a guide for safety surveillance during development. However, it must be
recognized that these documents may be subject to legal discovery. Therefore,
there are two actions which should be considered in the development of the
process. First, the company’s legal department should ensure appropriate lan-
guage is considered for a general statement that this is a working document.
Second, the company should ensure that processes including project manage-
ment are in place to make sure that action plans are followed through. The plan
should not be a place for speculation, theoretical explanations or potential ac-
tion plans. Any action which is written in the document should be followed.
The DRMP should include, at a minimum, the following sections;
many of them are expected to be addressed routinely in any development
plan, from the perspective of efficacy as well as safety:
(1) Introduction and Objectives
(2) Anticipated Product Profile
❏ Indications
❏ Intended population
❏ Expectations for new product (prevention vs symptomatic treatment vs
cure) and associated threshold for tolerating risk (see also Chapter 5)
❏ Anticipated benefit and/or risk advantages over existing therapies,
if any
(3) Epidemiology (see Section d. below for details)
❏ Definition of disease and diagnostic criteria
❏ Natural course of disease, including likely concurrent conditions
and concomitant medications
❏ Quantification of burden of disease (incidence, prevalence, morbid-
ity, mortality, percentage of patients diagnosed)
❏ Consideration of special populations, such as:
– pediatrics (ICH Guideline E11)
– elderly (ICH Guideline E7)
– ethnicity (ICH Guideline E5)
– women of child-bearing age, pregnancy5
– organ impaired patients (e.g., decreased hepatic or renal function)

5
The Committee for Human Medicinal Products (CHMP) in the EU issued (June 2004) a draft Note for
Guidance on the Exposure to Medicinal Products During Pregnancy: Need for Post-authorization Data (www.
emea.eu.int/pdfs/human/phvwp/188904en.pdf; document EMEA/CHMP/1889/04/Consultation). Comments
to the CHMP were due in December 2004. Although focused on the post-authorization period, this document,
still draft as of this writing, would be useful to consult regarding data requirements and other considerations.

64

group6_PH.indd 64 7.8.2007 12:19:36

 (4) Non-clinical safety experience6
❏ Pharmacokinetics/Pharmacodynamics
❏ Acute and chronic toxicity
❏ Developmental and reproductive toxicology
❏ Mutagenicity and carcinogenicity
❏ In vivo and in vitro drug interactions
❏ Special safety pharmacology studies (e.g., cardiac conduction,
neurotoxicity)
(5) Clinical safety experience (see footnote 6)
❏ Clinical pharmacology
– Absorption, distribution, metabolism, excretion (ADME)
– Drug interactions
– Dosing and dose-response information
– Efficacy
– Safety
– Safety profile of class
– Safety profile of new product
– Extent of exposure to date
– Evaluation of adverse events, including frequency
– Safety in demographic groups and special populations
– Effects on different body systems
❏ Benefit-risk profile of new product

(6) Identification and assessment of known or anticipated risks

❏ Known or anticipated adverse events might warrant special atten-
tion if special measures need to be taken. For example, if there is
the potential for gastrointestinal bleeding, it would be important to
define what would be considered a clinically significant bleed that
should be reported promptly to the sponsor even if not considered
serious for regulatory purposes (such an event could be considered
an “AE of special interest”). It would also be important to ensure
that informed consent documents include early signs and symptoms
for patients to be aware of so that bleeding can be detected early.
Consideration might also be given to developing coding guidelines
for adverse events of special interest.

6
The non-clinical and clinical sections should be consistent with the IB/DCSI (see Chapter 7), but may include
more. For example, there may be a discussion of an evolving, but still uncertain, safety issue that has not yet
reached the threshold for inclusion in the DCSI or IB.

65

group6_PH.indd 65 7.8.2007 12:19:37

 ❏ There may also be obvious compound-specific or known therapeutic
class-specific issues, such as drug-drug, food-drug or disease-drug
interactions. For biologics, the issue of immunogenicity should al-
ways be considered. There are also certain special populations that
should always be considered, for example, women of child-bearing
potential, pediatric patients, elderly patients or patients with renal
or hepatic insufficiency. Specific issues and special populations are
described below.
(7) Identification and assessment of potential new risks
❏ In developing a systematic approach to managing safety during de-
velopment, one can identify a handful of specific issues that should
always be “on the radar screen,” e.g., QT prolongation, hepato-
toxicity and potential for abuse. (See Section e. below.)
❏ Potential high risk populations or circumstances
❏ Potential for medication errors during treatment in clinical trials or
during general use once the product is approved/licensed
❏ Potential for off-label use once the product is on the market

(8) Actions and/or plans for evaluating and mitigating risk

Routine as well as compound- or protocol-specific steps should be de-
scribed, including data that will be monitored and the time frames for
conducting safety reviews.
As specific signals or issues are identified, action plans should be
made describing the specific activities that will be conducted in order
to assess and/or control them. An action plan for each issue will
generally include either a plan for further assessment (risk evalua-
tion), or a plan to decrease the risk to patients (risk minimization). The
action may range from relatively simple, e.g., monitoring during the
ongoing trials, to relatively complex, e.g., development of a special
data collection form or the conduct of a targeted study. From a practi-
cal point of view, this list will usually include actions developed by
the product team and be specific to the product, but may also include
a standard list that the company uses for all products. Examples of
actions include:
❏ Continuation of routine monitoring
❏ Communication to investigators, patients, IECs/IRBs, DSMBs,
regulators
❏ Protocol amendment(s)
– Specific monitoring and investigation

66

group6_PH.indd 66 7.8.2007 12:19:37

 – Alteration of patient population (inclusion and exclusion criteria)
– Change in dose or dose schedule
❏ Additional studies
❏ Temporary hold on one or more clinical trials in a program
❏ Termination of one or more clinical trials in a program
❏ Termination of program

d. Role of Epidemiology
There is a broad and long-standing recognition of the importance of
epidemiology to the development planning process, not only for defining
the natural history and burden of the disease being treated, but for antici-
pating the important confounding factors and background incidences of
concurrent illnesses.7 All of these factors must be taken into account in
planning the size and demographics of the safety database as well as in
evaluating case reports and case series when studies are still blinded.

(1) Patient Population, Natural History of Disease, Concurrent

Conditions and Background Rates of Adverse Events
The epidemiology of the disease being treated is an important compo-
nent of the planning of any clinical development program. The incidence
and prevalence will determine the size of the target population. For preven-
tive therapies, identification of populations at higher risk can help to define
the study population in a way that can reduce costs by requiring fewer
patients to show an effect.
In addition to aiding in the planning of trials to show efficacy, under-
standing the epidemiology and natural history of the disease is also impor-
tant for putting potential safety issues into proper context. An observation
in the study population of a “higher than expected” incidence of a particular
event when compared to the general population may actually be expected
when compared to the background rate in the target-disease population. For
example, patients with rheumatoid arthritis (RA) have a 2 to 4 fold or greater
incidence of lymphoma than the general population independent of therapy.8
Knowing this helps to put reports of lymphoma in clinical trials of RA into
proper perspective.

7
Guess H.A., Stephenson W..P., Sacks S.T., and Gardner J.S.. Beyond pharmacoepidemiology: The larger role
of epidemiology in drug development, Journal of Clinical Epidemiology, 41: 995-996, 1989.
8
Baecklund E., Askling J., Rosenquist R., et al. Rheumatoid arthritis and malignant lymphomas, Current Opinion
in Rheumatology, 16:254-61, 2004.

67

group6_PH.indd 67 7.8.2007 12:19:38

 Understanding the natural history of the disease is important for antic-
ipating certain high risk situations, such as when patients are more likely to
have concurrent renal or hepatic insufficiency. While exclusion of patients
with these conditions will likely lead to a greater chance for a successful
outcome in a clinical trial, it would be important to conduct studies of spe-
cial populations or high risk patients if they are likely to be treated with the
drug once approved. Likewise, if there are specific drugs or classes of drugs
that are likely to be used concomitantly in clinical practice, the possibility
of a drug interaction should be considered and plans made for separate
clinical pharmacology studies where appropriate.

(2) Sources of Data for Background Rates

Knowledge of population background event rates is an essential com-
ponent of the evaluation of any potential clinical trial safety signal, includ-
ing the results of aggregate analysis. The approach to obtaining the appro-
priate background information will vary depending on the adverse event,
the patient population and where the study is being conducted. There are
numerous sources of data; however, not all are relevant or necessary for
every new compound. If the sponsor has relevant prior experience for the
same or similar population, a review of their in-house historical data may
provide insights into the potential issues for the new product. If the relevant
clinical programs were large enough, the pooling of placebo patients can
provide background rates for some adverse events that are most relevant
for anticipated clinical trial populations. The applicability of the histori-
cal clinical trials will depend on the comparability of inclusion or exclu-
sion criteria and possible changes in the availability and use of concomitant
therapies.
If the organization has no direct experience, the literature may be a
good source of background incidence rates. It is important to consider care-
fully just how applicable the morbidity and mortality rates in the literature
are to the clinical trial population. Conversely, it is important to recognize
early the limitations of extrapolating incidence rates from clinical trials to
a broader target population.
It may be appropriate to perform analyses on data from external epi-
demiological databases. There are many sources for such data, which vary
in size, completeness, and medical specificity. Several relatively large da-
tabases are derived from North American populations, including US State
Medicaid databases, such as California, Ohio, and Tennessee, other large
US databases from health maintenance organizations and the Veterans’ Ad-

68

group6_PH.indd 68 7.8.2007 12:19:38

 ministration, and the database of the Saskatchewan Health Plan in Canada.
In Europe, there are also a number of databases, the most well-known being
the General Practice Research Database (GPRD) in the UK. Other small-
er databases in Europe include PHARMO Record Linkage System in the
Netherlands, the MEMO database in Scotland and a recently established
database in Spain. Appendix 9 includes a list of available databases from
an ongoing compilation by ISPE members.9 Details on a number of data-
bases in North America and elsewhere can also be found online at a website
maintained by DGI, Inc.10
Other potential sources of data are disease-specific registries such as
bone marrow and liver transplant registries.11,12,13,14 HIV disease is an ex-
ample of a disease for which registries have been useful for following the
evolving background adverse event profile as new medications have be-
come available, which effectively change the natural history of the disease.
Sweden has nation-wide registries, e.g., for cancer and birth defects, that
can be linked to a national death index.15
The Prescription Event Monitoring (PEM) program in the UK16 can be
of value to investigate certain safety related questions. Although PEM cap-
tures product-specific event rates for marketed medicines, these can be used
to estimate the expected rates of events for similar populations. Spontane-
ous reporting system databases such as the publicly available FDA AERS
database in the US, the Drug Analysis Prints available from the ADROIT
database at the MHRA in the UK, and the WHO ADR database (Uppsala,
Sweden) might provide insight into the types of reactions that have been
reported for similar drugs. However, they are not at all useful for determin-
ing background rates. Absence of denominators, delay in data availability,
sparseness of data, and varying degrees of underreporting are some of the
factors that limit the usefulness of such information.

9
The International Society for Pharmacoepidemiology (ISPE) Database Resource Document is an ongoing pro-
ject aimed at compiling a list of available databases that might be considered for the conduct of pharmacoepide-
miology studies. The databases listed (see Appendix 9) have been supplied by ISPE members. The list is posted
for informational purposes only. It is not intended to be comprehensive. Inclusion on the list is not an endorse-
ment by the Society, nor does the Society make any comments about size, validity, or other characteristics or
qualities of a specific database (see http://www.pharmacoepi.org/resources/summary_databases.pdf).
10
See http://www.dgiinc.org
11
Center for International Blood and Marrow Transplant Research (CIBMTR); see http://www.ibmtr.org
12
European Liver Transplant Registry; see http://www.eltr.org
13
Nordic Liver Transplant Registry; see http://www.scandiatransplant.org/liver01/liver01.htm
14
US Transplant – Scientific Registry of Transplant Recipients; see http://www.ustransplant.org/liver_primer.php
15
See http://www.sos.se/epc/epid
16
See http://www.dsru.org/pem2002.htm

69

group6_PH.indd 69 7.8.2007 12:19:38

 (3) Benefit-Risk Considerations
Important contributions that epidemiologists can make include follow-
ing the literature and evaluating the applicability of newly published studies
that may result in changes in the real or perceived benefits and risks of cur-
rent therapy. For example, postmenopausal hormone therapy was expected
to provide protection from both cardiovascular disease and osteoporosis in
addition to symptomatic relief. Therefore, infrequent serious adverse reac-
tions were initially considered acceptable because the overall benefit-risk
profile was considered very favorable. As subsequent data became avail-
able from large randomized clinical trials that were part of the Women’s
Health Initiative,17,18 the perception of benefit-risk changed substantially
and in some situations the risks may outweigh the benefits. Future Develop-
ment Risk Management Plans for products in this class or similar classes of
drugs would have to take into account such new information.

e. Specific Issues that Should Always be Considered

When planning for the development of virtually any new medicinal
product, there are certain toxicities that should always be explicitly consi-
dered. These include:
(1) Cardiac electrophysiology: Drug-induced prolongation of cardiac
repolarisation (measured as the QT or QTc (i.e., QT corrected for heart
rate) interval on the surface ECG) and subsequent development of life-
threatening ventricular arrhythmias of the torsade de pointes type has
caused post-marketing withdrawal of several drugs and stopped others
in different stages of clinical development. Regulatory authorities pay
considerable attention to effects on QT/QTc by drugs in development,
as QT prolongation is thought to increase the risk of torsade de pointes
and/or sudden death. Guidelines under development within ICH re-
flect common views and requirements on QT/QTc documentation in
preclinical19 and clinical20 development of new drugs. Market autho-
17
Writing Group for the Women’s Health Initiative Investigators. Risks and benefits of estrogen plus progestin in
healthy postmenopausal women: principal results from the Women’s Health Initiative randomized controlled
trial, Journal of the American Medical Association, 288:321-333, 2002.
18
Women’s Health Initiative Steering Committee. Effects of conjugated equine estrogen in postmenopausal
women with hysterectomy: the Women’s Health Initiative randomized controlled trial, ibid., 291:1701-1712,
2004.
19
ICH Guideline S7B, The Non-clinical Evaluation of the Potential for Delayed Ventricular Repolarization,
Step 3 as of June 2004. See http://www.ich.org.
20
ICH E14. The Clinical Evaluation of QT/QTc Interval Prolongation and Proarrhythmic Potential for Non-
Antiarrhythmic Drugs, Step 3 as of June 2004. (QT interval prolongation by human pharmaceuticals). See
http://www.ich.org.

70

group6_PH.indd 70 7.8.2007 12:19:39

 rization will be particularly challenging for drugs that significantly
prolong QT/QTc unless they have unique positive effects in life-
threatening conditions.
(2) Hepatotoxicity: Hepatotoxicity is considered a risk that should be as-
sessed in all new chemical entities prior to marketing. Similar to car-
diac conduction, the development of hepatotoxicity has resulted in the
post-marketing withdrawal of several products. Given the frequency
and impact of this event and the inability of preclinical data to clearly
predict or define the risk, hepatotoxicity should be considered as a po-
tential issue in all developmental pharmacovigilance/clinical develop-
ment plans. Several attempts are ongoing to define guidelines for bet-
ter identification of potential hepatic toxicity using preclinical models
as well as improved sensitivity and specificity for clinical monitoring.
The most recent regulatory guidance on the investigation of potential
hepatotoxicity is in an FDA discussion paper.21
(3) Drug-Drug and Food-Drug Interactions: Consideration should al-
ways be given to the potential for drug-drug interactions, based on
what is known about the drug’s metabolism, the mechanism of action
and the likely concomitant therapies. Depending on the situation, it
may be sufficient to analyze adverse events as they relate to concomi-
tant therapy within the planned pivotal clinical trials; or it may be nec-
essary to conduct targeted studies. Multiple issues remain, including
the predictability of in vitro work, the relevance to patients of inter-
action studies in healthy volunteers, and potential pharmacodynamic
interactions which are not predicted by classic pharmacology studies.
Food-drug interactions are also potentially important (e.g., the effect
of grapefruit juice on the kinetics of several drugs); available informa-
tion on experience with products in the same or related chemical and
pharmacologic classes should be sought.
(4) Immunogenicity: The assessment of potential immunogenicity re-
mains a significant issue. The development of antibodies may be a rare
event which is either not observed or is underestimated based on the
relatively short exposure seen in most clinical programs. A plan to
assess and monitor potential immunogenicity should be considered,
especially in the development of biologics. It is especially important to
consider factors such as formulation, stability, storage conditions and

21
FDA White Paper. CDER-PhRMA-AASLD Conference 2000 on Drug-Induced Liver Injury: a National and
Global Problem, November 2000. See http://www.fda.gov/cder/livertox.

71

group6_PH.indd 71 7.8.2007 12:19:39

 changes in the production process that may alter the tertiary structure
of a biologic molecule and hence the potential to induce antibodies.
Finally, the potential impact of neutralizing or other types of antibod-
ies needs to be considered, as evidenced by the recent finding of pure
red cell aplasia with some erythropoietin products.22
(5) Bone marrow toxicity: Agranulocytosis and aplastic anemia have
both been identified as potential adverse reactions of drug treatment.
Absence of reliable animal and in vitro models make it difficult to
find these potential side effects early. Agranulocytosis has a yearly in-
cidence of 5-10 per million in the general population while aplastic
anemia is even more rare (annual incidence of 2-5 per million in the
general population).23 Thus, these reactions are not likely to be ob-
served before the drug has been used in a large population, and inter-
pretation of their significance requires knowledge of the population
exposed to the product as well as the background incidence in a similar
unexposed population.
(6) Potential for reactive metabolite formation and hypersensitivity
reactions: Reactive groups and metabolites may be associated with
genotoxicity and hypersensitivity/idiosyncratic reactions, such as
serious cutaneous adverse reactions, hepatotoxicity or bone marrow
toxicity. At the earliest stages of drug development consideration
should be given to the identification of chemical structures suspected
to be associated with toxicity. The presence of an alerting structure
should initiate discussions with a toxicologist to evaluate the signifi-
cance and relevance of the alert and to prepare a clear rationale for
advancement of the compound.

f. Conclusion
The concepts and recommendations in this chapter are intended to pro-
vide guidance for managing the complex process of assessing and manag-
ing safety information in order to minimize risk to clinical trial subjects
during clinical development. They are also intended to ensure the availabil-
ity and assessment of as much safety information as is reasonably possible

22
Bennett C.L., Luminari S., Nissenson A.R., et al. Pure Red-Cell Aplasia and Epoetin Therapy, New Engl. J. of
Med., 351: 1403-8, 2004.
23
Kaufmann D., Kelly K.B., et al. The Drug Etiology of Agranulocytosis and Aplastic Anemia, Oxford University
Press, London 1991.

72

group6_PH.indd 72 7.8.2007 12:19:40

 prior to marketing to optimize the benefits and minimize the risks to future
patients. Ideally, the implementation of a procedure that assures a system-
atic approach to managing safety during development and the use of a De-
velopment Risk Management Plan to track progress and action plans along
the way will lead to more effective risk identification, risk evaluation and
risk minimization. These will go a long way toward protecting volunteers/
subjects who agree to participate in clinical trials as well as future patients
who will use the drug once it is marketed. Furthermore, it should serve as
a basis to define those issues which will require further evaluation in the
“real” world or for which specific actions are warranted to minimize risk.
The Development Risk Management Plan can thus serve as the basis for
developing post-marketing pharmacovigilance and risk minimization plans
to be included with new marketing authorization applications.

73

group6_PH.indd 73 7.8.2007 12:19:40

group6_PH.indd 74 7.8.2007 12:19:40
 IV

Collection and Management

of Safety Data during
Clinical Trials

group6_PH.indd 75 7.8.2007 12:19:40

group6_PH.indd 76 7.8.2007 12:19:40
 a. Introduction
Throughout the clinical development of a pharmaceutical product,
safety data are collected through the use of instruments such as case report
forms (CRFs), serious adverse event reporting forms and laboratory re-
ports. Collection may occur by use of paper, electronic or telephonic media.
Data collection methods utilized during the conduct of clinical trials are a
vital part of the process of safety monitoring and are of concern to investi-
gators, sponsors, regulators and patients.
Correct data elements must be collected to allow for the proper medi-
cal interpretation of individual cases as well as for the analysis of aggre-
gate data.1,2 The decision on what safety data to collect and when should
be carefully considered based on anticipated needs and concerns for the
compound under investigation.3,4 In an effort to be all-inclusive, sponsors
will frequently collect more data than is actually necessary for analysis.5
This may place an undue burden on the investigator and sponsor and divert
attention from more important matters during the conduct and monitoring
of the study. The aim should be to capture only data that are reasonably
expected to be analyzed and assessed. Nevertheless it is prudent to collect
more comprehensive safety data during Phase I through III studies in con-
trast to Phase IV studies, for which the collection of non-serious adverse
events and excessive laboratory data, especially for compounds with well
established safety profiles, may add little value to the existing knowledge
of the product.
Although global Good Clinical Practice standards exist (ICH Guide-
line E6), detailed standards for the types of data to be collected for safety
monitoring are lacking. While ICH Guideline E2A (Clinical Safety Data
Management: Definitions and Standards for Expedited Reporting) does
specify the key data elements for inclusion in expedited reports of seri-
ous unexpected adverse drug reactions, and ICH E2B specifies data ele-

1
Morse, M.A., Califf, R.M., and Sugerman, J. Monitoring and ensuring safety during clinical research, Journal
of the American Medical Association, 285:1201-1205, 2001.
2
Moody, L.E. and McMillan, S. Maintaining data integrity in randomized clinical trials, Nursing Research,
51(2):129-33, 2002.
3
Enas, G.G. and Goldstein, D.J. Defining, monitoring and combining safety information in clinical trials,
Statistics in Medicine, May 15-30, 1995.
4
Ioannidis, J.P.A. and Lau, J. Completeness of safety reporting in randomized trials, Journal of the American
Medical Association, 285: 437-443, 2001.
5
Salsburg, D. Deming Principles Applied to Processing Data from Case Report Forms, Drug Information
Journal, (36): 135-141, 2002.

77

group6_PH.indd 77 7.8.2007 12:19:41

 ments that are to be included in the electronic transmission of expedited
individual adverse reaction reports to regulators, these standards were not
intended to specify all safety data that might be needed during the conduct
of a clinical trial. Study protocols should be the most important tool in de-
fining the methods for study conduct but are not always sufficiently specific
and complete regarding safety data surveillance and collection. It is useful
to establish a standard template for the safety sections of protocols, which
can be amended or supplemented as needed.
A sponsor’s study monitors (e.g., Clinical Research Associates or
CRAs) have a significant influence on assuring accurate and proper ad-
verse event reporting from study sites. Among their responsibilities, CRAs
must assess the completeness and accuracy of safety information, identify
omissions, and bring appropriate safety reports to the attention of the phar-
macovigilance department in a timely fashion. A useful assessment of their
role has been published.6
Although this and other chapters focus on new product development
and therefore on Phase I-III trials, the role of Phase 4 studies in under-
standing a product’s safety profile should not be underestimated. Phase IV
trials are generally distinct from large, post-marketing surveillance (PMS)
and observational studies, but may form part of a commitment required by
regulators as a condition for approval to market a drug (post-authorization
study requirements). Phase IV studies make an important contribution in
expanding the clinical trial database. Although safety monitoring during
these types of studies may not require the same intensity as for Phase I-III
trials, the principles and ideas presented here remain applicable.
Phase IV studies that mimic clinical practice (reflecting routine ad-
ministration of the drug) and involve large numbers of patients, whether
of comparative design or not, may require some routine, general safety
monitoring. In contrast, more intense safety monitoring is critical in peri-
approval studies (initiated near the completion of registration studies and
often called Phase IIIb) where the parameters may be similar to those in-
cluded in Phase III trials. In some cases, specific focus may be needed for
a safety issue that requires exploration. This also applies to post-marketing
studies for orphan drugs, where although the primary objective may be fur-
ther assessment of efficacy, safety monitoring is also a critical element due

6
Nylen, R. A. The Impact and Responsibilities of the Clinical Research Associate (CRA) on the Accuracy of
Adverse Event Reporting, Regulatory Affairs Focus, p. 16-20, April 2000.

78

group6_PH.indd 78 7.8.2007 12:19:41

 to the limited patient exposure prior to approval. Similar and additional7
considerations apply to vaccines and to drugs given expedited marketing
authorization (e.g., anti-HIV and oncology medicines).
Sponsor requirements for investigator sites differ regarding collection
of signs and symptoms when a diagnosis is also specified, what data ele-
ments to collect, when collection begins and ends, how promptly safety
data must be reported to the sponsor, and how investigators should conduct
causality assessments. For an indication of the variability among sponsors
surrounding data collection, see items 3 through 8 of the survey results in
Appendix 3. Differing safety data terms, definitions and collection methods
requested by different sponsors may lead to confusion and inefficiency on
the part of investigators. One of the most important issues that is rarely
addressed is the manner in which safety experiences are actually elicited
during discussions with patients by the investigator and his/her staff during
visits or at other times.
Consistency in safety data collection and handling practices can con-
tribute to greater efficiencies in the conduct of clinical trials. This should
result in greater confidence in the data available for analysis and allow in-
vestigators, sponsors and regulators to focus more time on review of the
data, thereby promoting the health and wellbeing of clinical trial patients/
subjects as well as future patients who stand to benefit from the therapy.
The remainder of the Chapter discusses various approaches to those is-
sues by addressing Who is responsible for collecting the data, What should
be collected, How should the data be gathered, When, and some technical
considerations for managing the data once collected.

b. Who?
The collection of data originates with the patient/subject, caregiver
or legal representative of a patient in clinical trials. However, it is usually
the investigative site (investigator and his/her staff) that is responsible for
gathering data from the patient, recording the information properly, and
ultimately reporting to the sponsor. Even though patients may be collecting
data in diaries or in electronic format, our primary focus is on the collection

7
The populations in many vaccine pre-licensing programs are fairly large but still quite small in relation to
the intended general population (usually children). They are also traditionally of short monitoring duration.
Late sequelae are difficult to detect with reliability and precision. The design of post-authorization studies is
therefore critical (e.g., cluster designs where the program starts with a planned geographical distribution so
that comparative populations exist in different locations (seasonal and population controls)).

79

group6_PH.indd 79 7.8.2007 12:19:41

 of data at the investigative site. It is the responsibility of the investigator to
ensure that patient data are properly collected and reported to the sponsor.
In certain Phase I trials a sponsor may also act in the role of the investigator
and would assume these responsibilities.
In general the personnel at the site are the patient’s primary contact
during a trial. All site personnel and the investigator must ensure that safety
data are properly collected and forwarded to the sponsor. Although person-
nel other than the investigator may obtain adverse event information during
regular communication, even between visits, it is ultimately the responsi-
bility of the investigator to ensure that information is collected in accor-
dance with the study protocol. Study monitors representing the sponsor
will review source documents against case report form entries to check for
accuracy and completeness in recording of the data, and to ensure that there
is conformity with the protocol. Sponsors have a critical role in clearly de-
fining the data to be collected as well as the process the investigator should
use in recording these data. However, if an investigator becomes aware of
information that is considered to be important for safety reasons it should
be reported to the sponsor (immediately if judged critical), even if the pro-
tocol does not specifically state that the information must be collected. To
assure the investigator’s sensitivity to this point, one of the key responsibili-
ties of the sponsor includes proper training of the investigative site person-
nel regarding data collection and reporting.
Many studies involve collaboration with contract research organiza-
tions (CROs), public and private institutions, other collaborative groups,
and co-development partners. In all of these arrangements, data collection
is the responsibility of the investigator. Clear agreements must be reached
and documented among the collaborating partners as to who is responsible
for monitoring the study and retrieving and processing the data. Many of
the sponsor responsibilities involving data processing may be delegated to
a CRO.8
Studies not sponsored by the manufacturer of an approved medicine
can, of course, be conducted by independent investigators and their insti-
tutions (public or private), who take on the roles and responsibilities of
a sponsor in processing and analyzing safety data. However, if a compa-
ny provides any support for such an independent trial (supplies, research
grant, etc.), the company should still obtain at a minimum all reports of

8
For example, see US FDA Regulation 21CFR312. 52.

80

group6_PH.indd 80 7.8.2007 12:19:42

 serious suspected adverse reactions from the investigational site(s).9 Some
companies require reports on all serious events. Arrangements must be
made between the parties to ensure that the required obligation is fulfilled.
Investigators are expected to comply with any local regulations regarding
their reporting to authorities of adverse experiences during clinical trials.
However, once the relevant reports are received by the company, it should
document them in its own database and take them into account during their
ongoing safety assessments and when preparing appropriate periodic safety
reports (e.g., PSURs).10
Availability of the final results as a report and/or publication should also
be part of an agreement between the investigator and the manufacturer. Ani-
mal studies on an approved/marketed product may also be conducted inde-
pendently of a manufacturer; again, if support is provided (usually supplies
of product or active moiety), it is incumbent on the manufacturer to ensure
that results are made available. For more details, see Chapter 7, Section d.

c. What?
(1) General Principles
Data that should be collected and evaluated for safety will depend
on the design of the clinical trial but may include: adverse events (exper-
iences), laboratory values, pharmacokinetic data, results of mental and
physical examinations, special study data (e.g., Holter monitors, EEG,
ECG, audiology testing, pregnancy testing, etc.), pharmacogenetic data
and quality of life data. Patient demographics, study medication doses and
duration, a measure of medication compliance, concurrent medical con-
ditions, and concomitant medications are also extremely important in the
interpretation of safety data. Additional items such as exercise history
may be helpful in understanding changes in values such as CPK and liver
enzymes. However, investigators are frequently asked to collect data that
are never utilized, wasting the time and resources of both investigators and
sponsors. Therefore, sponsors must carefully pre-select the data elements

9
There is no standard definition of what constitutes “support” by a company. For example, does it include
medical and/or regulatory review of a protocol on request to a company by an independent investigator? Some
companies are known to consider any interaction of this sort to constitute support and therefore enter into an
agreement with the investigator to receive safety information.
10
The MHLW (Japan) is encouraging independent investigators to conduct research that manufacturers of
approved medicines may not wish to do on new uses (indications, for example); companies would be required
to provide drug supplies and to maintain awareness of important safety findings. This proposal is under con-
sideration.

81

group6_PH.indd 81 7.8.2007 12:19:42

 that are necessary for analyzing the safety (and efficacy) of all treatments
and clearly state them in the study protocol and/or case report forms. Re-
ports of serious adverse events typically require more detail than non-seri-
ous (see section c. (7) below and Appendix 6). In early phases of drug de-
velopment, it is generally necessary to collect more comprehensive safety
data than in post-marketing studies. In addition, certain drug types may
require longer routine follow-up as in the case of vaccines, immunothera-
pies and some biotechnology products.
The collection, monitoring and assessment of data from Phase 1 stud-
ies deserve special attention for two reasons: (a) with some exceptions
(e.g., oncology medicines, pharmacokinetic studies in subpopulations such
as the organ impaired), such studies are conducted in healthy volunteers for
whom there is no anticipated health benefit and (b) the results are critical
to the future development of the product and must be scrutinized and inter-
preted with great care. For prophylactic treatments and preventive vaccines,
the same considerations apply even to later stage clinical trials.
As explained by Salsburg,11 it is highly unlikely that a case report form
will ever contain data fields for all data that might ever be needed for evalu-
ation of all possible safety concerns. He also describes issues surrounding
the collection of “excessive” data and the negative impact on data qual-
ity. Therefore, case report form fields should be chosen based on the data
elements that will be analyzed and can be typically presented in tabular
compilations of study results. Safety data that cannot be categorized and
succinctly collected in predefined data fields should be recorded in the
comment section of the case report form when deemed important in the
clinical judgment of the investigator. Because comment sections are not
easily coded, they should be used in connection with a standard AE section
of the CRF and instructions as to their use given to the investigator during
pre-study training.
Prior to study initiation, consideration should be given to how certain
data are to be collected: adverse events; diagnoses with or without accom-
panying signs and symptoms; clinical outcomes; causality assessments; se-
rious and “medically significant” cases; as well as adverse events of special
interest (see section c.(4) below and Appendix 1). It is also helpful to de-
cide where in a CRF, if appropriate, non-protocol-related diagnostic and/or

11
Salsburg, D. Deming Principles Applied to Processing Data from Case Report Forms. Drug Information
Journal, (36): 135-141, 2002.

82

group6_PH.indd 82 7.8.2007 12:19:43

 treatment-emergent procedures should be captured. All of these items re-
quire definition and specification.
During clinical development, knowledge of the safety profile of the in-
vestigational product is limited. There are no definitive methods for distin-
guishing most adverse drug reactions (events that are causally attributable
to study therapy) from clinical adverse events that occur as background
findings in the population and have only a temporal association with study
therapy. The CIOMS VI Working Group thus recommends the following:
All adverse events, both serious and non-serious, should be collected
for any clinical trial during development, regardless of presumed re-
lationship to the study agent by the investigator or sponsor, in order
to allow for subsequent assessment of causality using standardized
methods for individual cases and aggregate data. This applies not only
to the experimental product but to placebo, no treatment, or active
comparator.
In studies initiated during the immediate post-approval period, it is
prudent to continue this practice. Once the safety profile of a marketed
product is judged to be well understood and established, it may be
acceptable to collect less data. While detailed information on serious
adverse events should always be collected, for well-established prod-
ucts it may be appropriate to collect non-serious adverse events only
if suspected by the investigator to be related to the compound. This
would be especially appropriate for large scale, simple post-marketing
trials when the population, indication, and doses are consistent with
those included in the approved use(s) of the drug.
In addition to the above recommendation, it may be of interest to col-
lect non-serious event reports that led to discontinuation from treatment;
this could be important in studies of slightly different populations than
studied during development, for example.
The collection of comprehensive laboratory chemistry data during
post-approval studies is usually not necessary. As for all studies, the protocol
should clearly specify what adverse event and lab data must be collected.
Finally, a commonly overlooked but potentially important aspect
of data collection relates to the possible use of herbal and other non-
traditional remedies by patients/subjects, who typically do not regard
such treatments as drugs or medicines. It is therefore important to inquire
specifically about their use since their concomitant use with study treatment

83

group6_PH.indd 83 7.8.2007 12:19:43

 can lead to adverse drug interactions.12 Recent classification and coding
schemes for herbal medicines are available.13 In addition, readers may be
interested in a recently organized information exchange process designed
to strengthen the scientific basis for standards governing the safety, quality
and efficacy of herbal medicines under the Western Pacific Regional Forum
for Harmonization of Herbal Medicines (FHH).14

(2) Causality Assessment

Investigators must inform the sponsor of serious adverse events as soon
as they become aware of them and, by using clinical judgement, should as-
sess the potential link to the drug treatment. Some hold the opinion that
causality determinations on individual case reports are a “waste of time”
especially for randomised studies. However, while individual case causality
assessment may be difficult for both investigators and sponsors, the inves-
tigator’s opinion contributes to the sponsor’s decision on the necessity for
expedited reporting to health authorities – a requirement that depends on
individual case attribution. Causality judgments based on analysis of mul-
tiple cases/aggregate data are almost always more meaningful and typically
have a greater impact on the conduct of clinical trials, including changes
to informed consent documents, study design, and core safety information.
However, while aggregate assessment of data is ultimately a more reliable
indicator of drug-event attribution, causality assessment of individual ad-
verse events by the investigator may play a role in the early detection of
significant safety problems, and contribute especially to understanding rare
events. The investigator is in the best position to judge any unusual changes
in the status of the patient that might be related to the administration of the
study medication or a study intervention. He/she should know the baseline
condition of the patient and therefore should be able anticipate the normal
clinical course that the patient is expected to follow. Therefore, the inves-
tigator’s opinion on the relatedness of the event to the study treatment or
intervention should be solicited when serious adverse events are reported.

12
See Brazier, N. C. and Levine, M. A. H. Understanding drug-herb interactions, Drug Information Journal,
12:427-430, 2003 and Willis, J. Drug interactions – when natural meets ethical, SCRIP Magazine, Issue 91,
pp. 25-27, June 2000.
13
See Guidelines for Herbal ATC Classification and Herbal ATC Index, the Uppsala Monitoring Centre,
Uppsala, Sweden, 2004. Also, see WHO Guideline on Safety Monitoring of Herbal Medicines, ibid. Herbal
substances are recorded in the WHO-Drug Dictionary. For detailed information, see www.umc-products.com
and www.who-umc.org.
14
The Forum provides an active means for regulatory authorities to share information, coordinate efforts and
transfer expertise. See http://www.fhhm.net.

84

group6_PH.indd 84 7.8.2007 12:19:44

 Collection of investigators’ relatedness assessments for non-serious events
adds little value and is not needed for routine regulatory reporting.
It is recommended that investigators not be asked routinely to indicate
causality information for non-serious adverse events. However, there
may be circumstances when such assessments are useful and impor-
tant, such as for non-serious adverse events of special interest.
Companies ask investigators to utilize various methods and terminologies for
categorizing the “likelihood” that a serious adverse event is caused by the drug.
Terms such as likely, unlikely, possible, probable, definite, definitely not, remote
likelihood, and cannot-be-ruled-out have been used. Although various compa-
nies have used several methods that imply different degrees of causality:
The CIOMS VI Working Group recommends that the investigator be
asked to use a simple binary decision for drug causality (related or not
related) for serious adverse events.
While there is rarely sufficient information and experience to assign
causality to an adverse event definitively as “yes” or “no”, the various gra-
dients of relatedness offer little or no advantage in data analysis or regula-
tory reporting. Initially, causality assignment is used mainly as a prioritiza-
tion tool for deciding on whether an individual case must be reported to
the regulators. Furthermore, there is very little agreement among differ-
ent people on the meaning and weight of the terms (probably vs. possibly
vs. likely, etc.) even within the same language, but is even more disparate
across languages. One possible approach that has been suggested is to ask
simply whether there is a “reasonable possibility” or “no reasonable pos-
sibility” that the study treatment caused the event; alternatively – Was there
a reasonable possibility? Yes or No. Finally, irrespective of any causality
assessments, aggregate analysis will be conducted with all the data. The use
of “unknown” or “cannot-be-ruled-out” also adds little value in early de-
termination of safety concerns. The use of “cannot-be-ruled-out” to imply
drug relatedness would lead to excessive over-reporting and excess noise
in the system. It is virtually impossible to completely rule-out the role of a
drug in causing an adverse event in single-case reporting.
While not unanimous in the above recommendation, the “binary”
decision choice was the method favored by a majority of the CIOMS VI
Working Group members.15

15
See the Glossary (Appendix 1) for more discussion under Adverse Drug Reaction.

85

group6_PH.indd 85 7.8.2007 12:19:44

 To facilitate the process by which an investigator judges the cause of a
serious adverse event, the Working Group advocates adoption of the recom-
mendation by the CIOMS III/V report on core safety information and the
DCSI (Development Core Safety Information), namely that on the CRF and
on any serious adverse event form there be included a standard list of poten-
tial causes from which the investigator must choose the most plausible one
in his/her opinion, specifically: medical history; lack of efficacy/worsening
of treated condition; study treatment; other treatment, concomitant or previ-
ous; withdrawal of study treatment (a withdrawal reaction could be considered
drug-related); erroneous administration of treatment; protocol-related proce-
dure; other – specify.16
The CIOMS VI Working Group recommends inclusion of the CIOMS III/V
checklist of potential causes of a serious adverse event on the reporting
forms used by the investigator. If an investigator considers that an event
is not drug related, the most likely other cause(s) should be indicated.
The CIOMS III/V report also provides criteria that can be helpful in
assessing causality for both individual cases and series of cases (aggregate
data) with a goal of deciding when the threshold has been reached for add-
ing new adverse drug reactions or other safety data to product information.
In the context of clinical trials, these same criteria with some additional
considerations is helpful for deciding when it is appropriate to add infor-
mation to the Investigator’s Brochure/Development Core Safety Informa-
tion (see Appendix 7). Investigators are usually asked to make causality
decisions on individual cases and study start-up should include training
for making such assessments. Much of the material in Appendix 7 can be
helpful in this regard.

(3) Diagnoses vs. Signs and Symptoms

An investigator’s expertise is important in aiding the sponsor to inter-
pret adverse events, especially in providing a diagnosis, if applicable. Some
sponsors request that an investigator record all signs and symptoms as well
as a diagnosis when possible. Others ask for just the diagnosis. If an inves-
tigator participates in trials involving different sponsors this may lead to
confusion and inconsistencies in how the data are recorded. The collection
of non-specific signs and symptoms rather than diagnoses or syndromes
often leads to extensive lists of these events in product information,

16
Guidelines for Preparing Core Clinical-Safety Information on Drugs, Second Edition, Including New
Proposals for Investigator’s Brochures, Report of CIOMS Working Group III/V, CIOMS, Geneva, 1999.

86

group6_PH.indd 86 7.8.2007 12:19:44

 resulting in limited usefulness to prescribers. Therefore, the CIOMS Working
Group recommends the following:
The investigator should be encouraged to evaluate the events of trial
patients and record on the case report form a diagnosis (when possible
and appropriate) rather than each individual sign and symptom. This
instruction should be clearly specified in the protocol. However, when
an investigator submits a serious adverse event report that includes a
diagnosis it is important that the signs and symptoms as well as any
other supporting information that led to the diagnosis also be recor-
ded, specifically as part of the narrative description of the case.
The advice to collect and document signs and symptoms for serious AEs
would seem to contradict the recommendation given earlier not to collect
extraneous or redundant information; it also may be in conflict with the
MedDRA® Points to Consider document (par. 2.5.3).17 However, knowledge
of signs and symptoms is especially important for serious adverse event cases
that may have to be reported to regulators promptly; frequently there may not
be enough information available to provide a confirmed diagnosis. As addi-
tional information becomes available, such as results of laboratory work and
diagnostic work-ups, the original presumed diagnosis may need to be changed.
A description of signs and symptoms may also be important in certain trials
such as Phase I studies or in a situation where an investigator is unable to
make a confirmed diagnosis. As an aid to making a diagnosis, it might be use-
ful to refer to an existing CIOMS guide that provides diagnostic standards for
adverse reactions, that would enhance accuracy and consistency in the use of
ADR terms.18 Training of the investigative site in the proper use of the relevant
diagnostic terms is important for consistent data collection.
Prior to study initiation, it is recommended that specific criteria for
identifying and defining significant, anticipated adverse events be
established and communicated to investigators involved in the detection,
assessment and reporting of adverse events.
An example might be significant liver function test elevation, defined as
three or more times the upper limit of normal, which is often used as a crite-
rion. Such definitions and criteria should be included in the safety section of
the protocol.
17
See http://www.ich.org/ichMedDRA_PTC.html
18
Reporting Adverse Drug Reactions: Definitions of Terms and Criteria for their Use, Edited by Z. Bankowski,
et al., Council of International Organizations of Medical Sciences, Geneva, 1999. This report comes with a
CD rom for ease of use. Also, see Venulet, J. and Bankowski, Z. Harmonizing Adverse Drug Reaction Termi-
nology, Drug Safety, 19(3):165-172 (1998).

87

group6_PH.indd 87 7.8.2007 12:19:45

 (4) Adverse Events of Special Interest
It is useful to consider a class of adverse events that may not be seri-
ous but have special meaning or importance for a particular drug or class
of drugs. Although it is ordinarily unnecessary to create specific defini-
tions or criteria for non-serious adverse events, it is important to do so
for apparently non-serious events that might be precursors (prodromes) of
more serious medical conditions; for example, muscle pain and elevated
CPK together may be indicative of potential rhabdomyolysis. Other types
of non-serious events may be important in and of themselves, such as
those that could affect quality of life in a meaningful way (e.g., impo-
tence, hair loss). Such events examples of what are often referred to as
adverse events of special interest, when there is evidence or suspicion of
their potential importance. For more detailed discussion, see the Glossary
(Appendix 1).
Toxicology studies and other non-clinical research may suggest the
potential for serious adverse events in humans. Prior to initiation of clini-
cal trials, the sponsor may identify adverse events of special interest from
these data, or from experiences with similar compounds, and require spe-
cial collection and reporting by the investigator. For example, if a com-
pound in development has been demonstrated to have the propensity to
cause tachycardia in pre-clinical studies or if this is a concern with other
compounds in the same class, it would be prudent to proceed with caution
in any human trials. ECGs should therefore be monitored and tachycardia
routinely reported by investigators to sponsors for all subjects/patients un-
til the risk to humans is delineated. While animal studies may or may not
be predictive of potential human toxicity, they cannot rule out all potential
toxicity.
It is important to define clearly “adverse events of special interest”
in the protocol and to specify close monitoring and prompt reporting
to the sponsor of these types of events, even if the event is considered
non-serious according to the usual regulatory criteria.

(5) Laboratory Chemistry Measurements

The use of clinical lab tests as surrogate markers for toxicity in early
clinical studies is critical. Laboratory assays such as haematopoietic (CBC
and cell differential), biochemistry panels (e.g., musculo-skeletal, renal,
hepatic, cardiovascular and lipid metabolism assays), and urinalysis results
should be collected in all early studies. More targeted laboratory investiga-
tions involving areas such as endocrine, coagulation, immunologic, and

88

group6_PH.indd 88 7.8.2007 12:19:45

 reproduction systems may be required based on results of early toxicology
studies. Certain laboratory parameters may also qualify as adverse events
of special interest and require more frequent testing and evaluation.

(6) Morbidity and Mortality as Efficacy Endpoints

In studies involving a disease state associated with significant morbid-
ity or mortality (for example, cancer, sepsis, AIDS), it may be appropriate
to collect certain medically anticipated clinical events only as clinical ef-
ficacy outcomes rather than adverse events. An example would be death as
the result of progression of breast cancer. In studies without such an antici-
pated clinical endpoint, any event resulting in death would be considered
a serious adverse safety event. Collection of clinical outcomes may allay
some of the burden for investigators in having to report all disease-related
events as serious safety adverse events in studies involving severe illness.19
The method of collection may differ from that of serious adverse events in
that it may be more streamlined (less data) and batched (sent in weekly, for
example, rather than immediately). The collection process should be clear-
ly delineated in the protocol. ICH Guideline E2A describes the conditions
for managing such situations. Under such a process, it may then be appro-
priate (although admittedly somewhat problematic) to enter the cases only
in the clinical trial database, but not into the separate safety database which
most companies maintain for serious clinical trial cases and all spontane-
ous reports from marketed products (see Sections d.(2) and f. below). On
the other hand, if a patient experiences a suspected serious adverse event
at the same time as the designated efficacy endpoint event, all the informa-
tion on both events should be included in both databases.
It is recommended that even when anticipated medically serious clini-
cal events are collected as clinical efficacy outcomes/endpoints, rather
than as adverse events, these data must be recorded by the investigator
and periodically reported to and reviewed by the sponsor or DSMB, on
a schedule specified in the protocol.
The protocol should also specify how promptly and frequent report-
ing should be. It should also be made clear how often these data will be
reviewed, how they will be reviewed (blinded or unblinded), and by whom,
including the use of Data and Safety Monitoring Boards, as needed. In the
course of reviewing such cases, it may be important to consider whether

19
Nichas, J. Clinical Trial Safety Surveillance in the New Regulatory and Harmonization Environment: Lessons
Learned from the “Fialuridine Crisis”, Drug Information Journal, (31): 63-70, 1997.

89

group6_PH.indd 89 7.8.2007 12:19:46

 study therapy could have had the paradoxical role of worsening the clini-
cal outcome (see footnote 19). Prior to trial initiation, agreement should be
reached with regulators in all countries where a study is conducted as to
how clinical endpoint data will be reported.

(7) Special Situations

Investigators should be sensitized to the concept that even when in-
formation is not considered adverse event data, it should be forwarded
promptly to the sponsor if it can possibly contribute to the overall knowl-
edge concerning safety of the compound. For example, any deviation from
specified doses as defined in the protocol (especially doses that are higher
than recommended) should be reported to the sponsor in the same time
frame as for serious adverse events even if there are no associated events.
Medication errors, including inappropriate route of administration, should
also be reported promptly. Companies may wish to use their serious AE
forms for convenience, or some other process for collecting the relevant
information.
Pregnancies occurring during clinical trials present a unique situation.
Any pregnancy that occurs in a female trial participant during a clinical trial
should be followed to termination or to term. Under special circumstances,
it may be necessary to monitor the development of the newborn for an ap-
propriate period post-delivery. There may also be special situations when it
will be necessary to monitor the pregnancy of a woman whose male partner
is the trial participant (e.g., class effects, evidence from animal reproduc-
tive studies). Partner privacy may become an issue in follow-up for these
situations. The protocol should describe in detail the process for monitoring
and managing pregnancy occurrences.
The collection of genetic data for safety purposes continues to gener-
ate much debate20 21 22 23. This topic is beyond the scope of this project but
has been considered by another CIOMS Working Group.24

20
Freund C.L., Wilfond, B.S.. Emerging ethical issues in pharmacogenomics, American Journal of Pharmaco-
genomics, 2(4): 273-281, 2002.
21
Roses, A.D. Pharmacogenomics and the future of drug development and delivery, Lancet , 355: 1358-1361,
2000.
22
Sander, C. Genomic medicine and the future of health care. Science, 287: 1977-1978, 2000.
23
Polymeropoulos, M.H.. Application of genetics and genomics in drug development, Drug Development
Research, 49: 43-45, 2000.
24
Pharmacogenetics – Towards Improving Treatment with Medicines. Report of a CIOMS Working Group,
CIOMS, Geneva, 2005.

90

group6_PH.indd 90 7.8.2007 12:19:46

 As part of recommendations for good follow-up practices, the
CIOMS V Working Group25 provided listings of data elements that should
be collected for various types of adverse event cases, depending on the
seriousness and expectedness of the case. While the CIOMS V formula-
tion primarily focused on post-marketing cases, the same elements, and
more, are important for clinical trial safety monitoring (see Appendix 6).
Consideration should be given to the collection of as many of these data
elements as possible, as part of the CRF for each patient. When a serious
case is reported, these elements should be collected, even if they are not
part of the CRF.
While different companies use different forms for collecting data on
serious and special adverse event cases from investigators, there has been
some interest in the development of a standard form that might be used by
all sponsors when possible. In the CIOMS VI survey, a majority of the re-
spondents (16 of 21) would support the use of a global form (see item 9 in
Appendix 3). It is recognized that the format and contents of data collection
forms are often dependent on the user’s internal standards and established
computer systems. However, as an illustration of what a prototype form
might look like, the Working Group presents an example in Appendix 8.
The Working Group is not proposing that this example become a standard
but is providing it for those who may wish to create their own form.
No matter what form is used, it is strongly recommended that the choice
of data elements and their definitions conform to those specified un-
der ICH Guideline E2B to facilitate the sponsor’s data processing and
eventual electronic transmission, as needed.

d. How?
(1) General Considerations
Various methods exist for collecting safety as well as efficacy data.
Most clinical trial data are collected on paper case report forms (CRFs)
or by electronic means.26 The method should be clearly defined in the pro-
tocol. The increasing use of wireless and Internet technologies by many
sponsors in an attempt to increase study and data management efficiency

25
Current Challenges in Pharmacovigilance:Pragmatic Approaches. Report of CIOMS Working Group V,
pp.128-130, CIOMS, Geneva, 2001.
26
Ruberg, S.J., McDonald, M. and Wolfred, M. Integrated electronic solutions, Applied Clinical Trials, 11(2):
42-9, 2002.

91

group6_PH.indd 91 7.8.2007 12:19:47

 introduces new issues in the move from paper-based to electronic records.27
Usually patients report their symptoms during examination by the investi-
gator or site staff at protocol-specified visits. During or after the examina-
tion, investigators or their assistants record adverse events along with other
relevant findings on the CRF. Except for in-patient studies, the existence
of acute, medically serious adverse events requiring emergency care will
usually first be learned by phone or sometimes through an emergency room
physician. Inter-institutional communication is quite important under such
circumstances. It is usual practice that for serious adverse events, the in-
vestigator be asked to obtain copies of relevant hospital records to supple-
ment the usual CRF and serious AE form information on the patient. For
cases involving deaths, coroner’s reports and any autopsy findings should
be obtained. However, it is important that emphasis be placed on the need
for the investigator to complete and submit the company’s special serious
AE form; the often voluminous supplemental records that may be obtained
can be uninterpretable.
Before further addressing the mechanism of how safety data should be
collected, it is important to consider the basis of all patient data collection:
the interaction and dialogue between patient and investigator site person-
nel. The measurement of objective parameters, such as lab tests, electro-
cardiograms, etc., is reasonably straightforward and generally does not
have a strong subjective component. In addition, site professionals should
be observant for signs and symptoms suggestive of adverse effects (rash,
etc.). However, there are many ways in which investigators and their staff
can and do solicit information and opinions from trial participants and they
are not all equivalent or consistent in their ability to elicit complete, mean-
ingful and unbiased data. For example, at each visit open-ended questions
might be asked, such as “Has the medication affected you in any way?”
(which might imply a suspicion that it could), or “Have you experienced
any ill effects from your treatment?” (a leading question which could
influence the patient to associate any untoward event with the treatment). In
some situations, the patients may be asked to keep a log of their experiences
between visits. In such a log, or even during face-to-face questioning, the
patient may be presented with a list of possible adverse experiences (“Have
you had any headaches, nausea,…?”). Other possibilities for soliciting this
type of information include the use of electronic, menu-driven interviewing
techniques.

27
Clinical Trials and the Internet, R&Directions, November/December 2001, p. 34-48.

92

group6_PH.indd 92 7.8.2007 12:19:47

 This area has not received much attention, but the CIOMS Working
Group believes that it can be very important and recommends the following:
The process used to solicit information from patients during clini-
cal trials should be consistent from site to site, and if possible from
program to program, and should be clearly outlined within study pro-
tocols, in the informed consent information, and during investigator
training. No matter what method or approach is used, it should be used
consistently throughout the trial, including at baseline (pre-treatment
information).
It is probably best to frame questions to the patients in general terms
rather than to invoke the possibility that study treatment may be re-
sponsible for ill effects. For example: “How have you felt since I saw
you last? Is there anything new that you wish to discuss?”
Although it is not advisable to read a specific list of possible ADRs
when soliciting the patient’s recent experience, patients should be
alerted to known signs and symptoms indicative of medically impor-
tant suspected or established ADRs in order to alert the investigator as
early as possible.
An example of the latter situation is muscle pain and/or tenderness in
trials of HMG CoA enzyme reductase inhibitors (i.e., statins) which could
possibly be associated with rhabdomyolysis. Patients can be advised during
the informed consent process or perhaps with a handout to be particularly
attentive to such important signs and symptoms and to mention them to the
investigator at the earliest opportunity. However, this type of “warning” to
patients should not be used routinely but only under special circumstances.
One particular difficulty in this area relates to the gathering of subjec-
tive data from patients who are unable to provide it, such as neonates and
infants, patients with Alzheimer disease, patients in a coma, and others for
whom a parent, home caregiver or other proxy represents and speaks for the
trial participant. To our knowledge, there are no international guidelines on
how such situations should be managed.28 However, as with more normal
circumstances, for studies involving such patients the process for obtaining
data should be described in the protocol and informed consent information.

28
For one region’s example, see Adults with Incapacity Act 2000 (Scotland; see http://www.scotland.gov.uk/
Topics/Justice/Civil/16360/4927#mod39793) and UK Department of Health’s Draft Guidance on Consent
by a Legal Representative on Behalf of a Person Not Able to Consent Under the Medicines for HumanUse
(Clinical Trials) Regulations 2003. For more discussion, see Chapter 14 of Medical Ethics Today, 2nd edition,
British Medical Journal Press, 2004.

93

group6_PH.indd 93 7.8.2007 12:19:48

 Many companies prepare study manuals to supplement protocols, in
which high levels of detail on processes and procedures are described; this
would be another place in which this subject can be covered.

(2) Serious and Other Important Adverse Events

Typically, sponsors require that investigators immediately notify them
when a serious adverse event occurs. This may be reported verbally by
telephone, by faxing a reporting form that is distinct from the CRF, or by
electronic means. As mentioned above, in an effort to simplify the process
and provide consistency and less confusion to investigators in reporting
serious adverse events to sponsors, a standardized form that could be com-
pleted by investigators might be considered (Appendix 8).
Laboratory, biopsy, ECG, EEG, audiology testing and other special
study data may be generated from local or central laboratories or clinics.
The investigator should arrange to receive immediate notification of any
alarming results. These should then immediately be brought to the attention
of the sponsor as well. Processes describing collection and notifications
should be specified in the protocol. Obviously, where appropriate, refer-
ence standards for laboratory values should be obtained by the investigator
and sponsor for proper interpretation of the data.
Most sponsors maintain two databases that contain safety data. One
contains serious adverse event cases that may require expedited regulatory
reporting as well as cases from ongoing surveillance activities on marketed
products (e.g., spontaneous reports). It would also be advisable to include
non-serious adverse events of special interest. This database (the “safety
database”) is used to accumulate safety data on the compound as it pro-
gresses through development and during marketing. The other contains all
of the safety, efficacy, and other data from the clinical trial, including seri-
ous and all non-serious adverse events. This clinical trial database, unlike
the usually separate safety database, is typically closed and “locked” for
analysis once the study is complete. It is important for the sponsor to have
clear policies and procedures for processing of these data and for ensuring
that the data within the two databases are consistent and any differences
reconciled when necessary.29 Attention must also be paid to the possibility
that information in the safety database may be updated after a study is

29
Some companies maintain a minimum set of data elements that must be reconciled between the two data-
bases, such as: project/protocol number, investigator number, patient initials and/or number, gender, birthdate,
verbatim AE terms, onset date of AE, severity of event (if used, e.g., mild, moderate, or severe), criteria for
serious if case is serious, and investigator’s causality assessment.

94

group6_PH.indd 94 7.8.2007 12:19:48

 completed and the clinical trial database has been frozen. Whether any
changes are needed to the final study report or in the analysis of the data (which
may already have been completed) will require judgment and depend on the
importance of the information in categorizing the safety profile (and
possibly the benefit-risk relationship) of the product.

e. When?
A period of observation must be defined in the protocol for each study.
Typically, the time that the informed consent is signed by the patient is
designated as the start of safety data collection (see survey results for item
4 in Appendix 3). This provides a clear starting point and helps to avoid
any selection bias. If a patient will not formally enter a trial until several
days or longer after the informed consent is signed, the day of random-
ization to treatment may be a more appropriate time to begin collecting
safety information. Adverse events occurring prior to randomization would
be considered as medical history or pre-existing conditions. It is impor-
tant to collect such information in order to place into perspective “study
treatment-emergent” findings; for example, the occurrence of nausea after
informed consent signing, but prior to administration of study treatment,
would be useful information.
In some studies it may be necessary to collect baseline safety data dur-
ing pre-drug therapy. If a washout period is included as part of the protocol,
with or without the use of placebo, safety data should be collected. This will
allow for assessment of any worsening of study treatment-emergent condi-
tions. It is also possible that an invasive procedure will be used as part of
screening prior to study inclusion (e.g., tissue biopsy) that carries the risk
of adverse events; such data should be collected and be part of the overall
safety experience for the trial population. The start of collection of safety
data should be clearly indicated in the protocol. Once data are recorded,
they should be forwarded to the sponsor in accordance with the protocol’s
requirements so that they are available for safety monitoring.
The protocol should specify the observation period for the patient
following the last dose of study medication and/or the last protocol-
specified visit. Survey results show that this observation period varies
widely from company to company (see item 5 in Appendix 3). The protocol
should clearly specify how and when collection of safety data should occur
during a post-study observation period. It may be accomplished by addi-
tional visits or via telephone, for example.

95

group6_PH.indd 95 7.8.2007 12:19:48

 It is recommended that in general, safety data event-collection should
continue after the last dose of the drug for at least an additional five
half-lives.
This time will vary depending on the type of compound studied and
its specific characteristics. While this general guideline may apply to most
compounds, due to the diversity of products in development and patient-
specific circumstances it is difficult to create a rule that would be appro-
priate in all cases. For example, compounds such as cytotoxic agents may
have delayed toxicity that needs to be monitored over longer time periods.
On the other hand, for compounds with extremely long biological half-
lives (e.g., several years, as with bis-phosphonates), post-study monitoring
can be considerably shorter than the half-life. Organ impairment may pro-
long the drug half-life. The biological effects of some drugs may continue
beyond five half-lives. Again, the time period for collection as well as a
description of what is to be collected must be defined in the protocol and
factored into the time-lines anticipated for the clinical program.
If a patient is withdrawn from treatment due to safety reasons or if
a patient has an ongoing serious event or adverse event of special
interest at the end of the study, the patient should be followed until
the event disappears, the patient’s condition has stabilized, or until a
pre-defined outcome is reached.
Any patient who voluntarily withdraws from a study should be care-
fully questioned for the possible occurrence of an adverse event.
Whenever possible, a patient should be followed through the last
scheduled study visit even if the patient is withdrawn from treatment,
in order to allow for appropriate intent-to-treat analysis. (See Chapter
6 for more discussion).
The sponsor should be informed if the investigator becomes aware of
any unusual safety information or any safety information that appears to be
drug related involving a patient who had participated in a study, even after
an individual patient completes the study. An investigator should always
be diligent in looking for possible latent safety effects that may not appear
until after a medication is discontinued. Sponsors should encourage this
practice. An example would be the discovery of a suspected hepatotoxic
effect three months after a patient had completed a two-year study (with no
other plausible cause).

96

group6_PH.indd 96 7.8.2007 12:19:49

 f. Safety Data Management Considerations
Collecting the best data in the world is of no use unless they are prop-
erly documented and made available in a consistent and accurate way for
examination, analysis, presentation, reporting and sharing with appropriate
stakeholders, within and outside the sponsor’s organization. A certain level
of expertise as well as judgment are needed to ensure that adverse events
and other data (e.g., laboratory findings) are properly named, classified, and
coded when creating a database. This section provides some guidance and
recommendations to that effect.
Generally, in order to assure standardized signal detection and evalua-
tion processes, data quality and completeness are paramount.
The CIOMS VI Working Group recommends the following principles for
this important objective:
❏ individual case safety reports from studies should be as fully docu-
mented as possible
❏ there should be diligent follow-up of each case, as needed
❏ the reporter’s verbatim AE terms must be retained within all
relevant databases
❏ if the reporter’s AE terms are not considered to be clinically
accurate or consistent with standard medical terminology used for
coding, attempts should be made to clarify the description of the
event with the investigator. If there continues to be disagreement,
the sponsor can code the AE terms according to its judgment on
the case, but should identify them as distinct from the investigator’s
terms. Reasons for the difference(s) should be documented.
❏ personnel with knowledge and understanding of both clinical
medicine and the dictionary used should review all codified terms
to ensure consistent and accurate codification of reported (“verba-
tim”) terms.
❏ primary analyses of AE data should be based on the investigator’s
assigned terms or diagnoses, carefully and properly coded by the
sponsor; additional analyses using the sponsor’s assignments if any
are different can be conducted, but explanations for any differences
between the two analyses must be given.
The rest of this section elaborates on these principles.

97

group6_PH.indd 97 7.8.2007 12:19:49

 (1) Clinical Description of Adverse Events
There are no universally accepted criteria and definitions for many widely
used terms commonly used in drug research such as abnormal LFTs, hepatitis,
hepatocellular damage, hepatic necrosis, and various clinical syndromes. Care
should be taken to ensure that adverse events are not misclassified with inap-
propriate or even erroneous clinical terms. This situation is often exacerbated
when the event affects a body system that is outside the investigator’s clinical
specialty. For instance, in one antibiotic development program, five cases were
reported as “LFT abnormalities” or “hepatitis” without any abnormalities in
the relevant laboratory values. Other examples include reporting of “increased
LFTs” to describe a patient with jaundice, “acute liver failure” without jaun-
dice or encephalopathy, “leucopenia” to describe a case of agranulocytosis,
“aplastic anaemia” without reduction in all haematopoietic lineages, etc.
When available, relevant laboratory data, in addition to signs and symptoms,
should form part of the clinical evaluation of reported events.
Another common example is skin reactions, most commonly reported
simply as “rash”, without further description or characterization. The sever-
ity of rashes may be either over- or underestimated, such as the reporting of a
benign morbilliform rash as erythema multiforme, or a case with mild signs
suggesting possible Stevens-Johnson syndrome but reported as just “rash”.
Inappropriate clinical characterization can potentially obscure the presence of
a real safety issue. The CIOMS publication on criteria for diagnosis of many
types of adverse events, especially serious events, can assist sponsors in estab-
lishing standards in this regard.30
Individual case safety reports (ICSRs) must be categorized and assessed
by the sponsor using trained individuals with broad expertise in both clinical
medicine and codification. Investigators should be encouraged to obtain spe-
cialist consultation for clinically important events that occur outside their own
areas of clinical expertise, so that sponsors can obtain all information required
for subsequent safety evaluation. Examples include behavioral changes in as-
sociation with antibiotic treatment, cardiac symptoms in patients treated for
depression or schizophrenia, and persistent skin rashes with systemic agents.
Sponsors should also consider the use of questionnaires based on diagnostic
standards to collect the detailed information needed for the analysis of specific
events of major importance, such as liver injury, bone marrow suppression,
or cardiac arrhythmias. In certain situations, the sponsor may wish to seek

30
Reporting Adverse Drug Reactions: Definitions of Terms and Criteria for their Use, Edited by Z. Bankowski,
et al., Council of International Organizations of Medical Sciences, Geneva, 1999. This report comes with a
CD rom for ease of use.

98

group6_PH.indd 98 7.8.2007 12:19:50

 external consultation with an independent clinical expert or a group of inde-
pendent experts, for appropriate categorization and interpretation of adverse
events. If a Data and Safety Monitoring Board (DSMB) is used for a study,
then depending on its membership, it could fulfill this function in addition to
its usual roles and responsibilities of independent safety (and efficacy) moni-
toring.
The investigator always retains the right not to modify a term with
which the sponsor disagrees. As already mentioned, the sponsor should re-
tain the investigator’s verbatim term and document thoroughly the reasons
for a different opinion. Such disagreements are accommodated, for exam-
ple, within ICH Guideline E2B, field B.5.3 (Sender’s diagnosis/syndrome
and/or reclassification of reaction/event), which may be used for docu-
mentation purposes. While such significant discrepancies will probably
be exceptional, they must be clearly documented for analysis and audit.
Nevertheless, there may be merit in analyzing and evaluating “as is” ver-
batim information from investigators during early stages of development,
when the safety profile of the medicinal product is not well-characterized
or understood. However, as safety information increases and improves,
standardization of terminology and communication with investigators
regarding the use of standard terms and definitions should be considered.
Depending on their purpose, adverse event tables can display both the
reported (investigator’s verbatim) term31 and the sponsor’s terms. However, as
discussed in Chapter 5, primary safety analyses (especially those used to de-
velop the DCSI and CCSI) should be based on investigator-assigned terms.
Most AE reports consist of one or more signs and symptoms with no
particular diagnosis possible or relevant (e.g., headache, nausea), especially
during the early stages of clinical drug development. The challenge is to know
when a symptom/sign complex might represent a diagnosis of a potentially
important medical condition. Such information has value in terms of signal de-
tection and evaluation. As described in section c.(3) above, investigators should
be encouraged to record a diagnosis or syndrome as the adverse event when-
ever possible. Even if they do not, when reported signs, symptoms, investiga-
tion results, and/or treatment strongly suggest a known clinical syndrome (e.g.,
chest pain, elevated CK-MB, and acute treatment with a thrombolytic agent),
a probable diagnosis, in this case myocardial infarction, may be assigned for

31
The original term(s) reported by an investigator may be in a language different from that used by the sponsor
in its day-to-day operations and in coding. “Verbatim” in this context is meant to refer to a properly translated
version of the original term(s) into the working language of the sponsor.

99

group6_PH.indd 99 7.8.2007 12:19:50

 analysis by the sponsor, even if not reported as such by the investigator; this
option is provided for in an ICH Guideline E2B data specification (field B.5.3).
As already recommended, it is essential that all AEs be collected, regardless
of presumed relationship to the study agent, for subsequent assessment using
standardized methods, such as those described in Chapters 5 and 6, in order to
determine causality from aggregate data.
Certain events may be anticipated based on the class of drug used for
treatment. There may also be other events that require special attention based
on knowledge of their background rates for the type of population under study.
It is useful to specify such events, along with the criteria for their diagnosis, in
the protocol and other instructional material for the study sites. For example,
the definitions of drug-induced liver injuries and blood disorders developed by
CIOMS can provide a useful standard (see footnote 30).
In addition, it is critical that “adverse events of special interest” be defined
for the purposes of consistent analysis, assessment, and evaluation of the safety
profile of the medicinal product. These definitions and the criteria for use of
particular terms should be described in detail in the clinical protocol and any
developmental safety plan for the medicinal product.
To avoid inclusion in the DCSI and ultimately in the CCSI of multiple
event terms that provide little or no medically useful information, individual
signs and symptoms (e.g., fever, rash, and nausea) should be codified for analy-
sis only when they are reported as isolated terms and are not consistent with
a clear, specific diagnosis. An aggregate analysis should attempt to ascertain
whether the individually codified symptoms and/or signs occur in isolation or
as frequently reported combinations, even if they do not initially comprise a
recognized clinical syndrome. This is especially important to avoid inappro-
priate categorization of relatively non-specific signs or symptoms, e.g., fever,
which may have multiple unrelated causes.
Some companies and health authorities maintain a list of event terms that
are always regarded as medically serious and important even if the specific case
might not satisfy the criteria for serious in a regulatory sense (require expe-
dited reporting, for example). Such “always serious” events are used routinely
to trigger special attention and evaluation. Although such lists were originally
created for post-marketing purposes, especially for spontaneous reports, they
might be useful for pre-approval clinical research purposes.32 We do not en-
dorse any particular list since it may be highly dependent on the treatment
32
For a full discussion of this concept and an extensive table of MedDRA® and WHO-ART terms that were
suggested as candidates for such a list, see Current Challenges in Pharmacovigilance: Pragmatic Approaches,
Report of CIOMS Working Group V, pp. 107-108 and Appendix 5, CIOMS, Geneva, 2001.

100

group6_PH.indd 100 7.8.2007 12:19:51

 and the specific population(s) under study, and can never be complete. Even
without formalizing such a list, remaining alert to the appearance of certain
medically important events is a key concept in protecting trial participants and
preventing future harm.

(2) Coding Procedures

Sponsors should have in place standard procedures for data codifica-
tion applicable to all products and projects; all personnel responsible for data
entry should be well trained in their use. As developed under ICH Topic M1,
there is an internationally agreed medical coding terminology-dictionary,
viz., MedDRA® (Medical Dictionary for Drug Regulatory Activities), which
the Working Group recommends and will refer to throughout this discussion.
MedDRA® codification principles, “Term Selection: Points to Consider,” 33
should be used as the basis for sponsor procedures. However, the principles
covered here are independent of the coding dictionary used.
While a complete description of the signs, symptoms and investigations
which led to a diagnosis should be obtained from the investigator, especially
for serious events, and while such data should be part of the overall study
database, these details should not usually be coded for describing the spe-
cific event, as outlined in the MedDRA® Points to Consider document. The
impact of codification on ultimate data output for clinical evaluation must be
considered during data entry, especially when dealing with AE terms that do
not have exact matches in the codification terminology used. Non-specific
“disorder” terms (e.g., the MedDRA® Preferred Terms (PTs) “blood disor-
der NOS”, “cerebral disorder”) should be avoided as they are not useful for
retrieval, clinical analysis, or display; reported events that are so ill-defined as
to require the use of such non-specific terms should be clarified with the re-
porter. When using MedDRA®, it is generally recommended that terms from
the Social Circumstances System Organ Class (SOC) should be used only for
medical history and not for coding AEs, even if a reported verbatim term is
an exact match for a MedDRA® Lowest Level Term in that SOC.34
33
See http://www.ich.org/ichMedDRA_PTC.html
34
The CIOMS VI Working Group endorses the recommendation in the MedDRA® Points to Consider, v. 3.3
(9 June 2004) that the SOC Social Circumstances generally not be used for coding ADRs/AEs, even if a reported
verbatim term is an exact match for a Lowest Level Term in that SOC, because of the potential impact on retrieval,
analysis, and reporting. The Social Circumstances SOC describes social factors, and as such is intended for use in
coding social history data, and is thus not included in the multi-axiality of the clinical disorder SOCs. Using it to
codify clinical concepts that are more appropriately reflected by terms in a clinical disorder SOC could therefore
adversely affect data retrieval and signaling by mismapping to an inappropriate SOC. For example, the term
“Aborted pregnancy” in the SOC Social Circumstances would not be grouped or retrieved together with the multi-
ple clinical terms reflecting various types of abortion in the SOC Pregnancy, puerperium and perinatal conditions,
and hence could lead to inappropriate omission from an analysis of abortions/ miscarriages of events so codified.

101

group6_PH.indd 101 7.8.2007 12:19:51

 Sponsors should avoid “excessive coding” of events reported in seri-
ous adverse event cases. Each such report should contain only the mini-
mum number of dictionary terms needed to ensure retrieval in the relevant
clinical context(s). Conversely, sponsors should take great care not to “un-
dercode” events, namely, assign codes that might downgrade the severity or
importance of an event term or terms.
Inconsistencies in clinical event classification and/or codification are
common not only among investigators within the same study and/or project,
but also among sponsors using different adverse event coding terminologies/
dictionaries, and even among sponsors using the same dictionary (including
MedDRA®). The current practice of coding AEs in safety databases in strict
adherence to the “verbatim” terms reported by investigators, in the absence of
clear and uniformly accepted definitions for many clinically important con-
ditions, may hamper subsequent retrieval and analyses. For example, clini-
cally distinct terms encompassing a single medical condition, e.g., hepato-
toxicity, may be spread across several SOCs and levels of the hierarchy.
Another challenge in generating the most accurate and useful informa-
tion is deciding what level of terminology (e.g., Lower Level or Preferred
term from a coding dictionary) should be used in presenting AE data (for
example, in summary tables).
The CIOMS VI Working Group suggests that AE data should gener-
ally be presented as Preferred Terms (e.g., from MedDRA®), organized
within the relevant System Organ Classes (SOCs). However, due to the
high granularity of MedDRA®, there may be several Preferred Terms
describing different AE/ADR cases that involve the same medical con-
cept within one SOC. Therefore, under some circumstances, it might
be useful to include data at more than one level of the hierarchy within
a SOC (e.g., High Level Terms (HLT) as well as Preferred Terms).
One approach to overcoming the various shortcomings discussed
above has been undertaken by a separate CIOMS Working Group on
“Standardized MedDRA® Queries (SMQs).” It has been operating for
several years as a collaboration between senior scientists from drug
regulatory authorities, pharmaceutical companies, the ICH MedDRA
Management Board, the MedDRA Management and Support Organiza-
tion (MedDRA® MSSO), and the WHO. The Group has developed SMQ
guidelines (proper database search strategies) for many defined medical
conditions, which are meant to aid in case identification from the various
signs, symptoms, diagnoses, syndromes, physical findings, laboratory

102

group6_PH.indd 102 7.8.2007 12:19:52

 and other physiological data within a database. Prior to their release for
general use, all SMQs are tested in databases of regulatory authorities and
pharmaceutical companies.35 They are subsequently made available by the
MedDRA MSSO to the user community and are being maintained and
updated as appropriate by that organization. Although SMQs for as many
important conditions as possible will be developed, it is unlikely that such
groupings will ever be available for all clinical conditions relevant to any
specific product.

(3) Dealing with Unblinded Data

Throughout the course of a clinical trial program, unless a waiver has
been granted (see section c.(6) above) the blind will be broken based on
ICH Guideline E2A for some individual AE cases in order to comply with
expedited regulatory reporting requirements, mainly for serious, unexpec-
ted ADRs. Companies struggle with a variety of choices with regard to
dealing with the newly available information on therapy assignment for
these patients, including whether certain personnel should have access to
the information. Questions include: Should the information be entered in
the clinical trial and/or safety database; if so, is entry independent of the
therapy (placebo, comparator, new product)? Should one wait until the
trial is over until entering the data? Should access to the information be
restricted to selected personnel (e.g., access permitted for all or some per-
sonnel in the safety department but not biostatistical or clinical personnel
involved in the conduct or analysis of the trial)? If unblinded reports are
sent to regulators, DSMBs and trial ethics committees, is it advisable or
appropriate to keep the cases blinded for investigators, as some companies
have chosen to do?
There is no one correct approach to this problem, and no regulatory
guidance. The solution will depend on many factors pertinent to a com-
pany, its organizational structure, its philosophy toward such matters, and
its technical systems for managing data. However, several members of the
CIOMS VI Working Group did express a preference for entering the new
information into the safety database without attempting to prevent access
by safety or other personnel involved in the conduct of the trial. The

35
See the first SMQ report: Development and Rational Use of Standardized MedDRA Queries (SMQs). Re-
trieving Adverse Drug Reactions with MedDRA, CIOMS, Geneva 2004. It covers torsades de pointes/QT
prolongation, rhabdomyolysis/myopathy, and hepatic disorders. For details and to monitor the progress of the
CIOMS Working Group’s efforts, see www.cioms.ch/What’sNew/WorkingGroups.

103

group6_PH.indd 103 7.8.2007 12:19:52

 rationale for making such information readily available is that it should be
taken into account in the course of ongoing monitoring and evaluation of
safety. However, guidance on this issue, for which there are varying opin-
ions, is beyond the scope of this Working Group.

(4) Data Processing Issues

The processing and interpretation of clinical trial safety data are rare-
ly straightforward and represent challenging activities for sponsors and
investigators. Part of this challenge stems from the fact that a comprehen-
sive review of safety data involves analysis of both individual reports as
well as aggregate data. This dual approach allows for both a qualitative
and quantitative understanding of the safety profile of a drug. An addition-
al challenge is that some important elements of safety information, such
as serious adverse event reports, must be reviewed within specified time
frames after the sponsor becomes aware of them, while aggregate data are
reviewed on a periodic basis, as well as at the end of a clinical trial or clini-
cal development program. These multiple aspects of safety data review
during clinical development demand that the data management processes
be both flexible and robust.
There are many activities involved in the management of clinical
trial safety data, and a full discussion is beyond the scope of this chap-
ter. Core activities include data entry; edit checks; data queries to resolve
discrepancies noted in the edit check process; coding of adverse events
using a standard dictionary, such as MedDRA®; and, in the case of data
from multiple trials, pooling datasets for a comprehensive analysis. Each
of these activities must be undertaken with care and precision, to insure
that the safety database is accurate and complete. However, once a study
or clinical development program is completed, there will be great pressure
to close (“lock”) the database, so that data analysis can begin and the final
reports can be written. While the analysis of safety data should proceed as
quickly as possible, there must also be mechanisms for investigators and
sponsors to handle suspected adverse drug reactions (ADRs) that may ap-
pear after the study is completed. In addition, there must a mechanism in
place for obtaining follow-up information on ADRs that were ongoing at
the time the study ended. Ideally, such issues should be covered in study
protocols.
Many sponsors use Contract Research Organizations (CROs) to
manage some or all aspects of their clinical trials, including data entry,
data management, and data analysis. In these cases, the CRO may hold

104

group6_PH.indd 104 7.8.2007 12:19:53

 the clinical database(s). It is important that the sponsor, as the responsible
party, have ready access to the data for prompt review and any required
action. Thus, agreements and methods for achieving ready access to data
must be in place. The same is true for other contractual relationships, such
as in co-development licensing agreements.
Ideally, a systematic, reproducible approach to detect, classify and
document adverse events would enable sponsors and investigators to de-
velop clinical as well as statistical understanding of the safety profile.
Different groups have approached this goal in different ways.36
New approaches to standardized data management techniques have
been in development to facilitate analysis and reporting of safety and other
clinical trial data. A recently formed open, non-profit organization which
involves the biopharmaceutical industry and regulatory authorities, Clini-
cal Data Interchange Standards Consortium (CDISC), is committed to the
development of worldwide industry standards to support the acquisition,
exchange, submission and archiving of electronic clinical trial data.37 Spe-
cific initiatives by CDISC include the following:
❏ a model for regulatory submission of data to support a marketing
application for a new product; this CDISC Submission Data Stan-
dard (SDS) includes data in the form of standard domains (e.g., de-
mographics, drug exposure, concomitant medications, laboratory
data, adverse events), which define the data elements for common
safety data and other data collected in clinical trials;38
❏ analysis dataset models (ADaM), which are being developed to
define standard ways to provide datasets for safety and efficacy
review and analysis by statisticians at regulatory agencies.

36
For example, see Tangrea, J. A., Adrianaza, M. E., and McAdams, M. A Method for the Detection and Man-
agement of Adverse Events in Clinical Trials, Drug Information Journal, 25:63-80, 1991; Gait, J. E., Smith,
S. and Brown, S. L. Evaluations of Safety Data from Controlled Clinical Trials: The Clinical Principles Ex-
plained, ibid., 34:273-287, 2000; and Hsu, P.-W., Pernet, A. G., Craft, J. C. and Hursey, M. J. A Method for
Identifying Adverse Events Related to New Drug Treatment, ibid., 26:109-118, 1992.
37
See www.cdisc.org or write to Dr. R. Kush at [email protected]. . Membership includes biotech and pharma-
ceutical companies, CROs, and academic medical centers in the EU, Japan, the US and India. Various CDISC
working groups have been established (e.g., in Japan, Europe and India). For an explanation of CDISC and
a report on laboratory data standards, see S. Bassion. The Clinical Data Interchange Standards Consortium
Laboratory Model: Standardizing Laboratory Data Interchange in Clinical Trials, Drug Information Journal,
37:271-281, 2003.
38
Submission Data Standards, Analysis Dataset Standards, Operational Data Model, and Laboratory Data Stan-
dards (see www.cdisc.org/standards/index/html).

105

group6_PH.indd 105 7.8.2007 12:19:53

 These and other standard CDISC models enable both clinicians and
statisticians to review and analyze a much richer, more comprehensive
and more accurate collection of safety data than is currently available
from post-marketing pharmacovigilance reports. CDISC also has a for-
mal working relationship with the international standards setting (ISO),
not-for-profit organization, Health Level 7 (HL7). Its members – health
care providers, vendors, payers, consultants, government groups and oth-
ers – have an interest in the development and advancement of clinical and
administrative standards for health care.39

39
For details, see http://www. Hl7.org/. A “Regulated Clinical Research and Information Management (RCRIM)
Technical Committee” (co-chaired by CDISC, HL7 and FDA) works toward accreditation of the CDISC
models described above and is involved in other standards-setting, such as: HL7 messages to support the
reporting of post-marketing pharmacovigilance data for safety surveillance; standards for submitting ECG
waveform data to regulatory agencies; and standard protocol representation, which includes standardization
of clinical trial protocol elements to support safety and efficacy assessments and statistical analyses. These
efforts are committed to harmonizing all of these standards and models to support regulated clinical research,
in addition to strengthening the link between healthcare and clinical trials.

106

group6_PH.indd 106 7.8.2007 12:19:54

 V

Identification
and Evaluation of Risk
from Clinical Trial Data

group6_PH.indd 107 7.8.2007 12:19:54

group6_PH.indd 108 7.8.2007 12:19:54
 a. Introduction
The ongoing evaluation of the safety profile of a drug during clinical de-
velopment is a dynamic process that serves several important purposes, first
and foremost of which is the protection of human subjects participating in
clinical drug trials. If new risks are identified, it would be important to put risk
management strategies in place to better understand and minimize the risk to
patients. But it is also important to gain an understanding of the safety profile of
the drug as early in its development as possible. Once new risks are identified, a
development program can be stopped if the risks are deemed unacceptable, or
modified to gain a better understanding of or better manage those risks.
The safety information that emerges at the end of clinical development
should be sufficiently rigorous to allow for comprehensive regulatory review
and determination of the benefit-risk profile of the drug to support marketing
approval. It should also be sufficiently comprehensive so that prescribers and
patients can be given adequate information for the safe use of the drug.
To the extent possible, the ongoing review of benefits is also very impor-
tant. Although benefits may be more difficult to assess early in a development
program, especially when studies continue to be blinded, ability to assess the
benefit-risk profile of a drug at least in a preliminary fashion, is essential. It
is not only important to terminate a program early when new risks are felt to
be unacceptable, but also to avoid premature termination of a program that
shows promise for potential value even in the face of certain risks. This will be
especially true in the development of life-saving therapies, particularly when
alternative therapies are not available.
The evaluation of clinical safety during drug development is dependent on
medical judgement as well as on an appreciation of descriptive and inferential
statistics. This chapter focuses on a clinical approach to the detection and eval-
uation of emerging risks. Quantitative and statistical concepts are presented
in Chapter 6. Some authors have suggested that existing methods of safety
data evaluation can be substantially improved by standardizing the approach to
early detection of safety signals.1,2,3,4

1
Morse, M.A., Califf, RM and Sugarman, J. Monitoring and Ensuring Safety During Clinical Research, J. Am.
Med. Assoc., 285:1201-1205, 2001.
2
Ioannidia, P.A. and Lau, J. Completeness of Safety Reporting in Randomized Trials. J. Am. Med. Assoc., 285:
437-443, 2001.
3
Wallander, M. The Way Towards Adverse Event Monitoring in Clinical Trials, Drug Safety. 251-262, 1993.
4
Lineberry, C. Approaches to Describing Common Adverse Events in the Integrated Safety Summary, Drug
Information Journal, 25:493-500, 1991.

109

group6_PH.indd 109 7.8.2007 12:19:54

 To ensure that complete and accurate safety information is collected
during a clinical trial, it is imperative that sponsors and investigators pay
careful attention to the overall development program, the design of each
clinical trial, and the process that is in place for the ongoing safety evalua-
tion during drug development. To minimize variability amongst investiga-
tors, it is critical that sponsors and investigators maintain consistency in the
identification and recording of adverse events and other safety data across
a development program. Different groups have approached this goal in dif-
ferent ways. Details on safety data collection in clinical trials can be found
in Chapter 4.
An important principle in the evaluation of safety data from clinical
trials is that while the data are designed to be analyzed in a comprehensive
fashion at the end of a trial or development program, they also must be
evaluated in an ongoing fashion, so that important safety signals can be
detected early and that trial participants are protected.
Several published ICH guidelines address the appropriate handling of
safety data in clinical trials. ICH E65 provides guidance on safety report-
ing for investigators (Section 4.11), ongoing safety evaluation for sponsors
(Section 5.16), reporting of adverse drug reactions to investigators, IRB(s)/
IEC(s), and regulatory authorities (section 5.17), and the assessment of
safety in a clinical trial protocol (Section 6.8). Section 12 of ICH guide-
line E3 (Structure and Content of Clinical Study Reports) and section 4 of
ICH M4 (Common Technical Document – Efficacy (Clinical Summary))6
contain useful recommendations for appropriate analyses and presentation
of safety data from completed clinical trials and integrated summaries of
safety. Other useful documents are the US FDA’s template on safety and
efficacy review of new submissions, and the detailed guide used by their
internal safety reviewers.7 The current chapter, which should be read in
conjunction with these guidelines, aims to complement and elaborate upon
their concepts with additional practical advice, including interpretation of
the results of the analyses suggested in E3 and M4. This will facilitate safe-
ty monitoring during ongoing clinical trials.

5
ICH Guideline E6, Guideline For Good Clinical Practice, 1 May 1996, http://www.ich.org/
6
ICH Guideline E3, Structure and Content of Clinical Study Reports, 30 November 1995, http://www.ich.org/
and ICH Guideline M4, The Common Technical Document, http://www.ich.org/
7
See Section 7.0 of FDA’s “Clinical Review Template,” CDER, Office of the Center Director, effective 9 July
2004 (http://wwwfda.gov/cder/mapp/6010.3.pdf), and Reviewer Guidance: Conducting a Clinical Safety Re-
view of a New Product Application and Preparing a Report on the Review (January 2005; see http://www.fda.
gov/cder/guidance/3580fnl.pdf).

110

group6_PH.indd 110 7.8.2007 12:19:54

 It is important to note that ongoing safety evaluation should occur
in clinical trials of all sizes and degrees of complexity, including single-
center clinical trials conducted by individual investigators to multi-center
trials conducted by a group of investigators and multi-center or multi-
national trials conducted by a pharmaceutical company. While the logistical
aspects of safety data collection and evaluation may vary from setting to
setting, the principle of protecting trial participants through early identi-
fication, evaluation and management of safety issues is paramount to all
clinical trials.
The CIOMS VI Working Group recommends that clinical trial sponsors
develop a process to assess, evaluate and act upon safety information
during drug development on a continuous basis in order to ensure the
earliest possible identification of safety concerns and to take appro-
priate risk minimization steps. Such steps can include modification of
study protocols to incorporate appropriate strategies to ensure that
clinical trial participants are not exposed to undue risk.

b. Expectations and Limitations in the Identification

and Evaluation of Safety Information
from Clinical Trials
It is important to recognize the limitations of clinical trial safety data
as they become available, and to have realistic expectations of what can
be learned from such data. It is also important to understand the inherent
limitations of clinical trials (see Chapter 1, section d.), while at the same
time maximizing usefulness by including safety considerations in their
design.
Critical for the interpretation of safety data is the number of subjects
exposed to the investigational product and for how long. The more subjects
are exposed for long durations, the greater will be the confidence in the
safety of the product. However, there is no standard rule for what that num-
ber or duration should be. Rather, the number of subjects and the duration
of treatment required to establish an acceptable safety profile of a product
depend on many factors, such as whether the drug represents a new chemi-
cal or therapeutic class, whether it is similar to other available products,
whether it has potential advantages over existing therapies, the character-
istics of the intended patient population (e.g., rare versus common disease
indication), and the intended duration of use (e.g., acute versus chronic

111

group6_PH.indd 111 7.8.2007 12:19:55

 conditions).8 While the sample size calculations for individual clinical trials
are usually dictated by efficacy considerations, the total number of persons
exposed to the investigational agent in a clinical development program and
for how long should also be influenced by safety considerations.
ICH Guideline E1 recommends specific minimum numbers of subjects
for the safety evaluation of drugs intended for long-term treatment of non-
life-threatening diseases.9 However it is important to note that there are cir-
cumstances in which these standardized subject exposures may not be suf-
ficient for safety data evaluation, e.g., when:
❏ specific safety concerns are identified (e.g., from animal studies or
chemically related products);
❏ the product has pharmacokinetic and/or pharmacodynamic prop-
erties known to be associated with adverse reactions, e.g., specific
metabolic pathways;
❏ there is a concern that a product may add to a significant background
rate of morbidity or mortality in the target patient population.
The number of persons exposed to the investigational agent should be
carefully considered, based on the above considerations, not only at the time
the clinical development program is planned but also during the program
itself, as the safety profile of the investigational drug becomes better under-
stood. Despite careful planning for the number of subjects , clinical develop-
ment programs are not able to identify all risks associated with a product.
Some risks occur so infrequently that they will become apparent only after
thousands or tens of thousands or more have been exposed to the product
– an extent of exposure usually achieved only after the product has been
marketed. For example, if the “true” frequency of a particular adverse event
is 1/1000, then administering the drug to 3,000 persons will result in a 95%
chance of observing at least one instance of the event. If the number of per-
sons exposed is decreased to 1,610, then the chance of observing at least one
instance of the event is reduced to 80%. If the “true” frequency of the event
is 1/10,000, then studying 10,000 persons will yield only a 63% chance of
observing at least one event.10 Observing an event is not the same as

8
FDA’s Guidance for Industry on Pre-marketing Risk Assessment, March 2005, http://www.fda.gov/cder/
guidance/6357fnl.htm
9
ICH Guideline E1, The Extent of Population Exposure to Assess Clinical Safety for Drugs Intended for Long-
Term Treatment of Non-Life-Threatening Conditions, 27 October 1994, http://www.ich.org/
10
These calculations are a reflection of the “rule of three,” which states that if no event of a particular type is
seen in x-individuals, one is 95% certain that the event occurs no more often than 3/x; e.g., if x=500, 95%
certainty that it occurs in less than 3 in 500 (0.6%).

112

group6_PH.indd 112 7.8.2007 12:19:55

 concluding that it is an adverse drug reaction. Thus, the ability to detect a
rare adverse drug reaction is a recognized limitation of clinical trials, where
rare is conventionally defined as equal to or less than 1 in a thousand.
For serious adverse events, a detailed understanding of the individual
case is important. For non-serious events, a high level of scrutiny of indi-
vidual events in early phase trials with small numbers of subjects is also
reasonable. But the same level of scrutiny may not be practical or useful
in large practice-based, post-approval studies, especially when the safety
profile is well-established. In this case, analysis of aggregate data is more
meaningful and practical.
For more commonly occurring events, the analysis of aggregate data is
both important and appropriate in order to explore the possible relationship
with the drug. A special challenge in the evaluation of aggregate safety data
is the application of appropriate statistical techniques, which have been
developed and used much more for efficacy determinations than for safety.
Chapter 6 provides a guide to currently accepted approaches for analyzing
and interpreting clinical trial safety data.
One of the goals of analysis and interpretation of safety data is to as-
sess the medical significance of one or more suspected adverse reactions
in order to develop appropriate and useful product information both during
the development program (through the Investigator’s Brochure) and after
the product is authorized (in the local data sheets/product information), and
to develop risk-management strategies to minimize them. It is therefore im-
portant that emphasis be placed on medical and scientific perspectives, ul-
timately on behalf of public health, rather than perfunctory data collection,
processing and regulatory reporting, as important as these activities may
be. There is a body of literature on risk assessment for pharmaceutical and
other sectors, but only recently have attempts been made to approach the
subject for drugs in a detailed and systematic way through new methodolo-
gies and regulations. Under a broad mandate covering risk management of
drugs under US legislation, the FDA has developed a series of draft guid-
ances for industry on (1) premarketing risk assessment, (2) development
and use of risk minimization action plans, and (3) good pharmacovigilance
practices and pharmacoepidemiologic assessment.11 The FDA defines risk
assessment as consisting of “identifying, and characterizing the nature,

11
U.S. Department of Health and Human Services, Food and Drug Administration, Centre for Drug Evaluation
and Research, Guidance Documents. http://www.fda.gov/cder/guidance/index.htm

113

group6_PH.indd 113 7.8.2007 12:19:56

 frequency, and severity of risks associated with the use of a product.” Simi-
lar developments are underway in the EU12,13 and Japan.

c. Points to Consider During Analysis

and Evaluation of Safety Information
There are several factors that need to be considered that influence the
evaluation and interpretation of safety information as well as the benefit-
risk assessment; they are important both in the analysis of individual cases
and of aggregate data. Some of the most important factors are considered
below.

(1) Patient Population Characteristics,

Including Natural History of Disease
The demographics of the population (e.g., age, gender, race, geogra-
phic regions, socioeconomic factors) should be considered during
safety data evaluation, for several reasons. Some adverse events oc-
cur more frequently in some groups than in others, even in the ab-
sence of treatment with an investigational drug. Older adults in
general have a higher incidence of cardiovascular disease than
younger persons. Thus, the occurrence of a myocardial infarc-
tion in a 75 year-old clinical trial subject may be evaluated and in-
terpreted differently than in a 25 year-old participant. It should also
be remembered that certain diseases are more prevalent in specific
populations (e.g., sickle cell anemia, Tay Sachs Disease) and sus-
pected adverse reactions of such types during clinical trials should be
reviewed with this in mind.
In addition to differences in background rates of certain adverse
events, there may be drug-demographic interactions that result in cer-
tain adverse reactions occurring more frequently in older patients than
in younger patients or more frequently in men than in women. For ex-
ample, the risk of gastrointestinal bleeding with an NSAID is higher
in older patients than in younger patients. In this case, the confound-
ing or interacting effects of a demographic factor can be determined

12
Establishing a European risk management strategy: Summary Report of the Heads of Agencies Ad Hoc Work-
ing Group, January, 2003. www.emea.eu.int
13
Handling by the CPMP of safety concerns for pre-and post-authorization applications submitted in accor-
dance with the centralized procedure. April 5, 2004. www.emea.eu.int

114

group6_PH.indd 114 7.8.2007 12:19:56

 only if aggregate data are analyzed since it is the incidence of a known
adverse reaction that is in question. It is also known that both safety
and efficacy results may be pharmacologically-mediated (how a drug
is “handled” by the body) and show ethnic and racial differences.14
Knowledge of the background incidence and prevalence of medically
important conditions in the target population is an important tool in
the evaluation of individual cases as well as in the analysis of aggre-
gate data. The information helps to provide a context for case reports
and incidence rates of AEs. However, caution should be exercised
when comparing the incidence in the patients treated with the inves-
tigational drug to the incidence from the literature or from historical
controls since clinical trials tend to include a highly selective popula-
tion. Ideally, a comparison should be made to concurrent controls,
which could serve to provide further context for the assessment of the
benefit-risk profile of the investigational product. Use of historical
clinical trial data, especially if there are sizeable similar trials with
placebo controls, may be particularly useful if a concurrent control is
not available.
Consideration of the natural history of the disease being treated can also
be very important in the evaluation and interpretation of safety data.
The distinction between a manifestation of the disease being treated
and an adverse drug reaction is a special challenge in the evaluation
of safety data. When certain adverse events are known manifestations
of the disease it is important to know the expected frequency, severity
and pattern of presentation. Their occurrence may not be of concern if
the incidence is in line with expectations based on the natural history.
On the other hand, it is important not to overlook the possibility of an
adverse drug reaction related to worsening of the disease treated, even
if the frequency is in line with expectations, if the nature, severity or
other presentation are not typical.

(2) Current Therapeutic Standards

In the ongoing evaluation of the evolving benefit-risk profile of a drug
in development, it is important to take into account what is know about
established therapies for the condition under study. The benefit-risk
profile for existing drugs can serve as a benchmark against which to

14
For discussion and recommendations, see ICH Guideline E5.

115

group6_PH.indd 115 7.8.2007 12:19:57

 weigh the acceptability of the emerging profile for a new product. In
addition, drugs taken concomitantly with the study drug may be con-
tributing to the incidence of adverse drug reactions either indepen-
dently or through a drug-drug interaction.
The standard treatment for a disease may change during a clinical
development program, especially if the program lasts several years.
If a new therapy reaches the market place during a development pro-
gram, it will be important to update the benchmark for an acceptable
benefit-risk profile. In this regard, it is important to note that standard
therapy need not refer exclusively to pharmacotherapy, but can refer to
non-pharmacological treatments, such as surgery, diet, exercise, psy-
chotherapy, physical therapy, or other treatment modalities.15

d. Timing of Safety Evaluation

The timely and thorough management and evaluation of safety infor-
mation is a shared responsibility among all parties involved in the clinical
trial process. Toward that end, it is critical that sponsors (including indepen-
dent sponsor-investigators) define and implement a system and schedule
for reviews of safety information. One can consider three general situations
requiring safety data review, which are independent of the size and com-
plexity of a clinical trial:
(1) Ad hoc for serious and special interest AEs; it is imperative that there
be a mechanism to review important safety data in a timely fashion.
(2) Routine, periodic, general review of all data; the frequency of an over-
all periodic review will vary from trial to trial and from development
program to development program and depend on such things as the
phase of clinical development, the amount of safety and other data
already known about the investigational product, the duration of treat-
ment, the amount of safety information known about drugs in the same
or a similar class, the number of patients exposed, the number of sites
and investigators in the trial, the level of concern over specific adverse
events, the anticipated benefit-risk profile of the drug, and the per-
ceived level of acceptable risk for the product.

15
www.clinicalevidence.com is an important website source of information in this respect.

116

group6_PH.indd 116 7.8.2007 12:19:57

 (3) Reviews triggered by specific milestones established for a trial or a
program (e.g., numbers of completed patients, end-of-trial, end-of-
program, preparation of integrated summary of safety and a marketing
application). See Chapter 3, section b.8. for discussion.
A recommendation to establish multidisciplinary teams and to use
advisors and advisory boards in the course of safety reviews is discussed
in Chapter 3.
In addition to the frequent review of serious adverse events (SAEs)
and those of special interest, overall assessment of all AEs, regardless of
seriousness, causality, or expectedness, should be performed periodical-
ly.16 Periodic and summary reviews should include both interval and cu-
mulative incidences (in relation to known subject exposure) of all AEs in
the study database.. When trying to draw conclusions and take any action
from evaluation of comparative safety data, it may be important to distin-
guish between placebo and active comparators.
Each time a study is completed and unblinded, all safety information,
not just clinical AEs but ideally emerging efficacy endpoints, vital signs,
and clinical investigation results, should be assessed and evaluated relative
to the previous information. As needed, the relevant product information
(investigator brochure, Development Core Safety Information (DCSI),
Company Core Safety Information (CCSI), local datasheets) should be
updated.

e. Safety-Signal Detection and Evaluation

The concept, definition and methods for signal detection have been
primarily associated with large, post-marketing databases, usually of spon-
taneous reports.17,18,19 While there is a growing body of published literature
on the advantages and limitations of various statistical methods to detect

16
Gait, J.E., Smith, S. and Brown, S.L. Evaluation of Safety Data From Controlled Clinical Trials: The Clinical
Principles Explained, Drug Information Journal, 34: 273-287, 2000.
17
Report of CIOMS Working Group V. Current Challenges in Pharmacovigilance: Pragmatic Approaches.
Council for International Organizations of Medical Sciences. Geneva 2001.
18
Brown, E.G. and Douglas, S. Tabulation and Analysis of Pharmacovigilance Data Using the Medical Diction-
ary for Regulatory Activities, in Pharmacoepidemiology and Drug Safety, 9: 479-489, 2000.
19
Jasmanda H., Wu, M.C., Fung, K.K., et al. Postmarketing Drug Safety Surveillance. Pharm. Dev. Regul., 231-
244, 2003.

117

group6_PH.indd 117 7.8.2007 12:19:57

 signals from post-marketing safety data, the field is still in its infancy.20,21,22
23, 24
As data accumulate, statistical methods for the evaluation of safety
signals in clinical trials may be possible (see Chapter 6). It is more likely,
however, that signal detection during early development will generally be
based on clinical judgment.25,26,27
While it is impossible to define standard criteria for the clinical evalu-
ation of safety, the CIOMS VI Working Group believes there are some
fundamental steps that can be taken to improve the process for detect-
ing signals. These include:
❏ prompt medical evaluation of all individual serious cases, regard-
less of attribution or expectedness, and adverse events of special
interest, whether serious or not
❏ periodic aggregate assessment (blinded, partially-blinded 28 or un-
blinded, depending on the status of the trials, as appropriate) of
all available clinical safety data (including clinical AEs, laboratory
data, selected physical data such as blood pressure), irrespective of
causality or seriousness; any relevant non-clinical data should also
be reviewed
❏ safety evaluation of completed unblinded studies, both individually
and combined where appropriate, principally from a clinical per-
spective, but also including relevant statistical analyses.

20
Wilson, A.,Thabane, T. and Holbrook A. Application of data mining techniques in pharmacovigilance, Br. J.
Clin Pharmacol., 57 (2): 127-134, 2003.
21
Szarfman A., Machado S.G. and O’Neill, R.T. Use of screening algorithms and computer systems to
efficiently signal higher-than-expeced combinations of drugs and events in the US FDA’s spontaneous reports
database, Drug Safety,25 (6): 381-392, 2002.
22
Van Puijenbroek E.P., Bate A., Leufkens, H.G., et al. A comparison of measures of disproportionality for
signal detection in spontaneous reporting systems for adverse drug reactions. Pharmacoepidem. Drug Safety,
11: 3-10, 2002.
23
Bate, A., Lindquist, M. and Edwards, I.R. A Bayesian neural network method for adverse drug reaction signal
generation, Eur. J. Clin Pharmacology; 54: 315-321, 1998.
24
Evans S.J., Waller, P.C. and Davis S. Use of proportional reporting ratios (PRRs) for signal generation from
spontaneous adverse drug reaction reports, Pharmacoepidem. Drug Safety, 10: 483-486, 2001.
25
Kock, G.G.. Discussion: Statistical Perspective, Drug Information Journal, 25: 461-464, 1991.
26
Enas, G.G.. Making Decisions about Safety in Clinical Trials – The Case for Inferential Statistics, Drug In-
formation Journal, 25: 439-446, 1991.
27
Huster, W.J. Clinical Trial Adverse Events: The Case for Descriptive Techniques, Drug Information Journal,
25: 447-456, 1991.
28
“Partially-blinded” refers to data that are categorized as groups A and B, e.g., without revealing the actual
treatment name for each arm.

118

group6_PH.indd 118 7.8.2007 12:19:58

 f. Consistent Causality Assessment – Adverse
Events vs Adverse Drug Reactions
The identification of a potential safety issue for a medicinal product re-
quires an ability to readily distinguish adverse drug reactions (ADRs) from
adverse events.29,30 While there are established definitions of AEs and ADRs,
there are no agreed criteria for reliably distinguishing between them, mak-
ing this key element of risk assessment relatively subjective (see Chapter 4,
section c.2.). Although investigator causality assessment is helpful in cat-
egorizing reports for regulatory reporting purposes, and for evaluating rare
or unusual events, it has very limited utility in the analysis of aggregate in-
formation. Algorithms for categorizing causality at the case level have been
developed for selected events, e.g., drug-induced liver injury31 but have not
gained wide acceptance.32
Previous CIOMS guidelines, current EU regulations, and proposed US
regulations recommend that core safety information, including the DCSI for
investigational products and the CCSI for approved products, should describe
adverse drug reactions, and exclude events that have no well-established
relationship to therapy. The purpose of DCSI is to provide the best safety
information available at every stage of development. Ideally we should have
consistent and reliable systems in place for causality assessments. However,
as stated previously in Chapter 4 and reflected in the survey results (Appen-
dix 3), there are no uniformly agreed criteria for determining whether there
is a causal association between a medicinal product and a given AE, and
therefore whether a given AE should be included in the DCSI and/or CCSI.
The decision to include information in the DCSI depends strongly on
the concept of threshold as outlined in the CIOMS III/V report.33 The con-
cept of threshold is discussed further in Chapter 7, section e., including a
proposed modification to the previous CIOMS recommendations.

29
Hsu, P.H. and Stoll, R.W. Causality Assessment of Adverse Events in Clinical Trials: I. How Good is the
Investigator Drug Causality Assessment?. Drug Information Journal, 27: 377-385, 1993.
30
Hsu, P.H. and Stoll, R.W. Causality Assessment of Adverse Events in Clinical Trials: II. An Algorithm for
Drug Causality Assessment, Drug Information Journal, 27: 387-394, 1993.
31
Danan, G. and Benichou, C. Causality assessment of adverse reactions to drugs – A novel method based
on the conclusions of international consensus meetings: application to drug-induced liver injuries, Journal
of Clinical Epidemiology, 46 (11): 1323-1330, and Benichou, C. Danan, G. and Flahault, A. Causality as-
sessment of adverse reactions to drugs – II. An original model for validation of drug causality assessment
methods: case reports with positive rechallenge, ibid., 46 (11):1331-1336, 1993.
32
Stephens, M.D.B.. From Causality Assessment to Product Labeling, Drug Information Journal, 31: 849-856, 1997.
33
Guidelines for Preparing Core Clinical-Safety Information on Drugs, Report of CIOMS Working Group III/V,
Council for International Organizations of Medical Sciences, Geneva, 1999.

119

group6_PH.indd 119 7.8.2007 12:19:58

 When combination medicinal products are used, either fixed (e.g., an
ACE inhibitor and a diuretic) or as part of a multi-drug regimen (e.g., for
cancer chemotherapy or HIV treatment), AE causality assessment for indi-
vidual components will be difficult for both individual cases and with aggre-
gate data, and in the absence of consistent and convincing data with regard
to a single agent, AE causality should be assessed for the combination.
The CIOMS VI Working Group suggests that the principal use of causality
assessment for individual serious adverse events (SAEs) is more relevant
for determining the regulatory reporting status, rather than for clinical
analysis. It recommends that determination of causality for ongoing sig-
nal detection and inclusion in the DCSI and eventually the CCSI should
be based on a combination of clinical judgement and aggregate data
analysis based on all reported cases. Investigator causality assessment
should be taken into account and may be particularly important when
evaluating rare or unusual events for which aggregate analytical methods
are not applicable.

g. Important Types of Analyses

While it is important to differentiate between serious and non-serious
events for the purposes of regulatory reporting, the practical medical sig-
nificance of the event(s) is of greater importance. Although it is appropriate
to apply greater scrutiny to what appear to be serious adverse events, the
true safety profile of a medicinal product throughout development can only
be assessed by careful evaluation of all AEs/ADRs. Serious AEs and AEs
of special interest (see Chapter 4, section c(4)) should be reviewed and as-
sessed individually and in aggregate on a continuous basis. Non-serious
AEs should also be critically appraised at regular intervals, in particular
those associated with discontinuation of study treatment.
Though non-serious AEs are generally not routinely reviewed individ-
ually, they should be given careful attention in preparing study reports and
integrated safety summaries. As already mentioned, contributions to evalu-
ating and understanding the safety experience can depend on information
other than AE reports, such as physical examination findings, vital signs,
clinical laboratory tests, cardiac electrophysiology, and other study or non-
study evaluations.
A similar approach should be considered for AEs associated with
treatment discontinuation. Investigators need to pay particular attention to

120

group6_PH.indd 120 7.8.2007 12:19:59

 the collection of all relevant information concerning these events, possibly
more than would be usual in actual clinical practice. In addition, all subjects
who withdraw from a study should be carefully questioned for the reason,
including the possible occurrence of an adverse event. It is the responsibil-
ity of the sponsor to critically evaluate these events promptly, obtain as
much follow-up information as required and available, and analyze them
against the known morbidity profile of the study population and any other
relevant information (e.g., comparable events with other members of the
drug class). The reasons for discontinuation, the time to study withdraw-
al, and other factors when examined against the same data for comparator
treatments (including placebo) can be very revealing.

h. Review of Individual Cases

The individual case safety report is the basic, fundamental unit of safe-
ty analysis. Throughout clinical development, review of individual cases,
especially cases of serious adverse events, can shed considerable light on
the safety profile of an investigational drug product. During early clinical
development, review of both serious and non-serious individual adverse
event case reports can potentially shed considerable light on the safety of
the drug. Medical review and judgment by an appropriately qualified pro-
fessional is critical for the evaluation of the individual case. The careful
and thoughtful evaluation of one or more serious adverse event reports or
reports of adverse events of special interest can be critical in detecting an
emerging safety signal. The evaluation of an individual case report requires
review of many elements of the report, and additional or follow-up infor-
mation is often necessary. The sponsor should work with the investigator
to insure that any additional information required to understand fully the
reported event(s) is made available. Patient-specific information, especially
co-morbid conditions, personal and relevant family medical history, con-
comitant treatments (including non-prescription medications, special diets,
surgery, physical therapy, dietary supplements, “natural”, herbal, homeo-
pathic, and other alternative medications and treatments) should be care-
fully reviewed, primarily to permit identification of possible confounding
factors, risk factors for an adverse drug reaction, possible drug-drug and
drug-disease interactions, and other potential causes of the event. The eval-
uation of individual cases should be done in the context of the patient popu-
lation, the indication for the investigational drug, the natural history of the
disease, current available therapies, and other benefit-risk considerations
(see section l. below).

121

group6_PH.indd 121 7.8.2007 12:19:59

 Investigators and sponsors should evaluate each case in as much detail
as possible to determine what factor or factors were responsible for the
event, and then to assess the causal role of the investigational product. The
use of specific report forms for particular types of events (e.g., liver injury,
bone marrow depression, severe skin events) can facilitate the capture of all
the information required for clinical evaluation.34

i. Considerations for Periodic Review and

Evaluation of Case Reports in Aggregate
Evaluation of aggregate data for understanding the evolving safety pro-
file, and especially for detecting potential safety signals, requires thorough
understanding of the existing safety data for the medicinal product, the
clinical study patient population, including relevant sub-populations (e.g.,
the elderly), and the risk factors for a particular adverse event. Although pe-
riodic review of aggregate data of all safety information is essential, special
attention must be paid to serious adverse event reports and adverse events
of special interest, and should include information known about that drug.
For example, in large and/or long-term studies, for events that are being
closely monitored (e.g., deep vein thrombosis, myocardial infarction), the
incidence of these events should be re-evaluated according to subject ex-
posure to determine whether the observed incidence is consistent with the
natural history, risk factors, and underlying morbidity of the study popula-
tion. In such situations, an external DSMB may be desirable, with appropri-
ate analytical methods and stopping rules in place (see Appendix 5).
Routine detection and evaluation of signals should be based on the
periodic review of aggregate data. Serious adverse event (SAE) reports and
adverse event reports of special interest should be readily retrievable from
the safety and clinical trial databases. For these reports, interval and cu-
mulative frequency tables should be generated at specified intervals. When
the treatment assignment is known (i.e., expedited reports unblinded, as
recommended in ICH E2A), SAE frequency should be compared by treat-
ment group. When, more commonly, treatment assignments are not known,
clinically important SAEs that occur with a notably higher incidence than
anticipated in the study population (background rates), a partially-blinded
analysis may be considered to determine whether there is any apparently

34
Some examples are provided in “Adverse Drug Reactions – A Practical Guide to Diagnosis and Manage-
ment”, Ed. C. Bénichou, John Wiley & Sons, 1994.

122

group6_PH.indd 122 7.8.2007 12:20:00

 clinically meaningful difference between groups. It is also important to
ensure that the intention to assess between-group differences be pre-
specified. Such an analysis should not usually be required, and would rarely
require unblinded analyses of individual studies and/or pooled safety data.
Consideration should also be given to reasons for discontinuation and to
the evaluation of data on marked laboratory abnormalities.
Although the relatively small number of subjects exposed to an inves-
tigational product may limit the utility of subgroup analyses, where pos-
sible data should be stratified for dose, duration, gender, age, and possibly
concomitant medications and concurrent diseases. In addition, when pos-
sible and where relevant, stratification based on other potential risk factors
should also be considered. These approaches are discussed in ICH E3 and
ICH M4. Also, see footnote 8.

j. Pooling of Data
In pooling data, the following points should be considered:
❏ It is most appropriate to combine data from studies that are of simi-
lar design (e.g., similar in dose, duration, methods of eliciting and
determining adverse events, and population).
❏ If the incidence for a particular adverse event differs substantially
across the individual studies in a pool, the pooled estimate is less
informative.
❏ Data from any study with an unusual adverse event pattern should
be presented separately.
❏ The appropriate extent of analysis depends on the seriousness of
the ADR and the strength of evidence for causality. Differences in
rates of drug-related, serious events or events leading to discontinu-
ation or dosage change deserve more investigation, whereas rates of
other ADRs may not merit elaborate analysis.
❏ Examination of subjects that have extreme laboratory value abnor-
malities (outliers) can be useful in identifying subgroups of indi-
viduals who are at particular risk for certain toxicity.
Groups of studies that could be used in pooled safety analyses include the
following.
❏ All controlled studies or subsets of controlled studies
❏ All placebo-controlled studies
❏ Studies with any positive control

123

group6_PH.indd 123 7.8.2007 12:20:00

 ❏ Studies with a particular positive control
❏ Studies of particular indications
❏ All studies, excluding short-term studies in healthy subjects. This
grouping is most useful for evaluating uncommon events
❏ All studies using a particular dose route or regimen, or a particular
concomitant therapy
❏ Studies by geographic region or ethnicity.

These groupings are considered the best source of information about

more common adverse events and may be able to distinguish drug-related
events from background events. Rates in control and experimental drug groups
should be compared. Cumulative dose, time of dosing and dose duration
should also be included. In general, subjects exposed to the medicinal product
in human pharmacology studies (Phase I) should not be included in the pooled
data, although there may be exceptions (e.g., certain oncology studies).
It is almost always useful to pool the first two categories of studies
listed above; the others chosen would vary from drug to drug and should
be influenced by inspection of individual study results. Whatever methods
are used, it should be recognized that, as for results of single studies, any
numerical rate is often only a rough approximation of reality.
Care should be taken not to pool studies in which adverse experiences are
elicited from patients in different ways (e.g., checklist vs direct questioning
vs purely volunteered; see Chapter 4, section d(1)). In addition, special forms
may be used to collect detailed information on serious or special interest cas-
es. Reports on those specific AEs should not be pooled with other types of AE
reports within studies, or between studies that do not use such forms.

k. Evaluation of Clinical Laboratory Data

Clinical laboratory tests are used in clinical trials for three main
purposes:
1) Screening of subjects for inclusion and exclusion
2) Protection of subjects by early detection of organ toxicity
3) Identification of physiological or potentially toxic effects of the
investigational agent.
However, the standard battery of laboratory tests assayed by clinical
laboratories was not developed for these purposes, but for diagnosis and
monitoring of diseases and their response to therapy in individual patients.

124

group6_PH.indd 124 7.8.2007 12:20:01

 All laboratory tests are not equally meaningful as potential indicators
of suspected adverse drug reactions. For example, results that measure an
aggregate of multiple constituents that vary independently (e.g., total white
cell count, serum total protein) are of limited value as indicators of spe-
cific safety problems. Some laboratory tests are meaningless outside the
total clinical context of a specific patient, e.g., serum and urine osmolality
(specific gravity, serum chloride, non-fasting serum glucose), while others
are subject to intra-subject or pre-analytical variability so large as to render
most individual or aggregate changes uninterpretable (e.g., triglycerides,
GGT, venous blood bicarbonate).
Methods for assessment of laboratory data have been discussed in var-
ious publications (see Chapter 6, section e.) and regulatory guidance (e.g.,
ICH E3) and are beyond the scope of this report. ICH M4E35 presents an
overview of how clinical laboratory data should be presented in the efficacy
section of the Common Technical Document.
There are three principal types of analyses that provide a brief summary
of changes in laboratory values in a clinical trial or in a clinical program:
1) measures of central tendency (e.g., group mean or median value), 2) the
range of values, along with the number of subjects without abnormal values
or with values beyond a certain limit (e.g., twice the upper limit of normal),
and 3) individual clinically important values. To characterize changes in
laboratory values over time, these analyses should be performed for each
study visit at which lab data were collected. To understand the context for
these measurements, the above analyses should also note those laboratory
values that were accompanied by adverse events and those that led to dis-
continuation of study medication.
In general, overall population trends are best identified statistically us-
ing large, integrated comparative datasets and are most likely to detect rela-
tively minor changes of limited clinical significance, but can rarely lead to
the identification of important safety concerns. Safety signals can be detect-
ed by analysis of Individual Clinically Significant Abnormalities (see ICH
E3), commonly termed “marked abnormalities”. A marked abnormality is
a laboratory value that meets predetermined criteria for degree of deviation
from the reference range and magnitude of change from a prior value, typi-
cally one obtained immediately prior to exposure to study therapy. When

35
ICH Guideline M4, The Common Technical Document For The Registration Of Pharmaceuticals For
Human Use. Efficacy – M4E – Clinical Overview And Clinical Summary, Module 5: Clinical Study Reports
(http://www.ich.org/).

125

group6_PH.indd 125 7.8.2007 12:20:01

 analyses of clinical laboratory data indicate a possible safety signal, further
analyses may be necessary. See Chapter 6 for more discussion.
Clinical laboratory values should not be analyzed for marked abnormali-
ties if they are clinically meaningless in isolation (e.g., serum chloride, urine
pH, urine specific gravity), if there are alternative methods of measuring the
same constituent (e.g., both hemoglobin and hematocrit as indicators of red
cell mass), if they are analytically unstable (e.g., bicarbonate, acid phospha-
tase), or if they represent aggregates of multiple constituents that vary inde-
pendently (e.g., total WBCs, total protein). Derived values (e.g., anion gap,
A/G ratio, red cell parameters) should generally not be subjected to marked
abnormality analysis. The leukocyte differential count should be analyzed
using absolute values only (i.e., number of cells/unit volume), and not rela-
tive values (percentage of the total), which can be spuriously influenced by
changes in other cell types.
In addition to the evaluation of individual laboratory values, there may
be important clinical correlates of patterns of abnormality of multiple labora-
tory measurements. For instance, there is published guidance on drug-induced
liver injury based in part, on combinations of different patterns of elevation of
ALT, alkaline phosphatase (ALP) and conjugated bilirubin.36 It has also been
suggested that hepatocellular injury manifesting as simultaneous elevation of
ALT and bilirubin but without elevation of ALP was associated with a mortal-
ity rate of 10-15%. This outcome was explained on the basis that hepatocellu-
lar injury sufficient to reduce bilirubin excretion must involve a large fraction
of liver cell mass.37 This principle, utilized for the evaluation of hepatotoxicity
for a new drug application by the FDA,38 has stated that even isolated simulta-
neous elevations of aminotransferase and bilirubin (even if sub-clinical) have
been predictive of serious liver injury in subsequent clinical use (e.g., bromf-
enac, troglitazone, and trovafloxacin). While the positive and negative predic-
tive value of this approach remain to be validated, it is nevertheless essential to
apply multivariate laboratory analyses to the differentiation of cholestatic and
hepatocellular drug toxicity, since the former is typically relatively benign,39
while the latter may be severe or fatal.

36
Benichou, C. Criteria of drug-induced liver disorders. Report of an international consensus meeting, Journal
of Hepatology, 11 (2): 272-276, 1990.
37
Zimmerman, H. Hepatotoxicity: The Adverse Effects of Drugs and Other Chemicals on the Liver, 2nd Ed.,
Lippincott Williams & Wilkins, 1999.
38
CDER-PhRMA-AASLD Drug Induced Liver Injury Clinical White Paper. November 2000; http://www.fda.
gov/cder/livertox/clinical.pdf
39
Velayudham, L.S. and Farrell, G.C. Drug-induced cholestasis: Expert Opinion. Drug Safety, 2 (3): 287-304,
2003.

126

group6_PH.indd 126 7.8.2007 12:20:01

 l. General Benefit-Risk Considerations
No one set of criteria for risk evaluation can or should be applicable to
every investigational agent; rather, the criteria will depend on the type of
product and many other factors:
• The intended use of the product
❏ Disease prophylaxis or modification of physiological function
(e.g., vaccines, hormonal contraceptives)
❏ Diagnosis (imaging agents, radio-isotopes)
❏ Symptomatic treatment (analgesics)
❏ Cure (antibacterial therapy)

• The nature of the illness

❏ Acute non-life-threatening illness (antibiotics)
❏ Chronic disease (antihypertensives, hypoglycemics)
❏ Acute life-threatening illness (biological response modifiers,
thrombolytics)
❏ Chronic life-threatening diseases (cytotoxic or antiretroviral
agents).
• Availability of alternative therapies
Once risks are evaluated and assessed as to their importance, it is useful
to attempt to place them in the context of acceptability, not only by the trial
subjects but by the anticipated target population. There are no standard ap-
proaches to evaluating or measuring an “acceptable level of risk”, but the is-
sue must be addressed throughout clinical development. A patient’s willing-
ness to accept a certain level of risk may be different from that of the sponsor
or the regulator acting on the patient’s behalf, and therefore, consideration
must be given to the disease indication, the safety profile of the drug, the
product’s efficacy and other therapeutic options that may be available. Risk
perception and acceptance, and benefit-risk weighing by individuals, can be
subject to strong cultural and ethnic influences. In spite of the lack of vali-
dated metrics or methods for its evaluation, the notion of “acceptable risk”
should be discussed with regulators, specialists in the disease indication, and
if possible with individual patients or patient/disease advocacy groups.

m. Aggregate Analysis and the DCSI

A consistent, logical rationale should be used for deciding when
adverse events observed in the study population should be considered at-
tributable to the medicinal product, and therefore included first in the DCSI

127

group6_PH.indd 127 7.8.2007 12:20:02

 and, if confirmed in the overall study database as an ADR, in the CCSI for
an approved product (see Chapter 4, section 2.). The results of each com-
pleted study should be compared, when appropriate, with the results of prior
studies, to determine whether consistent patterns or trends are observed. As
discussed in Chapter 6, inferential statistical approaches to AE data and nu-
merical comparisons between treatment groups may be of limited use and
clinical relevance; however, descriptive statistics may provide useful infor-
mation, especially where there are comparative data across multiple studies
or in a relevant control population, such as the pooled placebo groups from
similar studies.
Clinical judgment is essential to decide when the threshold for adding
information to the DCSI has been reached, based on aggregate data. The
weight of evidence will depend on multiple factors, including differences
between treatment groups (or compared to no treatment), information on
the properties of the medicinal product including animal toxicology and
pharmacokinetic data, experience with other products in similar chemical
and/or therapeutic classes, and epidemiological information on the relevant
population. In order to err on the side of caution, the greater the likelihood
that a given AE might have a significantly serious adverse outcome for the
subject or patient, the lower should be the inclusion threshold (i.e., fewer
and less stringent inclusion criteria need be met).

128

group6_PH.indd 128 7.8.2007 12:20:02

 VI

Statistical Analysis
of Safety Data
in Clinical Trials

group6_PH.indd 129 7.8.2007 12:20:02

group6_PH.indd 130 7.8.2007 12:20:03
 a. Introduction
The use of statistics in clinical trial design and analysis has largely
been focused on the establishment of efficacy using mostly inferential but
also descriptive methods. Demonstration of efficacy is usually the main
goal for individual trials and development programs, while wishing to show
that no unacceptable toxicity occurs. However, the role of a statistician and
the use of statistics in assessing safety data are also very important and
unfortunately do not receive as much attention. Chapter 5 has outlined the
principles for analysis of risk using data generated from clinical trials; use
of the most appropriate statistical techniques for analysis and display of the
data are essential for placing the absolute and relative safety of a medicinal
product in proper perspective. Early in drug development (Phase I and early
Phase II trials), much of the assessment of safety depends on individual
case assessment. However, as the database increases, aggregate analysis
tends to become more important, and that is where statistics play a crucial
role. The techniques and approaches to use of statistics for analysing safety
data have not been developed as fully as for efficacy and it is not uncommon
to find inappropriate or incomplete displays and analysis of adverse event
data, even in refereed publications.1 In regulatory submissions for drugs in
development the situation is often better, but consideration should be given
to ensuring that publications from trials meet high standards of reporting.2
A major new extension to the CONSORT (Consolidated Standards of
Reporting Trials) Group statement recognises that efficacy has been a
major focus and seeks to redress the balance.3
This Chapter is not intended to be a manual for statistical analysis
of safety data; the subject is much too broad and complex. However, it
does highlight key points that need attention when considering analysis,
and areas which we believe may not be adequately understood or appre-
ciated. There is at least one book that addresses biostatistical aspects of
clinical safety data specifically,4 and there are several general papers that

1
For example, see Ioannidis, J. P. A. and Lau, J. Completeness of safety reporting in randomized trials – an
evaluation of seven medical areas. Journal of the American Medical Association, 285:437-443, 2001, and
Ioannidis, J.P.A. and Contopoulos-Ioannidis, D.G. Reporting of safety data from randomised trials, Lancet
352:1752-3, 1998.
2
McPherson, K. and Hemminki, E. Synthesising licensing data to assess drug safety. Brit. Med. J., 328:518-20.
3
Ioannidis, J.P., Evans, S.J., Gotzsche , P.C., O’Neill, R.T., Altman, D.G., Schulz , K. and Moher, D. CONSORT
Group. Better reporting of harms in randomized trials: an extension of the CONSORT statement. Ann. Internal
Med., 2004;141:781-8. For the original statement see- http://www.consort-statement.org/ and for the revised,
see http://www.consort-statement.org/statement/revisedstatement.htm#app
4
Solgliero-Gilbert, G. (Ed.), Drug Safety Assessment in Clinical Trials, Marcel Dekker, New York, 1993.

131

group6_PH.indd 131 7.8.2007 12:20:03

 address the topic;5 recommended references on specific topics will be given
throughout this chapter. Books that concentrate on measures of efficacy can
also be helpful since most of the same principles apply to safety data.6
Of course, this material is not a substitute for professional statistical
involvement; however, it is meant to familiarise those not trained in sta-
tistics to understand some basic approaches and techniques. In addition, a
glossary of various statistical terms and definitions used in this Chapter is
provided in Appendix 1.
Professional statistical help is required for design, analysis and re-
porting of clinical trials; statistical issues related to safety should be
considered at each of those stages.
It is important to keep in mind when conducting statistical analyses
and assessing their results, that statistical association (P-values or other
measures) alone may or may not be of clinical value. In randomised trials
they have great strength in testing causality but they inevitably have uncer-
tainty attached to what can be said. The advantage is that this uncertainty is
capable of being quantified. Therefore, the issues also require the participa-
tion and insight of clinicians. While decision analysis is an important as-
pect of modern statistical thinking, its application to monitoring the safety
of medicines has not been fully developed and even if it were developed
further, medical judgement would still need to be applied rather than me-
chanical reliance on the magnitude of a P-value, for example. Statistical
methods are a tool in the process but are by no means the process itself.
Examination of both statistical and clinical significance must involve part-
nership. This chapter concentrates on unwanted, usually adverse effects,
but these effects must always be considered in the context of the benefits of
a medicine. Attempts have been made to consider the benefit-risk balance
in mathematical or statistical terms7 but these have not reached maturity or
even a consensus on their utility and are not covered here.
5
For example, O’Neill, R..T.. Statistical analyses of adverse event data from clinical trials. Special emphasis
on serious events. Drug Information Journal, 21:9-20, 1987; O’Neill, R. T. Assessment of Safety, Chapter 13
in Biopharmaceutical Statistics for Drug Development, Karl E. Peace, Ed., Marcel Decker, New York, 1988;
and Gait, J. E., Smith, S. and Brown, S. Evaluation of Safety Data from Controlled Clinical Trials: the Clinical
Principles Explained, Drug Information Journal, 34: 273-287, 2000.
6
Altman, D.G. Practical Statistics for Medical Research, Chapman and Hall, London, 1991; Pocock, S.J. Clini-
cal Trials: A Practical Approach, Wiley, Chichester, 1983; Piantadosi, S. Clinical Trials: A Methodologic
Perspective, Wiley, Chichester, 1997. For encyclopedic style works, see Day, S. Dictionary for Clinical Trials.
Wiley, Chichester, 1999 or Redmond, C. and Colton, T. (Eds). Biostatistics in Clinical Trials, Wiley,
Chichester, 2001.
7
E.g., a recent paper is Holden, W.L., Juhaeri, J. and Dai, W. Benefit-risk analysis: a proposal using quantitative
methods, Pharmacoepidemiol. Drug Safety, 12:611-6, 2003; and a review- Holden, W.L. Benefit-risk analy-
sis: a brief review and proposed quantitative approaches, ibid., 26:853-62, 2003.

132

group6_PH.indd 132 7.8.2007 12:20:03

 b. Uses of Statistics for Clinical Safety Data
Statistics has a key role in making comparisons and in reflecting uncer-
tainty. Correct use of inferential statistical methods helps to detect real prob-
lems and treats apparent effects with appropriate caution, allowing for the
possibility that chance has played the major part in creating the findings.
The purpose of a statistical analysis is to present the data in a way that
facilitates understanding of the effects of a drug and to make clear whether
variation in results is likely to be due to chance or whether substantial ef-
fects might be associated with a drug. It is necessary to acknowledge when
the data are insufficient to draw conclusions on safety, i.e., ‘absence of evi-
dence is not evidence of absence.’ In such situations, the use of descriptive
methods and well-designed graphics will be helpful in this process.
Uncertainty in any summary of a set of data is almost entirely depen-
dent on the numbers of individuals analysed. Therefore, the ability of a
study to detect causal effects in the face of variation within and between
individuals is dependent on sample size; the smaller or rarer an effect, the
larger the sample size required, if any degree of certainty is to be given to
the study conclusions. To cite one example, under typical standards of sta-
tistical analysis (α = 5%, power 90%),8 if the background rate of an event
in the population under study is 0.1% (1 in 1,000), then in order to detect
with confidence a relative risk of 2.0 (2 in 1,000) in the experimental drug,
a study would need about 31,000 patients in each of the placebo and drug
groups!
The different stages of clinical trials may require different applications
of statistics, but the fundamental principles related to variation and uncer-
tainty apply at every stage. In addition to inferential statistical approaches,
as mentioned above, descriptive statistical methods also have an important
role in assessing data, particularly with the use of graphical and other dis-
plays, and some pointers on good practice are covered here.
Statistical approaches have application at several stages of clinical trials:
Protocol design. The objectives of the intended statistical analyses
should be specified along with an analysis plan. This implies a requirement
for sufficient numbers of patients to be included in the trial (power and

8
Reminder: the α-level is the boundary for rejection of the null-hypothesis (that there are no real differences
between the data). At a 0.05 level, the chance of a false-positive finding is 5%; conversely the chance of a true
negative is 1 – α, or 95%. Finding a false-positive result when the null-hypothesis is actually true is referred
to as a Type I error.

133

group6_PH.indd 133 7.8.2007 12:20:04

 sample size), something usually predicated on the demonstration of effi-
cacy, with the proviso that unacceptable harm is not seen. It is possible to
carry out many different analyses on different safety variables, but it is easy
to be misled as to the true degree of uncertainty for the results unless it is
clear whether that analysis was pre-planned or not. The requirement for an
analysis plan helps resolve this, but care must be taken not to adhere too rig-
idly to pre-specified analyses lest safety issues be missed. The plan should
also describe proposed strategies for dealing with missing data. Special con-
siderations are needed and should be discussed in the protocol if any interim
analyses are planned, especially those that necessitate breaking the blinded
treatment code. Such analyses require special statistical approaches.9
Beware that unless a study is designed and planned to conduct certain
analyses, they may not be powered sufficiently to carry them out, or their
interpretation may be compromised by carrying out unplanned analyses.
There is a very wide range of possible unplanned analyses that if conducted
can affect statistical significance tests.
During a trial. As part of safety monitoring during trials, and for any
possible interim analyses, data must be assembled and presented clearly to
maximise the ability to detect any unexpected and unusual results. This is
particularly important if considerations arise for stopping or modifying the
trial so that any decisions made are soundly based and their impact on the
analysis for the final report is accounted for.
For the final analysis and writing of the trial report and any publi-
cation. Comparisons are made mainly between treatment groups (for labo-
ratory data, specific AEs, classes of AEs, numbers of discontinued patients,
times to occurrence of AE or withdrawal). Within patient changes, such as
baseline to follow-up, usually require a comparison group for their proper
interpretation, largely because the effect of treatment cannot readily be dis-
tinguished from effects of time and time-dependent phenomena (e.g., time
to onset of an AE or discontinuation). Sub-group analyses (e.g., AEs by age
and sex) must be treated with great caution; it is their misinterpretation that
is a major cause of misunderstanding in medical science.10

9
Jennison, B. W. and Turnbull, B. W. Group Sequential Methods with Applications to Clinical Trials, Chapman
& Hall/CRC Press, 1999; Whitehead, J. The Design and Analysis of Sequential Clinical Trials, Second
Edition, Chichester: Wiley, London, 1997. For a Bayesian approach, see Grossman J., Parmar, M.K.,
Spiegelhalter, D.J. and Freedman, L.S. A unified method for monitoring and analysing controlled trials, Stats.
in. Med., 13:1815-26, 1994.
10
Brookes, S.T., Whitley, E., Peters, T.J., Mulheran, P.A., Egger, M. and Davey Smith, G. Subgroup analyses
in randomised controlled trials: quantifying the risks of false-positives and false-negatives. Health Technol.
Assess.,5:1-56, 2001.

134

group6_PH.indd 134 7.8.2007 12:20:04

 When combining data across different trials. In order to provide an
overview of the safety experience, summaries across trials can be made us-
ing the techniques of meta-analysis11 for formal comparisons of treatments.
On the other hand, simpler descriptive or graphical summaries of the data
from several trials may also be revealing. The pattern of results across dif-
ferent body systems will be examined to search for patterns indicating pos-
sible effects that are not seen easily in a single trial.
Although most Phase II and III trials are randomised and usually dou-
ble-blind, generally the same statistical principles discussed here can apply
to non-randomised and non-comparative studies. Most comparative trials
are designed to have parallel-treatment groups, although cross-over designs
may also be used. Analysis of cross-over studies requires somewhat differ-
ent approaches, and great care has to be taken in the interpretation of data
relating to safety. Adverse reactions may be recognised some time after the
administration of the drug (e.g., latent effects), and might incorrectly be at-
tributed to the subsequent treatment in a cross-over treatment period. Even
with a wash-out period between treatment legs, the period may not be long
enough to eliminate any carryover effects. Thus, although immediate ef-
fects can be readily detected and compared in crossover trials, any delayed
effects are difficult to attribute to treatment.
Independent of a trial design or Phase, there are many kinds of com-
parisons that can and should be made within and between treatment groups
using the proper statistical tools, depending of course on the kind and
amount of data collected. Typical analyses involve such things as: compari-
sons between treatment groups of specific AEs, classes of AEs (different
organ systems, e.g.), and laboratory data; discontinuations from treatment;
sub-population results (age, sex, etc.); time-dependent phenomena (time
to onset of AE, time to discontinuation, etc.); combining data across trials.
Approaches to some of these analyses are covered in detail below. This
chapter refers to comparisons between two groups, but the same methods
can be extended to multi-arm trials. For studies in which different doses of
the same drug are compared, then it is possible to use methods that allow
for a trend with dose, but in general the interpretation of comparisons is
easiest when the possible comparisons are between two groups.
Recommendations on types of safety analyses as well as on how to dis-
play the data can be found in ICH Guidelines E3 (Structure and Content of

11
For example, Lee, M-L. T. and R. Lazarus. Meta-analysis of drug safety data with logistic regression. Drug
Information Journal, 31:1189-1193, 1997.

135

group6_PH.indd 135 7.8.2007 12:20:05

 Clinical Trial Reports) and M4 (The Common Technical Document or
CTD);12 another useful source is the US FDA’s guidance for internal Agen-
cy review of safety data in a New Drug Application.13 A guide to the cover-
age of statistical issues for safety data as discussed in ICH Guidelines E3
and E9 (Guidance on Statistical Principles for Clinical Trials) is given at
the end of this Chapter (section j.).14

c. Principle of Intention to Treat

The principle known as “intention to treat” (ITT; also called intent-
to-treat) is probably familiar to most readers in connection with analyses
of efficacy data from randomised trials, but it may be less familiar in its
application to safety data. Intention to treat means that the various study
groups are compared using the allocation of treatment to which they were
randomised, whether they received the randomised treatment or not, or
whether they continued to take the treatment or were withdrawn early from
the trial. Recommendations have been made that at least one ITT safety
analysis should be conducted.15 This reinforces the point made in Chapter
4 that the collection of data should continue whenever possible to obtain
study endpoints even in those who are prematurely withdrawn from treat-
ment, although this recommendation applies especially to analyses of ef-
ficacy. Recent proposals have been made on how to create and use a “full
analysis” data set for ITT analyses.16 The CONSORT group (referred to in
footnote 3 above) also recommend that an ITT analysis should be applied
to data on adverse events.
The rationale for using an ITT approach is to maintain the compa-
rability of the treatment groups attained by the original randomization.

12
For the complete documents, go to www.ich.org. Clinical safety issues in M4 are found under the Efficacy
heading.
13
See FDA’s Clinical Review Template (CDER, Office of the Center Director), Section 7.0 Integrated Review
of Safety (http://www.fda.gov/cder/mapp/6010.3.pdf; effective 9 July 2004) and the more detailed Reviewer
Guidance: Conducting a Clinical Safety Review of a New Product Application and Preparing a Report on the
Review (January 2005; see http://www.fda.gov/cder/guidance/3580fnl.pdf).
14
For more discussion on ICH E9, see Phillips, A., Ebbutt, A., France, L. and Morgan, D. The ICH Guideline
“Statistical Principles for Clinical Trials”: Issues in applying the guideline in practice, Drug Information
Journal, 34:337-348, 2000.
15
Peto R., et al., Design and analysis of randomized clinical trials requiring prolonged observation of each
patient. II. Analysis and examples, Br. J. Cancer 35:1-39, 1977.
16
Stewart, W. H. Basing Intention-to-Treat on Cause and Effect Criteria, Drug Information Journal, 38:361-
369, 2004.

136

group6_PH.indd 136 7.8.2007 12:20:05

 Since patients drop out for various reasons associated with one outcome or
another, excluding dropouts can bias the results. This would be considered
especially problematic for efficacy analyses where the results may be biased
in favor of the treatment. Hence ITT analyses are considered to be the most
conservative approach.
Although considered conservative for efficacy, ITT analyses are not
conservative for analyzing adverse event data where the results will be
biased, on average, towards finding no differences between the groups (in
other words, a tendency for the groups to be more similar to each other than
they should be). Other analyses, such as including only those who received
a minimal number of doses of the study drug, may be more appropriate for
analyzing safety. However, these may also be biased. The problem is that
the direction of the bias is unknown since the relationship, if any, between
the reasons for stopping treatment and the outcome of interest will gener-
ally not be known. They may exaggerate or minimise differences between
groups.
Because an ITT analysis tends to minimise differences between groups,
not only on efficacy variables but also on adverse effects, then using sur-
vival analysis methods (discussed below) and “censoring” some data (i.e.,
excluding patients in the analysis beyond the time when their outcome is
unknown) may be very useful.
It is customary to exclude from analysis of adverse effects those pa-
tients who do not take any doses of the study drug. This does not mean that
the absolute rate of an adverse event may be estimated more reliably; it
simply reflects the fact that leaving such patients in an analysis comparing
treatment and control groups can be biased. The reasons for non-adherence
to treatment should be examined carefully.

d. Some Key Problems in Safety Analyses

Perhaps it is obvious to state that the best possible generally accepted
statistical methods should be used to analyse and present safety data, but
as already pointed out the subject has not received as much attention as has
treatment of efficacy data. Knowledge of and experience with proper meth-
ods may therefore be inadequate.
Some of the more important problems associated with safety analyses
that require attention are as follows:

137

group6_PH.indd 137 7.8.2007 12:20:05

 ❏ power: usually the ability to find statistically significant differences
between treatment and control groups for important adverse effects,
which are usually relatively rare, is low; most trials, or even com-
binations of trials, are not large enough to detect or analyse such
adverse events reliably.
❏ multiplicity: multiple analyses can be and often are performed on
the same data set, such as on multiple time points and multiple
variables. Multiplicity affects the statistical analysis, especially the
calculation of P-values, because many different comparisons of ad-
verse effects are possible. Efficacy variables are pre-defined and
limited to a few effects on the disease being studied. The number of
different types of possible adverse effects can run into the hundreds
or even thousands and making a separate comparison for each of the
different possible effects results in a very large number of analyses.
Some pointers on the impact of multiple analyses are given below,
but this area generally requires expert statistical help.
❏ medical classification: the grouping of adverse effects into catego-
ries represents a challenge; if too narrow, it results in numbers of
event types that are too small for meaningful statistical compari-
son between groups, but if groupings are too wide (having larger
numbers in the groups to avoid the first problem), it could hide the
existence of a safety problem. This is so, because a specific effect
might be hidden within a large number of adverse events that have
nothing to do with the treatment being studied. Another difficulty
arises in deciding whether groupings of different event terms for a
patient can be formally regarded as a syndrome, for which a spe-
cific diagnosis might be possible. This requires medical judgement,
and the results of the analysis will need careful interpretation rather
than reliance on the result of a statistical test. Real adverse effects
are often described by several different terms and the effects may
occur in different organs; the problem can be described as multi-
dimensional and this makes statistical analysis more complex and
difficult to pre-specify. Chapter 4 discusses the use of coding dic-
tionaries to describe medical events; the use of different dictionar-
ies and different levels within those dictionaries leads to statistical
problems.
❏ time dependency: adverse effects should be examined carefully as
a function of time on drug; simple calculations of incidences (num-
ber of events such as AEs or discontinuations divided by number of

138

group6_PH.indd 138 7.8.2007 12:20:06

 patients treated) can be highly misleading and mask the true risks
associated with the treatments.
Currently used approaches to analyses of safety data are sometimes
over-simplified, and do not take the major characteristics of adverse reac-
tions into account. For example, some reactions have a rapid onset after
administration of a drug, and if they do not occur early, are much less likely
to occur later. Anaphylaxis is an obvious example. Some reactions will oc-
cur, or become obvious clinically, only after a long duration of treatment
or long after treatment is ended. Examples include onycholysis or cancers.
The most extreme example is carcinogenesis in the daughters of pregnant
women who took diethylstilbestrol. Therefore, different time profiles are an
important aspect of ADR classification17 and allowance is often not made
for this in statistical analyses.

e. Useful Approaches to Statistical Analysis

of Continuous Measurements:
Laboratory Chemistries
Analysis methods need to be as sensitive as possible, while taking into
account the problems related to multiplicity described above. Surrogate
variables for clinically relevant outcomes should be analysed with the best
methods. The laboratory tests that involve monitoring throughout a study
(such as liver function tests, LFTs) should be analysed using the continu-
ous data as well as binary data (data composed of only two categories, such
as present/absent, alive/dead, etc.). Converting continuous measures to a
binary indicator, e.g., a criterion for elevated LFTs that is greater than three
times the upper limit of normal, loses information; while it provides a use-
ful indicator in practice, it frequently will not be able to show statistically
significant differences between groups because of the rarity of such large
changes. It is likely that appropriate analysis of the continuous data (values
at multiple time points) will show statistically significant effects even when
there are very few extreme values.
It is best to analyse laboratory data using baseline values as a compari-
son whenever possible18. The most effective approach is usually to use the

17
Aronson, J. K. and Ferner, R. E. Joining the DoTS: New approach to classifying adverse drug reactions, Brit.
Med. J., 327:1222-5, 2003.
18
Frison, L. and Pocock, S. J. Repeated measures in clinical trials: analysis using mean summary statistics and
its implications for design, Statistics in Medicine, 11:1685-704, 1992.

139

group6_PH.indd 139 7.8.2007 12:20:06

 baseline value (or the mean where multiple measurements are made) as a
covariate. The post-treatment value (or their mean where multiple measure-
ments are made) is then the response that is analysed. It is also possible to
analyse the before-to-after treatment change using the baseline as a covari-
ate. This requires Analysis of Covariance (ANCOVA), as described in any
statistical textbook.19 In practice, the post-treatment value that may be stud-
ied in the context of looking for adverse effects may be the maximum (or
most adversely extreme) value. This is particularly true of things like liver-
function tests (e.g., AST or ALT enzymes) where the highest value seen in
a particular patient is the variable analysed. Analysis of covariance can also
be used in this context, but caution in interpretation is required; the analysis
of extreme values is more vulnerable to chance variation than might be sup-
posed. Using the continuous measures is not always a substitute for study-
ing the very high values that are associated with clinically relevant events
(e.g., ALT>5 times upper limit of normal). These binary (AE occurs or not)
measures will not usually have high statistical power for the comparison,
but may be important in drawing attention to a potential issue.
Analysis of laboratory measurements used for monitoring of adverse
effects should usually be done on binary measures of clinically rel-
evant values or changes, but should also be done comparing mean
values using analysis of covariance, since this is likely to have greater
sensitivity for detecting real adverse effects.
Graphical displays, such as scatter plots of baseline versus later values
for each trial participant, can help show both a shift from average and also
draw attention to outlying values, both in terms of absolute levels but also
large changes. An example is shown in Figure 1 below. The points can be
automatically labelled with a patient ID or the treatment group. This ex-
ample labels each point as 0 for control and X for experimental treatment
group patients. It shows that there is a slight tendency for the higher post-
treatment values to occur in the experimental treatment group. It is possible
to use simple t-tests to compare baseline and final values within a group,
though it will be more efficient to compare these changes between the treat-
ed and control groups using ANCOVA with baseline value as a covariate,
as noted above. In this example the t-test does give a statistically significant
result for the difference in mean values (P=0.01) and the ANCOVA gives
P=0.001 suggesting that it is more powerful. If one were to decide that a
clinically significant value were over 6 (see Figure 1), then there are 0/39

19
A classic text is: Snedecor, G. and Cochran, W. Statistical Methods, 8th ed., Iowa State University Press (1989).

140

group6_PH.indd 140 7.8.2007 12:20:07

 in the control group and 3/41 in the experimental group, a difference which
is far from statistically significant (P=0.24 using Fisher’s exact test). This
illustrates that just using a binary cut-off will usually have the least power
to detect differences between groups.

Figure 1: Laboratory Value Scatter Plot for Individual Patients

Points are shown for baseline and final values after treatment.
The diagonal line represents equal values for baseline and final measurements.

O = Control X = Experimental treatment

8 X

X
XX XX O
6 XO X X
X
X O O
XX X
O X
X O X
OX O O O O
Final X O XO O X
XX O
X X X OX X
O X O OX O O
X O X O
X O O XO
O O
O
X OXX
XO
X
O OX O
4 OO
O X
O

2
3 4 5 6
Baseline

Final v Baseline for a laboratory test

The ICH E3 guideline notes that analysis should be applied to continu-

ous data rather than just binary categorisations but does not offer sugges-
tions for statistical techniques.
Finding a significant difference between groups does not necessarily
prove causality based on a laboratory test result, but the most powerful
statistical analysis should be used so that early signs of organ damage are
detected. Trends in average values can be a surrogate for rare, clinically
important individual changes. If the only analysis conducted is on differ-
ences between groups in the proportion of individual patient changes that
are clinically relevant, such as 3 or 10 times the upper limit of normal, such
an analysis may have too little statistical power and can fail to detect real
problems. On the other hand, it is also useful to pay careful attention to

141

group6_PH.indd 141 7.8.2007 12:20:07

 extreme values since it is possible that while the vast majority of patients
have no meaningful changes, there could be some with very high values
caused by a drug.
Most of the remainder of this Chapter discusses the treatment of
binary data, that is whether an adverse event has occurred in an individual
or not. Further description of handling continuous lab data can be found in
Chuang-Stein, et al.20

f. Statistical Treatment of Binary Data

This section will cover several key topics related to how binary data
should be analysed: the statistical power needed, the difference between,
and use of, “one-sided” vs “two-sided” statistical tests, the problems in
conducting multiple analyses (multiplicity), general measures using binary
data, the meaning and value of confidence intervals, and the importance of
taking into account the effect of time on treatment.

(1) Power Considerations

Statistical power is the ability to detect as statistically significant an
effect that is real. If only a few patients are studied then it is very easy
to miss real effects. Similarly, if an effect occurs at a very low absolute
rate, i.e., the medical event is very rare, even in those treated, then the
power to detect it will be low. The power of a particular study will de-
pend on the size of the groups being studied; the baseline or background
rate of the adverse effect of interest, which is the rate expected in the
comparison group; and the change of interest in rates between groups
(for example a doubling or tripling). It also depends on the “P-value”
set as being statistically significant, which is usually 0.05. If allowance
is made for multiple testing (see below), then this P-value may be much
smaller and so the effects will be more difficult to detect as statistically
significant and therefore the statistical power will be lower.
Although guidelines are not absolute, the expectation for a typical
drug development program is that a minimum of about 1,500 patients
will have been treated with a new drug, with about 100 followed for
at least a year if treatment is intended for long term use in non-life
threatening chronic disease (see ICH Guideline E1). With such limited

20
Chuang-Stein, C., Le, V. and Chen, W. Recent advancements in the analysis and presentation of safety data.
Drug Information Journal, 35: 377-397, 2001.

142

group6_PH.indd 142 7.8.2007 12:20:08

 numbers, uncommon or rare ADRs (e.g., an incidence of less than 1
in a thousand) will not be detected as having statistically significantly
raised rates in the treated group. Therefore, the main use of statistical
analysis is to show that uncertainty exists over the true rate of a poten-
tial ADR. The introduction to Chapter 5 discusses criteria for numbers
of patients included and notes that some rare adverse events may be
observed in the complete sets of data prior to authorisation. The prob-
lem is that they may easily be ascribed to background or the disease
for which the drug is used.
For example, a trial of 1,500 per group will only have 56% power to
detect a statistically significant doubling in risk of an ADR from 1 in
100 to 1 in 50 (1% v 2%). In other words in 56% of similar trials it
will be concluded that there is a statistically significant difference, but
in 44% of trials it will be concluded that the difference is not statisti-
cally significant and a real adverse reaction could be dismissed as just
background based on the statistical analysis. However, a tripling from
1% to 3% will have good (97%) power. If the control group rate is
only 0.1% (1 in 1,000), then the power to detect 5 times that rate is
39% with 1,500 per group, while to detect an increase to 10 times that
rate (1%) the power is 87%. This assumes that the P-value used is the
conventional 0.05.
It is clear that the problem for statistical assessment of individual tri-
als with small numbers of patients is much greater since power will
be much less. For example, the power to detect a difference between
a 1% and a 2% rate will be less than 2.5% with 100 per group, while
power only approaches a reasonable level (75%) to detect a 10% vs
25% rate with 100 per group. This means that analysis of small trials
rarely provides useful results unless the event in question occurs at a
very high rate, or if not then the data can be pooled across trials either
in a meta-analysis or by using pooled control groups for serious, infre-
quent events (see below).

(2) One-Sided vs Two-Sided Testing

The use of P=0.05 is usually based on the assumption that a difference
between treatment groups in either direction of the effect is of equal
interest, which is called two-sided (two-tailed) hypothesis testing. In
other words, both an increase in rate and a decrease in rate with the
new drug relative to the control group are of interest. On the other
hand, if it is only an increase in the effect that is regarded as relevant,

143

group6_PH.indd 143 7.8.2007 12:20:08

 then a difference in only that direction is of interest and so the P-value
should then be 0.025 instead of 0.05; that is called one-sided testing.
The one-tailed test with P=0.025 is the same test as the two sided test
with P=0.05. However, the finding that an adverse effect that occurs in
the control group is prevented by a new drug would always be of great
interest. Therefore, a statistical test that allows for either an increase
or a decrease in the rate of an adverse effect should virtually always
be specified in the protocol. Hence, it is recommended that all statis-
tical testing on safety-related data be done on the basis of two-sided
hypothesis tests.

(3) The Consequences of Multiplicity

Protocols are designed to minimise Type I errors (concluding that
efficacy exists when it really does not), and the testing of multiple
hypotheses within a single study is discouraged. However, the numbers
of potential types of adverse events are very large, so that correction for
multiple testing in a conventional way will mean that it is impossible to
draw any conclusions. It is for this reason that corrections for multiple
testing are rarely done using a formal mechanism. There are over 20
“System Organ Classes” (SOC) in each of the major schemes of classi-
fication of medical events. A statistical test is usually carried out with a
P-value of 0.05 as the cut-off for a finding to be regarded as statistically
significant. This P-value applies when a single significance test is done.
If two independent tests are done on the same set of data (for example
looking at each of two types of AEs in two SOCs) then the probability
that one of them is significant is no longer 0.05, but is closer to 0.1 (2 x
0.05). In order to ensure that the overall probability of finding a signifi-
cant result remains at 0.05, then each test has to be done with a cut-off
of 0.025. This method is called a “Bonferroni correction”.
If AEs are grouped by SOC, and a test is done for each SOC, then 20
tests will be done. It is quite likely that at least one of these tests will be
significant if a cut-off of 0.05 is used. This implies that each test will
require testing at a level of 0.05/20 =0.0025, otherwise the probability
of finding a single one significant at 0.05 becomes high (0.64). If the
grouping is more specific than at the SOC level, with perhaps 100 dif-
ferent possible groups into which AEs are classified, and a statistical
test comparing the rate of AEs between treatment and control is done
for each of the 100 groups, then the potential for finding false positive
effects (Type I errors) is even higher. Unfortunately, by having a more
stringent cut-off for noting statistical significance, there is a penalty.

144

group6_PH.indd 144 7.8.2007 12:20:08

 This penalty is that it becomes harder to reach statistical significance
for real effects, which will then have to have very large differences
between groups in the rates of AE in order for them to be regarded as
statistically significant. So if the statistical cut-off level is reduced by
adjustment for multiple comparisons, then inevitably the probability
of false negatives rises. False negatives in this context are called Type
II errors. This means we have failed to find that a real adverse effect is
statistically significantly different between the groups, when in truth
there is a real difference. This is described as a situation in which there
is low power. The probability of a Type II error is (1- power). Like
power, it depends on the background rate of the effect and the magni-
tude of the difference that is regarded as being of interest.
In practice what usually happens is that no multiple significance test-
ing adjustment is made and the finding of a statistically significant
result at P<0.05 is a signal to explore the possibility of a real adverse
effect more carefully.
Since we will always be concerned about the lack of power in looking for
adverse effects, if adjustment for multiple comparisons is made, then it
should use a more sensitive method than Bonferroni, one which will not
decrease power as much. Discussion of these advanced approaches is
beyond the scope of this Chapter, but useful literature is available.21,22,23

(4) General Measures Using Binary Data

Participants are usually randomly allocated to experimental treat-
ment or control in parallel groups. In crossover trials, participants are
randomly allocated to one order of treatment, such as control, then
experimental. As pointed out earlier, special methods of analysis are
applicable to crossover trials, which will not be covered here.
Assume there are a participants in an experimental treatment group
who have an AE of interest, while b do not; the corresponding num-
bers for the control group are c and d.

21
Simes, R.J. An improved Bonferroni procedure for multiple tests of significance. Biometrika, 73:751-4, 1986
and Ludbrook, J. Multiple comparison procedures updated. Clinical and Experimental Pharmacology and
Physiology, 25: 1032-7, 1998.
22
Benjamini, Y. and Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to
multiple testing. J. Royal Stat. Soc. Series B, 57: 289–300, 1995.
23
Lin, K.K. and Rahman, M.A. Overall false positive rates in tests for linear trend in tumor incidence in animal
carcinogenicity studies of new drugs, J. Biopharm. Stat., 8: 1-15, 1998 (also, see discussion pp. 17-22).

145

group6_PH.indd 145 7.8.2007 12:20:09

 Table 1 shows the data in a 2 x 2 layout (a 2x2 contingency table).
Table 1: Occurrence of Adverse Events by Treatment Group

With AE Without AE Total

Experimental a b a+b
Control c d c+d

The proportion of patients on the experimental drug with that event

during the trial = a/(a+b). The control group proportion is c/(c+d).
Statistical significance tests can help to decide whether such differ-
ences can occur just by chance when there is no true difference. These
proportions are referred to as the risk of that adverse event. The dif-
ference in proportions is then called the risk difference, or absolute
difference in risk. The risks and the difference in risks may be given as
percentages but it is better to express them as proportions.
The main statistical test for the comparison of proportions is the rela-
tively familiar chi-square test. This is a statistical test of the hypothesis
that there is no difference between the experimental and control groups
in the proportions (risks) with the adverse event. This test uses the four
numbers in the four cells of the table – a,b,c,d – in a formula that al-
lows for a P-value to be calculated. It makes some assumptions about
the data that depend on their being quite large numbers in the table as
a whole. An alternative that does not make this assumption is called
Fisher’s exact test. These two methods are tests of the null hypothesis,
that the proportion with that AE is equal in the two groups – i.e., the
difference in proportions is zero. A significant result (P<0.05) occurs
when the differences between the groups in the proportions with the
AE is sufficiently large.
The magnitude of the difference between the experimental and control
groups may also be expressed in relative terms. Two of these relative
measures are the odds ratio (OR) and relative risk (RR). The odds of the
adverse event in the treated group are a:b, while the odds in the control
group are c:d.
The odds ratio (OR) is: (a/b)/(c/d) = ad/bc
The ratio of the proportions is a risk ratio also referred to as a relative
risk (RR). If there is no difference between groups, the ratio is there-
fore 1.These measures have become familiar in the post-marketing

146

group6_PH.indd 146 7.8.2007 12:20:09

 arena where analysis of spontaneous reports uses a 2 x 2 table of the
type shown above. Unlike clinical trial situations, for which the num-
bers of patients with adverse effects is accurately known, the only valid
calculation for spontaneous reports is a “reporting ratio,” which reflects
the number of reports received, but not the real number of patients with
the AE. One technique allows a determination of the “Reporting Odds
Ratio”24 while another yields a “Proportional Reporting Ratio” (PRR)25
which is similar arithmetically to a relative risk.
The RR for the example given above is (a/(a+b))/(c/(c+d)). It can be
seen that if a is much smaller than b and also that c is much smaller than
d (a rare adverse reaction), then (a+b ~ b) and (c+d ~ d); in such a situa-
tion, dealing with rare events, the OR and RR are approximately equal.
To take a specific example from a large trial, the comparison between
estrogen alone (E) and placebo in the Women’s Health Initiative (WHI)
study,26 generated the data in Table 2 for stroke.

Table 2: Occurrence of Stroke by Treatment Group (WHI Study)

With stroke Without stroke Total

E 158 5152 5310
Placebo 118 5311 5429

The proportions with stroke were:

(158/5310) = 0.029765 with E and
(118/5429) = 0.021735 with placebo.
The difference in proportions is 0.008030, which when rounded is
0.008 or 0.8%.
The odds of having stroke were:
(158/5152) = 0.03067 with E and
(118/5311) = 0.02222 with placebo.
The odds ratio is 1.38, and the relative risk is 1.37.

24
van Puijenbroek, E.P., Bate, A., Leufkens, H.G., Lindquist, M., Orre, R. and Egberts, A.C. A comparison of
measures of disproportionality for signal detection in spontaneous reporting systems for adverse drug reac-
tions. Pharmacoepidemiology and Drug Safety, 11:3-10, 2002.
25
Evans, S.J., Waller, P.C. and Davis, S. Use of proportional reporting ratios (PRRs) for signal generation from
spontaneous adverse drug reaction reports. Pharmacoepidemiology and Drug Safety, 10:483-6, 2001.
26
Anderson G.L., et al. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy: the
Women’s Health Initiative randomized controlled trial. J. Am. Med. Assoc., 291:1701-12, 2004.

147

group6_PH.indd 147 7.8.2007 12:20:10

 The chi-square test compares the observed numbers in each group
with the numbers that would be expected if there were no difference in
the proportions between the groups, but they had the proportion seen
in the trial as a whole. The simple chi-square value from this table of
data = 6.9, with an associated P-value = 0.0086. Therefore, the conclu-
sion from these data was that the difference between the groups in the
risk of stroke was statistically significant. It is recommended that exact
P-values are reported rather than just, for example, P <0.01. Fisher’s
exact test produces a P-value of 0.009. Usually the chi-square test will
have a smaller P-value than Fisher’s exact test, but with large numbers,
as here, they will be similar. However, the difference can be larger
when there are small values in any of the cells in the table. Fisher’s
exact test should be used whenever the expected numbers in any cell of
a 2x2 table are less than 5. An alternative is called “Yate’s Correction”,
which reduces the magnitude of a chi-square test and details can be
found in, e.g., the book by Altman mentioned in footnote 6.
Odds are always larger than proportions and the odds ratio for a given set
of data is always further from the null value of 1 than is the relative risk.

(5) Confidence Intervals

P-values have their uses, but it is usually better to report confidence inter-
vals (CIs). A confidence interval is a measure of the amount of statistical
uncertainty around a summary value known as the point estimate. This
estimate will not necessarily be the true value, which may only be known
if we have infinite knowledge about the parameter. What is needed is an
awareness of whether our estimate is likely to be close to the true value or
not. We construct confidence intervals for summaries of data such as pro-
portions or differences in proportions. Similarly, confidence intervals for
odds ratios or relative risks (by their nature comparative summaries) may
be obtained, as well as the perhaps more familiar CIs for means or differ-
ences in means. The confidence intervals for the RR and OR are based on
taking their logarithms and so the CIs will be symmetric on a log scale,
but asymmetric on the original scale of RR or OR. The null value, imply-
ing no difference between compared groups, is zero for both log (OR)
and log (RR), corresponding to ORs or RRs that are equal to 1.
Different approaches to statistical inference have different interpretations
(particularly the Bayesian approach), but it can be stated generally that
a confidence interval reflects statistical uncertainty in a summary value
(the point estimate).

148

group6_PH.indd 148 7.8.2007 12:20:10

 For the data from the WHI study discussed above, the risk difference
(RD) = 0.008, with a 95% confidence interval (CI) = 0.002 to 0.014;
the OR =1.38, with a 95% CI of 1.085 to 1.76; the RR =1.37, with a
95% CI=1.08 to 1.73. Each of these intervals excludes the null value
for the relevant summary (0 for RD and 1 for OR and RR). In other
words, with 95% confidence the data do show a difference between
groups.The differences or ratios are statistically significant with a P-
value <0.05. The general principle is that the 95% CI will exclude the
null value when the difference is P<0.05, and if it includes the null,
then P≥0.05. There is a danger of relying too much on whether by itself
P < 0.05 or P≥0.05; the CI gives more information. This is especially
true with non-significant differences between groups, where a large CI
will be compatible with the observed data. In a small study or one with
rare outcomes, the CI will be very wide and it then shows that even
substantial differences cannot be ruled out. This shows that CIs are
particularly useful for dealing with adverse event data.
Results of trials should show confidence intervals for a relevant
summary of the data rather than just quoting the P-value from a sig-
nificance test.
Details on how to perform the statistical tests and calculate confidence
intervals for odds ratios and risk ratios are given in intermediate level
textbooks on medical statistics or epidemiology.27 Currently available
statistical software programs are able to calculate confidence intervals
for rates or proportions reliably, even with small numbers.
A very simple example of a confidence interval is reflected in the “rule
of 3”. This was discussed in Chapter 5 to show the sample size re-
quired to detect even a single occurrence of an outcome in a trial. The
same principle can be used to construct an approximate 95% confi-
dence interval when zero events (of say a particular potential adverse
reaction) are observed. The rule can be restated as “when we saw zero
events (of any kind), we could have seen three”. Thus our uncertainty
in the observed rate of zero will depend on how many times we looked
for that event. If we actually saw zero occurrences in 10 patients, then
from a purely statistical perspective the approximate 95% upper con-
fidence level would be 3. If we had looked at 100 individuals and seen
zero, then the true rate might easily be 3/100, 3%; similarly if 1,000

27
Rothman, K.J. and Greenland, S. Modern Epidemiology, 2nd ed., Lippincott-Raven, Philadelphia, 1998 and
Altman, D.G. Practical Statistics for Medical Research, Chapman and Hall, London, 1991.

149

group6_PH.indd 149 7.8.2007 12:20:11

 were studied, the upper CI for the true rate would be 0.3%. This “rule”
is quite a good approximation when zero events are seen. It cannot
routinely be applied when one or more events are seen.

(6) Accounting for Time on or off Treatment

It is very important that the length of time that each patient is “at risk”
of having an adverse event be factored into any assessment of risk. The
length of time over which a patient is followed in order to determine
whether an AE occurred will not always be the same for every partici-
pant in a trial. This period may or may not be the same as the duration
of treatment, which in some settings such as single dose studies, may
well be the same for all patients. The planned time on treatment and
any post-treatment follow-up will have been set by the trial protocol as
the prescribed observation period. This period for each patient should
normally extend to at least 5 half-lives once treatment is stopped. Some
patients will not be followed for the planned time, especially in long-
term studies. As already recommended, even if treatment is ceased,
follow-up should continue so that delayed occurrences of new AEs or
changes in an existing AE be recorded, whether they are considered
due to a specific treatment or not.
Events that occur beyond the standard observation period can be dif-
ficult to include in a formal analysis, since unless all patients are fol-
lowed for the same length of time post-treatment, it will not be known
whether others also experienced the same or different events. Such
post-treatment events should be documented, of course, and discussed
in a trial report, but it is not usually appropriate to include them in
the formal statistical analysis since bias could result. It is essential
that the protocol clearly defines the end of the observation period. See
Chapter 4 for a discussion on how to handle post-treatment/post-study
event reports. Their clinical relevance may in some circumstances
be considerable, but formal statistical methods require that all those
included in the analysis are, in principle, treated equally. Follow-up
beyond the defined end of the study is not likely to be equal for all
those who reached the end of the study. However, ignoring such events
is not satisfactory either.
Calculating the sum of the total time at risk for all patients by treatment
group is useful, and this should be reported, often as person-time (e.g.,
person-years). The incidence rate is the total number of those having the
event divided by the person-years at risk, and the ratio of incidence rates

150

group6_PH.indd 150 7.8.2007 12:20:11

 between treatment and control groups is a rate ratio. This assumes
that the incidence rate is constant over time but this will often not be
true. Rates per person-time are recommended instead of just numbers
of patients with an event divided by the total numbers that were in the
relevant group. These values can be useful in carrying out a meta-
analysis. Other, possibly better methods are discussed below.
Rates per person-time for each treatment group should be reported
in addition to numbers of patients with an event divided by the total
number of patients in the relevant at-risk group. This is especially
important when combining data from studies involving different
treatment durations.
An interesting example comes from the Women’s Health Initiative
(WHI) randomised trial comparing estrogen + progestin (E + P, or
HRT) with placebo.28 There was a mean of 5.2 years of follow-up, so
that in the HRT group there were 44,075 person-years (p-years) and in
the placebo group there were over 41,289 p-years at risk. The rate per
10,000 p-years for coronary heart disease (CHD) was 38 and 30 in the
treated and placebo groups, respectively, which are averages over the
whole follow-up period of the trial. The rate ratio is 1.27 (the same as
the risk ratio to two significant figures). A risk has number of individu-
als as the denominator, whereas a rate has person-time as the denomi-
nator. The risk ratio is often referred to as a relative risk but both the
risk ratio and the rate ratio are described as relative risk measures.29
It should be noted that the assumption made when using person-years
as the denominator, is that the risk of having an adverse event is con-
stant at all times during the follow-up period. The risk per unit time is
called the hazard rate and using total person-years as the denominator
assumes that this rate is constant over time. With some types of ad-
verse reaction this assumption may be reasonable but often this is not
the case. For example, most hypersensitivity reactions are relatively
rapid in onset and if they do not occur early in treatment then their
likelihood of occurring later is very low. At the other extreme, any
causal effect on cancer is likely to take at least a year and usually at
least three years before it could be detected. This is illustrated by the

28
Rossouw, J.E., et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: princi-
pal results from the Women’s Health Initiative randomized controlled trial., J. Am. Med. Assoc., 288:321-33,
2002.
29
Rothman, K. and Greenland, S. Modern Epidemiology, 2nd Ed., Lippincott Raven Press, London, 1998, p. 49.

151

group6_PH.indd 151 7.8.2007 12:20:12

 data shown in Table 3 taken from the WHI report.30 A different as-
sumption is that the ratio of the hazard rates in two groups is constant.
This may be more realistic and analysis methods that utilise this as-
sumption are given below.
Table 3. Participant-years, Numbers of Cases of Breast Cancer, Rates and Rate Ratios
by Follow-up Year and Treatment Group in the WHI Trial

HRT Placebo HRT Placebo HRT Placebo HRT/placebo

Year
p-years p-years BC* BC* rate** Rate** rate ratio
1 8435 8050 11 17 13 21 0.62
2 8353 7980 26 30 31 38 0.82
3 8268 7888 28 23 34 29 1.17
4 7926 7562 40 22 50 29 1.72
5 5964 5566 34 12 57 22 2.59
6+ 5129 4243 27 20 53 47 1.13
Total 44075 41289 166 124 38 30 1.27
* BC= Number of cases of breast cancer / ** Rate per 10,000 participant-years

It could be argued that the expected effect of HRT on breast cancer

should only start to appear after two to three years, so using the total
person-time as the denominator is very misleading. However, as an
aside, it is also possible that HRT makes reading of mammograms
more difficult even after only a short period of use. It is often found, in
the summary of trials submitted for licensing of a new medicine, that
the total person-time in the treated group across all the trials, or even
the total number of patients treated, is the denominator used in deter-
mining the rate of occurrence of adverse events. This is rarely the best
way of presenting or summarising the data, and must be treated with
great caution. The correct ways of dealing with this issue have been
described31 but are often ignored.
The correct method is to use a “life-table” or survival analysis, even
though here it is not the time to death but an adverse event that is stud-
ied. ICH Guideline E3 mentions survival analysis methods for analys-
ing safety data (in section 12.2.3), but it appears that this has often
not been followed in practice. It is of greatest importance when there

30
Rossouw, J.E., et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: princi-
pal results from the Women’s Health Initiative randomized controlled trial., J. Am. Med. Assoc., 288:321-33,
2002.
31
O’Neill, R.T.. Statistical analyses of adverse event data from clinical trials. Special emphasis on serious
events. Drug Information Journal, 21:9-20, 1987.

152

group6_PH.indd 152 7.8.2007 12:20:12

 are notable numbers of patients who are lost to follow-up or withdraw
from the trial for any reason. Using these methods will lead to higher
estimates of the rates of an adverse effect than simply calculating the
total number of AEs divided by the number of patients treated; how-
ever, if drop-out rates are equal between groups it will apply equally
to both treatment groups. If the drop-out rate is higher on placebo, for
example, then it will lead to a higher rate in the placebo group. The
important point is that survival analysis gives a better, less biased esti-
mate than the crude analysis.
From the perspective of illustrating the course of an adverse event, it
is very much preferred to present the cumulative hazard and a good
example is shown in Figure 2 taken from the report on the Women’s
Health Initiative study (see footnote 30).
Figure 2: Example of Kaplan-Meier Estimates of Cumulative Hazard for Stroke
(derived from Rossouw, et al. See footnote 30)
Stroke
HR, 1.41
95% nCI, 1.07-1.85

E+P
Placebo

Time (years)

0 1 2 3 4 5 6 7
E+P 8506 8375 8277 8155 7032 4272 2088 814
Placebo 8102 8005 7912 7804 6659 3960 1760 524

Figure 2 illustrates, for stroke, the cumulative hazard in each of the

treatment groups. These curves show the rate at which new strokes are
occurring in the two groups, as time from start of the study increases

153

group6_PH.indd 153 7.8.2007 12:20:12

 along the X-axis. The Y-axis relates to the proportions having a stroke.
At each time point when an adverse event (stroke) occurs, the risk of
occurrence is calculated based on the number of adverse events oc-
curring at that time, divided by the number of participants still at risk
of having that event at that time. The numbers in each group still in
the trial at that time are also shown below the X-axis. Those who have
dropped out of the trial by that time point, for whatever reason, are not
counted in the denominator. The method was used for other adverse
effects and can also be used to examine benefits; the published figures
also showed benefit for two categories of clinical outcome, reduction
in colo-rectal cancer and hip fracture.
This and similar “survival” methods may be generally applied to ad-
verse events, though their original use was in looking at death rates.
The method is similar to that used for survival curves using a Kaplan-
Meier estimate of survival. Kaplan-Meier curves start at 100% (every-
one is alive) and move downwards over time; adverse events are best
shown as cumulative hazard plots which move upwards over time as
shown in Figure 2 above.
The calculation of cumulative survival is simple and is given in most
introductory medical statistics books.32 The curves, such as Figure 2,
are derived from more complex methods and require computer soft-
ware for their preparation.
The curves derived from the Kaplan-Meier or cumulative hazard
methods can themselves be misleading if too much attention is paid
to the data at longer times. This is where the estimates are at their
most uncertain, since the numbers “at risk” may be rather small. Good
practice truncates these curves so that data based on very few obser-
vations are not included. Figure 2, as we have noted above, gives the
numbers at risk (which is a good practice in presenting such figures)
but it can be seen that the numbers fall off sharply after 4 years of
follow-up, so that by year 6 less than 25%, and by year 7 less than
10% of those originally randomised are at risk of having events. There
is therefore much greater uncertainty in the position of the curves at
the time points beyond 6 years. For most clinical trials in new product
development programs, the periods of observation are usually much
shorter: for example, days or weeks for acute anti-infective treatment,
months for intermediate term therapy, and one or two years for chronic

32
For example, Altman, D.G. Practical Statistics for Medical Research, Chapman and Hall, London, 1991, p. 368.

154

group6_PH.indd 154 7.8.2007 12:20:13

 therapy. However, the same cumulative hazard method is appropriate;
the numbers at risk may not fall off as quickly or as much compared to
very long term observation periods, as in the WHI program.33

(7) Statistical Tests Using Time Since Start of Treatment

The Kaplan-Meier method does not directly provide significance tests
or confidence intervals for comparisons between groups. It is possible
to treat the data as comparisons of proportions as discussed above, but
these do not take into account differences over time and do not fully
utilise the data. The simplest method of comparing the curves is the log
rank test.34 Although the result of this test can be expressed as a chi-
square value, it is not the same as the simple chi-square test discussed
earlier. The log rank test treats the data in a similar way to calculating
a Kaplan-Meier estimate. At each time point where an adverse event
(a “failure”) occurs, it is assumed that the rate should be the same in
the treated as in the control group. An overall rate across both groups
is calculated so that an expected number of failures is obtained for
each group at that time point. The cumulative difference between the
observed number of failures (O) and the expected number (E) for the
whole time period under consideration is obtained and (O – E)2 /E can
be compared to a chi-square distribution on one degree of freedom for
testing the difference between the curves. This is a test of the null hy-
pothesis that the two curves are identical. It does not assume anything
about the hazard rate itself – it does not have to be constant, but it does
assume that the ratio of the hazards is always constant and equal to
one. There are various subtle modifications of the log rank test that ap-
ply different weights to the information at the beginning of follow-up
compared with that at the end of follow-up. Further details on survival
analysis can be found in Collett (1994)35.
A more complex method for comparing time to event data is a “pro-
portional hazards regression” or “Cox regression”. This, like the log
rank test, compares an entire survival curve without making assump-
tions about the form of the hazard rate at any particular time, but it
does assume that the ratio of the hazard rates between two groups

33
For an extensive discussion of “chronology bias” in general, see Chapter 7 in Feinstein, A. R., Clinical Bio-
statistics, C. V. Mosby, St. Louis, 1977. Also published as Clinical Biostatistics. XI. Sources of ‘chronology
bias’ in cohort statistics, Clin. Pharmacol. Ther., 12:864, 1971.
34
Peto R., et al., Design and analysis of randomized clinical trials requiring prolonged observation of each
patient. II. Analysis and examples, Br. J. Cancer, 35:1-39, 1977.
35
Collett, D., Modeling Survival Data in Medical Research, Chapman and Hall, London, 1994.

155

group6_PH.indd 155 7.8.2007 12:20:13

 is constant at all times. This method can be used to adjust for other
prognostic factors as well as for making a comparison between a treat-
ed and control group. It may be used for data from both randomised
trials and observational cohort studies. The result of the Cox model is a
hazard ratio, which is analogous to a relative risk averaged over all the
time points considered. It also allows for a confidence interval around
the hazard ratio to be calculated. The single value for the hazard ratio
assumes that it is constant over the time period studied.
In the WHI estrogen-alone study described above (e.g., see Table 2),
the estimated hazard ratio for stroke was 1.39 with a 95% confidence
interval of 1.10-1.77, derived from a Cox model analysis. This is simi-
lar to the point estimate of relative risk calculated above as 1.37 with
a CI of 1.08 to 1.73. The Cox model took into account age, prior dis-
ease and the treatment group in a simultaneous low-fat diet trial. These
adjustments will make less difference to the results in a randomised
clinical trial than in an observational study, but even in a randomised
trial, important explanatory variables measured at baseline should be
included in the analysis.
It is possible to use other statistical models that assume a particular
form for the hazard rate, and these are called parametric methods. For
example, the exponential model assumes a constant hazard rate. It is
possible to allow for hazard rates that increase or decrease or are even
J-shaped, such as the “Weibull” model. Some of these methods are de-
scribed by Collett.36 There are also methods available for checking the
assumptions of survival analysis and these should be used when exam-
ining the difference between groups in rates of occurrence of adverse
outcomes. This reflects the general principle that statistical tests make
assumptions; these assumptions should be checked for validity in the
particular sets of data under study.
When comparing rates using the number of cases with events as the
numerator and person-time as the denominator, the basic assumption
is that the number of cases follows a Poisson distribution. Analysis of
these rates uses Poisson regression.37 The results of these analyses can
be expressed as incidence rate ratios.
The results from a Cox model analysis are always presented as relative
measures of the effect rather than as absolute measures. It is not
36
Collett, D., Modeling Survival Data in Medical Research, Chapman and Hall, London, 1994.
37
Clayton, D. and Hills, M., Statistical Models in Epidemiology, Oxford University Press, Oxford, 1993.

156

group6_PH.indd 156 7.8.2007 12:20:14

 possible to obtain absolute measures of rates, or relative risks at a
specific time point directly from the analysis. With parametric
methods it is possible to obtain absolute measures, and this approach
may therefore be used more often in the future.
Adverse events that occur with sufficient frequency for formal analysis
should be analysed using “survival” type methods, and consideration
should always be given to showing graphs of cumulative hazards.

g. Combining Data from Several Trials:

The Role of Meta-analytical Techniques
A major problem with most individual clinical trials during drug de-
velopment and even for Phase IV studies is that they tend to be too small
to detect uncommon or rare ADRs. There are obvious benefits to be gained
from putting all the available information together to increase statistical
power, a process referred to as a meta-analysis (also called a standard “sys-
tematic review,” viz., the process of defining the problem, searching for all
data, combining, analyzing and presenting the data). In principle, this is
more important for analysis of ADRs than for analysis of efficacy. Howev-
er, most of the problems with individual trials are not solved by combining
data. Important problems which remain relate to the classification of ADRs
and in ensuring that all the relevant data have been captured. If the trials
have excluded those likely to be treated in clinical practice then meta-
analysis might give a false sense of reassurance. A major problem with
using a meta-analysis is that the data may be derived only from published
papers. Those data are prone to “publication bias,”38 namely that you may
never know how much relevant unpublished data exist; even if you are
aware of such data, access may not be possible. On the other hand, when
applying for marketing authorisation for a new drug, both regulators and
the company will have access to complete data on the drug and publication
bias is not an issue, even if some of the data may have been published.
No absolute criteria can be established for whether data from different
trials can be combined so as to yield a valid analysis. However, some points
that should be considered are listed here and the “QUOROM” guidelines39
are helpful in setting out some principles.
38
Egger, M., Davey, S.G. and Altman, D.G., Editors, (2001) Systematic Reviews in Health Care. Meta-analysis
in Context. [2nd Edition of Systematic Reviews], British Medical Journal Books, London, 2001.
39
Moher, D. Cook, D.J., Eastwood, S., Olkin, I., Rennie, D. and Stroup, D.F. Improving the quality of reports
of meta-analyses of randomised controlled trials: the QUOROM statement. Quality of Reporting of Meta-
analyses, Lancet, 354:1896-1900, 1999.

157

group6_PH.indd 157 7.8.2007 12:20:14

 Questions that must be considered for a meta-analysis:
a) Is the experimental drug the same in all trials?
Same dose? Same regimen?
Same formulation?
Same route of administration?
b) Is the comparator the same?
Placebo or active comparator?
Dose of active comparator?
c) Is the duration of treatment the same?
d) Are the protocols similar?
Are AEs sought in similar ways?
Inclusion and exclusion criteria?
A further set of questions relate to the specific patients included.
e) Is the patient population similar?
Age, sex, race, concurrent disease?
Disease state, duration and severity?
Even when the answer to some of these questions is “No” it does not
mean that a meta-analysis is impossible or inappropriate. It is important to
exercise judgement on what is sufficiently similar to shed light especially
on rare effects. The main purpose of a meta-analysis in this context is to
obtain sufficient data on a rare outcome for which a single trial does not
provide a sufficient answer. It may be helpful to use graphical methods in
meta-analyses, which can show similarities and differences for common
as well as rare effects across different trials in a clear way. They can also
be used to illustrate the uncertainty in effects so that apparently dissimilar
results may be seen to be simply different by chance.
The greatest strength of a meta-analysis of trials is that the results
which are combined are the within-study, between-treatment group differ-
ences. It means that the different studies themselves are not assumed to
have similar results, but it is assumed that the between-treatment differ-
ences are relatively similar across studies. One of the consequences is that
it is important that the scale on which the differences are measured is kept
consistent across studies. If the (absolute) baseline risk varies across stud-
ies, it may be that the (absolute) risk difference differs markedly across
studies, but the odds ratio is reasonably consistent. Therefore, pooling the
odds ratios across studies may be the best approach. Methods that assume

158

group6_PH.indd 158 7.8.2007 12:20:15

 the between-treatment differences are constant across trials are called
fixed-effect models; allowance may be made for some heterogeneity in the
between-treatment differences, and these are called random-effects models.
If the variation is very large, then even a random-effects model may not be
sensible, and the very idea of combining disparate results should be ques-
tioned. The detailed statistical methods are beyond the scope of this chapter
but are found in the literature.40
A frequently used but weaker, and in some instances flawed, approach
to combining data is simply to add up the numbers across all trials of all the
adverse events in the experimental group divided by the number of all the
patients randomised to treatment. The same is done for the control group
and the overall rates compared. In some instances this will give a similar
result to that from a proper meta-analysis, but in most cases it will have less
precision and may be biased. This is particularly likely when there has been
unequal randomisation to experimental and control groups in some of the
trials, and such a combination can be very misleading. Over- or under-
estimation of between-treatment rates of events can occur. The method
should not be used routinely. It also has problems when different durations of
treatment and/or follow-up are combined. It is possible to use meta-analysis
of individual patient data with survival analysis methods that does allow for
different follow-up but this is relatively complex and is not in routine use.41
A meta-analytic review should be a routine part of the drug develop-
ment process so that ADRs, and differences in ADR rates between treat-
ment groups, can be detected as readily as possible.42 Crude pooling of
adverse event numbers across different trials to compare experimental
and control groups should be avoided if possible.

40
Sutton, A.J., Abrams, K.R., Jones, D.R., Sheldon, T.A. and Song, F. Methods for Meta-Analysis in Medical
Research, Chichester-Wiley, London, 2000 or the book by Egger, et al. (see footnote 38).
41
Higgins, J.P., Whitehead, A., Turner, R.M., Omar, R.Z. and Thompson, S.G. Meta-analysis of continuous
outcome data from individual patients. Statistics in Medicine, 20:2219-2241, 2001; Stewart, L.A. and Parmar,
M.K. Meta-analysis of the literature or of individual patient data: is there a difference?, Lancet, 341:418-422,
1993; and Duchateau, L., Pignon, J.P., Bijnens, L., Bertin, S., Bourhis, J. and Sylvester, R. Individual patient-
versus literature-based meta-analysis of survival data: time to event and event rate at a particular time can
make a difference, an example based on head and neck cancer, Controlled Clinical Trials, 22:538-547, 2001.
42
Lee, M-L. and Lazarus, R. Meta-Analysis of drug safety data with logistic regression, Drug Information Jour-
nal, 31: 1189-1193, 1997; Temple, R. Meta-analysis and Epidemiologic Studies in Drug Development and
Postmarketing Surveillance, J. Amer. Med. Assoc., 281: 841-4, 1999; and Koch, G.G., Schmid, J.E., Begun,
J.M. and Maier, W.C. Meta Analysis of Drug Safety Data in Solgliero-Gilbert, G (Ed.), Drug Safety Assess-
ment in Clinical Trials, pp. 279-304, Marcel Dekker, New York, 1993.

159

group6_PH.indd 159 7.8.2007 12:20:15

 h. Analysis of Rare Events
When rare events occur in a trial it will be difficult to conclude anything
about relative rates of occurrence, or possibly causality, from an analysis based
solely on that trial’s randomised groups. A conventional statistical test to com-
pare proportions as described above, Fisher’s exact test, will find a count of
5 events in one group compared with zero events in the other as not being a sta-
tistically significant difference, when the sample size is 25 per group or larger.
The chi-square test will give a similar result because it assumes that the over-
all observed numbers in total give the “expected” numbers for each group. If
there are 25 subjects in each of two treatment groups and the observed number
of events is 5 and 0, the chi-square test assumes that 2.5 events are expected
in each group. In this situation, from a statistical standpoint the difference
between 5 observed and 2.5 events expected (or between 0 observed and 2.5
expected) is not seen to be very great. However, if other reasonably reliable
evidence indicates that the control group rate actually is expected to be very
close to zero, then the count of 5 vs 0 may be very significant medically, even
if not statistically. Clearly medical judgment would prevail to see that the trial
is stopped even before 5 events of a very serious nature, such as liver failure,
occurred in basically healthy patients treated for a headache.
On statistical grounds, as has been noted using the “rule of 3”, the up-
per 95% confidence limit on zero events occurring in 25 patients is 3/25 =
12%. If there are data derived from some other, larger source that give a very
different bound on the true rate of that event in an approximately similar
population, we can use those data. If, for example, we are sure that the rate
does not exceed 0.1%, then we can apply statistical methods used in other
contexts, notably in spontaneous reporting, to calculate the ratio of observed
to expected occurrences and derive a very different P-value. If we are con-
fident that the expected rate is 0.1% (1 in a thousand), the ratio of the ob-
served to the expected is 5/(25x0.001) = 200. This is very different from an
observed/expected ratio based on the conventional statistical test described
above, which yields a ratio of 5/2.5. An appropriate statistical test gives a
P-value of <0.00001. Even a single observed event with an expected per-
centage rate of 0.1% in such circumstances gives a statistically significant
result. Hence if external information is available about the rate of an event in
a similar population and it is known to be reasonably precise, careful analy-
sis of rare events may then be amenable to statistical analysis rather than be-
ing dependent only on subjective judgment.
This type of methodology has generally not been applied routinely in
clinical trials but there is an opportunity for it to be used more than it is.

160

group6_PH.indd 160 7.8.2007 12:20:16

 Background rates can be obtained from a number of population-based data
sources, as described in Chapter 3. Some of these databases are listed in Ap-
pendix 9. Pooling of control groups from historical clinical trials is another
potential source of background rates. Optimum use of such data would
require that either regulatory authorities or companies with large databases
make them available, for example, to determine the expected rates of very
serious events like liver failure or cancers. Further research is needed to ob-
tain rates of occurrence of events that are extremely serious and rare. These
rates should be published so that interpretation of a single or a few such
events can be accomplished more objectively than is currently possible.
When a greater number of a serious but rare events is observed compared
to that expected on the basis of background data, this will always be cause for
careful scrutiny. However, it is important to keep in mind that the unexpected
can and does occasionally occur by chance. Multiple cases can occur close
together in time or space (sometimes referred to as a cluster of cases) which
may not be caused by the drug. Chance can be an explanation, or there may
be some external factor not associated with the trial treatment that is pro-
ducing multiple cases. This is why a within-trial comparison will always be
more reliable for deciding on causality. The use of background data, whether
from population-based data or from pooled control groups from many trials,
is subject to more uncertainty than the comparison of randomised groups.
Further research is required, through examination of large databases
of completed clinical trials, to attempt to obtain rates of occurrence
of events that are serious and rare. These rates should be published
so that interpretation of a single or a few such events can be accom-
plished more objectively than is currently possible.

i. Measuring and Expressing Effects in Ways

Relevant to Public Health
When an effect of a medicine is genuinely causal, then relative mea-
sures like the odds ratio, relative risk, rate ratios, and hazard ratios are often
fairly high. At the same time of course, some effects are accepted as causal
but do not have very high relative risks (in the sense of being far from
1), such as the effect of a statin seen in the MRC/BHF Heart Protection
Study43 on coronary death rate, which was 0.83. These small effects need
43
Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of cholesterol lowering with
simvastatin in 20,536 high-risk individuals: a randomised placebo-controlled trial. Lancet, 360:7-22, 2002.

161

group6_PH.indd 161 7.8.2007 12:20:16

 to be studied in very large numbers; the Heart Protection Study had over
20,000 patients randomised and followed up. However, they do not give
the clinical or public health impact of the results which should always also
be expressed on an absolute scale – the risk difference. It is important that
both absolute and relative risk information are reported clearly.
A relative risk (RR) of 2 can be a difference between a 10% rate and a
20% rate of occurrence of an AE. It can also be the difference between one
in a million and two in a million. The public health impact clearly varies
enormously. Benefit-risk balance should always be based on both absolute
and relative effects, as when comparing the benefits and risks of a new
treatment to no treatment or alternative treatment.
A way of describing absolute effects related to benefits is to consider
how many people are treated with a drug in order for a single “event” to be
prevented.44 The NNT (“number needed to treat” – the number of patients
who have to be treated with the drug in order that one of them gets a benefit
they would not otherwise have had) has become popular in articles in some
medical journals. NNT would appear to be an absolute number, but this
is not so. The time period for the treatment and follow-up of the outcome
must also be given. “NNT” is implicitly the NNT to obtain benefit. For
example, in the Heart Protection Study, the risks of a coronary death over
the 5 years of treatment were 5.7% on the statin versus 6.9% on placebo,
a difference of 1.2%. This difference suggests that 83 people need to be
treated for 5 years to prevent one coronary death (there are other benefits of
course). If the difference were constant over time then the NNT for 1 year
of treatment would be 5 x 83= 415 people needing to be treated for 1 year
to prevent 1 coronary death.
Some authors have used “NNH” as the “number needed to harm;”
however, it actually is the number of patients who have to be treated with
the drug, in order that one of them has an adverse effect (harm) they would
not otherwise have had. It is recommended to use “NNT/H”, to make it
clear that it is the number needed to be treated (and not, for example, the
number harmed). For example, the data from the WHI trial shown in Table
3 above gives a total of 166 women on HRT and 124 on placebo who con-
tracted breast cancer in the 43,909 and 41,165 women, respectively, fol-
lowed for about 5 years. This is a difference of about 8 per 10,000 over the
5 years. The NNT/H is expressed as 1310 women treated for 5 years who
44
Cook, R.J. and Sackett, D.L. The number needed to treat: a clinically useful measure of treatment effect, Brit.
Med. J.,310:452-4, 1995 and an erratum, ibid., 310:1056, 1995.

162

group6_PH.indd 162 7.8.2007 12:20:17

 will contract breast cancer they would not otherwise have had. There have
been statistical objections to these numbers that are largely related to the is-
sue of time.45 These absolute measures of effect are relevant in public health
terms, but may need modification in different patient groups and different
settings. They are very dependent on the actual absolute value of the rate of
the adverse effect in the control group.
Absolute measures of risks (and benefits) associated with treatments
should be presented along with the explicit time periods to which the
results apply.

j. Comments on ICH Guidelines E3 and E9:

Discussion of Statistical Aspects
of Clinical Safety Data
ICH Guideline E3 makes a number of suggestions for presentation
and analysis of data that pertain to safety of a newly studied medicine. It
sets out three levels: 1) extent of exposure, 2) common adverse events and
laboratory results that can be compared between treatment groups, and 3)
serious adverse events, both those in the conventional ICH/regulatory sense
but also other “significant” adverse events that require narrative statements
and listings that relate to individual patients.
The comments on analysis are good but limited in the ICH document.
Little mention is made of graphical summaries and most attention is paid
to providing listings so that all events can be easily examined by regula-
tors. There is mention of dividing events by whether the treating physician
thought that they were causally related to the treatment, perhaps on a grad-
ed scale (even though the actual treatment allocation might be blinded).
Individual case causal assessment is fraught with difficulty, and while it has
its place for serious events, the use of such assessments in overall analyses
when comparing rates of AEs between treatment groups is limited. It is
the randomisation that allows for causal inference when comparing groups
between randomised arms, and consideration of the investigator’s causality
assessment is not helpful for aggregate analyses.
The E3 guideline does mention life-table approaches to analysis, which
is to be recommended strongly, but it does not mention data presentation in

45
Hutton, J..L., Number needed to treat: properties and problems. J.Royal Statist. Soc.A., 163:403-19, 2000.

163

group6_PH.indd 163 7.8.2007 12:20:17

 a graphical form. It is important that account be taken of those who are no
longer in the study, especially towards the later time period in a trial.
One graphical summary mentioned is that of baseline versus final (or
intermediate) scatterplots for laboratory measurements, so that shifts in
values, even if they do not lead to clinically significant changes in many
individuals, can be detected (see ICH E3, section 12.4.2.2 III). It does not
mention that statistical tests, such as simple paired t-tests, can be employed
to study these changes; such tests are not often used in practice, but should
be. There are no clear guidelines on the problem of multiple significance
testing, but as discussed in this chapter they cannot be ignored. Using a P-
value of 0.05 means that with many tests 1 in 20 are expected to be “signifi-
cant” even if there are no true differences between groups being compared,
but this does not mean that these differences can be ignored; at the least
they require more detailed investigation.
There is emphasis given to “shift” tables and to “treatment emergent
signs and symptoms” (TESS). A shift table is essentially a tabular form of
a scatter plot with the baseline and follow-up values for a continuous mea-
surement grouped into categories. Such a table shows the numbers of in-
dividuals at different times, who have normal or abnormal test results. The
numbers who started with “normal” laboratory values that subsequently
became higher; the number with normal values at all times; the numbers
with high values at the start but normal values later, etc. The post-baseline
value included should be either the value at the end of the study, or, espe-
cially if there are drop-outs, the maximum value during the study. Tables
of this nature are useful but their analysis is usually limited. One test that
should definitely not be used as the only one is the chi-square test, since
this does not take into account the ordered nature of the categories. “TESS”
tables are similar but are done for categorical measures that are not based
on an underlying continuous measure or have been converted to a binary
variable. Analysis of newly emergent adverse effects can be done using the
methods for binary data described in the Chapter.
ICH Guideline E9 has limited coverage of “safety data”. It does em-
phasise the use of survival analysis methods to examine patterns of occur-
rence of events over time, but it appears that this technique is not commonly
applied, whereas it should be. It also mentions that pooling of data across
similar trials can be helpful in studying rare effects, but notes the problems
when this is done without the availability and inclusion of a comparator
group. Further details on the best methods for combining data across trials
are given in section g. in this Chapter.

164

group6_PH.indd 164 7.8.2007 12:20:17

 VII

Regulatory Reporting
and Other Communication
of Safety Information
from Clinical Trials

group6_PH.indd 165 7.8.2007 12:20:18

group6_PH.indd 166 7.8.2007 12:20:18
 a. Introduction
Traditionally, safety reporting from clinical trials has focused on in-
dividual case reports and has been viewed as a somewhat routine activity
mandated by various regulations. Some aspects of the various regulations
(e.g., expedited reporting of suspected adverse drug reactions that are both
serious and unexpected) are fairly well-defined. CIOMS I Working Group
was responsible for the successful introduction of standard criteria, format
and timing for the expedited reporting of suspected adverse reactions to
marketed drugs.1 The ICH Guideline E2A focused on extending the harmo-
nization of the regulatory requirements for expedited reporting of suspected
adverse reactions from clinical trials in the pre-approval environment.2 The
more recently adopted ICH Guideline E2D adapted E2A for expedited post-
marketing reporting.3 As a result, similar criteria are now more likely to be
applied through the life cycle of a medicinal product, from pre-approval to
post-approval. CIOMS III/V made recommendations for determination of
“expectedness” for clinical trial case reporting based on Development Core
Safety Information (DCSI).4 These recommendation have been taken up in
varying degrees by sponsors of clinical trials, as evidenced by the responses
to question 15 of the industry survey (See Appendix 3).
While requirements for expedited reporting to regulatory authorities
have been largely harmonized on a global basis, other aspects of the regu-
lations, such as reporting by sponsors to investigators and/or ethics com-
mittees, reporting by investigators to ethics committees and DSMBs, when
appropriate, and how these reports translate into information for study
subjects (via informed consent) are not as clearly defined and vary widely
across regions. Companies’ practices also vary widely, as can be seen from
the responses to survey questions 19-26 (Appendix 3). In addition, the use-
fulness of some of the information that is currently shared with investiga-
tors and ethics committees on a routine basis has been questioned.

1
International Reporting of Adverse Drug Reactions, Final Report of CIOMS Working Group. Council for
International Organizations of Medical Sciences, Geneva, 1990.
2
ICH E2A Guideline for Industry: Cinical Safety Data Management: Definitions and Standards for
Expedited Reporting, developed by the Expert Working Group (Efficacy) of the International Conference of
Harmonisation of Technical Requirments for Registration of Pharmaceuticals for Human Use (ICH), Step 5
as of October 1994 (http://www.ich.org).
3
ICH E2D. Post-Approval Safety Data Management: Note for Guidance on Definitions and Standards for
Expedited Reporting, Step 5 as of November 2003 (http://www.ich.org).
4
Guidelines for Preparing Core Clinical-Safety Information on Drugs. Second Edition, Including New Pro-
posals for Investigator’s Brochures. Report of CIOMS Working Groups III and V. Council for International
Organizations of Medical Sciences, Geneva, 1999.

167

group6_PH.indd 167 7.8.2007 12:20:18

 For purposes of this discussion, we use the term “reporting” to refer
to the submission of individual and multiple case reports, line listings or
tabulations in compliance with regulations, and the term “communication”
to refer to the broader concept of notification of safety information to appro-
priate stakeholders (investigators, regulators, IEC/IRBs, DSMBs, patients).
In this chapter the CIOMS VI Working Group considers the reporting
of cases and the communication of important new safety information by (1)
sponsors to regulators, investigators, ethics committees and data and safety
monitoring boards and (2) investigators to ethics committees. We consider
the following questions: What is (or should be) the intended purpose of
regulatory reporting requirements? What do the existing and/or proposed
regulations, directives and guidance documents say should be the practice?
Do the current regulations and practice adequately address the intended pur-
pose? What alternative approaches might better meet the information needs
of regulators, investigators and patients? We also consider whether one set
of rules is appropriate for all clinical trials and whether information needs
may change during the life cycle of an investigational or marketed product.
The following precepts form the basis for the CIOMS VI Working
Group recommendations.
❏ Ongoing safety monitoring of the experimental drug is an opera-
tional as well as intellectual task requiring scientific, medical, epi-
demiological and statistical expertise. It is a responsibility allocated
to the trial sponsor, overseen by regulatory authorities (fulfilling
their obligation to protect public health). In some circumstances, it
also warrants utilisation of an independent safety monitoring com-
mittee or other outside consultants.
❏ The ongoing evaluation of safety information involves judgement
and is based on clinical expertise that takes into account all avail-
able information on the drug. This expert assessment may result in
the identification of a new risk, which needs to be communicated to
relevant ethics committees, investigators, regulatory authorities and
patients. Ad hoc reporting of individual case safety reports is gener-
ally not considered an effective way of communicating important
new information to investigators and ethics committees.5

5
We note that the US FDA held a public hearing (21 March 2005) on the various problems associated with re-
porting of individual case and other clinical trial safety information to IRBs, as the basis for possible changes
to current regualtions and practices (see http://www.fda.gov/OHRMS/DOCKETS/98fr/oc04297.pdf).

168

group6_PH.indd 168 7.8.2007 12:20:18

 ❏ Undue harm can arise when the potential benefits are outweighed
by possible risks, based on a balanced assessment of all available
safety and efficacy data on the investigational product. Sponsors,
regulators, ethics committees and investigators have a joint respon-
sibility to put newly identified risks into such context.
It is important to point out that many of the recommendations included
in this Chapter are proposals only. Although the CIOMS VI Working
Group as well as its external panel of reviewers agree that these propos-
als represent a meaningful way forward, none of the recommendations
should be interpreted as superseding current regulations. Rather, the
recommendations are intended to inform discussions for future regu-
lations, as has been the case with prior CIOMS proposals. Until such
time as these proposals may be implemented, sponsors are expected to
maintain compliance with all existing regulations.

b. Expedited Reporting from Clinical Trials

(1) Expedited Reporting to Regulatory Authorities
Most of the regulations that describe safety reporting from clinical
trials focus on the expedited reporting of individual case safety re-
ports (ICSRs). ICH Guideline E2A which is generally considered the
standard for what information to send, stipulates that sponsors should
submit suspected adverse drug reactions that are both serious and un-
expected to regulators within 7 (if fatal or life-threatening) or 15 cal-
endar days in an appropriate format.6
Expedited single case reports from clinical trials are accepted by the
majority of regulatory authorities on the CIOMS I or similar form.
With the adoption of ICH Guideline E2B7 and then E2B(M)8 which
define standard data elements for electronic individual case safety
reports, some regulatory authorities have begun to require the elec-
tronic submission of expedited reports in the post-marketing environ-
ment. More recently, the EU and Japan have begun requiring electronic
submission of expedited reports from clinical trials as well.
6
ICH E2A Guideline for Industry: Cinical Safety Data Management: Definitions and Standards for Expedited
Reporting, Step 5 as of October 1994 (http://www.ich.org).
7
ICH E2B Guidance on Data Elements for Transmission of Individual Case Safety Reports, Step 5 as of
July 1997 (http://www.ich.org).
8
E2B (M) Clinical Safety Data Management: Data Elements for Transmission of Individual Case Safety Reports,
Step 5 as of November 2000 (http://www.ich.org).

169

group6_PH.indd 169 7.8.2007 12:20:19

 Time frames for expedited reporting, i.e., seven (7) and fifteen (15)
calendar days, are for the most part consistent across regions. Neverthe-
less, while authorities generally accept the same format and most have
incorporated the 7 and 15 day timeframes into regulation, there conti-
nues to be some divergence from the ICH recommended criteria for what
constitutes an expedited report from a clinical trial. For example, while
most authorities require the expedited reporting of suspected adverse
drug reactions that are both serious and unexpected, consistent with ICH
Guideline E2A, other authorities require expedited reporting of suspect-
ed adverse drug reactions that are serious, regardless of expectedness.
Some authorities will ask sponsors to report events of special interest in
an expedited fashion regardless of causality or expectedness.
The CIOMS VI Working Group endorses the ICH Guideline E2A and
thus recommends the harmonization of criteria for expedited report-
ing to regulatory authorities, to include suspected adverse drug reac-
tions that are both serious and unexpected. Only under exceptional
circumstances and on an ad hoc basis (e.g., when close scrutiny and
monitoring of a specific adverse reaction is warranted) should spon-
sors be expected to report, on an expedited basis, suspected adverse
drug reactions that are considered expected. If there is a need to report
events without regard to causality, this should generally be on a peri-
odic basis with the periodicity and format, e.g., line listing, agreed in
advance with the concerned authority.
(2) Expedited Reporting: Causality
The definition of a suspected adverse reaction, incorporating the con-
cept of relatedness, may be found in ICH E2A.9 This definition has
been adopted by most regions; however, its meaning has been inter-
preted inconsistently. The difficulty appears to lie in the use of both
the phrase “a reasonable possibility of a causal relationship” and the
phrase “a causal relationship cannot be ruled out”. While intended by
the authors of ICH E2A to be synonymous, they are subject to in-
terpretation, with the former phrase suggesting a threshold based
on clinical judgment but the latter implying something broader and
more inclusive, with less room for judgment. Most sponsors currently
follow the approach of using clinical judgment to determine if there
is a reasonable possibility of a causal association. If sponsors were to
report based on whether or not a causal relationship can be definitively
9
ICH E2A Guideline for Industry: Cinical Safety Data Management: Definitions and Standards for Expedited
Reporting, Step 5 as of October 1994 (http://www.ich.org).

170

group6_PH.indd 170 7.8.2007 12:20:19

 ruled out, then the number of expedited reports would likely increase
dramatically. The impact would be especially significant if some coun-
tries continue to require the expedited reporting of serious suspected
adverse drug reactions regardless of expectedness. Even with full har-
monization, excluding expected cases, the impact on managing the
significantly increased number of expedited reports would be great.
CIOMS VI Working Group does not believe that increasing the num-
ber of expedited reports, by lowering the threshold for considering
an adverse event a suspected adverse reaction, would contribute to
the protection of trial subjects or to the overall assessment of safety.
To the contrary, individual case reports are generally not an effec-
tive means of communicating important new safety information. The
CIOMS VI Working Group recommends that regulators adopt the
phrase “a reasonable possibility of a causal relationship” and con-
sider dropping the phrase “a causal relationship cannot be ruled
out” from the definition of suspected adverse drug reaction.
See Chapter 4, section c.2. and Appendix 1 (Glossary) for more dis-
cussion of this issue.

(3) Expedited Reporting: Expectedness

The CIOMS III/V report defines expectedness for clinical trials
based on “listedness” in the DCSI for investigational drugs.10 Since
the DCSI, as part of the Investigators Brochure, will apply to all re-
gions where clinical trials are being conducted, its use for determin-
ing expectedness facilitates harmonization of reporting. Once a drug
is approved, regulations may require the use of the local datasheet
(e.g., US Package Insert, EU Summary of Product Characteristics)
for determining expectedness in a particular country. The CIOMS
III/V and CIOMS V Working Groups also made the recommenda-
tion that local datasheets be used for determining expectedness once
a drug is marketed, “for reports from all sources, including clinical
trials”.11,12 However, there are likely to be circumstances where, for the
sake of uniform reporting from clinical trials to regulators, ethics com-
mittees and investigators, it would be preferable to report based on a
10
Guidelines for Preparing Core Clinical-Safety Information on Drugs. Second Edition, Including New Pro-
posals for Investigator’s Brochures. Report of CIOMS Working Groups III and V. Council for International
Organizations of Medical Sciences, Geneva, 1999.
11
Ibid.
12
Current Challenges in Pharmacovigilance: Pragmatic Approaches. Report of the CIOMS Working Group V.
Council for International Organizations of Medical Sciences, Geneva, 2001.

171

group6_PH.indd 171 7.8.2007 12:20:20

 single reference. For example, if there are large Phase IV international
trials being conducted in several regions, local datasheets may vary
due to varying stages of approval of labelling changes, thus resulting in
variability in reportability by country. In such circumstances, the
CIOMS VI Working Group recommends use of the CCSI rather than the
local label for post-marketing reporting from clinical trials, analagous
to the use of the CCSI for determining listedness in the PSUR. Hence
the CIOMS VI Working Group makes the following recommendation,
which diverges somewhat from the earlier CIOMS recommendation.
In order to maintain global consistency of clinical trial reporting, the
Working Group recommends that once a drug is marketed, the CCSI
effectively become the reference safety information for the purpose
of determining expectedness for regulatory reporting from Phase IV
clinical trials. For clinical trials of new indications, new populations
or new dosage forms for a marketed drug, every attempt should be
made to align the DCSI and the CCSI, but the DCSI should be used
if it is different from the CCSI.
Some sponsors determine reportability of a case at the event lev-
el (i.e., the case would be reportable if there is a suspected adverse
reaction that is both serious and unexpected) and some do so by the case
level (i.e., the case has at least one suspected adverse reaction that is seri-
ous and at least one suspected adverse reaction that is unexpected). The
latter situation results in erroneous reporting when the serious adverse
reaction is expected and the unexpected adverse reaction is not serious.
As with spontaneous reports, the CIOMS VI Working Group recom-
mends that the determination of reportability for case reports from clin-
ical trials be determined at the event level. That is, a case would meet
the criteria for expedited reporting only if there is a suspected adverse
reaction that is both serious and unexpected.
(4) Expedited Reporting: Unblinding
Blinded clinical trials bring specific requirements for unblinding ex-
pedited single case reports. This process has been defined in the ICH
guidelines and CIOMS recommendations13,14,15 and is reiterated here.
13
ICH E2A Guideline for Industry: Cinical Safety Data Management: Definitions and Standards for Expedited
Reporting, Step 5 as of October 1994. http://www.ich.org
14
ICH E6 Good Clinical Practice: Consolidated Guideline, Step 5 as of May 1996. http://www.ich.org
15
Guidelines for Preparing Core Clinical-Safety Information on Drugs. Second Edition, Including New Pro-
posals for Investigator’s Brochures. Report of CIOMS Working Groups III and V. Council for International
Organizations of Medical Sciences, Geneva, 1999.

172

group6_PH.indd 172 7.8.2007 12:20:20

 Suspected adverse drug reactions that are both serious and un-
expected, and thus subject to expedited reporting, should generally
be unblinded. However, there are likely to be special circumstances
where an exception to this rule would be appropriate, for example,
where the efficacy endpoint is also a serious adverse event (SAE). In
this case, the circumstance and the process to be followed should be
clearly defined in the protocol and the sponsor should seek agreement
from the relevant regulatory authorities. Such exceptions should be
clearly described in the protocol and Investigator Brochure.
Exceptions to unblinding are not always clear cut. Therefore it would be
important to establish, in advance, clear criteria for the diagnosis and
agreement from all concerned authorities for the exception. Even with
clear criteria, it may still be necessary to report a case while awaiting
further information. For example, if the endpoint of a study is myocardial
infarction, the diagnosis may not be confirmed at the time of the report.
In that case the blind should be maintained until the endpoint can be ruled
out. In circumstances where the endpoints are not clear cut, there should
be a mechanism established for making decisions regarding unblinding
and it should be described in the protocol. For example, a “committee”
of two or three physicians might be established to review each poten-
tially reportable case and decide on whether or not the exception applies.
Defining the criteria and establishing a procedure for making decisions
should go a long way toward maintaining consistency and conformity to
the exception.
Sponsors may elect to establish an independent DSMB with responsibil-
ity for the ongoing review and assessment of safety data from one or
more clinical trials. (See Appendix 8 on DSMBs.) One possibility is for
the sponsor to obtain agreement from relevant authorities that the use of
a DSMB might obviate the need to unblind and report individual cases.
Instead the DSMB would be responsible for notifying the sponsor of any
significant safety issues which would in turn be reported by the sponsor
in an expedited fashion to regulators and ethics committees. The exact
nature and content of the information would depend on the situation.
Sponsors should discuss with regulators the use of a DSMB in lieu
of expedited reporting.

(5) Expedited Reporting: Comparators

Once a case is unblinded the question of whether or not to report com-
parator or placebo cases arises, especially given that the expectedness

173

group6_PH.indd 173 7.8.2007 12:20:21

 decision upon which the unblinding was based generally relates to the
experimental drug. Individual cases which, when unblinded, are found
to involve patients on placebo will usually not be reported as expedited
reports. With regard to comparators, the CIOMS VI Working Group
felt that the sponsor of the clinical trial has the duty to send the report
to the company that is the marketing authorization holder (MAH) for
the medicinal product or directly to regulators when the other com-
pany is not known or when information on the nature of the trial is
considered proprietary. When the report is sent to the company that
is the marketing authorization holder (MAH) for the drug, the spon-
sor should inform the MAH of the regulatory reporting status. If the
sponsor has chosen to forward the report only to the MAH, the MAH
would be expected to report the case to regulatory authorities where
applicable. When the sponsor chooses to send the report only to the
concerned authority, the most current, up to date reference safety in-
formation (e.g., EU SPC) may not be readily available to the sponsor.
Even if it is, it may be not be in agreement for the suspected ADR in
question with the data sheets in all countries where the case may have
to be reported. In that case, reporting by the sponsor to the regulator
should be made regardless of expectedness. Independent of the method
chosen by the sponsor it should be determined in advance and applied
consistently throughout a particular clinical trial program.
The CIOMS VI Working Group recommends that unblinded placebo
cases should generally not be reported to regulatory authorities
on an expedited basis. On the other hand, it is recommended that
unblinded comparator cases be reported to regulatory authorities
and/or the company owning the comparator on an expedited basis,
regardless of expectedness. Likewise, serious suspected adverse re-
actions for open-label comparators should be sent on an expedited
basis to the appropriate regulatory authorities and/or company
regardless of expectedness.
It should be noted that this proposal may be in conflict with at least
one regulatory guidance, namely that of the EU Clinical Trial Direc-
tive. The Directive suggests that comparator reports be expedited to
both the regulatory authority and the MAH. The CIOMS VI Working
Group felt that there needed to be some flexibility in this regard, per-
haps closer to the FDA requirement that stipulates that either the sponsor
or the MAH needs to report the case to the regulatory authority. The
April 2004 guidance for the EU Directive suggests that expectedness

174

group6_PH.indd 174 7.8.2007 12:20:21

 for the comparator be based on an arbitrarily chosen datasheet to be
included in the protocol or the IB. The CIOMS VI Working Group
believes that reporting without regard to expectedness is more appro-
priate than determining expectedness in a somewhat arbitrary manner.
An exception is possible when the sponsor can identify a reasonable,
broadly applicable representation of the reference safety information
for the suspected ADR(s) under consideration.

(6) Expedited Reporting: Spontaneous Reports

Early in the clinical development of a new investigational product
when little is known about its safety, a heightened level of awareness
and scrutiny of serious adverse events is especially important. Hence
the recommendation in the ICH Guideline E2A that unexpected sus-
pected adverse drug reactions from clinical trials which are fatal or
life-threatening be reported within 7 calendar days. As the safety pro-
file becomes better understood, and once a drug is approved for mar-
keting anywhere in the world, it should not be necessary to apply
the same 7-day time frame for spontaneous reports from the post-
marketing environment.
The CIOMS VI Working Group proposes that, as a general rule,
7-day reporting be limited to reports from clinical trials and not in-
clude those from the spontaneous reporting environment. This should
generally apply to reporting in countries where the drug is not yet
approved as well as in countries where the drug is approved.
(Note that this may be in conflict with and does not supersede cur-
rent regulation.)

(7) Prompt Reporting Other than Case Reports

ICH Guideline E2A and some national and regional regulations define
other types of information that would warrant an expedited report by
the sponsor to the regulatory authority. Examples include: non-
clinical safety information having implications for the potential
for serious adverse reactions in human subjects (including but not
limited to findings of mutagenicity, carcinogenicity or teratogenicity);
an increased frequency (see the CIOMS V report for a discussion of
“increased frequency”) or severity of a previously recognized serious
adverse reaction; an incidence of a serious adverse event that is sig-
nificantly higher for the experimental drug than for a comparator; a
greater than expected incidence of a serious adverse event compared

175

group6_PH.indd 175 7.8.2007 12:20:22

 to the relevant background rate in the general population; a significant
drug interaction observed in a pharmacokinetic study; a protocol
procedure-related adverse event unrelated to treatment.
In addition, ICH Guideline E6 states “The sponsor should promptly
notify all concerned investigators/institutions and the regulatory au-
thorities of findings that could affect adversely the safety of subjects,
impact the conduct of the trial, or alter the IRB/IEC’s approval/
favourable opinion to continue the trial.”
In the various circumstances described above, where the information
that should be expedited is not an individual case report, the current
standard is to report within 15 days; however, it may not be clear when
the reporting clock starts. The CIOMS V working group suggested
such reports might be referred to as prompt notifications rather than
expedited reports.16 In the absence of a well-defined policy, one ap-
proach would be to have the clock start when the study co-ordinator
(non-clinical or clinical) becomes aware of the potentially important
safety finding. However, this is more easily said than done, since it
is not always clear when the awareness actually begins. Another ap-
proach might be for the sponsor to establish a decision-making com-
mittee, define the timeframe within which the committee would meet
once there is a possible finding that might constitute a 15-day report,
and have the clock start the day the committee decides the information
is reportable.
For circumstances other than individual case reports, the CIOMS VI
Working Group recommends that sponsors define their internal deci-
sion-making process in a standard operating procedure (SOP), including
how the clock start date will be determined for prompt notifications.
In addition to the usual criteria for an expedited report, adverse
events that are not deemed to be drug-related but are considered to
be protocol related should also be reported in an expedited fashion
if they are serious.
There is no established format for reporting adverse events considered
to be protocol-related. One reasonable approach would be to use the
CIOMS I report form and explain the situation in the narrative. An
example would be the occurrence of stroke following a significant rise

16
Current Challenges in Pharmacovigilance: Pragmatic Approaches. Report of the CIOMS Working Group V.
Council for International Organizations of Medical Sciences, Geneva, 2001.

176

group6_PH.indd 176 7.8.2007 12:20:22

 in blood pressure during the washout phase of a clinical trial for a new
treatment for hypertension.

(8) Expedited Reporting: Investigators and IECs/IRBs

While ICH Guideline E2A resulted in the relatively successful har-
monization of reporting to regulatory authorities, it did not specifi-
cally address the reporting of events to investigators and ethics com-
mittees. Rather, it refers the reader to ICH E6 (GCP Guideline) which
states, “the sponsor should expedite the reporting to all concerned
investigator(s)/institution(s), to the IRB(s)/IEC)s), where required, and
to the regulatory authority(ies) of all adverse drug reactions (ADRs)
that are both serious and unexpected.” Prior to ICH, some countries
with a specific regulatory requirement for sending expedited reports
to investigators required that individual case safety reports be sent to
all investigators conducting trials registered under the same clinical
trial authorization (e.g., IND in the US). Subsequent to the adoption of
ICH Guideline E6, other countries now require that the same reports
that are expedited to the regulatory authorities are also sent to each and
every investigator in that country that is conducting a clinical trial with
the investigational drug that is the subject of the report.
The CIOMS VI Working Group recommends replacing the current prac-
tice of sending large numbers of individual case reports to investigators
and ethics committees with a more reasonable approach to commu-
nicating important safety information to all who need to know. Such
an approach would involve periodic and ad hoc communications to
investigators and ethics committees that include an update of important
safety information as well as the evolving benefit-risk profile.
The EU Clinical Trial Directive does not include a requirement for
sending individual case safety reports to investigators. National au-
thorities within the EU have the option of instead requiring the spon-
sor to provide the investigator with periodic listings and a concise
summary of safety. The periodicity of the listing to investigators is
not specified. However, the Directive does introduce the possibility of
quarterly line listings to ethics committees in lieu of individual case
reports from other regions. Some countries in Europe, e.g., Germany
and Austria, have nevertheless incorporated into new national legis-
lation a continued requirement for submitting individual case safety
reports to investigators and ethics committeees within their countries
as well as to the national authoritiy.

177

group6_PH.indd 177 7.8.2007 12:20:22

 When considering expedited reporting to IRBs/IECs, the international
rules are even less well-defined and more inconsistent. Even ICH E6
(GCP Guideline) is noncommital in this regard. While specifying that
reports should be expedited to investigators, ICH E6 only specifies
that reports be expedited to IECs/IRBs “where required”. Some coun-
tries’ regulations, including those of the U.S. FDA, leave it up to the
responsible IEC/IRB to define what information it must receive from
the investigator. On the other hand, the sponsor is responsible for en-
suring that the investigator is following GCP, including compliance
with rules defined by the IEC/IRB . Hence most sponsors will instruct
the investigator to forward all expedited reports to the respective IRB/
IEC, regardless of the country in which the investigator resides.
The new EU Clinical Trial Directive clearly places the responsibility
with the sponsor for reporting to the IEC(s).17 However, the revised
final guidance issued April 2004 leaves open the possibility of send-
ing to the IEC only those expedited reports that originate in the IEC’s
own country, with a quarterly line listing on cases from other places.
In addition, any significant new safety information that would affect
adversely the safety of subjects or the conduct of the trial would be
reported to ethics committees within 15 days.18
With the growing number of trials that are multinational and the ex-
panding size of the typical development program, from a couple of
hundred subjects to thousands or sometimes tens of thousands of sub-
jects, the volume of reports that an investigator or IEC/IRB may have
to process and deal with can be staggering. As sponsors have adopted
the CIOMS III/V report recommendations to use a higher threshold
than previous practice for considering events expected,19 the volume
of reports to regulators and investigators has increased even further.
If regulators institute less strict criteria for considering events to be

17
Directive 2001/20/EC of the European Parliament and of the Council of 4 April 2001 on the approximation of
the laws, regulations and and administrative provisions of the Member States relating to the implementation
of good clinical practice in the conduct of clinical trials on medicinal products for human use, Article 17:
Notification of serious adverse reactions. http://www.emea.eu.int/
18
Detailed guidance on the collection, verification and presentation of adverse reaction reports arising from
clinical trials on medicinal products for human use, issued April 23, 2004 (http://www.emea.eu.int/)
19
CIOMS III/V introduced the concept of threshold for deciding when to add a new adverse reaction to the
DCSI, which includes a reasonable degree of suspicion of a causal relationship. Many sponsors were more
likely to include events in the IB sooner and thus consider them expected. While the higher threshold (do NOT
include events so soon) is considered an improvement from the standpoint of the value of the information in
the DCSI, it has resulted in many more expedited reports being sent to regulators, investigators and ethics
committees since they remain unlisted/unexpected. For more discussion, see Section e. below.

178

group6_PH.indd 178 7.8.2007 12:20:23

 “possibly drug related” (i.e., if the ICH E2A definition of adverse drug
reaction is taken literally and includes anything that cannot be defini-
tively ruled out as causally related; see Appendix 1 for further discus-
sion), the number of “suspected reactions” will increase yet again. As
the volume of reports increases so does concern about overwhelming
investigators/IECs/IRBs as well as concern over the integrity of the
trials by unblinding a significant number of study subjects.
While sponsors have become accustomed to reporting in an expedited
fashion to regulatory authorities based on a well-established set of cri-
teria, it is questionable whether it is useful to disseminate the same
information to the scores and sometimes hundreds of investigators and
in turn to IECs/IRBs. Sponsors and regulatory authorities generally
have computerized databases at their disposal for storing, cataloguing,
coding and analyzing the information. Investigators and IECs/IRBs
generally do not and are often overwhelmed with the amount of pa-
perwork that comes their way. Even if the resources were available for
each investigator to manage, maintain and analyze the data, the value
of such redundancy is questionable. Likewise, while certain IECs/IRBs
will continue to have the need to receive and review individual case
reports from their own sites, they are ill-equipped to manage and in-
terpret the many other case reports originating from other sites, often
from other parts of the world, and to place them into proper perspec-
tive. The responses to survey question 24 (Appendix 3) demonstrate
the increasing level of frustration among IECs and IRBs in dealing
with the information they currently receive.
Unfortunately, while based on a well-intentioned desire to improve the
protection of human subjects, the system has become a resource inten-
sive activity that does not necessarily result in effective communica-
tion of useful safety information to those who need to know and act.
The CIOMS VI Working Group believes that individual case reporting
should not be considered synonymous with communication of impor-
tant new safety information. When compliance is the goal, sponsors
tend to err on the side of conservative assessments of causality and
expectedness. In addition, it is well recognised that the investigator
assessment of causality is a crude and imprecise tool. As a result, indi-
vidual case reports do not always (and often do not) include important
new safety information. Conversely, important new information that is
best derived from an overall analysis of reports in aggregate may not
be effectively conveyed through sporadic case reporting.

179

group6_PH.indd 179 7.8.2007 12:20:23

 Although contrary to established regulations, the CIOMS VI Working
Group proposes that routine expedited case reporting by sponsors to
investigators and IECs/IRBs be eliminated. Instead, sponsors should
provide regular updates of the evolving benefit/risk profile and high-
light important new safety information. Significant new information,
occasionally a single case report, that has implications for the conduct
of the trial or warrants an immediate revision to the informed consent
would be communicated on an expedited basis. More commonly, im-
portant new safety information would be communicated periodically,
based on the assessment of accumulating information in aggregate, as
delineated in Chapter 5.
See Section c.(3) below for recommendations on update reports to in-
vestigators and IECs/IRBs.

c. Periodic Communication of Safety Information

from Clinical Trials
(1) Development Safety Update Report (DSUR)
Regulatory requirements for periodic reporting of safety from clinical
trials vary widely. Some authorities (e.g., Switzerland, EU and USA)
require a periodic report of safety in clinical trials during develop-
ment. Until the recently implemented EU Clinical Trial Directive, the
vast majority of countries had no such requirement. Those that do have
a requirement tend to define the format, content and timing of such
periodic reports differently. Post-marketing, countries that require
PSURs in the ICH E2C format would also receive an update of safety
in clinical trials as part of that report.
In the U.S., the FDA IND regulations define an “annual IND report”
which includes line listings of the most serious and the most frequent
adverse events as well as reasons for discontinuation.20 The new Euro-
pean Clinical Trial Directive, which became effective May 2004, for
the first time defines a periodic safety reporting requirement that ap-
plies to clinical trials both pre- and post-approval.21 In addition to an

20
U.S. Code of Federal Regulations 21CFR312.33: Investigational New Drug Application Annual Reports,
Revised as of April 1, 2004. http://www.accessdata.fda.gov/
21
Directive 2001/20/EC of the European Parliament and of the Council of 4 April 2001 on the approximation of
the laws, regulations and and administrative provisions of the Member States relating to the implementation
of good clinical practice in the conduct of clinical trials on medicinal products for human use, Article 17:
Notification of serious adverse reactions. http://www.emea.eu.int/

180

group6_PH.indd 180 7.8.2007 12:20:24

 annual report to regulators, there is the possibility of quarterly line
listings with a brief safety summary to inform investigators and ethics
committees.
The CIOMS VI Working Group recommends defining a single Develop-
ment Safety Update Report (DSUR) for submission to regulators on an
annual basis, with a consistent format and content which are yet to be
defined. In this regard, the CIOMS VI Working Group endorses the con-
cept published in the EU Clinical Trial Directive Guidance document
“Detailed guidance on the collection, verification and presentation of
adverse reaction reports arising from clinical trials on medicinal prod-
ucts for human use.” However, it is strongly recommended that the re-
ports be based on an entire development program and not per protocol.
Consideration should be given to establishing a common internation-
al birthdate which would be the date of first authorization to begin
clinical trials anywhere in the world.22 The CIOMS VI Working Group
recommends the use of MedDRA preferred terms for line listings.
The DCSI should be attached to the annual DSUR with an explana-
tion of any changes since the last update, with any significant new
safety information highlighted.
For products with a well-established safety profile and when most
clinical trials are Phase IV in the approved indication(s), it is
strongly recommended that the PSUR replace the annual DSUR.
A detailed proposal for the content, format and timing of DSURs was
felt to be beyond the scope of CIOMS VI. However, this topic has been
adopted by a new drug safety working group, CIOMS VII.

(2) Investigator Brochure and DCSI Updates

One common and very important method for informing investigators
of new safety findings is through periodic updates to the Investigators
Brochure (IB). ICH Guideline E2A says “In general, the sponsor of a
study should amend the Investigator’s Brochure as needed, and in ac-
cordance with any local requirements, so as to keep the description of
safety information updated.” Some national regulations also refer to
keeping investigators informed through periodic updates to the IB, but
the periodicity of such updates is generally not specified.

22
The concept of a single DSUR for submission to regulators with an international birthdate has also been pro-
posed by FDA, Docket No. 00N-1484, CDER 199665. Safety Reporting Requirements for Human Drug and
Biological Products. http://www.accessdata.fda.gov/

181

group6_PH.indd 181 7.8.2007 12:20:24

 The CIOMS III/V report introduced the concept of Development Core
Safety Information (DCSI) and made the recommendation that the
DCSI should be the part of the IB that defines the company’s position
on suspected adverse reactions.23 Some sponsors have established the
DCSI as an attachment to the IB. The advantage of making the DCSI
an attachment is that it makes it possible to update safety information
more frequently than the usual annual update to the IB by updating
only the attachment.
The CIOMS VI Working Group endorses the previous recommenda-
tions of CIOMS III/V. Sponsors should establish a policy of incorpo-
rating DCSI into every IB, either as a special section of the IB or as
an attachment to the IB. The DCSI should clearly identify the events
for which the company believes there is sufficient evidence to sus-
pect a drug-relationship. These would be the events that would be
considered expected (“listed”) from the standpoint of pre-approval
regulatory reporting criteria.
Consistent with the previous CIOMS III/V Working Group recommen-
dations, the CIOMS VI Working Group recommends that sponsors
review the IB and DCSI at least annually and update them as appro-
priate. If there are no changes to the IB or DCSI, then the investiga-
tors and ethics committees should be so informed at a convenient
time, such as with a periodic update.

(3) Other Periodic and Ad Hoc Communications

to Investigators and IECs/IRBs
As noted above, the CIOMS VI Working Group recommends that in-
dividual case safety reports not be reported to investigators or to IECs/
IRBs on a routine basis. Instead, it is recommended that there be pe-
riodic communications to investigators and IECs/IRBs, the timing of
which might depend on the stage of development.
For unapproved products, and in lieu of expedited reports, the CIOMS
VI Working Group recommends periodic reports to investigators and
IECs/IRBs that include a line listing of unblinded clinical trial cases
that were expedited to regulatory authorities since the last periodic
report, a copy of the current DCSI along with an explanation of any

23
Guidelines for Preapring Core Clinical-Safety Information on Drugs. Second Edition, Including New Pro-
posals for Investigator’s Brochures. Report of CIOMS Working Groups III and V. Council for International
Organizations of Medical Sciences, Geneva, 1999.

182

group6_PH.indd 182 7.8.2007 12:20:25

 changes, a statement if there are no changes, and a brief summary of
the emerging safety profile. Although it is recommended that the de-
fault would be quarterly updates, there may be circumstances when
a more immediate communication would be appropriate. Likewise,
there may be circumstances when less frequent updates should be
sufficient.
For approved products, the timeframe for periodic reports to inves-
tigators and IECs/IRBs would depend on the extent to which new
indications are being developed. For a product undergoing Phase
III trials, continuation of the quarterly reports would be advisable.
For well-established products, less frequent updates would be ap-
propriate and at some point, there should only be a need to update
investigators and IECs/IRBs when there is significant new informa-
tion to report.
When updates are provided by the sponsor to investigators or IECs/
IRBs, whether for unapproved or approved products, line listings
should include only unblinded expedited reports from clinical trials.
The line listings should include interval data, i.e., only cases expe-
dited since the last update; however, the summary of the emerging
safety profile should take into account all of the accumulating data.
The use of MedDRA preferred terms is recommended. The line list-
ings generally should not include spontaneous reports; instead, sig-
nificant issues arising from spontaneous reports can be described in
narrative form in the update.
For Phase IV investigators and their associated IECs/IRBs, com-
munication of changes to the CCSI should be sufficient and periodic
reports or line listings should no longer be necessary.
In addition to the periodic reports to investigators, there are circum-
stances when it would be appropriate to communicate important infor-
mation on a more immediate basis. This will of necessity be based on
clinical judgement, the seriousness of the event and the strength of the
evidence for causality. Although such safety alerts will most likely be
based on an assessment of several reports in aggregate, there may be
a single case report that warrants communication to investigators as
well as regulators on an expedited basis. For example, a single report
of severe hepatotoxicity for which a causal relationship is likely may
trigger an expedited communication to investigators if it is the first
such report early in development.

183

group6_PH.indd 183 7.8.2007 12:20:25

 If a significant safety issue is identified, either from an individual
case report or review of aggregate data, then the sponsor should
issue a prompt notification to all parties, namely regulatory authori-
ties, investigators and IECs/IRBs. A significant safety issue could
be defined as one that has a significant impact on the course of the
clinical trial or programme (including the potential for suspension
of the trial programme or amendments to protocols) or warrants
immediate update of informed consent.

(4) Safety Management Process

As described in more detail in Chapters 3 and 5, the CIOMS VI Work-
ing Group strongly recommends that sponsors define and implement a
system for regular reviews of safety information during development.
It is important to identify clear roles, responsibilities and accountabil-
ity for making sure that safety information is reviewed at pre-defined
intervals and that there is a mechanism for triggering an ad hoc review
whenever there is a special concern.
The CIOMS VI Working Group recommends that sponsors define a
Safety Management Team to review all available safety information
on a regular basis so that decisions can be made with cross-
functional input. It is further recommended that these reviews take
place quarterly pre-approval and be coordinated with the PSUR
schedule (six-monthly or annually) post-approval. In addition, ad
hoc Safety Management Team meetings may be warranted to ad-
dress urgent safety issues or significant safety signals.
Safety Management Team meetings should be used to review the over-
all evolving safety profile during development, to make changes to
the DCSI and/or informed consent and to determine if any changes in
the conduct of the trials need to be considered. The outcome of these
meetings may then provide the basis for the brief summary of safety
in the periodic reports to investigators and IECs/IRBs.

d. Other Reporting Considerations

The CIOMS VI Working Group felt that periodic summaries of safety,
with the occasional alert report when warranted, should be sufficient for
keeping investigators and IECs/IRBs informed of the emerging safety pro-
file of a new drug in development. However, it was also recognized that
existing regulations will make the continued submission of individual case
reports to investigators and IECs/IRBs a necessity in some regions.

184

group6_PH.indd 184 7.8.2007 12:20:25

 DSMBs are most commonly employed for a single large clinical trial
and are not usually charged with providing oversight of an entire clini-
cal program. It would therefore be important to ensure that important new
safety information is communicated to the DSMB even if the information
did not originate from the DSMB-monitored study.
Often the developer or manufacturer of a product is not the sponsor
of a particular clinical trial, but rather has agreed to support an external
investigator-sponsor financially or by providing drug supplies. In this situa-
tion the investigator as the sponsor is responsible for upholding good clini-
cal research practices and complying with all regulations that apply to the
sponsor in the region or regions where the investigator-sponsor is conduct-
ing the study. However, it would be important for the developer/manufac-
turer to ensure ready access to important safety information in order to meet
its own obligations for overall assessment, reporting and communication. (See
also Chapter 4, section b.)
The CIOMS VI Working Group recommends that a standard provision
of any agreement with an outside investigator-sponsor, whether for a
clinical or non-clinical study, should be the prompt reporting by the in-
vestigator-sponsor to the Company of all serious suspected adverse drug
reactions as well as any suspicion of a previously unrecognized hazard to
patients. Timely access to the final study report should also be included.

e. Informed Consent
In the conduct of clinical trials, the number one goal must always be to
ensure the safety of patients who consent to participate. Suspected adverse
drug reactions should be evaluated in a timely manner during the course
of clinical trials in order to assess what, if any, actions may be warranted
with respect to the continued conduct of the trial. This includes, but is not
limited to, assessing the need to communicate important new safety infor-
mation to participating patients.
The principal means of communication to the patient is via the inves-
tigator. The informed consent form (ICF), agreed and approved by the IEC/
IRB, should describe the risks and benefits of participating in the trial in a
way that the subject or the subject’s guardian can understand. The informa-
tion must be current and balanced.
International Guidelines for Biomedical Research Involving Human
Subjects, published by CIOMS in 2002, should be consulted for a

185

group6_PH.indd 185 7.8.2007 12:20:26

 comprehensive compilation of points to consider and rules to follow in de-
veloping and communicating informed consent for research subjects.24 The
Guidelines cover the process of obtaining as well as renewing informed
consent, the importance of ensuring understanding on the part of the sub-
ject, and situations where ethics committees might approve a waiver from
obtaining informed consent, e.g., in emergency situations, cultural consid-
erations and confidentiality.
A key area for consideration that has not been previously addressed in any
real depth relates to the determination of which adverse reactions should be
added to the consent form and when. In fact, a common complaint among in-
vestigators and ethics committees is their inability to make that determination
based on the individual expedited case reports that they currently receive.
In its introduction of the DCSI concept, CIOMS Working Group III/V
proposed that a relatively high threshold (compared to the CCSI, for exam-
ple) be used for adding new, serious AEs; once added, they become “listed”
and therefore any subsequent cases no longer require expedited reporting
to regulators, investigators and IECs/IRBs. The rationale for this position
(i.e., a high threshold for adding to DCSI) was as follows: careful attention
to new signals of serious ADRs demands ongoing monitoring and atten-
tion; by changing the DCSI on the basis of one case of a given event, for
example, there might be a tendency not to pay as much attention to that
event if new cases arise. CIOMS Working Group VI believes that the higher
threshold approach is no longer deemed appropriate for two reasons: (1)
by implementing a systematic process (Chapter 3) during development for
pharmacovigilance and risk management, there will be ongoing oversight
of all safety issues, and (2) there is heightened sensitivity to ensuring that
trial patients are fully informed of any new important information, even if
somewhat tenuous (see Chapter 2). Therefore, CIOMS Working Group VI
believes that the threshold for adding a new serious AE to the DCSI should
be the same as that for the CCSI.
In addition, it is recommended that the same criteria be applied to in-
formed consent information. Applying the CIOMS III/V threshold concept
to informing patients ensures consistency of information. A lower thresh-
old for inclusion in the consent form is unlikely to bring clarity of risk and
more likely to detract from what is already known.

24
International Ethical Guidelines for Biomedical Research Involving Human Subjects, Prepared by the
Council for International Organizations of medical Sciences (CIOMS) in collaboration with the World Health
Organization (WHO) , CIOMS, Geneva, 2002.

186

group6_PH.indd 186 7.8.2007 12:20:26

 The CIOMS VI Working Group recommends applying the same concept
and level of threshold to the DCSI and the informed consent form as
has been previously recommended for the CCSI. Thus, communication
to investigators and IECs/IRBs of an update to the DCSI may indicate
the need to update the informed consent form, the final decision for
which rests with the IEC/IRB.
The CIOMS VI Working Group believes that such a policy will be a
welcome aid to investigators and ethics committees who are currently in the
daunting position of deciding what to add to the informed consent informa-
tion based on individual case reports. It will also be advantageous to the
patient, who will be presented with the most important information and not
a long list of reported events of questionable relevance.
Informed consent should be renewed whenever there is new informa-
tion that could affect subjects’ willingness to participate, including new
information about risks. CIOMS International Ethical Guidelines suggest
that under certain selected circumstances, e.g., for long-term studies, in-
formed consent should be renewed at predefined intervals whether or not
there is new information. In most cases, when updating informed consent,
it should be sufficient to do so for continuing study participants at the next
scheduled visit. However, there may be circumstances where a more imme-
diate communication would be more appropriate. This would be the case,
for example, if a new risk has been identified that is life-threatening, even
if the benefit-risk relationship is still considered a favourable one. Commu-
nication between visits is also advisable for less alarming situations if there
is a long time interval between visits.

f. Other Communication Considerations

This chapter covers the reporting and communication of safety infor-
mation by sponsors of clinical trials to other stakeholders, namely regula-
tors, ethics committees, investigators, DSMBs and subjects. What has not
been covered, and yet is an area of increasing attention and scrutiny, is the
communication of important safety as well as efficacy results of clinical
trials to treating physicians and patients who may not be directly involved
with the conduct of the trials but for whom the information may be impor-
tant for making informed treatment decisions. This subject is discussed at
greater length in Chapter 2, section c, including the Working Group’s ratio-
nale for considering this topic out of scope for CIOMS VI.

187

group6_PH.indd 187 7.8.2007 12:20:27

 g. Conclusion
This chapter introduces several new proposals for the reporting of
safety information to regulators, investigators and IECs/IRBs. In addition,
suggestions are made regarding the content and timing of informed con-
sent. Although some of the proposals are not in accordance with current
regulations in most countries, the CIOMS VI Working Group believes these
recommendations have several advantages:
❏ Useful and informative safety information would be provided to
ethics committees and investigators without barraging them with
large numbers of case reports that they may not be equipped to
handle or effectively interpret in the context of the overall develop-
ment program.
❏ Sponsors would be encouraged to enhance their systems and proce-
dures for maintaining a proactive stance toward the monitoring of
safety during development.
❏ Reinforcing the use of a consistent approach for adding informa-
tion to the DCSI (and hence IB) will result in greater consistency
in determining expectedness for reporting to regulatory authorities
and for inclusion of ADRs in informed consent forms.
If these proposals are accepted and implemented through regulations,
the CIOMS VI Working Group believes that the result will be a much more
effective system for managing safety information from clinical trials, and
more importantly, for identifying and communicating important new safety
information to all who need to be informed and need to take appropriate
action in a timely manner.

188

group6_PH.indd 188 7.8.2007 12:20:27

 VIII

Summary of Concepts
and Proposals

group6_PH.indd 189 7.8.2007 12:20:27

group6_PH.indd 190 7.8.2007 12:20:27
 a. Introduction and Overview
• Patients and trial subjects will have different “acceptable” levels of
harm or risk, and in that sense risk and harm are relative concepts
to the individual.
• Although general responsibilities for managing drug safety issues
are usually covered in GCP regulations or guidances, the details
and increasing complexity of the field would benefit from the de-
velopment of more specific, internationally based Good Pharmaco-
vigilance Practices (GPP).
• Although there are some important differences between pre-
marketing and post-marketing safety monitoring and management,
there should be a much stronger and closer relationship between
them.
• The CIOMS VI Working Group has developed proposals based on
scientific principles for harmonizing many aspects of the collec-
tion, monitoring, analysis, evaluation/interpretation, and commu-
nication to all relevant parties of clinical trial safety information.
The general principles proposed apply to Phase I, II, III and IV
trials.
• Practical guidance for the design and execution of a rational drug
safety surveillance plan during any clinical research program
should be directed not only to pharmacovigilance/clinical safety
specialists, but to all those involved in the design, planning and
execution of the clinical research process for the development of
new medicines or diagnostic substances, as well as new uses and
preparations of already available products.
• This report is also directed at independent clinical researchers and
others not involved in commercially-based medicines develop-
ment, since the pursuit of enhanced safety standards is principally
concerned with the protection of patients.
• In any development program, the ultimate goal is to evaluate and
provide a measure of the benefit-risk relationship for the anticipat-
ed conditions of use. However, this report only indirectly addresses
the benefit side of the relationship and does not deal in a major way
with the evolving methodologies for qualitative and quantitative
aspects of benefit-risk weighing.

191

group6_PH.indd 191 7.8.2007 12:20:27

 b. Ethical Considerations for Clinical Trial
Safety Management
• For anyone designing and conducting a clinical trial, the fundamen-
tal principle should be that any study that is not scientifically sound
can be considered unethical.
• There is growing importance and sensitivity not only for patient
rights generally, but for clinical trials in non-industrialized, devel-
oping countries, vulnerable and socially underprivileged patients,
transparency (including on payments to investigators and to trial
subjects), and the availability of results of all trials, including those
with “negative” findings.
• Although gaining informed consent is the cornerstone of all human
subject clinical research, there are situations where it may not be
possible or appropriate, such as in the use of anonymized tissue
samples, in some types of epidemiological research, certain kinds
of survey research (to avoid biased results), and in emergency-
treatment study protocols (at least initially).
• The CIOMS Working Group VI endorses the concept of transpar-
ency of results and outcomes for all clinical research, especially
safety data; however, concrete proposals or recommendations on
this continuously evolving topic involve many complicated factors
which are beyond the scope of the Group at this time.

c. Systematic Approach to Managing Safety

During Drug Development
• Although the term “pharmacovigilance” has traditionally been as-
sociated with post-marketing activities, the CIOMS VI Working
Group recommends that the term be applied to the pre-marketing
process for collecting, managing and assessing safety information
during development. Likewise, the concepts of risk assessment and
risk minimization, together comprising risk management, are terms
that are as applicable to the pre-marketing environment as they are
to the post-marketing environment.
• It is important for sponsors to ensure that a well-defined and
well-structured process is in place that will allow them to readily

192

group6_PH.indd 192 7.8.2007 12:20:28

 identify, evaluate and minimize potential safety risks relative to
potential benefits for study subjects in pre-approval trials. Such
a process should start before initiating the first Phase I study and
continue through post-approval use of the drug or biologic in
the general population. It is important to consider and define, in
advance, the roles and responsibilities of individuals within the
organization who are expected to participate.
• A formal Development Risk Management Plan (DRMP) should be
created and modified as needed during a clinical program. In the
initial planning stages of a new clinical development program, one
goal is to gather the necessary knowledge and information to ad-
equately plan the optimum program from the standpoint of safety.
The plan should include early documentation of known, anticipated
and potential risks along with plans for addressing them during
development and, where appropriate, the DRMP would eventually
evolve into a post-marketing risk management plan that will ac-
company the registration application.
• Sponsors should establish standard operating procedures that define
a framework for a process that can be applied consistently across all
development programs, but which allows enough flexibility to meet
the needs of what will inevitably be a diversity of products and a
broad range of safety issues associated with them. In some cases it
may be appropriate to supplement standard operating procedures
with product-specific procedures.
• A dedicated Safety Management Team (SMT) should be formed for
each development program, to review all the available safety infor-
mation on a regular basis so that decisions on safety can be made in
a timely manner. It also recommends that these reviews generally
take place at least quarterly pre-approval and be coordinated with
pre-approval and, if applicable, post-approval periodic reporting.
Quarterly and ad hoc safety reviews should consider the overall
evolving safety profile of the investigational product, make neces-
sary changes to the IB/DCSI and informed consent, determine if
any changes to the conduct of the trials need to be considered, and
initiate prompt communications to investigators, ethics committees
and regulators when appropriate.
• Roles and responsibilities should be clearly defined for the Safety
Management Team as well as for each individual on the team. Each

193

group6_PH.indd 193 7.8.2007 12:20:28

 member of the team must have responsibility and accountability
for raising issues, in particular those emanating from their respec-
tive disciplines. The team should be empowered to make decisions
that will accomplish the goal of minimizing risk while maximizing
benefits to subjects in clinical trials, as well as anticipating the use
of the product once marketed.
• When licensing partners are involved, a joint safety management
process, including clear roles and responsibilities of the respec-
tive companies, should be defined in advance with timelines for
exchange and joint review of data. Ideally the terms should be part
of the initial contract, but at the very least should be incorporated
into a follow-on agreement on safety matters.
• Key to the successful implementation of a consistent and system-
atic approach is the establishment of a mechanism for scheduling
meetings, tracking issues and timelines, and assuring comple-
tion of action items. The CIOMS VI Working Group recommends
establishing a project management function to manage these tasks,
document any decisions, and ensure compliance with internal
procedures.
• All pertinent data must be readily available to the safety team from
the clinical trial and safety databases as well as from other relevant
sources, such as the pre-clinical toxicology department (e.g., carcino-
genicity and development and reproductive toxicology), in vitro
mutagenicity studies, and pharmacokinetic and drug-interaction
studies.
• It is important to incorporate epidemiology into the development
planning process, not only for defining the natural history of the
disease being treated, but for anticipating important confounding
factors and background rates of occurrence of concurrent illnesses.
Understanding these will help to put the evolving safety profile into
proper perspective.
• When planning for the development of virtually any new medici-
nal product, there are certain categories of potential toxicities that
should always be considered. These include abnormalities in car-
diac conduction, hepatotoxicity, drug-drug interactions, immuno-
genicity, bone marrow toxicity and reactive metabolite formation.

194

group6_PH.indd 194 7.8.2007 12:20:28

 d. Collection and Management of Safety Data
During Clinical Trials
• If an investigator becomes aware of information that is considered
to be important for safety reasons it should be reported to the spon-
sor (immediately if judged critical), even if the protocol does not
specifically state that the information must be collected. To assure
the investigator’s sensitivity to this point, one of the key responsi-
bilities of the sponsor includes proper training of the investigative
site regarding data collection and reporting.
• The collection of “excessive” data can have a negative impact on data
quality. Therefore, case report form fields should be chosen based on
the data elements that will be analyzed and can be typically present-
ed in tabular compilations of study results. Safety data that cannot be
categorized and succinctly collected in predefined data fields should
be recorded in the comment section of the case report form when
deemed important in the clinical judgment of the investigator.
• Safety monitoring during Phase 4 studies, which can make an im-
portant contribution in expanding the clinical trial database, may
not require the same intensity as for Phase I-III trials, but the same
principles and practices remain applicable.
• Although personnel other than the investigator may obtain adverse
event information during regular communication, even between vis-
its, it is ultimately the responsibility of the investigator to ensure
that information is collected in accordance with the study protocol.
• If a company provides any support for an independent trial it does
not sponsor (e.g., supplies, research grant, etc.), the company should
still obtain at a minimum all reports of serious suspected adverse
reactions from the investigational site(s). Once the relevant reports
are received by the company, it should conduct its own causality as-
sessment and decide whether they should be sent to the appropriate
regulatory authority (ies), even if it is known that the investigator
has already done so on his/her own.
• In early phases of drug development, it is generally necessary to
collect more comprehensive safety data than in post-marketing
studies. In addition, certain drug types may require longer routine
follow-up as in the case of vaccines, immunotherapies and some
biotechnology products.

195

group6_PH.indd 195 7.8.2007 12:20:29

 • The collection, monitoring and assessment of data from Phase 1
studies deserve special attention for two reasons: (a) with some
exception (e.g., oncology medicines, pharmacokinetic studies in
subpopulations such as the organ impaired), such studies are con-
ducted in healthy volunteers for whom there is no anticipated health
benefit and (b) the results are critical to the future development of
the product and must be scrutinized and interpreted with great care.
For prophylactic treatments and preventative vaccines, the same
considerations apply even to later stage clinical trials.
• There are no definitive methods for distinguishing most adverse
drug reactions (events that are causally attributable to study therapy)
from clinical adverse events that occur as background findings in the
population and have only a temporal association with study therapy.
The CIOMS VI Working Group thus recommends the following:
❏ All adverse events, both serious and non-serious, should be col-
lected for any clinical trial during development, regardless of
presumed relationship to the study agent by the investigator or
sponsor, in order to allow for subsequent assessment of causality
using standardized methods for individual cases and aggregate
data. This applies not only to the experimental product but to
placebo, no treatment, or active comparator.
❏ In studies initiated during the immediate post-approval period
it is prudent to continue this practice. Once the safety profile of
a marketed product is judged to be well understood and estab-
lished, it may be acceptable to collect less data. While detailed
information on serious adverse events should always be collect-
ed, it may be appropriate for well established products to collect
only those non-serious adverse events suspected by the investi-
gator to be related to the compound. This would be especially
appropriate for large scale, simple post-marketing trials when
the population, indication, and doses are consistent with those
included in the approved use(s) of the drug.
• A commonly overlooked but potentially important aspect of data
collection relates to the possible use of herbal and other non-
traditional remedies by patients/subjects, who typically do not
regard such treatments as drugs or medicines. It is therefore
important to inquire specifically about them, since their concomi-
tant use with study treatment can lead to adverse drug interactions.

196

group6_PH.indd 196 7.8.2007 12:20:29

 Recently developed classification and coding schemes for herbal
medicines are available.
• Causality judgments based on analysis of multiple cases/aggregate
data, rather than on individual cases, are almost always more mean-
ingful and typically have a greater impact on the conduct of clini-
cal trials, including changes to informed consent documents, study
design, and core safety information. However, causality assessment
of individual adverse events by the investigator may play a role in
the early detection of significant safety problems, and are the only
source of information on rare events.
• The CIOMS VI Working Group recommends that the investigator
be asked to use a simple binary decision for drug causality (related
or not related) for serious adverse events. One possible approach
that has been suggested is to ask simply whether there is a “rea-
sonable possibility” or “no reasonable possibility” that the study
treatment caused the event. Alternatively – Was there a reasonable
possibility? Yes or No.
• It is virtually impossible to completely rule-out the role of a drug
in causing an adverse event in single-case reporting. Therefore, the
use of “unknown” or “cannot-be-ruled-out” adds little value in early
determination of safety concerns. The use of “cannot-be-ruled-out”
to imply drug relatedness would lead to excessive over-reporting
and excess noise in the system.
• The Working Group advocates adoption of the recommendation by
the CIOMS III/V report on core safety information and the DCSI
(Development Core Safety Information), namely that on the CRF
and on any serious adverse event form there be included a standard
list of potential causes from which the investigator must choose the
most plausible one in his/her opinion, specifically: medical history;
lack of efficacy/worsening of treated condition; study treatment;
other treatment, concomitant or previous; withdrawal of study
treatment (a withdrawal reaction could be considered drug-related);
erroneous administration of treatment; protocol-related procedure;
other – specify.
• It is recommended that investigators not be asked routinely to indi-
cate causality information for non-serious adverse events. However,
there may be circumstances when such assessments are useful and
important, such as for non-serious adverse events of special interest.

197

group6_PH.indd 197 7.8.2007 12:20:30

 • Investigators should be encouraged to provide a diagnosis (when
possible and appropriate) on the CRF rather than each individual
sign and symptom. This instruction should be clearly specified in
the protocol. However, when an investigator submits a serious ad-
verse event report that includes a diagnosis it is important that the
signs and symptoms as well as any other supporting information
that led to the diagnosis also be recorded, specifically as part of the
narrative description of the case.
• Prior to study initiation, it is recommended that specific criteria for
identifying and defining significant, anticipated adverse events be
established and communicated to investigators involved in the de-
tection, assessment and reporting of adverse events.
• Although it is ordinarily unnecessary to create specific definitions
or criteria for non-serious adverse events, it is important to do so for
apparently non-serious events that might be precursors (prodromes)
of more serious medical conditions; for example, muscle pain and
elevated CPK together may be indicative of potential rhabdomy-
olysis. Such prodromes are an example of what are often referred
to as adverse events of special interest, when there is evidence or
suspicion of their potential importance.
• It is important to define clearly “adverse events of special interest”
in the protocol and to specify close monitoring and prompt report-
ing to the sponsor of these types of events, even if they are consid-
ered non-serious according to the usual regulatory criteria.
• It is recommended that even when anticipated medically serious
clinical events are collected as clinical efficacy outcomes/endpoints,
rather than as adverse events, these data must be recorded by the
investigator and periodically reported to and reviewed by the spon-
sor or DSMB, on a schedule specified in the protocol.
• The process used to solicit information from patients during clini-
cal trials (i.e., how they are asked questions about their experiences)
should be consistent from site to site, and if possible from program
to program, and should be clearly outlined within study protocols,
in the informed consent information, and during investigator train-
ing. No matter what method or approach is used, it should be used
consistently throughout the trial, including at baseline (pre-
treatment information).

198

group6_PH.indd 198 7.8.2007 12:20:30

 • It is probably best to frame questions to the patients in general
terms rather than to invoke the possibility that study treatment may
be responsible for ill effects. For example: “How have you felt since
I saw you last? Anything new that you wish to discuss?” Although
it is not advisable to read a laundry list of possible ADRs when
soliciting the patient’s recent experience, patients should be alerted
to known signs and symptoms indicative of medically important
suspected or established ADRs in order to alert the investigator as
early as possible.
• Typically, the time that the informed consent is signed by the patient
is designated as the start of safety data collection. This provides a
clear starting point and helps to avoid any selection bias. Whenever
possible, a patient should be followed through the last scheduled
visit even if the patient is withdrawn from treatment, in order to al-
low for appropriate intention-to-treat analysis.
• As a general rule, it is recommended that safety data event-
collection should continue after the last dose of the drug for at least
an additional five half-lives of the experimental product. In addi-
tion, investigators should be instructed to always be diligent in
looking for possible latent safety effects that may not appear until
after a medication is discontinued.
• In order to assure standardized signal detection and evaluation pro-
cesses, data quality and completeness are paramount. The CIOMS
VI Working Group recommends the following principles for this
important objective:
❏ individual case safety reports from studies should be as fully
documented as possible
❏ there should be diligent follow-up of each case, as needed
❏ the reporter’s verbatim AE terms must be retained within all rel-
evant databases
❏ if the reporter’s AE terms are not considered to be clinically ac-
curate or consistent with standard medical terminology used for
coding, attempts should be made to clarify the description of the
event with the investigator. If there continues to be disagreement,
the sponsor can code the AE terms according to its judgment on
the case, but should identify them as distinct from the investiga-
tor’s terms. Reasons for the difference(s) should be documented.

199

group6_PH.indd 199 7.8.2007 12:20:30

 ❏ personnel with knowledge and understanding of both clinical
medicine and the dictionary used should review all codified
terms to ensure consistent and accurate codification of reported
(“verbatim”) terms.
❏ primary analyses of AE data should be based on the investiga-
tor’s assigned terms or diagnoses; additional analyses using the
sponsor’s assignments can be conducted, but explanations for
any differences between the two analyses must be given.
• Individual case safety reports (ICSRs) must be categorized and assessed
by the sponsor using trained individuals with broad expertise in both
clinical medicine and codification. Investigators should be encouraged
to obtain specialist consultation for clinically important events that oc-
cur outside their own areas of clinical expertise, so that sponsors can
obtain all information required for subsequent safety evaluation.
• Depending on their purpose, adverse event tables can display both
the reported (investigator’s verbatim) term and the sponsor’s terms.
However, primary safety analyses (especially those used to develop
the DCSI and CCSI) should be based on investigator-assigned terms
which are consistently defined. Clinically discrepant terms should
be appropriately identified to ensure transparency of the process
used to derive the final data.
• Some companies and health authorities maintain a list of event
terms that are always regarded as medically serious and important
even if the specific case might not satisfy the criteria for serious in
a regulatory sense (require expedited reporting, for example). Such
“always serious” events are used routinely to trigger special atten-
tion and evaluation. Although such lists were originally created for
post-marketing purposes, especially for spontaneous reports, they
might be useful for pre-approval clinical research purposes. The
CIOMS VI Working Group does not endorse any particular list
since it may be highly dependent on the treatment and the specific
population(s) under study, and can never be complete.
• Sponsors should avoid “excessive coding” of events reported in seri-
ous adverse event cases. Each such report should contain only the
minimum number of dictionary terms needed to ensure retrieval in
the relevant clinical context(s). Conversely, sponsors should take
great care not to “undercode” events, namely, assign codes that might
downgrade the severity or importance of an event term or terms.

200

group6_PH.indd 200 7.8.2007 12:20:31

 e. Identification and Evaluation of Risk
from Clinical Trial Data
(1) Ongoing Safety Evaluation
• The purpose of ongoing safety evaluation during drug development
is to ensure that important safety signals are detected early and to
gain a better understanding of the benefit-risk profile of the drug.
• Clinical trial sponsors should develop a system to assess, evaluate
and act on safety information during drug development on a con-
tinuous basis in order to ensure the earliest possible identification
of safety concerns and allow appropriate risk minimization, such as
modification of ongoing study protocols, to ensure that clinical trial
participants are not exposed to undue risk.
• Safety monitoring, evaluation and analysis should be performed in
such a manner as not to compromise the integrity of the individual
studies or the overall development program. Study sponsor should
be fully aware at every stage of development of the potential risks
of the investigational product and the morbidities characteristic of
the study population.

(2) Safety Data Management

• Consistent standards and criteria for the diagnosis and recording of
adverse events and other safety data must be established.
• Sponsors must ensure that activities involved in the management of
clinical trial safety data (e.g., data entry, edit checks, data queries,
coding of adverse events using a standard dictionary, etc.), are un-
dertaken with care and precision in order to ensure that the safety
database is accurate and complete.
• Attempts should be made to individualize safety evaluation criteria
for an investigational drug product based on an assessment of ac-
ceptable risk. Since there are no standard approaches to evaluating
or measuring an “acceptable level of risk”, the issue must be ad-
dressed throughout clinical development. Where possible, the risk
associated with a given medicinal product can be compared with
the established benefit-risk profile of an existing product used for a
similar population and indication, or it can be compared to the risk
of the disease itself, if no therapeutic options are available.

201

group6_PH.indd 201 7.8.2007 12:20:31

 (3) Review of Safety Information
• Review of safety data should involve an analysis of both individual
reports as well as aggregate data in order to allow for both a qualita-
tive and quantitative understanding of the safety profile of the drug.
• The evaluation of serious adverse events requires a detailed under-
standing of the individual case. However, for commonly occurring
events, the analysis of aggregate data is appropriate in order to ex-
plore a possible relationship with the drug.
• All serious adverse event reports must be reviewed within specified
time frames, whereas aggregate data should be reviewed on a peri-
odic basis.
• Prompt medical evaluation of all individual serious cases and adverse
events of special interest (irrespective of causality), and the periodic
aggregate assessment of all available clinical safety data, are critical
for improving the process of signal detection in drug development.
• The evaluation of individual cases should be done in the context of
the patient population, the indication for the investigational drug,
the natural history of the disease, currently available therapies and
other benefit-risk considerations.
• The determination of causality of adverse events should be based
on a combination of clinical judgement and aggregate data analy-
sis based on all reported cases. Investigator causality assessment
should be taken into account and may be particularly important
when evaluating rare or unusual events for which aggregate ana-
lytical methods are not applicable.
• Adverse events of special interest should ideally be identified in the
developmental safety plan and protocols for handling by investiga-
tors and sponsors as if they were serious, even though they do not
necessarily meet the regulatory definition of serious.
• A review of the non-serious adverse events reported in clinical tri-
als should be conducted to look for adverse events of special inter-
est. This could assist in the tracking of those events that may be
predicted but could also capture unexpected potential signals.
• Non-serious AEs should be critically appraised at regular intervals,
in particular those associated with discontinuation of study treat-
ment. In addition, although non-serious AEs may generally not be

202

group6_PH.indd 202 7.8.2007 12:20:32

 reviewed individually, they must be addressed carefully in study
reports and integrated safety summaries.

(4) Frequency of Review of Safety Information

• The frequent review of serious and special interest adverse events,
as well as overall assessment of all AEs, regardless of seriousness,
causality, or expectedness, should be performed periodically: (1)
ad hoc, for serious and special interest AEs, (2) routine, periodic,
general review of all data, whose frequency will vary from trial
to trial and from development program to development program
and depend on many factors, and (3) reviews triggered by specific
milestones established for a trial or a program (e.g., numbers of
completed patients, end-of-trial, end-of-program, preparation of in-
tegrated summary of safety, and a marketing application).
• Appropriate analyses should also be conducted periodically for
safety-related information other than AEs, including physical ex-
amination findings, vital signs, clinical laboratory tests, cardiac
electrophysiology, and other evaluations.
• Aggregate safety data should be monitored and evaluated periodi-
cally during the course of the overall developmental program, dur-
ing each study, and at the end of every study to provide an ongoing
appraisal of benefit-risk balance.
• Each time a study is completed and unblinded, all safety informa-
tion, not limited to clinical AEs but ideally including emerging effi-
cacy endpoints, vital signs, and clinical investigation results, should
be assessed and evaluated relative to previous knowledge; product
information should be updated as needed (investigator brochure,
development core safety information, informed consent, company
core safety information, local datasheets).

(5) Analysis and Evaluation

• Although the relatively small number of subjects exposed to an inves-
tigational product may limit the utility of subgroup analyses, where
possible data should be stratified for dose, duration, gender, age, and
possibly concomitant medications and concurrent diseases.
• When pooling data, it is most appropriate to combine data from
studies that are of similar design (e.g., similar in dose, duration,
methods of determining adverse events, and population).

203

group6_PH.indd 203 7.8.2007 12:20:32

 • If the duration of treatment with the investigational agent varies
widely among participants in a clinical trial, data on the effect of
treatment duration on adverse events may be available. Such an
analysis can be important for the detection of adverse reactions that
occur only with prolonged treatment compared to adverse reactions
that tend to occur early in the course of treatment.
• Groups of studies that are useful in pooled safety analyses include
all controlled studies or subsets of controlled studies, placebo-
controlled studies, studies with any positive control, studies with a
particular positive control, and studies of particular indications.

f. Statistical Approaches for Treating

Clinical Safety Data
• The techniques and approaches to use of statistics for analysing
safety data have not been developed as fully as for efficacy and it
is not uncommon to find inappropriate or incomplete displays and
analysis of adverse event data, even in refereed publications.
• Statistical approaches have application at several stages of clinical
trials: protocol design, during a trial, for final analysis and writing
of the trial report or publication, and when combining data across
different trials. Professional statistical help is required and should
be obtained at each of those stages.
• Statistical association (P-values or other measures) alone may or
may not be of clinical value. In randomised trials they have great
strength in testing causality but they inevitably have uncertainty.
Examination of both statistical and clinical significance must in-
volve a partnership.
• It is necessary to acknowledge when the data are insufficient to
draw conclusions on safety, i.e., ‘absence of evidence is not evi-
dence of absence.’ In such situations, the use of descriptive methods
and well-designed graphics will be helpful in this process.
• The ability of a study to detect causal effects in the face of variation
within and between individuals is dependent on sample size; the
smaller or rarer an effect, the larger the sample size required, if any
degree of certainty is to be given to the study conclusions.

204

group6_PH.indd 204 7.8.2007 12:20:32

 • A statistical test that allows for either an increase or a decrease in
the rate of an adverse effect should virtually always be specified in
the protocol. Hence, it is recommended that all statistical testing on
safety-related data be done on the basis of two-sided (two-tailed)
hypothesis tests.
• Although most Phase II and III trials are randomised and usually
double-blind, generally the same statistical principles discussed
here can apply to non-randomised and non-comparative studies.
• Independent of a trial design or Phase, there are many kinds of
comparisons that can and should be made within and between treat-
ment groups using the proper statistical tools, depending of course
on the kind and amount of data collected. Typical analyses involve
such things as: comparisons between treatment groups of specific
AEs, classes of AEs (different organ systems, e.g.), and laboratory
data; discontinuations from treatment; sub-population results (age,
sex, etc.); time-dependent phenomena (time to onset of AE, time to
discontinuation, etc.); combining data across trials.
• It is recommended that at least one intention-to-treat (ITT) safety
analysis should be conducted. As a consequence, the collection of
data should continue whenever possible to obtain study endpoints
even in those who are prematurely withdrawn from treatment. ITT
analyses are considered to be the most conservative approach.
• Some of the more important problems associated with safety analy-
ses that require attention are:
❏ power: The power of a particular study will depend on the size of
the groups being studied; the baseline or background rate of the
adverse effect of interest, which is the rate expected in the compar-
ison group; and the change of interest in rates between groups (for
example a doubling or tripling). It also depends on the “P-value”
set as being statistically significant, which is usually 0.05. If al-
lowance is made for multiple testing (see below), then this P-value
may be much smaller and so the effects will be more difficult to
detect as statistically significant and therefore the statistical power
will be lower. Most trials, or even combinations of trials, are not
large enough to detect or analyse rare adverse events reliably.
❏ multiplicity: multiple analyses can be and often are performed
on the same data set, such as on multiple time points and multiple

205

group6_PH.indd 205 7.8.2007 12:20:33

 variables. Multiplicity affects the statistical analysis, especially
the calculation of P-values, because many different comparisons
of adverse effects are possible. Clinical trials are designed to
minimise Type I errors (concluding that efficacy exists when it
really does not), and the testing of multiple hypotheses within a
single study is discouraged. However, the numbers of potential
types of adverse events are very large, so that correction for mul-
tiple testing in a conventional way will mean that it is impossible
to draw any conclusions. It is for this reason that corrections for
multiple testing are rarely done using a formal mechanism.
❏ medical classification: if the grouping of adverse effects into cat-
egories is too narrow, it results in numbers of event types that are
too small for meaningful statistical comparison between groups,
but if too wide (having larger numbers in the groups to avoid the
first problem), it could hide the existence of a safety problem. An-
other difficulty arises in deciding whether groupings of different
event terms for a patient can be formally regarded as a syndrome,
for which a specific diagnosis might be possible. This requires
medical judgement, and the results of the analysis will need care-
ful interpretation rather than reliance on the result of a statistical
test. The use of different coding dictionaries and different levels
within those dictionaries can lead to statistical problems.
❏ time dependency: adverse effects should be examined carefully
as a function of time on drug; simple calculations of incidences
(number of events such as AEs or discontinuations divided by
number of patients treated) can be highly misleading and mask
the true risks associated with the treatments.
• Currently used approaches to analyses of safety data are sometimes
over-simplified, and do not take the major characteristics of adverse
reactions into account. For example, some reactions have a rapid
onset after administration of a drug, and if they do not occur early,
are much less likely to occur later.
• It is best to analyse laboratory data using baseline values as a com-
parison whenever possible. The most effective approach is usually
to use the baseline value (or the mean where multiple measure-
ments are made) as a covariate. The post-treatment value (or their
mean where multiple measurements are made) is then the response
that is analysed.

206

group6_PH.indd 206 7.8.2007 12:20:33

 • Analysis of laboratory measurements used for monitoring of ad-
verse effects is done on binary measures of clinically relevant val-
ues or changes, but should also be done comparing mean values
using analysis of covariance (ANCOVA), since this is likely to have
greater sensitivity for detecting real adverse effects.
• Graphical displays, such as scatter plots of baseline versus later
values for each trial participant, can help show both a shift from
average and also draw attention to outlying values, both in terms of
absolute levels but also large changes.
• Finding a significant difference between groups does not necessar-
ily prove causality based on a laboratory test result, but the most
powerful statistical analysis should be used so that early signs of
organ damage are detected. Trends in average values can be a sur-
rogate for rare, clinically important individual changes.
• Results of trials should show confidence intervals for a relevant sum-
mary of the data rather than just quoting the P-value from a signifi-
cance test. A confidence interval is a measure of the amount of statisti-
cal uncertainty around a summary value known as the point estimate.
This estimate will not necessarily be the true value, which may only
be known if we have infinite knowledge about the parameter. What is
needed is an awareness of whether our estimate is likely to be close
to the true value or not. We construct confidence intervals for sum-
maries of data such as proportions or differences in proportions. CIs
are particularly useful for dealing with adverse event data.
• It is very important that the length of time that each patient is “at
risk” of having an adverse event be factored into any assessment of
risk. The length of time over which a patient is followed in order to
determine whether an AE occurred will not always be the same for
every participant in a trial. This period may or may not be the same
as the duration of treatment, which in some settings, such as single
dose studies, may well be the same for all patients.
• Events that occur beyond the standard observation period can be
difficult to include in a formal analysis, since unless all patients
are followed for the same length of time post-treatment, it will not
be known whether others also experienced the same or different
events. Such post-treatment events should be documented and dis-
cussed in a trial report, but it is not usually appropriate to include
them in the formal statistical analysis since bias could result.

207

group6_PH.indd 207 7.8.2007 12:20:34

 • Calculating the sum of the total time at risk for all patients by treat-
ment group is useful, and this should be reported, often as person-
time (e.g., person-years). The incidence rate is the total number of
those having the event divided by the person-years at risk, and the
ratio of incidence rates between treatment and control groups is a
rate ratio. The risk per unit time is called the hazard rate and using
total person-years as the denominator assumes that this rate is con-
stant over time. These calculations assume that the incidence rate is
constant over time but this will often not be true.
• Rates per person-time for each treatment group should be reported
in addition to numbers of patients with an event divided by the total
number of patients in the relevant at-risk group. This is especially
important when combining data from studies involving different
treatment durations.
• The total person-time in the treated group across all the trials, or
even the total number of patients treated, is often the denominator
used in determining the rate of occurrence of adverse events. This
is rarely the best way of presenting or summarising the data, and
must be treated with great caution. The correct method is to use
“life-table” or survival analysis, even though here it is not the time
to death but an adverse event that is studied. Survival analysis gives
a better, less biased estimate than the crude analysis.
• The method is similar to that used for survival curves using a
Kaplan-Meier estimate of survival. Kaplan-Meier curves start at
100% (everyone is alive) and move downwards over time; adverse
events are best shown as cumulative hazard plots which move
upwards over time.
• Adverse events that occur with sufficient frequency for formal anal-
ysis should be analysed using “survival” type methods, and con-
sideration should always be given to showing graphs of cumulative
hazards.
• The Kaplan-Meier method does not directly provide significance
tests or confidence intervals for comparisons between groups. It is
possible to treat the data as comparisons of proportions, but these do
not take into account differences over time and do not fully utilise
the data. The simplest method of comparing the curves is the log
rank test.

208

group6_PH.indd 208 7.8.2007 12:20:34

 • A more complex method for comparing time to event data is a “pro-
portional hazards regression” or “Cox regression”. This, like the
log rank test, compares an entire survival curve without making as-
sumptions about the form of the hazard rate at any particular time,
but it does assume that the ratio of the hazard rates between two
groups is constant at all times. This method can be used to adjust
for other prognostic factors as well as for making a comparison
between a treated and control group. It may be used for data from
both randomised trials and observational cohort studies.
• When comparing rates using events as the numerator and person-
time as the denominator, the basic assumption is that the number of
cases follows a Poisson distribution. Analysis of these rates uses a
Poisson regression. The results of these analyses can be expressed
as incidence rate ratios.
• A meta-analytic review should be a routine part of the drug de-
velopment process so that ADRs, and differences in ADR rates
between treatment groups, can be detected as readily as possible,
especially for uncommon or rare events. Crude pooling of adverse
event numbers across different trials to compare treated and control
groups should be avoided if possible.
• No absolute criteria can be established for whether data from differ-
ent trials can be combined so as to yield a valid analysis. However,
some points should be considered: Is the experimental drug the same
in all trials (dose, regimens, formulation, route of administration)?
Is the comparator the same (placebo, active; dose of comparator)?
Is duration of treatment the same? Are the protocols similar (inclu-
sion and exclusion criteria; ages, sex, race; duration and severity of
disease; concurrent disease)?
• It may be helpful to use graphical methods in meta-analyses, which
can show similarities and differences for common as well as rare
effects across different trials in a clear way. They can also be used
to illustrate the uncertainty in effects so that apparently dissimilar
results may be seen to be simply different by chance.
• When a greater number of a serious but rare events is observed com-
pared to that expected on the basis of background data, this will
always be cause for careful scrutiny. However, multiple cases can
occur close together in time or space (sometimes referred to as a
cluster of cases) which may not be caused by the drug. Chance

209

group6_PH.indd 209 7.8.2007 12:20:34

 can be an explanation, or there may be some external factor not as-
sociated with the trial treatment that is producing multiple cases.
A within-trial comparison will always be more reliable for decid-
ing on causality. The use of background data, whether from popu-
lation-based data or from pooled control groups from many trials,
is subject to more uncertainty than the comparison of randomised
groups.
• Further research is required, through examination of large data-
bases of completed clinical trials, to attempt to obtain rates of
background occurrence of events that are serious and rare. These
rates should be published so that interpretation of a single or a
few such events can be accomplished more objectively than is
currently possible.

g. Regulatory Reporting and Other Communication

of Safety Information from Clinical Trials
It is important to point out that many of these recommendations
are only proposals. Although the CIOMS VI Working Group as well
as its external panel of reviewers agree that these proposals represent
a meaningful way forward, none of the recommendations should be
interpreted as superseding current regulations. Rather, the recom-
mendations are intended to inform discussions for future regulations,
as has been the case with prior CIOMS proposals. Until such time as
these proposals may be implemented, sponsors are expected to main-
tain compliance with all existing regulations.

• The CIOMS VI Working Group endorses ICH Guideline E2A

and thus recommends the harmonization of criteria for expedited
reporting to regulatory authorities, to include suspected adverse
drug reactions that are both serious and unexpected. Only under
exceptional circumstances and on an ad hoc basis (e.g., when close
scrutiny and monitoring of a specific adverse reaction is warrant-
ed) should sponsors be expected to report, on an expedited basis,
suspected adverse drug reactions that are considered expected. If
there is a need to report events without regard to causality, this
should generally be on a periodic basis with the periodicity and
format, e.g., line listing, agreed in advance with the concerned au-
thority.

210

group6_PH.indd 210 7.8.2007 12:20:35

 • It is recommended that regulators adopt the phrase “a reasonable
possibility of a causal relationship” and consider dropping the
phrase “a causal relationship cannot be ruled out” from the ICH
E2A definition of suspected adverse drug reaction.
• In order to maintain global consistency of clinical trial reporting,
the Working Group recommends that once a drug is marketed, the
CCSI effectively become the reference safety information for the
purpose of determining expectedness for regulatory reporting from
Phase IV clinical trials. For clinical trials of new indications, new
populations or new dosage forms for a marketed drug, every at-
tempt should be made to align the DCSI and the CCSI, but the
DCSI should be used if it is different from the CCSI.
• As with spontaneous reports, the determination of reportability for
case reports from clinical trials should be determined at the event
level. That is, a case would meet the criteria for expedited reporting
only if there is a suspected adverse reaction that is both serious and
unexpected.
• Suspected adverse drug reactions that are both serious and unex-
pected, and thus subject to expedited reporting, should generally
be unblinded. However, there are likely to be special circumstances
where an exception to this rule would be appropriate, for example,
where the efficacy endpoint is also a serious adverse event (SAE). In
this case, the circumstance and the process to be followed should be
clearly defined in the protocol and the sponsor should seek agreement
from the relevant regulatory authorities. Such exceptions should be
clearly described in the protocol and Investigator Brochure.
• Unblinded placebo cases should generally not be reported to regu-
latory authorities on an expedited basis. On the other hand, it is
recommended that unblinded comparator cases be reported to regu-
latory authorities and/or the company owning the comparator on
an expedited basis, regardless of expectedness. Likewise, serious
suspected adverse reactions for open-label comparators should be
sent on an expedited basis to the appropriate regulatory authorities
and/or company regardless of expectedness.
• As a general rule, 7-day reporting should be limited to reports from
clinical trials and not include those from the spontaneous reporting
environment. This should generally apply to reporting in countries
where the drug is not yet approved as well as in countries where the

211

group6_PH.indd 211 7.8.2007 12:20:35

 drug is approved. (Note that this may be in conflict with, and does
not supersede, current regulation.)
• For circumstances other than individual case reports, where prompt
notification to authorities is warranted (e.g., non-clinical safety in-
formation having implications for the potential for serious adverse
reactions in human subjects; an increased frequency of a previ-
ously recognized serious adverse reaction; an incidence of a seri-
ous adverse event that is significantly higher for the experimental
drug than for a comparator; a greater than expected incidence of
a serious adverse event compared to the relevant background rate
in the general population; a significant drug interaction observed
in a pharmacokinetic study), sponsors should define their internal
decision-making process in a standard operating procedure (SOP),
including how the clock start date will be determined for prompt
notifications.
• In addition to the usual criteria for an expedited report, adverse
events that are not deemed to be drug-related but are considered to
be protocol related should also be reported in an expedited fashion
if they are serious.
• Although contrary to established regulations, the CIOMS VI Work-
ing Group proposes that routine expedited case reporting by spon-
sors to investigators and IECs/IRBs be eliminated. Instead, sponsors
should provide regular updates of the evolving benefit/risk profile
and highlight important new safety information. Significant new
information, occasionally a single case report, that has implications
for the conduct of the trial or warrants an immediate revision to the
informed consent would be communicated on an expedited basis.
More commonly, important new safety information would be com-
municated periodically, based on the assessment of accumulating,
aggregate information.
• It is proposed that there be a single Development Safety Update Re-
port (DSUR) for submission to regulators on an annual basis, with
a consistent format and content which are yet to be defined. It is
strongly recommended that DSURs be based on an entire develop-
ment program and not per protocol. Consideration should be given
to establishing a common international birthdate which would be
the date of first authorization to begin clinical trials anywhere in the
world. The DCSI should be attached to the annual DSUR with an

212

group6_PH.indd 212 7.8.2007 12:20:35

 explanation of any changes since the last update, with any signifi-
cant new safety information highlighted.
• For products with a well-established safety profile and for which
most clinical trials are Phase IV studies in the approved indication(s),
it is recommended that the PSUR replace the annual DSUR.
• Sponsors should establish a policy of incorporating Development
Core Safety Information (DCSI) into every IB, either as a special
section of the IB or as an attachment to the IB. The DCSI should
clearly identify the events for which the company believes there is
sufficient evidence to suspect a drug-relationship. These would be
the events that would be considered expected (“listed”) from the
standpoint of pre-approval regulatory reporting criteria.
• Sponsors should review the IB and DCSI at least annually and up-
date them as appropriate. If there are no changes to the IB or DCSI,
then the investigators and ethics committees should be so informed
at a convenient opportunity.
• For unapproved products, instead of sending individual expedited
clinical trial case reports to investigators and IECs/IRBs, as men-
tioned above, the CIOMS VI Working Group recommends periodic
reports to investigators and IECs/IRBs. It is recommended that
such reports include a line listing of unblinded clinical trial cases
that were expedited to regulatory authorities since the last periodic
report, a copy of the current DCSI along with an explanation of any
changes, a statement if there are no changes, and a brief summary
of the emerging safety profile. Although it is recommended that
the default would be quarterly updates, there may be circumstances
when a more immediate or less frequent communication would be
appropriate.
• For approved products, the timeframe for periodic reports to inves-
tigators and IECs/IRBs would depend on the extent to which new
indications are being developed. For a product undergoing Phase III
trials, continuation of the quarterly reports would be advisable. For
well-established products, less frequent updates would be appropri-
ate and at some point, there should only be a need to update investi-
gators and IECs/IRBs when there is significant new information to
report. For Phase IV investigators and their associated IECs/IRBs,
communications of changes to the CCSI should be sufficient.

213

group6_PH.indd 213 7.8.2007 12:20:36

 • When updates are provided by the sponsor to investigators or IECs/
IRBs, whether for unapproved or approved products, line listings
should include only unblinded expedited reports from clinical
trials. The line listings should include interval data, i.e., only
cases expedited since the last update; however, the summary of the
emerging safety profile should take into account all of the accumu-
lating data. The use of MedDRA preferred terms is recommended.
The line listings generally should not include spontaneous reports;
instead, significant issues arising from spontaneous reports can be
described in narrative form in the update.
• If a significant safety issue is identified, either from an individual
case report or review of aggregate data, then the sponsor should
issue a prompt notification to all parties, namely regulatory authori-
ties, investigators, IECs/IRBs, and if relevant DSMBs. A significant
safety issue could be defined as one that has a significant impact on
the course of the clinical trial or programme (including the potential
for suspension of the trial programme or amendments to protocols)
or warrants immediate update of informed consent.
• Sponsors should define a Safety Management Team to review all
available safety information on a regular basis so that decisions can
be made with cross-functional input. It is further recommended that
these reviews take place quarterly pre-approval and be coordinated
with the PSUR schedule (six-monthly or annually) post-approval.
In addition, ad hoc Safety Management Team meetings may be war-
ranted to address urgent safety issues or significant safety signals.
• Safety Management Team meetings should be used to review the
overall evolving safety profile during development, to make chang-
es to the DCSI and/or informed consent and to determine if any
changes in the conduct of the trials need to be considered. The out-
come of these meetings may then provide the basis for the brief
summary of safety in the periodic reports to investigators/IECs/
IRBs.
• DSMB’s are most commonly employed for a single large clinical
trial and are not usually charged with providing oversight of an
entire clinical program. It would therefore be important to ensure
that important new safety information is communicated to a DSMB
even if the information did not originate from the DSMB-moni-
tored study.

214

group6_PH.indd 214 7.8.2007 12:20:36

 • Often the developer or manufacturer of a product is not the spon-
sor of a particular clinical trial, but rather has agreed to support an
external clinical or non-clinical investigator-sponsor financially or
by providing drug supplies. A standard provision of any agreement
with an outside investigator-sponsor should be the prompt report-
ing to the Company of all serious suspected adverse drug reactions,
or significant findings from, say, an animal study, as well as any
suspicion of a previously unrecognized hazard to patients. Timely
access to the final study report should also be included.
• The same previously recommended concept and level of threshold
for changes to the CCSI (CIOMS III/V report) should be applied
to the DCSI and informed consent information. Thus, communica-
tion to investigators and IECs/IRBs of an update to the DCSI may
indicate the need to update the informed consent form, the final
decision for which rests with the IEC/IRB.
• Informed consent should be renewed whenever there is new infor-
mation that could affect the subjects’ willingness to participate, in-
cluding new information about risks. In most cases, when updating
informed consent, it should be sufficient to do so for continuing
study participants at the next scheduled visit. However, there may
be circumstances where a more immediate communication would
be more appropriate (e.g., when a new risk has been identified that
is life-threatening or when there is a prolonged time interval be-
tween visits).

215

group6_PH.indd 215 7.8.2007 12:20:37

group6_PH.indd 216 7.8.2007 12:20:37
 Appendix 1

Glossary and Abbreviations

This glossary contains key terms used in this report. As the reader will
appreciate, most of these terms have been in common use for some time,
but in spite of initiatives under ICH, WHO and CIOMS, internationally
agreed and uniformly adopted definitions do not necessarily exist for many
of them. In some cases, the differences between different definitions – such
as those within different countries’ regulations – are relatively insignifi-
cant, but in others they may be important. It must also be acknowledged
that most countries in the world have not participated formally in the ICH
process or in CIOMS activities and therefore traditionally rely on WHO
for guidance on terminology, especially in the area of pharmacovigilance.
However, clinical trials for new drug development are conducted in many
such countries and CIOMS Working Group VI is advocating the acceptance
and use of the definitions given here. Whenever possible, the definitions for
pre- and post-approval conditions should be identical.
It is recognised that many country health authorities are more famil-
iar with, and may have incorporated into regulations, previously published
WHO definitions1 which may differ from some of those given here.
Throughout this Appendix and the full report, unless indicated other-
wise, the word “drug” is meant to include all medicines (drugs, vaccines,
biotechnology products) for prevention, prophylaxis or treatment of a dis-
ease or medical condition, and possibly for use in diagnosis.
In most cases, the definitions are taken from ICH guidelines that have
reached Step 4. Other definitions come from CIOMS, WHO, or elsewhere
and a few have been created specifically for this work. Some of the terms
are accompanied by a commentary to help clarify the definition in the con-
text of safety in clinical trials, or to recommend that the existing, official
definition be modified at the next opportunity (e.g., by ICH). Unless indi-
cated otherwise, all definitions are quoted verbatim from their sources.

1
WHO Guidelines for Good Clinical Practice (GCP) for Trials on Pharmaceutical Products, WHO Technical
Report Series, No. 850, 1995, pp. 97-137 (see http:/www.who.int/medicines/).

217

group6_PH.indd 217 7.8.2007 12:20:37

 Many of the terms covered in this Glossary are also defined in the
European Union Clinical Trial Directive which became legally binding on
all 25 Member States in May 2004. Therefore, the corresponding EU defi-
nitions are also included for reference, whether or not they differ from the
definitions recommended here.2
Although this Glossary covers many terms and definitions related to
drug safety and pharmacovigilance, it also addresses many abbreviations,
terms and concepts associated with biostatistics and risk, particularly those
used in Chapter 6. The origins of statistical terms and definitions are not
provided, since they are commonly found in many reference works; the
specific terminology presented here was prepared by a senior statistician
member of the CIOMS VI Working Group. For more detailed discussion,
see Chapter 6.
Readers may be interested in the following general reference sources:
“Dictionary for Clinical Trials,” by Simon Day, Wiley Interscience, 1999;
“Dictionary of Pharmacoepidemiology,” by B. Begaud, John Wiley and Sons,
2000; “Pharmacovigilance from A to Z,” by Barton L. Cobert and Pierre Bi-
ron, Blackwell Science, Malden, MA (USA), 2002; and Medilexicon, the
world’s largest online database of medical and pharma-related abbrevia-
tions – over 70,000 (see Medilexicon.com or http://eu.xmts.net/34683).
A special comment is important with regard to abbreviations used in
connection with adverse event or reaction reports, especially those that
are described as “serious.” There are many abbreviations and acronyms in
current use that unfortunately are not completely standardized across the
regulatory world and have different meanings. They include the following,
some of which are defined below within the Glossary:
ICSR – Individual Case Safety Report (used in ICH Guideline E2B to
refer to an electronic report on a single patient)
SADR – Suspected Adverse Drug Reaction (proposed by FDA in its
March 2003 proposed Rule on Safety Reporting Requirements for Hu-
man Drug and Biological Products)
SAE – Serious adverse event (commonly used in industry)

2
For details on interpretation and application of key terms defined in the EU Clinical Trial Directive, see
“Detailed guidance on the collection, verification and presentation of adverse reaction reports arising from
clinical trials on medicinal products for human use”, April 2004 (http://pharmacos.eudra.org/F2/pharmacos/
docs/Doc2004/). Corresponding definitions for other regulatory bodies (e.g, in Japan and the US) are not in-
cluded primarily because new definitions and interpretations were pending as of the beginning of 2005 when
this report was completed.

218

group6_PH.indd 218 7.8.2007 12:20:37

 SSAR – Suspected Serious Adverse Reaction or Serious Suspected
Adverse Reaction (commonly used in the EU)
SUSAR – Suspected Unexpected Serious Adverse Reaction (included
in the EU Clinical Trials Directive, effective May 2004)
Clearly, the use of the letter S for both “Serious” and “Suspected”
can only lead to confusion, especially with the current profusion of dif-
ferent abbreviations. The CIOMS Working Group strongly encourages the
harmonization of conflicting terms, abbreviations and definitions, logically
through the ICH process.

* * *

Absolute Risk
The number of people in a group who experience an adverse effect divided
by the number in that group who could experience that adverse effect.

Acceptable Risk
We do not provide a definition for this concept.
Commentary: Although this term is often used, especially in connec-
tion with benefit-risk considerations, it has proven impossible to define
(acceptable to whom and under what circumstances, for example?).
Readers are advised that they should be aware of this concept but
that acceptable risk may mean many different things depending on the
context and from whose perspective. If sponsors or regulators wish
to invoke the concept in assessing the value or use of a product dur-
ing development, they should base their judgments on the particular
circumstances of the clinical program. See Chapter 5 for more discus-
sion. Attempts have been made to define and measure acceptable risk
based on the concept of “utility” (e.g., see Lane, D.A. and Hutchinson,
T. The Notion of “Acceptable Risk”: The Role of Utility in Drug Man-
agement, J. Chron. Dis., 40:621-625, 1987).

Adverse Drug Reaction (ADR)

In the pre-approval clinical experience with a new medicinal product or its
new usages, particularly when the therapeutic dose(s) may not be estab-
lished:

219

group6_PH.indd 219 7.8.2007 12:20:38

 All noxious and unintended responses to a medicinal product related to any
dose should be considered adverse drug reactions. The phrase “responses to
a medicinal product” means that a causal relationship between a medicinal
product and an adverse event is at least a reasonable possibility, i.e., the
relationship cannot be ruled out.
ICH Guideline E6: Good Clinical Practice
In the EU: “Adverse Reaction” – all untoward and unintended responses
to an investigational medicinal product related to any dose administered.
Commentary: As shown, the current ICH definition includes the phrase
“i.e., the relationship cannot be ruled out.” The CIOMS Working Group
believes that it is virtually impossible to rule out with any certainty the
role of the drug on the basis of a single case. Therefore, we recom-
mend elimination of that phrase and prefer the ICH E2A elaboration
of “reasonable possibility” to mean that there are facts, evidence, or
arguments to support a causal association with the drug.

Adverse Event/Adverse Experience

Any untoward medical occurrence in a patient or clinical investigation sub-
ject administered a pharmaceutical product and which does not necessarily
have a causal relationship with this treatment. An adverse event (AE) can
therefore be any unfavourable and unintended sign (including an abnormal
laboratory finding), symptom, or disease temporally associated with the
use of a medicinal (investigational) product, whether or not related to the
medicinal (investigational) product.
ICH Guideline E6: Good Clinical Practice
In the EU: “Adverse Event”: any untoward medical occurrence in
a patient or clinical trial subject administered a medicinal product
and which does not necessarily have a causal relationship with this
treatment.

Adverse Event of Special Interest

An adverse event of special interest (serious or non-serious) is one of scien-
tific and medical concern specific to the sponsor’s product or program, for
which ongoing monitoring and rapid communication by the investigator to
the sponsor may be appropriate. Such events may require further investiga-
tion in order to characterize and understand them. Depending on the nature
of the event, rapid communication by the trial sponsor to other parties may
also be needed (e.g., regulators).

220

group6_PH.indd 220 7.8.2007 12:20:38

 This definition is proposed by the CIOMS VI Working Group.
Commentary: An adverse event of special interest is a noteworthy event
for the particular product or class of products that a sponsor may wish
to monitor carefully. It could be serious or non-serious (e.g., hair loss,
loss of taste, impotence), and could include events that might be poten-
tial precursors or prodromes for more serious medical conditions in
susceptible individuals. Such events should be described in protocols
or protocol amendments, and instructions provided for investigators
as to how and when they should be reported to the sponsor.

Analysis of Covariance (ANCOVA)

A statistical method for making comparisons between groups, while taking
into account different variables measured at the start of a trial. It is a form
of multiple regression.

Bayesian
A theorem in probability named after Reverend Thomas Bayes (1702-
1761). It is used to refer to a philosophy of statistics that treats probability
statements as having degrees of belief, in contrast to classical or Frequen-
tist statistics that regards probability strictly as being based on frequencies
of occurrence of events.

Binary Analysis
An analysis involving only two categories (e.g., baseline vs final values,
in contrast to analysis of multiple values from continuous measurements,
as for a progression of laboratory values). The latter can be turned into a
binary analysis by setting a single cut-off point so the data are split into just
two possible values (e.g., baseline vs highest post-baseline value).

Bonferroni Correction
A correction to allow for the probability of many events that are indepen-
dent, named after Carlo Emilio Bonferroni (1892-1960). In statistical sig-
nificance testing, it allows, for example, 10 different significance tests to be
made on a data set (e.g., 10 different laboratory parameters) but still have
an overall significance for one of the 10 tests at a probability of P=0.05, by
carrying out each of the 10 tests by using a more stringent probability level
of P=0.005 (thus, 0.05/10).

221

group6_PH.indd 221 7.8.2007 12:20:38

 Case Report Form
A printed, optical, or electronic document designed to record all of the proto-
col required information to be reported to the sponsor on each trial subject.
ICH Guideline E6: Good Clinical Practice

Censored, or Censoring of Data

The act of eliminating data from analyses. Observations on certain patients,
particularly the time until an event occurs, may be missing or incomplete.
That is, the person has been followed for a known length of time but the
event of interest for analysis has not yet occurred. Such observations are
called “censored” observations, and the process is called “censoring”.

Chi-square
This can refer to a statistical significance test or to the theoretical distri-
bution to which a chi-square test refers (i.e., chi-square distribution). The
test is usually a comparison of proportions. In its simplest form, with a
2 x 2 contingency table, it is described as a one degree of freedom test. For
example, a statistical comparison of the proportions of adverse reactions in
two groups of patients is made using a chi-square test. The test results in
a chi-square value from which a P value is obtained. This gives the prob-
ability of finding a difference in proportions as large as or larger than the
difference observed, even when there is no true difference in those propor-
tions. The data can have more than two treatments, and also more than two
categories of response. Chi-square tests of data from larger size tables have
higher numbers of degrees of freedom.

Company Core Safety Information (CCSI)

All relevant safety information contained in the company core data sheet
prepared by the MAH [Marketing Authorization Holder] and that the MAH
requires to be listed in all countries where the company markets the drug,
except when the local regulatory authority specifically requires a modi-
fication. It is the reference information by which listed and unlisted are
determined for the purpose of periodic reporting for marketed products,
but not by which expected and unexpected are determined for expedited
reporting.
ICH Guideline E2C: Periodic Safety Update Report for Marketed Drugs
Commentary: The CIOMS VI Working Group believes that for drugs
on the market in some places while under investigation in others, con-

222

group6_PH.indd 222 7.8.2007 12:20:39

 sideration should be given to using the CCSI as the basis for expedited
reporting on cases arising in post-marketing (Phase 4) clinical trials.
See Chapter 7, section b.(3).

Confidence Interval (CI)

An interval which shows the range of uncertainty in a measured summary
value, such as a relative risk (RR). It is typically expressed as a 95% CI but
it can be 99% or other value. If a 95% CI is from 0.26 to 0.96, it implies
that the treated group shows evidence of a reduction in the event rate, but
that the data are compatible with a large reduction (RR = 0.26) and also a
small reduction (RR = 0.96). Strictly speaking, a 95% CI implies that 95%
of such intervals, will, in the long run, contain the true value of the sum-
mary (in this example, the RR). The boundaries are the lower (0.26) and the
upper (0.96) confidence interval. If the boundary includes the null value,
such as an RR of 1, it means the difference is not statistically significant
(e.g., a CI of 0.5 to 1.8).

Contingency Table
A table of data arranged in categories in rows and columns. The simplest
is a two-by-two (2 x 2) table with 4 cells, but it could have any number of
rows and columns.

Contract Research Organization (CRO)

A scientific organization (commercial, academic or other) to which a spon-
sor may transfer some of its tasks and obligations. Any such transfer should
be defined in writing.
ICH Guideline E6: Good Clinical Practice

Correlation
A measure of the relationship between two (or more) variables. A correla-
tion coefficient, which measures the strength of a linear relationship, can
range from -1 (perfect negative linear relationship) through zero (no linear
relationship) to +1, a perfect positive relationship.

Covariance
The statistical measure of the way that two variables vary in relation to each
other. It is used in calculations of correlation and regression coefficients.

223

group6_PH.indd 223 7.8.2007 12:20:39

 Covariate
This is a variable that is examined as to how it relates to another variable. It
usually refers to an explanatory (influential) variable, while the variable of
interest is the response or outcome variable.

Cox Model
A form of multivariable regression used in survival analysis, named after
Sir David Cox who suggested the method in 1972. It can examine the effect
of several explanatory variables on the time to occurrence of some outcome
event such as an adverse reaction. It makes some assumptions about the ef-
fect of these explanatory variables on the outcome.

Development Core Safety Information (DCSI)

An independent section of an Investigator’s Brochure (IB) identical in
structure to the Company Core Safety Information (CCSI) that contains a
summary of all relevant safety information that is described in more detail
within the main body of the IB. It is the reference safety document that
determines whether an ADR is listed or unlisted.
Based on: Guidelines for Preparing Core Clinical Safety Information
on Drugs. Second Edition. Report of CIOMS Working Groups III and V,
CIOMS, Geneva, 1999

Development Pharmacovigilance and Risk Management Plan

A plan to conduct activities relating to the detection, assessment, under-
standing, reporting and prevention of adverse effects of medicines during
clinical trials. This plan should be initiated early and modified as necessary
throughout the development process for a new drug or drug-use.
This term and definition are proposed by the CIOMS VI Working Group.

Development Safety Update Report (DSUR)

A periodic summary of safety information for regulators, including any
changes in the benefit-risk relationship, for a drug, biologic or vaccine un-
der development, prepared by the sponsor of all its clinical trials.
This term and definition are proposed by the CIOMS VI Working Group.
Commentary. A DSUR should serve as a summary of the safety experi-
ence in all clinical trials for a drug in development, including trials for
new uses of an already approved drug (e.g., new dosage forms, indica-

224

group6_PH.indd 224 7.8.2007 12:20:40

 tions, populations). In practice, it can serve as the foundation for any
changes in the Investigator’s Brochure and/or Development Core Safe-
ty Information (DCSI). The CIOMS VI Working Group believes that the
DSUR can serve as a platform for reconciling and harmonizing the
currently different periodic reporting requirements for clinical trials
in the US (IND Annual Report) and the EU (Annual Safety Report).
For details, see Chapter 7. CIOMS Working Group VII, in progress as
of this report, is dedicated to proposing details on the format, content
and timing of such reports.

Effectiveness
Effectiveness is a measure of the effect a medicine (or medical technol-
ogy) is purported, or is represented, to have under conditions for the use
prescribed, recommended or labeled.
Benefit-Risk Balance for Marketed Drugs. Report of CIOMS Working
Group IV, CIOMS, Geneva, 1998
Commentary: The standard definition usually given in medical
dictionaries is similar: the ability of an intervention to produce the
desired beneficial effect in actual use.

Efficacy
Efficacy is the ability of a medicine or medical technology to bring about
the intended beneficial effect on individuals in a defined population with a
given medical problem, under ideal conditions of use.
Benefit-Risk Balance for Marketed Drugs. Report of CIOMS Working
Group IV, CIOMS, Geneva, 1998
Commentary: Efficacy refers to how well a particular medicine causes
the desired effect under ideal or near ideal conditions, as in a clinical
trial setting. A drug is “efficacious” if it demonstrates the intended
therapeutic effect under standardized/experimental conditions.

Expected and Unexpected Adverse Drug Reaction

An expected ADR is one for which its nature or severity is consistent with
that included in the appropriate reference safety information (e.g., Inves-
tigator’s Brochure for an unapproved investigational product or package
insert/summary of product characteristics for an approved product).
Based on: Current Challenges in Pharmacovigilance: Pragmatic Approaches,
Report of CIOMS Working Group V, CIOMS, Geneva, 2001, p. 109.

225

group6_PH.indd 225 7.8.2007 12:20:40

 An unexpected ADR is defined as: An adverse reaction, the nature or sever-
ity of which is not consistent with the applicable product information (e.g.,
Investigator’s Brochure for an unapproved investigational product or pack-
age insert/summary of product characteristics for an approved product).
ICH Guideline: E6 Good Clinical Practice
[Note: ICH does not define “expected” ADR.]
In the EU: “Unexpected Adverse Reaction” – an adverse reaction,
the nature or severity of which is not consistent with the applicable
product information (e.g., investigator’s brochure for an unauthorised
investigational product or summary of product characteristics for an
authorised product).
Commentary: The concept of “expectedness” refers to events which
may or may not have been previously observed and documented. It
does not refer to what might have been anticipated (expected in a differ-
ent sense) from the known pharmacological properties of the medicine.
Depending on the context, expected and unexpected can refer to la-
beled vs unlabeled (for official data sheets/package inserts for marketed
products) or listed vs unlisted (for the Investigator’s Brochure, Develop-
ment Core Safety Information (DCSI), or Company Core Safety Informa-
tion (CCSI)). These other terms are also defined within this Glossary.

Fairweather Rules
Rules used by the FDA to analyze carcinogenicity studies. See Fairweather,
W.R., et al., Biostatistical Methodology in Carcinogenicity Studies. Drug
Information Journal, 32: 402-421 (1998).

False Positive
Usually used in connection with diagnostic testing, when a test result is
positive in someone who does not have the disease. It is also applied to sta-
tistical test results where a significant test result occurs but the null hypoth-
esis (no real difference) is in fact true. The probability of this happening
can be set in advance by the analyst.

False Negative
Usually used in connection with diagnostic testing, when a test result is
negative in someone who actually does have the disease. It is also applied
to statistical test results where a non-significant test result is found, whereas
the null hypothesis (that there is no difference) is in fact false. The prob-

226

group6_PH.indd 226 7.8.2007 12:20:40

 ability of this happening depends on the magnitude of the true difference.
This magnitude can be assumed and the sample size in a study adjusted in
order to ensure that the probability of a false negative is low. In studies of
adverse reactions it will often be high because the usual low incidence of
ADRs makes finding significant differences difficult.

Fisher’s Exact Test

An alternative to a chi-square test that is used when numbers in some cells
are small. It gives a P value as its result.

Independent Data-Monitoring Committee (IDMC),

Data and Safety Monitoring Board (DSMB),
Monitoring Committee,
Data Monitoring Committee
An independent data-monitoring committee that may be established by the
sponsor to assess at intervals the progress of a clinical trial, the safety data,
and the critical efficacy endpoints, and to recommend to the sponsor wheth-
er to continue, modify, or stop a trial.
ICH Guideline E6: Good Clinical Practice
Commentary: Data monitoring committees/boards are referred to by
several names and they may have different roles and responsibilities
depending on the particular circumstances. For convenience and con-
sistency, the CIOMS Working Group favors the term Data and Safety
Monitoring Board (DSMB). DSMBs are responsible for monitoring
and reviewing both safety and efficacy data, not just “critical study
endpoints.” For detailed discussion on DSMBs, see Appendix 5 in this
report and the references cited in Chapter 2, Section b.

Independent Ethics Committee (IEC)

(Also, see Institutional Review Board)
An independent body (a review board or a committee, institutional, regional,
national, or supranational), constituted of medical/scientific professionals and
non-medical/non-scientific members, whose responsibility it is to ensure the
protection of the rights, safety, and well-being of human subjects involved in a
trial and to provide public assurance of that protection, by, among other things,
reviewing and approving/providing favourable opinion on, the trial protocol,
the suitability of the investigator(s), facilities, and the methods and material to
be used in obtaining and documenting informed consent of the trial subjects.

227

group6_PH.indd 227 7.8.2007 12:20:41

 The legal status, composition, function, operations and regulatory require-
ments pertaining to Independent Ethics Committees may differ among
countries, but should allow the Independent Ethics Committee to act in
agreement with GCP as described in this guideline.
ICH Guideline E6: Good Clinical Practice
In the EU: “Ethics Committee” – an independent body in a Member
State, consisting of healthcare professionals and non-medical mem-
bers, whose responsibility it is to protect the rights, safety and well
being of human subjects involved in a trial and to provide public
assurance of that protection, by, among other things, expressing an
opinion on the trial protocol, the suitability of the investigators and the
adequacy of facilities, and on the methods and documents to be used to
inform trial subjects and obtain their informed consent.

Inference, Inferential
Statistical inference is the process of inferring conclusions about data based
on the uncertainty in summary data, as opposed to descriptive statistics.
This process is inferential.

Informed Consent
A process by which a subject voluntarily confirms his or her willingness to
participate in a particular trial, after having been informed of all aspects of
the trial that are relevant to the subject’s decision to participate. Informed
consent is documented by means of a written, signed and dated informed
consent form.
ICH Guideline E6: Good Clinical Practice
In the EU: “Informed Consent” – decision, which must be written,
dated and signed, to take part in a clinical trial, taken freely after be-
ing duly informed of its nature, significance, implications and risks
and appropriately documented, by any person capable of giving con-
sent or, where the person is not capable of giving consent, by his or her
legal representative; if the person concerned is unable to write, oral
consent in the presence of at least one witness may be given in excep-
tional cases, as provided for in national legislation.
Commentary: As specified in the Declaration of Helsinki (see Appen-
dix 4, paragraph 22), a physician should obtain a subject’s freely-
given consent preferably in writing. If the consent cannot be obtained
in writing, “non-written consent must be formally documented and

228

group6_PH.indd 228 7.8.2007 12:20:41

 witnessed.” Informed consent as applied to children and incapacitated
participants requires special consideration; see the EU Clinical Trial
Directive (Article 2J, 2001/20/EC), the Declaration of Helsinki
(Appendix 4), and the International Ethical Guidelines for Biomedical
Research Involving Human Subjects, CIOMS, Geneva, 2002.

Institutional Review Board (IRB)

(See also Independent Ethics Committee – IEC)
An independent body constituted of medical, scientific, and non-scientific
members, whose responsibility is to ensure the protection of the rights,
safety and well-being of human subjects involved in a trial by, among other
things, reviewing, approving, and providing continuing review of trial pro-
tocol and amendments and of the methods and material to be used in ob-
taining and documenting informed consent of the trial subjects.
ICH Guideline E6: Good Clinical Practice
Commentary: IEC (EC) and IRB are generally used synonymously.
However, depending on country or region, the term IRB may be used
instead of IEC (or EC), especially if the term is specified in regulations
or may be legally binding (e.g., IRB in the U.S.). There also may be
slight differences between Ethics Committees and Institutional Review
Boards. For detailed discussion, see Chapter 2 of this CIOMS report.

Investigational Product
A pharmaceutical form of an active ingredient or placebo being tested or
used as a reference in a clinical trial, including a product with a marketing
authorization when used or assembled (formulated or packaged) in a way
different from the approved form, or when used for an unapproved indica-
tion, or when used to gain further information about an approved use.
ICH Guideline E6: Good Clinical Practice
In the EU: “Investigational medicinal product” – a pharmaceutical form
of an active substance or placebo being tested or used as a reference in a
clinical trial, including products already with a marketing authorisation
but used or assembled (formulated or packaged) in a way different from
the authorised form, or when used for an unauthorised indication, or
when used to gain further information about the authorised form.
Commentary: For purposes of this CIOMS report, for drugs in develop-
ment the term “investigational product” refers to the experimental (un-
approved) product.

229

group6_PH.indd 229 7.8.2007 12:20:41

 Kaplan-Meier
Named after two statisticians who developed a graphical and tabular
method of analysing survival-type data, which is relevant to ADR data.

Labelled or Unlabeled (Also, see Expected and Unexpected)

For a product with an approved marketing application, any reaction which
is not mentioned in the official product information is unlabeled. If it is
included it is termed labeled.
Current Challenges in Pharmacovigilance: Report of CIOMS Working
Group V, CIOMS, Geneva, 2001.

Listed or Unlisted (Also, see Expected and Unexpected)

Any reaction which is not included in the Company Core Safety Informa-
tion within a company’s core data sheet for a marketed product is unlisted.
If it is included it is termed listed.
Current Challenges in Pharmacovigilance: Report of CIOMS Working
Group V, CIOMS, Geneva, 2001.
Commentary: The terms listed and unlisted were purposely adopted
in ICH Guideline E2C (Periodic Safety Update Reports for Marketed
Drugs) for use with internal company safety information documents,
so as to distinguish them from the terms labeled and unlabeled, which
should only be used in association with official “labeling,” i.e., the SPC,
Package Insert, and generally the regulator-approved data sheets for
marketed products. The usage of listed/unlisted has been extended to
the Development Core Safety Information (DCSI) as recommended in
Guidelines for Preparing Core Clinical-Safety Information on Drugs,
Second Edition, CIOMS Working Group III/V, CIOMS, Geneva, 1999.

Meta-analysis
The process of summarising data from more than one study to obtain a sin-
gle answer. There are various different statistical techniques to accomplish
this, each of which makes slightly different assumptions.

Multiplicity
The statistical problem caused by making multiple comparisons with a single
set of data. Significance tests are affected by how many such tests are made.

230

group6_PH.indd 230 7.8.2007 12:20:42

 Null Hypothesis
A statistical hypothesis that usually implies no difference between groups.
For rates of adverse reactions this may imply a relative risk of 1.

Number Needed to Harm (NNH)

The number of individuals needed to be treated for some specified pe-
riod of time in order that one person out of those treated would have one
harmful event (again, during some specified time period). See NNT for
calculation.

Number Needed to Treat (NNT)

The number of individuals needed to be treated for some specified period
in order that one person out of those treated should have the desired
benefit/outcome, such as the prevention of a medical event under treatment
(MI, e.g.). NNT is the reciprocal of the difference in rates of the measured
benefit, between a treated and a control group. For example, if the rate of
death is 1% in the experimental group as opposed to 2% in a control group
over one year of treatment, the difference is 1%. Thus, 100 people would
need to be treated for 1 year to prevent 1 death (1/100 = 1%).

Odds Ratio (OR)

The odds of an event (such as death) in one group compared to the odds in
a reference group. Odds are used in betting but have useful mathematical
properties in analysis of binary data. For example, if there are 10 individu-
als studied and 2 experience an event, the probability is 2/10 = 0.2. The
odds are 2:8 (2 have the event compared with 8 who do not). Therefore, the
odds = 0.25. If these odds are compared with another group in whom the
odds are different, say 0.125, then the odds ratio is 2 (0.25/0.125). With rare
events the OR approximates the relative risk.

One-sided vs Two-sided Testing

One-sided testing (also called one-tailed testing) refers to an analysis that
allows for/examines an effect in one direction only (e.g., an increase over
a comparator). Two-sided testing accounts for changes in either direction.
In most instances, as with comparisons of risk between different pro-
ducts, two-sided testing is preferred. For more details, see section f.(2) of
Chapter 6.

231

group6_PH.indd 231 7.8.2007 12:20:42

 Parametric
A form of statistical analysis that makes assumptions about the type of dis-
tribution of the data. E.g., a t-test assumes a normal distribution of the data,
and is referred to as a parametric test.

Pharmacovigilance
The science and activities relating to the detection, assessment, understand-
ing and prevention of adverse effects or any other drug-related problem.
The Importance of Pharmacovigilance – Safety Monitoring of Medicinal
Products, World Health Organization 2002 (ISBN 92 4 1590157), and ICH
Guideline E2E, Pharmacovigilance Planning (Step 4, November 2004).
Commentary: There is some uncertainty concerning the phrase “any
other drug related problem.” At least in the present context, the CIOMS
Working Group understands the phrase to refer to issues that could af-
fect the safety and safe use of medicines, such as medication errors
and potential product quality issues (e.g., glass particles in ampoules).
The CIOMS Working Group endorses the use of the term pharmaco-
vigilance for clinical safety activities during drug development as well
as for marketed products.

PHASES OF CLINICAL STUDIES (I – IV)

Phase I (Human Pharmacology)
Initial trials provide an early evaluation of short-term safety and tolerabil-
ity and can provide pharmacodynamic and pharmacokinetic information
needed to choose a suitable dosage range and administration schedule for
initial exploratory therapeutic trials.
Phase II (Therapeutic Exploratory)
Phase II is usually considered to start with the initiation of studies in which
the primary objective is to explore therapeutic efficacy in patients.
Phase III (Therapeutic Confirmatory)
Phase III usually is considered to begin with the initiation of studies in which
the primary objective is to demonstrate, or confirm therapeutic benefit.
Phase IV (Therapeutic Use)
Phase IV begins after drug approval. Therapeutic use studies go beyond the
prior demonstration of the drug’s safety, efficacy and dose definition. Studies in

232

group6_PH.indd 232 7.8.2007 12:20:42

 Phase IV are all studies (other than routine surveillance) performed after drug
approval and related to the approved indication. They are studies that were not
considered necessary for approval but are often important for optimising the
drug’s use. They may be of any type but should have valid scientific objectives.
Commonly conducted studies include additional drug-drug interaction, dose-
response or safety studies, and studies designed to support use under the ap-
proved indication, e.g., mortality/morbidity studies, epidemiological studies.
For all the above definitions – ICH Guideline E8: General Considerations
for Clinical Trials
Commentary: As delineated above, ICH Guideline E8 has proposed that
studies be categorized according to their objectives (human pharmacol-
ogy, therapeutic exploratory, therapeutic confirmatory, and therapeutic
use), as distinct from the traditional concept based strictly on temporal
phases of drug development. For example, human pharmacology stud-
ies (traditionally referred to as Phase I) can be and often are conducted
throughout a product’s lifetime (even though they are referred to as “Ini-
tial studies ..” in the definition above). In some settings, other terms are
used to categorize study types; for example, Phase IIA studies are some-
times referred to as “proof of concept studies,” Phase IIB can refer to
studies that establish proper dosing, and Phase IIIB refers to “peri-
approval” studies (Phase 4-like studies initiated prior to drug approval).
Depending on the product and nature of the program, there may not be a
sharp or distinct division between the various Phases of trials.
Phase IV studies may be required as a condition of regulatory approval.
The CIOMS Working Group believes that the ICH definition of Phase
IV studies needs modification by deleting the expression “(other than
routine surveillance),” which is not accurate, and by emphasizing that
such studies should be limited to uses and conditions specified within
the approved data sheet (SPC, Package Insert, etc.).

Point Estimate
The best estimate of a summary of data such as a mean or a relative risk. The
value of this figure on its own does not indicate how precisely it is estimated.

Poisson Distribution
A distribution of numbers, as in a normal distribution, but which applies to
counts of numbers of events rather than to continuous values and is asym-
metric. Negative values cannot occur.

233

group6_PH.indd 233 7.8.2007 12:20:43

 Power
In statistical terms, a measure or indication of whether an analysis that is
conducted is good at detecting differences. A powerful analysis is one that
finds differences to be statistically significant. Power largely depends on
how many events are observed, which therefore depends both on how many
individuals are studied (the more studied, the greater the power) and on the
rarity of the event (the less there are, the less powerful).

Rank
The order of a value in a set of values. Some statistical methods (non-
parametric tests) use the order rather than the actual value. In survival
analysis the ordering of times is important and a “log rank test” is able to
compare times to an event that occurs in different groups.

Regression
A statistical technique that examines relationships between a response vari-
able and one or more explanatory variables. This can be done for continu-
ous measurements but also for binary measures and survival times.

Relative Risk (RR)

A multiplicative factor applied to a reference risk associated with an expo-
sure. It is the risk of an outcome (event) measured in an exposed population
(absolute risk) divided by the risk (reference risk) of the same outcome
(event) in an unexposed group (the reference population).
Based on: Benefit-Risk Balance for Marketed Drugs. Report of CIOMS
Working Group IV, CIOMS, Geneva, 1998 and Dictionary of Pharmaco-
epidemiology, by B. Begaud, John Wiley & Sons, 2000.
Commentary: The relationship between two risks, generally estimated
in different populations, is often referred to as the “risk ratio” as well
as relative risk. There is a need to ensure that the two populations that
are compared are “comparable” (i.e., same/similar kinds of patients,
age, gender, disease state, exposure time, etc.). Example: risk of ADR
is 10/100,000 in drug-treated population and 5/1,000,000 in a compa-
rable but untreated population. Relative risk = 20.

Risk
As used in the context of adverse experiences, it is the proportion of in-
dividuals who have an event out of all those who could possibly have that

234

group6_PH.indd 234 7.8.2007 12:20:43

 event. Two groups can be compared either by taking their ratio (relative
risk) or by subtracting the two risks. The latter is called an absolute risk
difference.

Risk Assessment
Risk assessment is subdivided into risk estimation and risk evaluation. It is
defined as the integrated analysis of the risks inherent in a product, system
or plant and their significance in an appropriate context. Risk estimation
includes the identification of outcomes, the estimation of the magnitude of
the associated consequences of these outcomes and the estimation of the
probabilities of these outcomes. Risk evaluation is the complex process
of determining the significance or value of the identified hazards and esti-
mated risks to those concerned with or affected by the decision. It therefore
includes the study of risk perception and the trade-off between perceived
risks and perceived benefits. It is defined as the appraisal of the significance
of a given quantitative (or where acceptable, qualitative) measure of risk.
Risk analysis, perception and management. The Royal Society UK, 1992

Risk Management
Risk Management is the making of decisions concerning risks and their
subsequent implementation, and flows from risk estimation and risk evalu-
ation. It is defined as the process whereby decisions are made to accept a
known or assessed risk and/or the implementation of actions to reduce the
consequences or probability of occurrence.
Risk analysis, perception and management. The Royal Society, UK, 1992.
Commentary: In the field of drug safety there is no accepted, universal
definition of “risk management,” but in current usage, it refers to the
overall process for the technical and communication activities needed
to understand and prevent or minimize risk/harm, including the as-
sessment of any programs put in place. The US FDA refers to risk
management as the combination of risk assessment and risk minimi-
zation (see the Guidance for Industry. Development and Use of Risk
Minimization Action Plans, FDA, March 2005 (http://www.fda.gov/
cder/guidance/6358fnl.htm).

Scatterplots
Graphical diagrams that show the variation of individual continuous values
for two variables in a set of data. Different symbols can be used for the

235

group6_PH.indd 235 7.8.2007 12:20:44

 points themselves to distinguish between different groups. They are often
used to show before and after treatment values of the same variable (e.g.,
the liver enzyme value for each patient plotted as a function of time).

Sensitivity
This can have two meanings in statistical terms. The first is whether an anal-
ysis has high power (sensitive) or not. It can also mean sensitivity to the
assumptions made for an analysis, i.e., a test of whether the results of the
analysis change when assumptions about effects (parameters) are changed.

Serious Adverse Event or Reaction: Standard Criteria

Any untoward medical occurrence that at any dose:
❏ results in death,
❏ is life-threatening,*
❏ requires inpatient hospitalisation or prolongation of existing hospi-
talisation,
❏ results in persistent or significant disability/incapacity, or
❏ is a congenital anomaly/birth defect.

Medical and scientific judgement should be exercised in deciding whether

expedited reporting is appropriate in other situations, such as important
medical events that may not be immediately life-threatening or result in
death or hospitalisation but may jeopardise the patient or may require inter-
vention to prevent one of the other outcomes listed in the definition above.
These should also usually be considered serious. Examples of such events
are intensive treatment in an emergency room or at home for allergic broncho-
spasm; blood dyscrasias or convulsions that do not result in hospitalization;
or development of drug dependency or drug abuse.
*
Note: the term “life-threatening” refers to an event or reaction in which the
patient was at risk of death at the time of the event or reaction; it does not refer
to an event or reaction which hypothetically might have caused death if it were
more severe.

ICH Guideline E2A: Definitions and Standards for Expedited Reporting

In the EU: “Serious Adverse Event or Serious Adverse Reaction” –
any untoward medical occurrence or effect that at any dose results in
death, is life-threatening, requires hospitalisation or prolongation of
existing hospitalisation, results in persistent or significant disability or
incapacity, or is a congenital anomaly or birth defect.

236

group6_PH.indd 236 7.8.2007 12:20:44

 Commentary: The ICH definition of a serious AE or ADR has been ad-
opted for postmarketing applications in ICH Guideline E2D. The EU
definition given above is considered by the CIOMS Working Group as
incomplete without the paragraph beginning with “Medical and scien-
tific judgment ….” in the ICH definition.

Signal
A report or reports of an event with an unknown causal relationship to treatment
that is recognized as worthy of further exploration and continued surveillance.
Based on: Benefit-Risk Balance for Marketed Drugs. Report of CIOMS
Working Group IV, CIOMS, Geneva, 1998 and Dictionary of Pharmaco-
epidemiology, by B. Begaud, John Wiley & Sons, 2000.
Commentary: A signal can arise from non-clinical as well as clinical
sources. It should be based on data and not theory, and can refer not
only to a new (unexpected) and potentially important event, but also to
an unexpected finding for an already known event, such as information
on an ADR related to the nature (specificity), intensity, rate of occur-
rence or other clinically relevant finding that represents a meaning-
ful change from that expected in the subject/patient population under
investigation or treatment. A signal is not a confirmed finding, but is
generally referred to as an hypothesis-generating situation that must
be validated (“signal strengthening”) or disproved.
An older definition of a signal by the WHO Collaborating Centre for
International Drug Monitoring (British Medical Journal, 304:465,
22 February 1992) focused on post-marketing conditions and predated
the new definitions of adverse event and adverse reaction introduced
under ICH: “Reported information on a possible causal relationship
between an adverse event and a drug, the relationship being unknown
or incompletely documented previously. Usually more than a single re-
port is required to generate a signal, depending upon the seriousness
of the event and the quality of the information.”

Significance, Significant, Significantly

These terms refer to the quantitative interpretation of statistical tests. These
tests produce levels of probabilities (P-values) that indicate whether the differ-
ences measured are low (significant) or high (non-significant) if there are no
true differences. The conventional cut-off for “significant” is usually P=0.05
(5%), but reliance only on P values or “significance” can be misleading.

237

group6_PH.indd 237 7.8.2007 12:20:44

 Adverse reactions are often rare so that power is low and statistically significant
results may not be seen even in the presence of clinically important effects.

Simes
A method similar to a Bonferroni correction (see above) but with greater
power.

Sponsor
An individual, company, institution, or organization which takes responsi-
bility for the initiation, management, and/or financing of a clinical trial.
ICH Guideline: E6 Good Clinical Practice
In the EU: Identical to the above definition.

Survival Analysis
A statistical analytical technique originally developed for studying time
until death (survival time) following an intervention (or no intervention),
such as in cancer treatment trials. However, it is applicable to studying time
to some other type of event such as an adverse reaction or a non-fatal myo-
cardial infarction. Some types of survival analyses use non-parametric tests
such as the Log Rank Test, others can be “semi-parametric” such as the Cox
model (see above), or parametric (exponential or Weibull (see below)).

Suspected Unexpected Serious Adverse Reaction (SUSAR)

This term and acronym were introduced within one of the guidances to the
EU Clinical Trial Directive in connection with expedited reporting: “All
suspected adverse reactions related to an IMP (the tested IMP and com-
parators) which occur in the concerned trial that are both unexpected and
serious (SUSARs) are subject to expedited reporting.” [Note: IMP = inves-
tigational medicinal product]
Detailed guidance on the collection, verification and presentation of ad-
verse reaction reports arising from clinical trials on medicinal products for
human use, April 2004 (http://pharmacos.eudra.org/F2/pharmacos/docs/).

Systematic Error
An error that is not random/haphazard, but which will occur in the same
direction within one or many studies. For example, studying treatments for
too short a duration will systematically underestimate long-term effects.

238

group6_PH.indd 238 7.8.2007 12:20:45

 Type I and Type II Errors
A Type I error in statistical testing is a false positive (see above). A Type
II error is a false negative (see above), usually arising by studying too few
individuals.

Weibull Distribution
A distribution of data that is relevant to parametric survival analyses.

Yate’s Correction
A correction applied to data in a 2 x 2 contingency table when carrying
out a chi-square test. With modern computer software, however, a Fisher’s
exact test is generally preferred.

239

group6_PH.indd 239 7.8.2007 12:20:45

group6_PH.indd 240 7.8.2007 12:20:45
 Appendix 2

Membership and Process of

CIOMS Working Group VI
CIOMS Working Group VI on the Management of Safety Informa-
tion from Clinical Trials met in a series of eight formal meetings in Europe
and North America from March 2001 until October 2004. Listed below,
Name Organization* Part-time / full-time
Mary Couper WHO (Geneva) Part-time
Gerald Dal Pan FDA Part-time (Second-half)
Gaby Danan Aventis Full-time
Brian Davis MCA/MHRA (UK) Part-time (First-half)
Stephen J.W. Evans MCA /London School of Hygiene Full time
and Tropical Medicine
Hylar Friedman Pfizer Full-time
Trevor G. Gibbs GlaxoSmithKline Full-time
Arnold J. Gordon Pfizer/Consultant Full-time
Philip Harrison MCA/MHRA Part-time (Second-half)
Mohammed Hassar Institut Pasteur du Maroc Full-time
Linda S. Hostelley Merck & Co., Inc. Full-time
Martin Huber Hoffman-La Roche Part-time
Leonie Hunt Therapeutic Goods Administration Full-time
(TGA, Australia)
Juhana E. Idänpään-Heikkilä CIOMS/University of Helsinki Full-time
Sidney N. Kahn Bristol-Myers-Squibb/Consultant Full-time
Marianne Keisu AstraZeneca Full-time
Gottfried Kreutz BfArM (Germany) Full-time
Tatsuo Kurokawa Pharmaceuticals and Medical Full-time
Devices Agency (PMDA) (Japan)
Edith La Mache EMEA (London) Part-time
Hani Mickail Novartis Full-time
Siddika Mithani Health Canada Full-time
Jeff Powell Eli Lilly Full-time
Vicktor Raczkowski FDA (US) Part-time (First-half)
Patricia Saidon ANMAT (Argentina) Part-time
Wendy Stephenson Wyeth Full-time
Hugh Tilson University of North Carolina Part-time
Akiyoshi Uchiyama Yamanouchi/GlaxoSmithKline Full-time
Bozidar Vrhovac University of Zagreb (Croatia) Full-time
Ernst Weidmann Bayer Part-time
* Some members had more than one affiliation during the project. Some suggestions were contributed in
writing by Barbara Sickmueller (BPI, Germany).

241

group6_PH.indd 241 7.8.2007 12:20:45

 followed by a chronology of their work, are 29 senior scientists from drug
regulatory authorities, pharmaceutical companies and academia who par-
ticipated in the project.
At the first official meeting held at WHO in Geneva, Switzerland, in
March 2001, the Group agreed on the outline of the project and the topics
to be addressed. Some of the candidate topics had been identified during
the CIOMS Working Group V exercise (Current Challenges in Pharmaco-
vigilance: Pragmatic Approaches, Report of CIOMS Working Group V,
CIOMS 2001) and additional topics were identified by senior colleagues of
pharmaceutical companies and drug regulatory authorities.
CIOMS Working Groups I, II, III, IV and V had addressed pharmacovigi-
lance issues mostly for the post-authorization phase. It was obvious, however,
that there were many issues regarding management of the basic safety infor-
mation on pharmaceutical products collected during developmental research
prior to marketing authorization. In fact, a need was foreseen to introduce
the concept of a comprehensive safety plan for implementation throughout
Phase I through III trials that would evolve into a post-authorization pharma-
covigilance plan. Based on its early discussions, Working Group VI decided
to place its main focus on good practices surrounding the collection, assess-
ment and reporting/communication of safety data during clinical trials prior
to marketing. Special emphasis was placed on the ethical underpinnings of
conducting medical research and clinical trials in human subjects.
In 2000-2003, drug regulatory authorities, pharmaceutical companies
and clinical investigators were challenged by several new national, regional
and international guidelines and regulations, including those dealing with
ethical aspects of biomedical research. Implementation of ICH Guideline
E6 on GCP was completed, the World Medical Association’s Declaration
of Helsinki was revised in 2000 (and subsequently clarified in 2002 and
2004), the European Commission published the Clinical Trials Directive
in 2001 and its guidances in 2003, and CIOMS published the revised In-
ternational Ethical Guidelines for Biomedical Research Involving Human
Subjects in 2002. Moreover, the Group reviewed new developments in drug
safety regulations and concepts and in risk-management put forth by the
US FDA and the EU EMEA. Similarly, it was also kept up-to-date on new
initiatives in Japan, Australia and South America. All these aspects are re-
flected or referred to in the final report of the Group.
Individual topic chapters and other sections of the CIOMS VI report
were assigned for consideration and drafting to subgroups early in the

242

group6_PH.indd 242 7.8.2007 12:20:46

 project, but many participants served on multiple subgroups. The draft texts
and concepts were subsequently reviewed, discussed and debated several
times within the entire Working Group which led to revisions and refine-
ments of the texts. A survey of pharmaceutical companies on their safety
practices during clinical trials was conducted in early 2003; the results of
that survey helped inform the Group’s deliberations.
G. Kreutz and W. Stephenson acted as chairs and L. Hostelley served
as secretary of the Working Group with occasional assistance of M. Keisu
and L. Hunt. After the first meeting in Geneva in March 2001, the subse-
quent meetings were as follows: November 2001 (Philadelphia, PA),
May 2002 (Visby, Sweden), November 2002 (Montreal, Canada), May 2003
(Cologne, Germany), October 2003 (Washington, DC), May 2004
(Lucerne, Switzerland) and September-October 2004 (New York, NY).
During 2003 and 2004, the appointed editorial group for the report (A. J.
Gordon, M. Keisu, S. Mithani, Victor Raczkowski followed by Gerald Dal
Pan, and W. Stephenson) held teleconferences and meetings to coordinate
and design the overall report.
Outside experts were invited to critique a late draft of the report; they
included pharmacovigilance and related specialists from the pharmaceuti-
cal industry, academia, and health authorities (see Acknowledgements at
the beginning of this report). Their valuable input was incorporated into the
final document.
A. J. Gordon accepted the role of chief editor and compiled and edited
the draft consolidated reports and prepared the final manuscript for publi-
cation by CIOMS.

243

group6_PH.indd 243 7.8.2007 12:20:46

group6_PH.indd 244 7.8.2007 12:20:46
 Appendix 3

CIOMS VI Working Group Survey

on Pharmaceutical Company
Safety-Management Practices
During Clinical Trials

Introduction
In order to ascertain prevailing practices in the industry for many of the
areas under consideration in this publication, a survey was conducted via
the Internet during February and March 2003. The topics covered in the
survey included broad organization and policy issues as well as case pro-
cessing and data management. The results of this survey were helpful to the
CIOMS Working Group in formulating its proposals. The Working Group
gratefully acknowledges the help of the companies that responded to the
survey. None of the companies is identified in the results.

Results
[Note: Any answers appearing in italics were provided by the respondents and were not choices in
the original questionnaire.]

Of 5 Japanese, 19 European, and 35 US companies sent the question-

naire, there were 21 responses: Japanese = 4, European = 9, U.S. = 8 (based
on headquarters location of the parent company at the time the question-
naire was disseminated).
Abbott Merck (US)
Astra Zeneca Merck AG (Germany)
Aventis Novartis
Berlex Pharmacia
Boehringer Ingelheim Pfizer
Daiichi Procter & Gamble
Eisai Sanofi-Synthelabo
Eli Lilly Schering Plough
GSK Shionogi & Co.
Hoffman LaRoche Yamanouchi
Wyeth

245

group6_PH.indd 245 7.8.2007 12:20:47

 1. Is responsibility for the overall management of the safety of pre-
marketed (investigational) vs. marketed compounds divided between
separate areas within your company?

Yes 8
No 13

2. Recently, companies and regulators have been expanding the practice

of pharmacovigilance/drug safety to incorporate the concept of “risk
management,” which includes detection, assessment, management,
and communication of product safety issues both during development
and marketing of a drug.
a. Has your company developed, or is it currently developing, such an
all-encompassing approach to drug safety?

Yes 20
No 1

b. If yes, do you have a distinct group that is responsible for clinical

safety risk management?

Yes 7
No 10

c. If yes to 2.b., what are the responsibilities of this group?

(7 respondents)

Day to day safety issues (review of serious event 4

reports, lab data, etc.)

Strategic decisions related to compound 4

development and approval
Management of safety crises 4
Development and implementation of specific risk 1
mitigation/management programs

d. If yes to 2.b., what department is primarily responsible?

(7 respondents)

Clinical development 3
Pharmacovigilance/safety 4

246

group6_PH.indd 246 7.8.2007 12:20:47

 3. Some companies rely on standard medical texts or the various
CIOMS publications, such as the compendium, “Reporting Adverse Drug
Reactions: Definitions of Terms and Criteria for Their Use” (CIOMS,
Geneva, 1999), for strict definitions of medical events such as conditions/
diagnoses. Such terms are used to code the event in the database only if the
reported signs and symptoms are consistent with the predefined criteria for
the diagnosis (i.e., all the specified criteria must be met before a term such as
“acute hepatic failure” is applied to the case). Thus, both the reported term(s)
AND the term(s) assigned by the company to ensure consistent coding will be
included in a case report data file and summary prepared by the company.
a. Do you always code and enter into your pharmacovigilance data-
base only the exact term(s) used by the investigator?
Yes 15
No 5

b. Do you use standard definitions (e.g., CIOMS, other recognized-

medical sources) as predetermined criteria for assigning a term for
a diagnosis or condition, even if the result disagrees with the investiga-
tors term(s)?
Yes 4
No 17

c. If yes to 3.b., do you enter into your database both your choice of
term(s) and the investigators?
Yes 4
No 0

4. When does your company begin collecting adverse events in a clinical

trial? (21 respondents)
Informed consent is signed 14
Subject is randomized 0
Protocol specified interventions begin 0
First administration of the treatment 2
(placebo-washout, placebo, active RX)
Protocol specific 4
SAE’s at time of consent / AEs with first RX 1

247

group6_PH.indd 247 7.8.2007 12:20:47

 5. How long after a subject receives the last dose of study drug do you
purposely (by plan) continue to collect information on all adverse
events (non-serious and serious)? (20 respondents)
Protocol specific 16
30 days 8
5 half-lives of the compound 5
14 days 2
Completion of all visits 2
At least 42 days for live attenuated vaccines 1
No company standard 1
2 days 0
7 days 0

6. Does your company require that the investigator record all signs and
symptoms for an adverse event on the clinical case report form (CRF)
even if a diagnosis is made by the investigator?

Yes 4
No 17

7. Are signs and symptoms that are collected along with a diagnosis on a
CRF coded in your company database for the clinical trial?

Yes 13
No 8

8. Are signs and symptoms that are collected along with a diagnosis
coded in your company safety (pharmacovigilance) database for
serious cases?

Yes 12
No 9

9. Would your company support the use of an industry-standardized global

form for collection of serious adverse event data from investigators?
Yes 16
No 5

248

group6_PH.indd 248 7.8.2007 12:20:47

 10. If a serious adverse event case includes more than one event, at what
level do you assess drug relatedness? (21 respondents)
Every event (serious or not) 11
Serious adverse event ONLY 8
Diagnosis 4
Case as a whole 1

11. If you conduct causality assessments for each serious event/case, what
method(s) do you use for this assessment? (21 respondents)
“Introspection” 12
No specific method 4
Specific algorithm, home grown 2
Naranjo 1
Algorithm using earlier similar cases 1
Temporal relationship + alternative explanations 1
Suspected/non-suspected 1
Two degrees: excluded and cannot be ruled out 1

12. Does your company take the investigator’s causality assessment into
account in choosing AEs for analysis or inclusion in product safety
information (whether the Investigators Brochure, core data sheet, or
official product data sheet/SPC/Package Insert)?
Yes 20
No 1

13. a. Does your company currently collect investigator causality assessment

for adverse events on the clinical case report form or serious adverse
event report form?
Yes 20
No 0

For which adverse events?

All (serious and non-serious) 14
Only serious 4
Only non-serious 0

249

group6_PH.indd 249 7.8.2007 12:20:48

 b. When asking for an investigator’s causality assessment do you allow:
Yes/No 9
Qualitative answer using a scale* 11
* 3 to 6 degrees depending on company, including unclassifiable and from not related to definitely.

c. Are the investigator’s causality assessments (no matter what scale,

if any) utilized in data analysis or for regulatory reporting:
For data analysis only 0
For regulatory reporting only 2
Both analysis and reporting 19
Not used at all 0

d. Would your company support a single, global standard that requires

the investigator to indicate either “Yes” or “No” as the only choices
for the investigator’s opinion for a causal association between an
event and a study treatment?
Yes 14
No 7

14. a. Do you add a medical comment (company’s interpretation of the

case, including causality and/or appropriateness of the medical
terms/diagnosis provided by the investigator) to your record of
every serious event report?
Yes 14
No 7

b. If no, which reports contain such a comment? (5 respondents)

All unexpected related events 2
All possibly related events 0
All unexpected events (serious and non-serious) 0
Alternative etiologies, missing data 1
Serious, Unexpected, Related 1
None 1

15. a. Has your company adopted the Development Core Safety Informa-
tion (DCSI) concept proposed by CIOMS III/V?

250

group6_PH.indd 250 7.8.2007 12:20:48

 Yes: 7 No: 14
• Is it used to determine expectedness? • Do you have plans to adopt this approach?
Yes: 6 Yes: 7
No: 1 No: 6
• Has this increased the number of expedited
reports?
Yes: 2
No: 2

16. What is (are) the method(s) your organization uses to determine AE

attribution (causality assessment) to study agents in aggregate analy-
ses (study reports and integrated summaries of safety) for subsequent
inclusion in marketing application submissions and ultimately in
prescribing information? (21 respondents)
Comparison versus historical controls 6
Investigator assessment 14
Medical plausibility (company) 16
Comparison vs. Placebo 16
• With test of statistical significance 7
• Without test 8
Comparison vs. Active comparator 15
• With test of statistical significance 6
• Without test 6
Introspection 1
Use of significance test depends of the nature 1
of the event
Relative risk and risk over time 1

17. a. Does your company currently include in its clinical development

plans for most compounds predefined criteria for identifying potential
safety signals and strategies for the evaluation of such potential
signals?
Yes 12
No 9

b. If yes to 17.a., do they include prior assessments of population back-

ground event rates or plans to conduct such assessments if required?
Yes 9
No 3

251

group6_PH.indd 251 7.8.2007 12:20:48

 18. a. Are all AEs (including laboratory abnormalities), serious and non-
serious, across an entire development program regularly aggregated
and reviewed?

YES 16
How often?
Every 6 months 1
Yearly 6
End of each phase of development 10
At submission 8
“Depends” 6
Pre-specified points 8
Every 3 months for serious AEs 1
Quarterly 1
NO 5

b. For these reviews, if data are included from blinded studies, are
these data unblinded for the reviewer? (16 respondents)

Yes No NA
Every 6 months 0 10 2
Yearly 1 10 1
End of development phase 5 6 1
Submission 7 3 2
Depends 4 3 1
Pre-specified points 3 7 0

19. a. What new safety information does your company communicate to

investigators? (21 respondents)

Alert mailings (expedited reports to regulators) 19

Investigator brochure updates 20
Periodic updates 3
Ad hoc letters for a particular issue 1
DCSI updates 1
Serious-unexpected cases whose causality 1
is not ruled out (unblinded first;
placebo reports not sent)

252

group6_PH.indd 252 7.8.2007 12:20:49

 Method Electronic Paper Fax Phone Other
Alert mailings 5 16 7 1 3*
IB updates 0 19 0 0 3*
Periodic reports 1 3 1 0 0
* By hand delivery, site visit, or respondent uncertain how they are delivered.

b. Are these communications sent to? (21 respondents)

[Note: A company could provide more than one answer to this question, thus, the numbers add up
to more than 21. For example a company may send IB updates to all phase 1-3 investiga-
tors and alert mailings only to those under an IND, EU process or other formal regulation.]

All Investigators worldwide studying the compound 17

in Phase 1-3
All Investigators worldwide studying the compound 9
in Phase 4
Only Investigators conducting studies under an IND, 7
EU process, or other formal regulation
Only Investigators taking part in the specific devel- 1
opment program (e.g., indication)
Only Investigators involved in the study protocol 1
from which the data come

c. If periodic reports are sent to investigators (other than IB updates),

how frequently does this occur? (10 respondents)

No periodic reports 6
Yearly 1
Monthly 1
Quarterly 0
Every 6 months 0
As needed 1
Protocol specific 1
Specific issue 1
DCSI update 1

d. If single case alert reports are sent to investigators, what format is

used? (19 respondents)

253

group6_PH.indd 253 7.8.2007 12:20:49

 CIOMS I form 14
MedWatch form 9
Letter format 10
Japanese form 2

e. If single case alert reports are sent to investigators, when is this

done? (21 respondents)
Simultaneously with submission to regulators 12
Within a short time after submission to regulators 9
1-2 days but <15 days
Twice a month (in Japan)
Not defined
Within 15 days
Several days

20. What is your company policy for the frequency of updating safety
information in the Investigator Brochure (IB)? (21 respondents)
Yearly 19
When significant new information is discovered 12
Every six months 1
Quarterly 1
No standard policy 0
Start of a new Phase 1
DCSI quarterly update or IB yearly 1

21. Does your company utilize the same IB for a given product for all countries?
Yes 17
No 4

22. An issue under consideration is whether study subjects already par-

ticipating in a trial should always be informed of any new, important
safety information that the investigator receives (e.g., a new serious
ADR that is added to the investigator brochure.)
a. Does your company have a policy or practice with regard to this issue?
Yes 10
No 11

254

group6_PH.indd 254 7.8.2007 12:20:49

 b. If yes, how is this done? (10 respondents)

Company requests that investigator change the official 9

informed consent and discuss it with subject for reconsent
Rely on IRB/ERC for each study site to decide 2
Company requests investigators to inform subjects but no 0
reconsent done
Change informed consent if ADR significant 1
If significant issue added to the IB 1
Only when a serious ADR is added to DCSI 1

23. If permitted by regulation, would your company support the concept

of periodic reporting to investigators of aggregate analyses of safety
information in place of multiple expedited individual case reports?

Yes 19
No 2

24. The roles and responsibilities of Institutional Review Boards (IRBs),

Ethics Review Committees (ERCs), and Data and Safety Monitoring
Boards, and their interaction with site-investigators and study sponsors
are under increased attention by health authorities in many countries.
Has your company experienced any issues in this area, such as new
requirements for any of the following: changing/updating informed
consent on the basis of new safety data during a study, providing IRBs/
ERCs special/customized data from trials, or any other new require-
ment with regard to safety data and study progress?

Yes 15*
No 6
*
Examples of specific responses:
❏ Varies by Country
❏ Requests to update ICs, inquiries regarding specific events
❏ Changing/updating informed consent
❏ IRB specific request for updating the patient’s information judged
not relevant by other IRBs or DSMB

255

group6_PH.indd 255 7.8.2007 12:20:50

 ❏ Informed consents have been updated with significant new risk
information
❏ Some IECs making individual requests for particular data, and/or
refusing data prepared by the company
❏ More questions on individual ADRs
❏ We have had several instances of IRBs asking for detailed infor-
mation for specific case reports that they have received as part of
the investigator mailings
❏ Submit customized data to IRB
❏ Infrequent
❏ Changing informed consent, changing IB, providing specific infor-
mation to IEC
❏ Increase level of awareness of volume of SAE reports and greater
likelihood or request for overall assessment from company, con-
cerns raised by some regarding perception that may need to add
information to ICF whenever new SAE report sent to investigator
❏ General unclarity in some areas, i.e., no standardized process across
all studies

25. How does your company transmit safety information to IRBs/ERCs

while a study is ongoing? (21 respondents)
Sent directly by company 9
Sent to Investigators who are expected to forward it 12
Varies from country to country 10
Varies from study to study 2
Depends on site 1

256

group6_PH.indd 256 7.8.2007 12:20:50

 26. Has any regulator requested that your company send periodic sum-
mary reports of clinical trial safety data to investigators, IRBs/ERCs
or to the regulator (other than IND annual reports in the U.S. and under
the rules of the EU Clinical Trial Directive)?

Yes 5
Portugal and Spain
MCA for CTX
EU countries applying CT directive
No 16

257

group6_PH.indd 257 7.8.2007 12:20:50

group6_PH.indd 258 7.8.2007 12:20:50
 Appendix 4

World Medical Association

Declaration of Helsinki

ETHICAL PRINCIPLES FOR

MEDICAL RESEARCH INVOLVING HUMAN SUBJECTS
Adopted by the 18th WMA General Assembly
Helsinki, Finland, June 1964
and amended by the
29th WMA General Assembly, Tokyo, Japan, October 1975
35th WMA General Assembly, Venice, Italy, October 1983
41st WMA General Assembly, Hong Kong, September 1989
48th WMA General Assembly, Somerset West, Republic of South Africa, October 1996
and the 52nd WMA General Assembly, Edinburgh, Scotland, October 2000
Note of Clarification on Paragraph 29 added by the WMA General Assembly, Washington 2002

a. Introduction
1. The World Medical Association has developed the Declaration of
Helsinki as a statement of ethical principles to provide guidance to
physicians and other participants in medical research involving human
subjects. Medical research involving human subjects includes research
on identifiable human material or identifiable data.
2. It is the duty of the physician to promote and safeguard the health of
the people. The physician’s knowledge and conscience are dedicated to
the fulfillment of this duty.
3. The Declaration of Geneva of the World Medical Association binds
the physician with the words, “The health of my patient will be my first
consideration,” and the International Code of Medical Ethics declares
that, “A physician shall act only in the patient’s interest when provid-
ing medical care which might have the effect of weakening the physi-
cal and mental condition of the patient.”
4. Medical progress is based on research which ultimately must rest in
part on experimentation involving human subjects.
5. In medical research on human subjects, considerations related to the
well-being of the human subject should take precedence over the inter-
ests of science and society.

259

group6_PH.indd 259 7.8.2007 12:20:50

 6. The primary purpose of medical research involving human subjects
is to improve prophylactic, diagnostic and therapeutic procedures and
the understanding of the aetiology and pathogenesis of disease. Even
the best proven prophylactic, diagnostic, and therapeutic methods must
continuously be challenged through research for their effectiveness,
efficiency, accessibility and quality.
7. In current medical practice and in medical research, most prophylac-
tic, diagnostic and therapeutic procedures involve risks and burdens.
8. Medical research is subject to ethical standards that promote respect
for all human beings and protect their health and rights. Some research
populations are vulnerable and need special protection. The particular
needs of the economically and medically disadvantaged must be recog-
nized. Special attention is also required for those who cannot give or
refuse consent for themselves, for those who may be subject to giving
consent under duress, for those who will not benefit personally from the
research and for those for whom the research is combined with care.
9. Research Investigators should be aware of the ethical, legal and regula-
tory requirements for research on human subjects in their own coun-
tries as well as applicable international requirements. No national
ethical, legal or regulatory requirement should be allowed to reduce
or eliminate any of the protections for human subjects set forth in this
Declaration.

b. Basic Principles for all Medical Research

10. It is the duty of the physician in medical research to protect the life,
health, privacy, and dignity of the human subject.
11. Medical research involving human subjects must conform to generally
accepted scientific principles, be based on a thorough knowledge of
the scientific literature, other relevant sources of information, and on
adequate laboratory and, where appropriate, animal experimentation.
12. Appropriate caution must be exercised in the conduct of research
which may affect the environment, and the welfare of animals used for
research must be respected.
13. The design and performance of each experimental procedure involving
human subjects should be clearly formulated in an experimental pro-
tocol. This protocol should be submitted for consideration, comment,
guidance, and where appropriate, approval to a specially appointed

260

group6_PH.indd 260 7.8.2007 12:20:51

 ethical review committee, which must be independent of the investiga-
tor, the sponsor or any other kind of undue influence. This independent
committee should be in conformity with the laws and regulations of
the country in which the research experiment is performed. The com-
mittee has the right to monitor ongoing trials. The researcher has the
obligation to provide monitoring information to the committee, espe-
cially any serious adverse events. The researcher should also submit to
the committee, for review, information regarding funding, sponsors,
institutional affiliations, other potential conflicts of interest and incen-
tives for subjects.
14. The research protocol should always contain a statement of the ethical
considerations involved and should indicate that there is compliance
with the principles enunciated in this Declaration.
15. Medical research involving human subjects should be conducted only
by scientifically qualified persons and under the supervision of a clini-
cally competent medical person. The responsibility for the human sub-
ject must always rest with a medically qualified person and never rest
on the subject of the research, even though the subject has given con-
sent.
16. Every medical research project involving human subjects should be
preceded by careful assessment of predictable risks and burdens in
comparison with foreseeable benefits to the subject or to others. This
does not preclude the participation of healthy volunteers in medical
research. The design of all studies should be publicly available.
17. Physicians should abstain from engaging in research projects involving
human subjects unless they are confident that the risks involved have
been adequately assessed and can be satisfactorily managed. Physi-
cians should cease any investigation if the risks are found to outweigh
the potential benefits or if there is conclusive proof of positive and
beneficial results.
18. Medical research involving human subjects should only be conducted
if the importance of the objective outweighs the inherent risks and
burdens to the subject. This is especially important when the human
subjects are healthy volunteers.
19. Medical research is only justified if there is a reasonable likelihood
that the populations in which the research is carried out stand to bene-
fit from the results of the research.

261

group6_PH.indd 261 7.8.2007 12:20:51

 20. The subjects must be volunteers and informed participants in the re-
search project.
21. The right of research subjects to safeguard their integrity must always
be respected. Every precaution should be taken to respect the privacy
of the subject, the confidentiality of the patient’s information and to
minimize the impact of the study on the subject’s physical and mental
integrity and on the personality of the subject.
22. In any research on human beings, each potential subject must be ad-
equately informed of the aims, methods, sources of funding, any pos-
sible conflicts of interest, institutional affiliations of the researcher, the
anticipated benefits and potential risks of the study and the discomfort
it may entail. The subject should be informed of the right to abstain
from participation in the study or to withdraw consent to participate
at any time without reprisal. After ensuring that the subject has under-
stood the information, the physician should then obtain the subject’s
freely-given informed consent, preferably in writing. If the consent
cannot be obtained in writing, the non-written consent must be for-
mally documented and witnessed.
23. When obtaining informed consent for the research project the physi-
cian should be particularly cautious if the subject is in a dependent
relationship with the physician or may consent under duress. In that
case the informed consent should be obtained by a well-informed phy-
sician who is not engaged in the investigation and who is completely
independent of this relationship.
24. For a research subject who is legally incompetent, physically or men-
tally incapable of giving consent or is a legally incompetent minor, the
investigator must obtain informed consent from the legally authorized
representative in accordance with applicable law. These groups should
not be included in research unless the research is necessary to promote
the health of the population represented and this research cannot in-
stead be performed on legally competent persons.
25. When a subject deemed legally incompetent, such as a minor child,
is able to give assent to decisions about participation in research, the
investigator must obtain that assent in addition to the consent of the
legally authorized representative.
26. Research on individuals from whom it is not possible to obtain con-
sent, including proxy or advance consent, should be done only if the

262

group6_PH.indd 262 7.8.2007 12:20:52

 physical/mental condition that prevents obtaining informed consent is
a necessary characteristic of the research population. The specific rea-
sons for involving research subjects with a condition that renders them
unable to give informed consent should be stated in the experimental
protocol for consideration and approval of the review committee. The
protocol should state that consent to remain in the research should be
obtained as soon as possible from the individual or a legally authorized
surrogate.
27. Both authors and publishers have ethical obligations. In publication
of the results of research, the investigators are obliged to preserve the
accuracy of the results. Negative as well as positive results should be
published or otherwise publicly available. Sources of funding, insti-
tutional affiliations and any possible conflicts of interest should be
declared in the publication. Reports of experimentation not in accor-
dance with the principles laid down in this Declaration should not be
accepted for publication.

c. Additional Principles for Medical Research

Combined with Medical Care
28. The physician may combine medical research with medical care, only
to the extent that the research is justified by its potential prophylactic,
diagnostic or therapeutic value. When medical research is combined
with medical care, additional standards apply to protect the patients
who are research subjects.
29. The benefits, risks, burdens and effectiveness of a new method should
be tested against those of the best current prophylactic, diagnostic,
and therapeutic methods. This does not exclude the use of placebo, or
no treatment, in studies where no proven prophylactic, diagnostic or
therapeutic method exists.1

1
The WMA hereby reaffirms its position that extreme care must be taken in making use of a placebo-controlled
trial and that in general this methodology should only be used in the absence of existing proven therapy.
However, a placebo-controlled trial may be ethically acceptable, even if proven therapy is available, under the
following circumstances:
• Where for compelling and scientifically sound methodological reasons its use is necessary to determine
the efficacy or safety of a prophylactic, diagnostic or therapeutic method; or
• Where a prophylactic, diagnostic or therapeutic method is being investigated for a minor condition and the
patients who receive placebo will not be subject to any additional risk of serious or irreversible harm
All other provisions of the Declaration of Helsinki must be adhered to, especially the need for appropriate
ethical and scientific review.

263

group6_PH.indd 263 7.8.2007 12:20:52

 30. At the conclusion of the study, every patient entered into the study
should be assured of access to the best proven prophylactic, diagnostic
and therapeutic methods identified by the study.
31. The physician should fully inform the patient which aspects of the care
are related to the research. The refusal of a patient to participate in a
study must never interfere with the patient-physician relationship.
32. In the treatment of a patient, where proven prophylactic, diagnostic
and therapeutic methods do not exist or have been ineffective, the phy-
sician, with informed consent from the patient, must be free to use un-
proven or new prophylactic, diagnostic and therapeutic measures, if in
the physician’s judgement it offers hope of saving life, re-establishing
health or alleviating suffering. Where possible, these measures should
be made the object of research, designed to evaluate their safety and
efficacy. In all cases, new information should be recorded and, where
appropriate, published. The other relevant guidelines of this Declara-
tion should be followed.

264

group6_PH.indd 264 7.8.2007 12:20:52

 Appendix 5

Data and Safety Monitoring Boards

(DSMBs)

A Data and Safety Monitoring Board (DSMB), or Data Monitoring

Committee (DMC) is an independent committee established by the sponsor
and consists of a group of individuals with pertinent expertise that reviews
on a regular basis, accumulating data from an ongoing clinical trial. The
purpose of such a committee is to protect the safety of trial participants, the
credibility of the study and the validity of study results.1,2 DSMBs are of
value in the following situations:3
❏ large, randomized multi-center high morbidity/mortality trials;
❏ studies where data could justify early study termination or where
the design or expected data accrual is complex;
❏ early studies of a high-risk intervention;
❏ studies carried out in emergency situations in which informed con-
sent is waived;
❏ studies involving vulnerable populations; or,
❏ studies in the early phases of a novel intervention with very limited
information on clinical safety or where prior information may have
raised safety concerns.
The function of a DSMB is primarily to advise trial sponsors or their
Steering Committee about the continuation or curtailment of the trial,
recommend modifications to the study protocol and/or investigative proce-
dures, and to verify the continuing validity and scientific merit of the trial.4,5

1
Ellenberg, S.S., Fleming, T.R., DeMets, D.L. Data Monitoring Committees in Clinical Trials – A Practical
Perspective. Edited by V. Barnett. John Wiley and Sons Ltd., Chichester, England, 2002, pp. 12-15.
2
U.S. Department of Health and Human Services, Food and Drug Administration, Center for Biologics
Evaluation and Research (CBER), Center for Drug Evaluation and Research (CDER) Center for Devices
and Radiological Health (CDRH), November 2001, Guidance for Clinical Trial Sponsors, February 2004,
http://www.fda.gov/cber/gdlns/clindatmon.htm
3
UNICEF, UNDP, World Bank, World Health Organization Special Program for Research and Training in
Tropical Diseases (TDR), 31 March 2004, version 0.61, Operational Guidelines for the Establishment
and Functioning of Data & Safety Monitoring Boards (draft), http://www.acrpnet.org/chapters/belg/
whotdr_guidelines.doc
4
Slutsky, A.S. Data Safety and Monitoring Boards. The New England Journal of Medicine, 350: 1075-7083, 2004.
5
Ellenberg, S.S. Monitoring Data on Data Monitoring. Clinical Trials, 1: 6-8, 2004.

265

group6_PH.indd 265 7.8.2007 12:20:53

 Recommendations made by the DSMB are based on a sequential benefit-
risk assessment;6 consideration is given to whether the potential benefits of
the investigational intervention have been established or whether the risks
appear greater than previously anticipated.
A DSMB should comprise a defined scientific membership with ap-
propriate expertise for the task at hand, and where possible, should repre-
sent the cultural territories involved in the studies.7,8 DSMBs should have a
clear scientific remit and this should be clearly outlined in appropriate doc-
umentation sometimes referred to as a Charter or Terms of Reference.9,10
They may also be required to follow standard operating procedures that
may address issues such as conflicts of interest, confidentiality, remuner-
ation, member replacements and co-options.11 The Board should ideally
have a clear written review and communication policy that relates to the
review of blinded, partially-blinded and unblinded data and how this may
be disclosed within sessions. Some of these may be open to the sponsor
(open sessions) or held entirely in closed sessions.12 Partially-blinded and
unblinded data should not be available to the sponsor or the Steering Com-
mittee members unless this should be required to justify appropriate action
to modify or terminate a study. 13 In such cases, only the minimum infor-
mation to support such recommendations should be released until the data
integrity is secured.14
In certain trials, interventions can exert effects both on morbidity and/
or mortality, or may reduce the risk of a major adverse health outcome,
e.g., cardiovascular events, recurrence of cancer. DSMBs are used in trials

6
Committee for Proprietary Medicinal Products (CPMP) (2004), Concept Paper on the Development of a
Committee for Proprietary Medicinal Products (CPMP), Points to Consider on Data Monitoring Committees,
The European Agency for the Evaluation of Medicinal Products, February 2004, http://www.emea.eu.int/
pdfs/human/ewp/245902en.pdf
7
Ellenberg, S.S., Fleming, T.R. and DeMets, D.L. Data Monitoring Committees in Clinical Trials – A Practical
Perspective. Edited by V. Barnett. John Wiley and Sons Ltd., Chichester, England, 2002, pp. 45-55.
8
Walker, A.E.. and McLeer, S.K.. Small Group Processes Relevant to Data Monitoring Committees in Con-
trolled Clinical Trials. Clinical Trials, 1 (3): 282-296, 2004.
9
Hemmings, R. and Day, S. Regulatory Perspectives on Data Safety Monitoring Boards: Protecting the Integ-
rity of Data. Drug Safety. 27 (1): 1-6, 2004.
10
Ellenberg, S.S., Fleming, T.R., DeMets, and D.L. Data Monitoring Committees in Clinical Trials – A Practical
Perspective. Edited by V. Barnett. John Wiley and Sons Ltd., Chichester, England, 2002, 175-183.
11
Sydes, M.R., et al. Systematic Qualitative Review of the Literature of Data Monitoring Committees for Ran-
domized Controlled Trials. Clinical Trials. 1: 60-79, 2004.
12
O’Neil, R.T. Regulatory Perspectives on Data Monitoring, Statistics in Medicine, 21: 2831-2842, 2002.
13
Fleming, T.R. , Ellenberg, S. and De Mets, D.L. Monitoring clinical trials: issues and controversies regarding
confidentiality, Statistics in Medicine, 21: 2843-2851, 2002.
14
Ellenberg, S.S., Fleming, T.R. and DeMets, D.L. Data Monitoring Committees in Clinical Trials – A Practical
Perspective, Edited by V. Barnett. John Wiley and Sons Ltd., Chichester, England, 2002, 74-86.

266

group6_PH.indd 266 7.8.2007 12:20:53

 where interim monitoring of the study data is essential to protect clinical
trial subjects. Wherever possible, they should always be independent of the
sponsor, whether or not trials fulfill a commercial or non-commercial re-
mit. Thus, due to the independent nature of these committees, an objective
review of the interim data for any emerging concerns can be assured.
Study oversight by way of a DSMB provides further protection of
study participants over and above the pharmacovigilance functions tradi-
tionally undertaken by sponsors adhering to international standards. A trial
that is large, of long duration, and multi-center raises a greater potential for
concerns about safety due to greater overall treatment exposures and be-
cause this may be associated with adverse effects not previously identified
in shorter duration clinical development programs.
DSMBs are most frequently established for controlled trials that have
mortality and major morbidity as primary or secondary outcomes, since
monitoring of accumulating data may be critical in such trials. As part of
its function, the DSMB should evaluate the accuracy and timeliness of the
data, study recruitment and whether it is adequate to answer the questions
in the appropriate timeframe. It should also ensure that the reporting of the
patient records, adverse events and study endpoints is done in accordance
with the protocol.
It is important that DSMBs keep abreast of changes in the external
environment (for example, reporting of efficacy and safety data from other
relevant large trials). In addition, DSMBs must have access to data that ar-
eas up-to-date as possible, especially when recommendations on stopping
a trial might have to be made. Sponsors should ensure that the DSMB is
kept updated (particularly during the open sessions at meetings) about the
evolving benefit and risk profile of the drug under development.
DSMBs should formulate the rules for monitoring the trial before
reviewing any study data on treatment effects. Such rules should include
statistical boundaries for stopping of the trial. These stopping boundar-
ies should generally be asymmetrical – a less conservative boundary for
adverse effects than for beneficial treatment effects. It should be noted,
however, that statistical stopping boundaries are not absolute rules, but
are guidelines to aid the DSMB in its deliberations.15,16 Results from other

15
Kiri, A. , Tonascia, S. and Meinert, C.L. Treatment effects monitoring committees and early stopping in large
clinical trials, Clinical Trials, 1: 40-47, 2004.
16
Wilhelmsen, L. Regulatory Perspectives on Data Monitoring Committees (DSMC), Statistics in Medicine,
21: 2823-2829, 2002.

267

group6_PH.indd 267 7.8.2007 12:20:54

 trials as well as other circumstances can offset the exact statistical stop-
ping rules. In certain situations, a negative trend (not just a clear benefit or
clear harm), that rules out a clinically important benefit if the study should
continue to its scheduled end, may be sufficient for recommendation of
early termination, as such a trend may demonstrate that the probability of
documenting a benefit of the treatment under study is extremely low. There
may be situations in a trial where the primary and secondary endpoints do
not consistently favor the same treatment during a trial. The primary end-
point may indicate a benefit and one or several of the secondary endpoints
may indicate adverse effects. Caution must be exercised to ensure that trials
are not terminated prematurely, and particularly in these circumstances, a
balanced view on the safety issues is important.
It is recommended that DSMBs should also consider efficacy data
evaluation,17 and not limit their evaluation to safety data. The lack of ef-
ficacy as it relates to safety may pose a risk to clinical trial participants.
For example, in certain sub-populations within the clinical trial, evalu-
ation of efficacy data (and the lack of efficacy thereof), would result in
the modification of the patient population enrolled, and therefore reduce
the risk of enrolling the types of patients where interim evaluation has not
shown a favorable benefit-risk relationship. Relative to the above, certain
adverse event patterns might be acceptable if the treatment were effective,
and might be completely unacceptable if the observed data were suggesting
lack of benefit.
Recommendations about early trial termination or continuation must
be based on a global consideration of all available data from the trial in-
cluding information on primary and secondary efficacy measures, adverse
effects, and quality of trial conduct, along with relevant information exter-
nal to the trial.
Safety monitoring in long-term outcome studies can be particularly
challenging for DSMBs. For example, in such studies the primary efficacy
endpoint can have serious safety implications. If an early interim review
of the data suggests participants treated with the investigational product
are at a higher risk for the outcome of interest than those in the control
arm, the DSMB may consider recommending early termination on safety
grounds. Such assessments have potential implications for falsely conclud-
ing that there is an adverse effect. Statistical considerations for early stop-

17
Ellenberg, S.S., Fleming, T.R. and DeMets, D.L. Data Monitoring Committees in Clinical Trials – A Practical
Perspective. Edited by V. Barnett. John Wiley and Sons Ltd., Chichester, England, 2002, pp. 93-94.

268

group6_PH.indd 268 7.8.2007 12:20:54

 ping should be considered, and it is usually appropriate to demand less
rigorous proof of harm to justify early termination. In some cases, how-
ever, it may be appropriate to establish a harmful effect more definitively
– for example, if a positive effect on the primary endpoint has been demon-
strated or appears to be emerging, a precise assessment of a negative trend
on a potentially important safety endpoint may be required for benefit-risk
considerations. Another example would be when the product being tested
is already in wide use and more definitive data may be required to support
practice change recommendations.
The DSMB is also responsible for the interim review of serious and
occasionally non-serious adverse events observed in the study, such as sig-
nificant trends in laboratory parameters that may portend serious conse-
quences. The sponsor may be requested to provide the DSMB with sum-
maries of specific or all adverse events observed.18 This is particularly
important when the event may result from the disease being treated as well
as the study intervention itself. If an imbalance between groups emerges,
concerns will arise that the adverse event may be due to the intervention
rather than the disease itself. Since a potentially large number of adverse
event categories may be observed and compared between the study arms,
the interpretation of safety findings by the DSMB must be sensitive to the
issues of multiplicity.
Although a DSMB should always review summary adverse event data,
it will not usually review in detail every adverse event reported, or even ev-
ery serious adverse event. This responsibility generally lies with the sponsor
who reviews such events promptly and has the responsibility for reporting
and communicating this information appropriately. The involvement of a
DSMB in the review of individual adverse event reports will vary from case
to case. The DSMB should always be prepared to review any individual
event thought to be of major significance by the sponsor or study’s medical
monitor. The DSMB should learn in a timely manner of any cases for which
unmasking of treatment code at the clinical site or by the treating clinician
is thought to be necessary to provide an appropriate intervention.

18
The DSMB may define specific serious or non-serious events which it may consider important for it to monitor.
These might involve known background morbidity (serious adverse events) or known serious expected adverse
drug reactions (as identified in the Investigators Brochure or product label). Although some regulatory authori-
ties may grant ‘waivers’ for the expedited reporting of serious related and unexpected possibly drug-related
events, the existence of a DSMB does not automatically preclude the need for reporting in all countries (the EU
for example, where ‘SUSARS’ are considered to be generally unpredictable). Where such serious events are not
addressed in the protocol, Investigator’s Brochure or DSMB charter, they are still required to be reported to the
regulatory authorities within the standard 7 or 15 day time frame). A careful consideration of these documents
can therefore minimise the need to unblind critical trial data when a DSMB is involved.

269

group6_PH.indd 269 7.8.2007 12:20:54

 DSMBs are less likely to be established for small, short-term studies
of interventions to relieve symptoms.19 The need for an outside group to
regularly monitor data to consider questions of early stopping for efficacy
or protocol modification is less compelling in this situation. For such prod-
ucts, however, an expert group to oversee all studies at all stages of develop-
ment, monitor the developing safety database and make recommendations
for the design of successive studies based on early results may be useful.
Such a group may be particularly valuable when the patient population is
at relatively high risk of serious events. The external group would indepen-
dently evaluate individual events and overall event rates in ongoing studies
and advise the sponsor of emerging concerns – monitoring considerations
of this type are clearly more clinical than statistical.

19
Ellenberg, S.S. Monitoring Data on Data Monitoring, Clinical Trials. 1: 6-8, 2004.

270

group6_PH.indd 270 7.8.2007 12:20:55

 Appendix 6

Data Elements
that Should Be Considered for
Individual Adverse Event Reports

It is difficult to decide in advance exactly what specific data elements

are necessary or sufficient to include on case record forms or serious AE
report forms for adverse events occurring during clinical trials. ICH Guide-
lines E2A and E2B have indicated the most common data elements for
routine data collection, as well as those that might be useful, especially for
serious or special adverse events. The goal in collecting safety data is to
learn as much about the drug as possible without overburdening investiga-
tors with unnecessary requests. Ultimately, the data will be converted into
product information (package insert, SPC, etc.) that will be important for
practitioners in advising their patients in the use of the product. A simple
example of the kind of detail that can be useful is the following adverse
reaction: early onset of nausea that disappears after one week on the drug.
CIOMS Working Group V developed a recommended set of data ele-
ments that should be sought during follow-up of post-marketing, mostly
spontaneous, cases. For that purpose, the elements were prioritized ac-
cording to three categories of cases: non-serious expected cases; serious-
expected and non-serious unexpected cases; and serious unexpected and
“special interest” cases.*
For safety monitoring during clinical trials, however, in practice it is
desirable that most or all the data elements recommended be collected; if
initially missing from the CRF, attempts should be made to determine them
through follow-up. This is particularly important for suspected serious and
special interest cases. The list given here is consistent with the entries in the
sample serious AE reporting form in Appendix 8.
It is worth noting that in processing clinical trial safety data, non-serious
AE cases are usually not examined in detail or analyzed until study-end

*
See Current Challenge in Pharmacovigilance:Pragmatic Approaches, Report of CIOMS Working Group V
(2001). Council for International Organizations of Medical Sciences, Geneva, 2001.

271

group6_PH.indd 271 7.8.2007 12:20:55

 (or perhaps during an interim analysis). At that stage, all relevant CRF and
other data will be incorporated into the analysis and evaluation (AE, demo-
graphic, etc.). It is thus critical that systems be in place to extract and merge
data from the sponsor’s various computer records and databases.
The recommended list of data elements given below is based on the
CIOMS V list. Some of the elements have been modified, or new ones added
to make them more pertinent for clinical trial safety monitoring (shown
in italics). Several of the data elements listed are administrative in nature
and would be collected for any clinical trial for general use (core study
information), not for safety monitoring alone (e.g., the first 8 in the list).
The other types of data elements might be classified as (a) core AE data
(typical entries in a CRF), and (b) data for serious AEs or AE’s of special
interest (e.g., additional information that might be found on serious AE
reporting forms used by investigators). Many if not most of the same data
elements often occur on both the CRF and a special form for serious AEs.
However, the data elements listed below are not sorted in any special way
or presented in any order of priority or importance, since that will depend
on the situation. It is a check-list for consideration but should not be con-
sidered exhaustive.
It is useful to recall that the minimum data elements (four) necessary
to qualify a clinical trial (or spontaneous) case as valid and potentially
reportable to drug regulators are: identifiable patient, identifiable reporter,
one or more adverse reactions (or outcome, such as death), and a suspect
drug.
❏ Country of occurrence
❏ An identifiable reporter
❏ An identifiable patient
❏ Patient demographics (e.g., age, sex, body weight, height)
❏ Source type (physician or other medical professional)
❏ Study drug or drugs (name or code, as appropriate)
❏ Study code or protocol number
❏ Setting (e.g., hospital, outpatient clinic, home, nursing home)
❏ Daily dose of suspected medicinal product and regimen
❏ Route of administration
❏ Indication(s) for which suspect medicinal product is administered
❏ For concomitant medications: Daily dose and regimen
❏ Starting date of trial treatment (and if relevant, time of day of treat-
ment, e.g., for an acute hypersensitivity reaction)
❏ Stopping date and time of treatment, or total duration of treatment

272

group6_PH.indd 272 7.8.2007 12:20:55

 ❏ One or more adverse events (investigator’s verbatim term(s))
❏ Severity of AE (mild, moderate, or severe)1
❏ Any action taken as a result of the AE (discontinuation from trial,
dose reduction, etc.)
❏ If serious, criterion or criteria for regarding the case as serious
❏ Full description of reaction(s) including body site and severity
❏ Starting date of onset of reaction (or time to onset)
❏ If date of onset not available, best available date or treatment duration
❏ Time lag from end of treatment if ADR occurred after cessation of
treatment
❏ Date AE disappeared/ended; if date not available, duration of AE
❏ Patient outcome (at case level and, when possible, at event level);
should include information on recovery and any sequelae
❏ Dechallenge information (if any)
❏ Rechallenge information (if any)
❏ Other etiologic information
❏ For a fatal outcome, cause of death and a comment on its possible
relationship to the suspected reaction(s)
❏ Any autopsy or other post-mortem findings (and/or indication if
autopsy report and/or death certificate is available)
❏ Causal relationship assessment by the investigator 2
❏ Is the AE present/ongoing in the patient at the end of the trial (or
after the patient’s last dose of drug)
❏ Specific tests and/or treatment required as a result of AE, and their
results
❏ Whether or not the hospital discharge summary is available if the
patient was hospitalized
❏ Anything relevant to facilitate assessment of the case such as medi-
cal history (especially concurrent disease(s)), relevant drug history
including allergies, drug or alcohol abuse, family history, pregnancy

1
Some study sponsors may use a different gradation of severity, or none at all.
2
Some sponsors may require this for non-serious events, or for adverse events of special interest, as well as for
serious events.

273

group6_PH.indd 273 7.8.2007 12:20:56

group6_PH.indd 274 7.8.2007 12:20:56
 Appendix 7

Causality Criteria
and Threshold Considerations
for Inclusion of Safety Data in
Development Core Safety Information
(DCSI)

Whether assessing an individual AE case or a series of related cases

for causality, there are basic criteria that can brought to bear on the as-
sessment. Aside from regulatory reporting criteria, causality assessments
contribute to the information needed to decide when the threshold has been
reached for adding new adverse reactions and other safety data to the rel-
evant reference safety document, in this case the Investigator’s Brochure
(IB) or Development Core Safety Information (DCSI).
The following lists are derived from three key documents: (1) Guide-
lines for Preparing Core Clinical-Safety Information on Drugs, Second
Edition, Including New Proposals for Investigator’s Brochures, Council
for International Organizations of Medical Sciences, Geneva, 1999, p. 29
(CIOMS Working Group III/V), (2) FDA’s Clinical Review Template
(CDER, Office of the Center Director), Section 7.0 Integrated Review of
Safety (http:/www.fda.gov/cder/mapp/6010.3pdf; effective 9 July 2004),
and (3) the more detailed Reviewer Guidance: Conducting a Clinical Safety
Review of a New Product Application and Preparing a Report on the Re-
view (January 2005; see http://www.fda.gov/cder/guidance/3580fnl.pdf).
The CIOMS III/V Group proposed a ranking of importance for the criteria
related to causality, based on a survey of its members; however, that work
dealt mainly with post-marketing AE reports. Although it is tempting to
prioritize the various items (which ones are the most important or most
convincing), such an ordering may depend on the specific product develop-
ment program and trial conditions.
Furthermore, not all reports of a given event will necessarily meet
all, or the same, criteria. Rather, multiple reports will manifest a spectrum
of applicable causality criteria of varying strengths. Clinical judgment is
essential to assess the weight of evidence. An important factor in this

275

group6_PH.indd 275 7.8.2007 12:20:56

 judgment is to ensure that fewer and less stringent criteria for including
new ADR information in the IB/DCSI are applied for events that might
have a significant adverse outcome for the trial participants.
Although the lines of evidence listed below are useful for either (or
both) individual case or aggregate data causality assessments, the CIOMS
Working Group strongly believes that whenever possible, analysis of aggre-
gate data should be used for robust determination of product-event relation-
ships and changes in the safety profile.

Evidence from Individual Cases

Positive rechallenge
Definitive (i.e., clearly defined, well documented specific case histories)
Time to onset plausible
Positive dechallenge
Lack of confounding risk factors
Amount and duration of exposure consistent/plausible with cause and effect
Corroboration of the accuracy of the case history
Case clear-cut, easily evaluated
Co-medication unlikely to play a role
Investigator’s causality assessment
Lack of alternative explanation

Evidence from Multiple Cases

Positive outcome in targeted safety study(ies)
Consistently higher incidence vs placebo or active comparator (whether
statistically significant or not)
Positive dose-response (fixed or escalating dose studies)
Higher incidence vs comparator(s) of event-specific patient discontinuations
Earlier onset and/or greater severity in active vs comparator group(s)
Consistency of pattern of presenting symptoms
Consistency of time to onset
Consistent trends across studies
Consistent pattern of clinical presentation and latency

Previous Knowledge of AE or Drug/Class, Including Metabolites

Recognized consequence of overdose
Rarity of event in comparable untreated populations or indications
Event is commonly drug-related (e.g., neutropenia, Stevens-Johnson Syndrome)

276

group6_PH.indd 276 7.8.2007 12:20:57

 Pharmacokinetic evidence (e.g., interactions)
Known mechanism
Recognized class effect
Similar findings in animal or in vitro models
Closeness of drug characteristics to those of other drugs known to cause the AE

277

group6_PH.indd 277 7.8.2007 12:20:57

group6_PH.indd 278 7.8.2007 12:20:57
 Appendix 8

Sample Serious AE Report Data

Collection Form for Investigators

COMMENTARY:
The CIOMS Working Group is not proposing this form as a standard.
It is a format and content presented for possible consideration by those
who would like a sample template. When developing any form for AE data
collection, careful consideration should be given to what are the necessary
and desirable data elements for the situation. For example, the form below
does not include a field for “race”, a concept that is subject to debate and
possible privacy/confidentiality restrictions, but might be pertinent in cer-
tain settings. Also, please note that the choices for causal relationship to
study treatment in section C.3 (“No reasonable possibility” or “Reasonable
possibility”) are preferred by some to the more traditional “Not related” or
“Related”; the former terms imply more judgment and less certainty, which
comports with the known difficulty of establishing causality for individual
cases; see chapters 4 and 7 for more discussion. Any form used should be
designed to support electronic regulatory reporting of cases, especially un-
der ICH Guideline E2b. See Appendix 6 of this report for more discussion
of specific data elements.

279

group6_PH.indd 279 7.8.2007 12:20:57

 A. STUDY INFORMATION

280

group6_PH.indd 280
1. Study Drug Name/Code: 2. Protocol No: 3. p Initial 4. Country:
p Follow up
B. SUBJECT INFORMATION
1. Initials: 3. Age: 5. Sex: 6. Weight: 7. Height: 8. Pregnancy: 10. At the onset of initial AE:
p Male pkg pcm p Yes p No p Outpatient
2. Number: 4. Date of Birth: p Female plb pin 9. ( ) wks p Inpatient
(MM/DD/YYYY) 11. Inpatient Admission date:
/ / (MM/DD/YYYY)
/ /
13. Date of AE diagnosis:
12. Indication for use
(MM/DD/YYYY) / /

14. Concom. Disease

15. Relevant Past

Medical History,
including allergies/
risk factors

7.8.2007 12:20:57
 C. EVENT INFORMATION

group6_PH.indd 281
1. Event(s) 2. Onset date 3. Severity *4. Seriousness 5. Outcome 6. Dechallenge 7. Rechallenge 8. Relationship to **9. Possible
(MM/DD/YYYY) (See below) (Date: MM/DD/YYYY) If study drug If study drug Study Drug Cause of SAE
stopped, did restarted, other than
event improve did event study drug
or disappear? reappear? (see below)

p Mild p1 p5 (Date: / / ) p Yes p Yes p No reasonable p1 p5

p Moderate p2 p6 p Recovered p Death p No p No possibility p2 p6
p Severe p3 p7 p Not recovered p Not applicable p Not applicable p Reasonable p3 p7
/ /
p4 p8 p Unknown p Unknown p Unknown possibility p4 p8
p Recovered Rechallenge Dose
with sequelae ( )

p Mild p1 p5 (Date: / / ) p Yes p Yes p No reasonable p1 p5

p Moderate p2 p6 p Recovered p Death p No p No possibility p2 p6
p Severe p3 p7 p Not recovered p Not applicable p Not applicable p Reasonable p3 p7
/ /
p4 p8 p Unknown p Unknown p Unknown possibility p4 p8
p Recovered Rechallenge Dose
with sequelae ( )

p Mild p1 p5 (Date: / / ) p Yes p Yes p No reasonable p1 p5

p Moderate p2 p6 p Recovered p Death p No p No possibility p2 p6
p Severe p3 p7 p Not recovered p Not applicable p Not applicable p Reasonable p3 p7
/ / p4 p8 p Unknown p Unknown p Unknown possibility p4 p8
p Recovered Rechallenge Dose
with sequelae ( )
*4: Seriousness: Check the applicable numbers in the above “Seriousness” box
1. Death 2. Life-threatening 3. Involved hospitalization 4. Prolonged hospitalization 5. Involved persistent disability 6. Congenital anomaly
7. Important medical event (e.g., seizure, blood dyscrasias) 8. Non-serious (does not meet above criteria)
**9: Possible cause of serious AE other than study drug: If “No reasonable possibility” checked in box C.8 (Relationship to Study Drug), specify other possible cause of AE.
1. Disease under study 2. Medical condition(s) 3. Lack of efficacy/worsening of treated condition 4. Study drug withdrawal-effect
5. Concomitant or previous medication 6. Erroneous administration of treatment 7. Protocol-related procedure 8. Other, specify.

281

7.8.2007 12:20:58
 C. EVENT INFORMATION (continued)

282

group6_PH.indd 282
9. Description of the event: Include signs, symptoms and treatment of the event, as well as any relevant laboratory data.
Whenever possible, assign a diagnosis supported by signs and symptoms.

10. Did the AE result in discontinuation of the subject from the study? p No p Yes
11. Date of dropout: (MM/DD/YYYY / / )
If outcome of event was death: 12. Date of death: (MM/DD/YYYY) / / 13. Cause of Death: ( )
14. Autopsy: p No p Yes 15. Autopsy date: (MM/DD/YYYY) / /
16. Autopsy results: : Details Attached? p No p Yes
17. Has the subject previously experienced this sign(s)/symptom(s)/disease(s) during the study? p No p Yes
18. If so, provide details:
19. Has the subject experienced this sign(s)/symptom(s)/disease(s) prior to this study? p No p Yes
20. If so, provide details:
21. Has a serious AE report been made previously for this patient? p No p Yes
22. If so, for what event(s) and when? ( )
21. Date of AE occurrence: (MM/DD/YYYY / / )
23. Hospital discharge summary attached? p No p Yes p Not applicable

7.8.2007 12:20:58
 D. DRUG INFORMATION

group6_PH.indd 283
1. Drugs 2. Total 3. Frequency 4. Route Date of administration 7. Indication 8. Is drug
daily dose (Dose regimen; 5. Start date 6. Treatment stop date suspected
(include if IV, duration (MM/DD/YYYY) (MM/DD/YYYY). If study to cause
units) of infusion) (For IV, include treatment continuing AE?
time of day for write C; if IV, duration of Yes (Y) or
infusion) infusion) No (N)
Study treatment: This information is
(Lot: ) already captured
in Box A.12.

(Lot: )

(Lot: )
Other drugs
(tradename/generic name):

283

7.8.2007 12:20:58
 E. LABORATORY TESTS RESULTS (Please attach lab data or describe here)

284

group6_PH.indd 284
(Specify if results relate to Baseline, During administration, At the time of AE, follow-up to AE, etc.)

F. REPORTING INVESTIGATOR’S COMMENT ON CAUSALITY

Explain your reasoning for attributing the event(s) to the cause chosen.

7.8.2007 12:20:59
 G. INFORMATION ON REPORTING INVESTIGATOR

group6_PH.indd 285
1. Name: 2. Title: 3. Specialization: 4. Location (study site):

5. Address: 6. Telephone:

7. Investigator’s signature: 8. Date signed: (MM/DD/YYYY)

/ /

FOR COMPANY USE ONLY

a. Date this report initially received: (MM/DD/YYYY) b. Database No: c. Local No:

285

7.8.2007 12:20:59
group6_PH.indd 286 7.8.2007 12:20:59
 Appendix 9

Databases for Epidemiology

and Pharmacoepidemiology
The list of database sources shown below is derived from a standard list
under continuous development by the International Society for Pharmacoepidemiology
and is provided with their permission. Some URLs were not available.
(see http//www.pharmacoepi.org/resources/summary_databases.pdf)

For additional information on various databases, see BRIDGE

(Benefit-Risk Information for Drug Evaluations) at www.dgiinc.org.

Database Resources - Databases

Country URL
(List contributed by IPSE members)
British Columbia Healthcare Utilization Canada http://www.gov.bc.ca/healthservices
Population Research Unit Canada http://www.phru.medicine.dal.ca
Saskatchewan Health Databases Canada http://www.health.gov.sk.ca/
Odense University Pharmacoepidemiologi- Denmark http://www.sdu.dk/health/research/units/
cal Database (OPED) clinpharm.php
Pharmacoepidemiological Prescription Denmark http://www.clin-epi.dk
Databases of North Jutland (PDNJ)
Finland Medical Record Linkage System Finland
PEDIANET Italy http://www.pedianet.it
Sistema Informativo Sanitario Regionale Italy
Database-FVG Region (FVG)
Health Insurance Review Agency Database Korea http://www.hira.or.kr
Integrated Primary Care Information Netherlands http://www.ipci.nl
Database
InterAction Database (IADB) Netherlands
PHARMO Records Linkage System Netherlands http://www.pharmo.nl
Medicines Monitoring Unit (MEMO) Scotland http://www.dundee.ac.uk/memo
Primary Care Clinical Informatics Unit- Scotland http://www.abdn.ac.uk/general_practice/research/
Research (PCCIU-R) special/pciu.shtml
Base de datos para la Investigacion Farma- Spain http://www.bifap.org/
coepidemiologica en Atencion Primaria
(BIFAP)
Swedish Centre for Epidemiology Sweden http://www.sos.se/epc/epceng.htm#epid
General Practice Research Database (GPRD) UK http://www.gprd.com/
GPRD through Boston Collaborative Drug UK http://www.bcdsp.org
Surveillance Program (BCDSP)
IMS Disease Analyzer (MediPlus) UK http://research.imshealth.com
Prescription Event Monitoring (PEM) UK http://www.dsru.org/main.html
Database
The Health Improvement Network (THIN) UK http://www.epic-org.

287

group6_PH.indd 287 7.8.2007 12:20:59

 BRIDGE Database of Databases US / Europe http://www.dgiinc.org/html/frameset.htm
Case-Control Surveillance Study US http://www.bu.edu/slone/
Constella He s US http://www.constellagroup.com/health_sciences/
Framingham Heart Study Database US http://www.nhlbi.nih.gov/about/framingham/
index.html
Group Health Cooperative of Puget Sound US http://www.centerforhealthstudies.org/
Harvard Pilgrim Health Care US http://www.harvardpilgrim.org
Healthcare Cost & Utilization Project (HCUP) US http://www.ahrq.gov/data/hcup/
Healthcore (Wellpoint/Blue Cross/Blue US http://www.healthcore.com
Shield)
Henry Ford Health Systems (HFHS) US http://www.henryfordhealth.org
HMO Research Network (HMORN) US http://www.hmoresearchnetwork.org
IMS LifeLink US
IMS National Disease and Therapeutic US http://www.imshealth.com/ims/portal/front/articleC/
Index 0,2777,6599_44000160_44022368,00.html
United Health Care - Ingenix Epidemiology US http://www.epidemiology.com
Integrated Healthcare Information Solutions US http://www.ihcis.com/information_services/
(IHCIS) National Managed Care Benchmark databases/
Kaiser Permanente Medical Care Programs US http://www.dor.kaiser.org/
Kaiser Permanente Northwest US http://www.kpchr.org/public/studies/studies.aspx
Lovelace Center for Pharmacoeconomic and US http://www.lrri.org/cr/cpordata.html
Outcomes Research (CPOR)
Managed Care Database
MarketScan US http://www.medstat.com/1products/marketscan.asp
US Medicaid and Medicare Databases US http://www.cms.hhs.gov/data/contacts.asp
Medical Expenditure Panel Survey (MEPS) US http://www.ahrq.gov/data/mepsix.htm
National Ambulatory medical Care Survey US http://www.cdc.gov/nchs
National Death Index US http://www.cdc.gov/nchs/r&d/ndi/ndi.htm
National Health and Nutrition Examination US http://www.cdc.gov/nchs/nhanes.htm
National Health Care Survey US http://www.cdc.gov/nchs/nhcs.htm
National Health Interview Study US http://www.cdc.gov/nchs/nhis.htm
National Hospital Discharge Survey US http://www.cdc.gov/nchs/about/major/hdasd/nhds.htm
National Natality Survey US http://www.cdc.gov/nchs
National Nursing Home Survey US http://www.cdc.gov/nchs/about/major/nnhsd/nnhsd.
htm
NDC Health’s Intelligent Health Repository US http://www.ndchealth.com/index.asp
Nurses Health Study US http://www.channing.harvard.edu/nhs/
PharMetrics US http://www.pharmetrics.com
Pregnancy Health Interview Study US http://www.bu.edu/slone/
Slone Survey US http://www.bu.edu/slone/
Solucient Databases US http://www.solucient.com/solutions/Solucient_
Databases.shtml
Surveillance Epidemiology & End Results US http://seer.cancer.gov/
Vaccine Safety Datalink US http://www.cdc.gov/nip/vacsafe/
Veterans Administration Databases US http://www.virec.research.med.va.gov

288

group6_PH.indd 288 7.8.2007 12:20:59

 Index

A aggregate assessment of data 84

“adverse events of special interest” 27, aggregate data 28, 77, 83, 84, 86, 100,
88, 100, 198 104, 113, 114, 115, 120, 122, 128,
“always serious” events 100, 200 184, 196, 197, 202, 214, 276
abbreviations 27, 218, 219 aggregate data analysis 28, 120, 202
absence of harm 21 Agranulocytosis 72
absolute and relative risk information 162 analysis and interpretation of safety data
absolute and relative safety 131 113
absolute difference in risk 146 analysis dataset models 105
Absolute Risk 219, 234, 235 Analysis of Covariance (ANCOVA) 140,
221
absolute risk difference 235
academic clinical researchers 33 analysis of extreme values 140
acceptable benefit-risk profile 23, 116 Analysis of Rare Events 10, 160
acceptable level of risk 127, 201 analysis of small trials 143
acceptable risk 116, 127, 201, 219 analysis plan 133, 134
accessibility of data 8, 60 Animal studies 81
access by the public 49 animal studies 88, 112
accreditation of IECs 42 animal toxicology 60, 128
acronyms 218 annual DSUR 181, 212, 213
ACRP 43 annual IND report 61, 180
action plans 64, 66, 73, 113 anonymization of data 45
adding information to the DCSI 128, 188 antibodies 71, 72
adjustment for multiple comparisons 145 anticipated risks 60, 65
ADROIT database 69 aplastic anemia 72
Adverse Drug Reaction (ADR) 219 appropriate statistical techniques 113, 131
Adverse Event/Adverse Experience 220 approving clinical trials 42
adverse events of special interest 9, 27, ARENA 43
65, 82, 85, 88, 89, 94, 96, 100, 118, assessment of laboratory data 125
121, 122, 197, 198, 202, 220, 273 Association for the Accreditation of
adverse event tables 99, 200 Human Research Protection
advisory board 62 Programs (AAHRPP) 42
Advisory bodies 8, 62 auditing of ethics committees 42
Ad hoc reporting of individual case audits 40
safety reports 168 autonomy 37
AERS database 69
AE of special interest 65
Agency review of safety data 136 B
aggregate analysis 10, 68, 85, 100, 127, background data 8, 29, 60, 161, 209, 210
131 background incidences 67

289

group6_PH.indd 289 7.8.2007 12:21:00

 background rate 67, 68, 69, 100, 112, causality 26, 28, 79, 82, 83, 84, 85, 86,
114, 122, 133, 142, 145, 161, 176, 100, 118, 119, 120, 141, 161, 163,
194, 205, 212 170, 179, 183, 275, 276, 279
background rate of morbidity or mortality causality assessments 79, 82, 84, 85,
112 119, 249, 250, 275, 276
Bayesian 118, 134, 148, 221 Causality Criteria 12, 275
Bayesian approach 134, 148 causality method 26
beneficence 37, 38 causal association 119, 170, 220, 250
Benefit-Risk 10, 15, 33, 70, 127, 225, causal relationship cannot be ruled out
234, 237, 287 170, 171, 211
benefit-risk-cost analyses 32 CCSI 16, 99, 100, 117, 119, 120, 128,
172, 183, 186, 187, 222, 224, 226
benefit-risk assessment 23, 114, 266
Censored 222
benefit-risk balance 132, 162, 203
censoring 137, 222
benefit-risk profile 22, 23, 58, 60, 65, 70,
Censoring of Data 222
109, 115, 116, 177, 201
centralized IRBs 41, 42
benefit-risk relationship 21, 27, 33, 95,
187, 191, 224, 268 central database 50
central laboratories 94
benefit-risk weighing 33, 127, 191
changes in laboratory values 125
benefit and/or risk advantages 64
Chi-square 222
between-group differences 123
chi-square distribution 155, 222
Binary Analysis 221
chi-square test 146, 148, 155, 160, 164,
binary data 10, 29, 139, 142, 164, 231
222, 227, 239
binary measures 140, 207, 234
chronology bias 155
biochemistry panels 88
CIOMS International Ethical Guidelines
biotechnology products 21, 32, 82, 195, 187
217 CIOMS I Working Group 15, 167
Blinded clinical trials 172 CIOMS VI Survey 7, 25
blinded data 43 class-specific issues 66
blinded studies 28, 252 class effect 277
Bone marrow toxicity 8, 72 clinically relevant outcomes 139
Bonferroni correction 144, 221, 238 clinical concepts 101
BRIDGE 287, 288 Clinical Data Interchange Standards
burden of disease 64 Consortium (CDISC) 105
business processes 55 Clinical Description of Adverse Events
9, 98
Clinical Development Plan 56, 63
C clinical efficacy endpoints 28
cannot-be-ruled-out 85, 197 clinical efficacy outcomes 89, 198
Cardiac electrophysiology 8, 70 clinical endpoint data 90
carryover effects 135 clinical evaluation of reported events 98
case report forms 77, 82, 91, 222 Clinical Evidence 50

290

group6_PH.indd 290 7.8.2007 12:21:00

 clinical holds 40 comparative safety data 117
clinical judgement 82, 84, 118, 120, 128, Comparators 11, 173
170, 183, 195, 202, 275 comparisons of proportions 146, 155,
Clinical Laboratory Data 10, 124 208, 222
clinical lab tests 88 compliance 23, 42, 59, 81, 168, 169,
clinical outcomes 82, 89 178, 179, 194, 210, 261
Clinical pharmacology 65 comprehensive regulatory review 109
clinical pharmacology studies 68 comprehensive review of safety data 104
Clinical Research Associates 78 concise summary of safety 177
clinical research ethics committees 41 concurrent control 115
clinical safety departments 22 confidence intervals 142, 148, 149, 155,
clinical syndromes 98, 99, 100 207, 208, 223
clinical trial database 50, 78, 89, 94, 95, confidentiality 8, 27, 40, 43, 45, 186,
195 262, 266, 279
Clinical Trial Information Disclosure 50 conflicts of interest 38, 39, 47, 261, 262,
clock start date 176, 212 263, 266
cluster of cases 161, 209 confounding factors 67, 121, 194
Cochrane Central Registry of Clinical confounding or interacting effects 114
Trials 49 confounding risk factors 276
Coded Private Information 45 consent documents 40, 65, 84, 197
codified symptoms and/or signs 100 CONSORT (Consolidated Standards of
codified terms 97, 200 Reporting Trials) 47, 131, 136
contingency table 146, 222, 223, 239
coding AEs 101, 102
continuous data 29, 139, 141
coding guidelines 65
Continuous Measurements: Laboratory
coding of adverse events 28, 104, 201
Chemistries 139
Coding Procedures 9, 101
contractual obligations 60
combination medicinal products 120
contractual partners 44
combining data across different trials
contractual relationships 105
135, 204
contract research organizations (CROs)
comment section of the case report form
80, 104, 223
82, 195
Copernicus Group 42
commonly occurring events 113, 202
core study information 272
Common Technical Document 110, 125,
136 correction for multiple testing 144, 206
Communications to Investigators and Correlation 223
IECs/IRBs 182 correlation coefficient 223
Communication between visits 187 cost effectiveness 32
Communication Of Clinical Trial Re- Council of Europe 37
sults 50 Covariance 140, 221, 223
communication to the patient 185 Covariate 224
Company Core Safety Information Cox model 156, 224, 238
(CCSI) 117, 222, 224, 226 Cox regression 155, 209

291

group6_PH.indd 291 7.8.2007 12:21:01

 CPK 81, 88, 198 degrees of causality 85
criteria for diagnosis 98 delayed effects 135
criteria for risk evaluation 127 delayed occurrences of new AEs 150
cross-over designs 135 delayed toxicity 96
Cumulative dose 124 demographics 67, 81, 105, 114, 272
cumulative hazard 153, 154, 155, 208 descriptive and inferential statistics 109
cumulative survival 154 descriptive methods 131, 133, 204
Current Therapeutic Standards 9, 115 descriptive or graphical summaries 135
descriptive statistics 29, 128, 228
detecting potential safety signals 122
D detecting signals 118
Databases for Epidemiology and Phar- determination of reportability 172, 211
macoepidemiology 287 Development Risk Management Plan
database search strategies 102 (DRMP) 63
Data and Safety Monitoring Boards 3, 7, Development Core Safety Information
12, 43, 89, 99, 227, 255, 265 (DCSI) 12, 16, 26, 56, 86, 117, 167,
data codification 101 182, 197, 213, 224, 225, 226, 230,
data collection 28, 66, 79, 80, 83, 87, 275
91, 92, 95, 110, 111, 113, 195, 199, Development Pharmacovigilance and
271, 279 Risk Management Plan 224
data collection form 66 Development Risk Management Plan 8,
data elements 12, 16, 28, 77, 79, 81, 82, 63, 73, 193
91, 94, 105, 169, 195, 271, 272, 279 Development Safety Update Report
data entry 101, 104, 201 (DSUR) 11, 61, 180, 181, 212, 224
data management processes 104 diagnoses 9, 82, 86, 97, 102, 200, 247
Data Monitoring Committee 227, 265 diagnosis 21, 25, 79, 86, 87, 98, 99, 100,
data output for clinical evaluation 101 101, 124, 138, 173, 198, 201, 206,
Data Processing 9, 104 217, 247, 248, 250, 280, 282
data quality 82, 97, 195, 199 diagnostic standards for adverse
data queries 104, 201 reactions 87
DCSI 10, 11, 12, 16, 26, 62, 65, 86, 99, discontinuation from treatment 83, 120,
100, 117, 119, 120, 127, 128, 167, 202
171, 172, 181, 182, 184, 186, 187, disease-specific registries 69
188, 224, 225, 226, 230, 275, 276 divergence from the ICH 170
DCSI Updates 11, 181 dose-response 65, 233, 276
death 41, 69, 70, 89, 152, 154, 161, 162, dose duration 124
208, 231, 236, 238, 272, 273, 282 drop-out rates 153
dechallenge 276 drug-demographic interactions 114
decision-making 8, 57, 62, 176, 212 drug-drug and drug-disease interactions
decision rules 63 121
Declaration of Helsinki 12, 27, 37, 38, drug-drug interaction 71, 116, 194, 233
46, 47, 51, 228, 229, 242, 259, 263 drug-event attribution 84

292

group6_PH.indd 292 7.8.2007 12:21:01

 drug-induced liver injuries 71, 100, 119, Epidemiology 8, 12, 44, 60, 64, 67, 119,
126 149, 151, 156, 287, 288
drug half-life 96 erroneous clinical terms 98
drug interactions 61, 65, 66, 71, 84, 116, established therapies 115
176, 194, 196, 212, 233 ethical concepts 38
drug metabolism studies 61 ethnicity 64, 124
drug related 86, 96, 179, 232 ethnic and racial differences 115
DSMBs 12, 23, 27, 28, 29, 43, 44, 63, EU Clinical Trial Directive Guidance
66, 103, 167, 168, 173, 185, 187, 181
214, 227, 265, 266, 267, 268, 270
EU CT Directive 45
DSMB in lieu of expedited reporting
EU Pharmacovigilance Guidelines 43
173
evaluating and mitigating risk 66
duration of use 111
Evaluation of aggregate data 122
Evaluation of Safety Information 9, 109,
E 111, 114
“excessive” data 82, 195 evaluation of signals 13, 122
“excessive coding” 102, 200 events of special interest 27, 65, 82, 85,
“expectedness” 167, 226 88, 89, 94, 100, 118, 121, 122, 170,
early detection of safety signals 109 197, 198, 202, 273
edit checks 104, 201 evolving benefit-risk profile 115, 177
Effectiveness 225 Exceptions to unblinding 173
effects of time 134 excluding dropouts 137
Efficacy 9, 65, 89, 110, 125, 136, 138, exercise history 81
167, 225 Expectedness 11, 171
efficacy endpoints 28, 117, 203, 227 expected (“listed”) 182, 213
elderly 29, 64, 66, 122 expected ADR 225
electronic clinical trial data 105 Expedited Reporting 11, 17, 77, 167,
electronic records 92 169, 170, 171, 172, 173, 175, 177,
electronic regulatory reporting 279 236
electronic submission of expedited expedited reporting to IRBs/IECs 178
reports 169 expedited reporting to regulatory au-
electronic transmission 78, 91 thorities 167, 170, 210
elevations of aminotransferase and experimental (unapproved) product 229
bilirubin 126 explanatory (influential) variable 224
emergency-case patients 46 external advisor 62
emergency care 92 external consultation 99
emerging profile 116 extrapolating incidence rates 68
epidemiological databases 68 extreme laboratory value abnormalities
epidemiological information 128 123
epidemiologists 44, 60, 70 extreme values 139, 140, 142

293

group6_PH.indd 293 7.8.2007 12:21:02

 F hepatocellular injury 126
Fairweather Rules 226 Hepatotoxicity 8, 66, 71, 102, 126
false negatives 145, 226 herbal and other non-traditional remedies
false positive effects 144, 226 83, 84, 196, 197
final analysis 134, 204 High Level Terms (HLT) 102
First in class 58 high risk populations 66
Fisher’s exact test 141, 146, 148, 160, high risk situations 68
227, 239 historical clinical 68, 115, 161
fixed-effect models 159 historical clinical trial data 115
follow-up information 104, 121 historical data 68
follow-up practices 91 hospital records 92
Food-drug interactions 71 hypersensitivity reactions 72, 151
formal risk management plan 55 hypothesis-generating 237
forms 24, 77, 82, 86, 90, 91, 122, 124,
172, 188, 211, 224, 271, 272
I
Frequentist statistics 221
ICH E2B 77, 169
frequent review of serious adverse
events 117 ICH GCP Guideline (E6) 17, 31, 172,
177, 178
ICH Guideline E1 61, 112, 142
G ICH Guideline E2A 77, 89, 103, 167,
GCP regulations 22, 191 169, 170, 175, 177, 181, 210, 236
gene-therapy research 32 ICH Guideline E2B 16, 91, 99, 100, 218,
General Practice Research Database 279
(GPRD) 69, 287 ICH Guideline E2D 167, 237
genetic data 90 ICH Guideline E3 11, 110, 136, 163
genotoxicity 72 ICH Guideline E5 64, 115
global project team 59 ICH Guideline E8 32, 233
good clinical trial safety practices 27 ICH Guideline E9 11, 136, 163
Good Pharmacovigilance Practices 17, ICH process 21, 217, 219
22, 56 ICH Topic M1 101
Graphical displays 140, 207 ICSR 218
graphical summary 164 ideal conditions 225
graphics 133, 204 identification of a new risk 168
grouping of adverse effects 138, 206 identification of safety concerns 111, 201
guideline for DSMBs 43 IECs and IRBs 7, 41, 42, 179
immediate update of informed consent
184, 214
H immunogenicity 8, 66, 71, 194
hazard rate 151, 155, 156, 208, 209 immunotherapies 82, 195
healthy volunteers 71, 82, 196, 261 Important Adverse Events 9, 94
Health Level 7 (HL7) 106 important signs and symptoms 93

294

group6_PH.indd 294 7.8.2007 12:21:03

 Inappropriate clinical characterization 98 International Society for Pharmacoepi-
inappropriate route of administration 90 demiology (ISPE) 69, 287
incapacitated participants 229 Internet technologies 91
incidence rate 150, 151, 156, 208, 209 interpretation of safety information 114
incidence rate ratios 156, 209 interval and cumulative frequency tables
increased frequency 175, 212 122
indemnification 38 interval and cumulative incidences 117
independent clinical expert 99 interval data 183, 214
independent clinical researchers 33, 191 Introspection 26, 249, 251
Independent Ethics Committee (IEC) invasive procedure 95
41, 227 Investigational Product 229
independent investigators 80, 81 investigator’s assigned terms or diagno-
independent tests 144 ses 97, 99, 200
independent trial 80, 195 Investigator’s Brochure 25, 29, 86, 113,
181, 224, 225, 226, 269, 275
individual cases 26, 77, 83, 86, 114, 115,
120, 121, 173, 196, 197, 202, 279 investigator’s causality assessment 26,
28, 94, 119, 163, 179, 249, 250
individual case report assessment 28
investigator’s checklist 42
Individual case safety reports (ICSRs)
investigator’s verbatim term 99, 273
98, 200
Investigators Brochure 56, 171, 181,
Individual Clinically Significant Abnor-
249, 269
malities 125
Investigator Brochure 11, 173, 181, 211,
individual signs and symptoms 100
254
Inference 228
IRB (Institutional Review Board), ERC
Inferential 118, 228
(Ethics Review Committee) 41
inferential statistical approaches 128,
isolated terms 100
133
ITT safety analysis 136
Informed Consent 8, 40, 46, 185, 228
informed consent form 185, 187, 215,
228 J
informed consent information 93, 186, journalists 44, 48
187, 198, 215
justice 37
Institutional Review Board (IRB) 229
integrated summary of safety 61, 110,
117, 203, 251 K
integrity of the trials 179 Kaplan-Meier 153, 154, 155, 208, 230
intended target population 57, 64 Kaplan-Meier estimate 153, 154, 155, 208
intent-to-treat analysis 10, 96, 136 known risk 60
interaction and dialogue between patient
and investigator 92
interim analyses 134 L
internal safety committee 62 Labelled 230
international birthdate 181, 212 laboratory chemistry data 83

295

group6_PH.indd 295 7.8.2007 12:21:03

 Laboratory Chemistry Measurements MedDRA® Points to Consider 87, 101
9, 88 MedDRA MSSO 103
Laboratory Value Scatter Plot 141 MedDRA preferred terms 181, 183, 214
latent safety effects 31, 96, 135, 199 Medicaid databases 68
legislation 18, 41, 48, 113, 177, 228 medically important events 101
liability issues 49 medical classification 138, 206
licensing partners 59, 194 medical concept 102
licensors/licensees 44 medical devices 33
life-saving therapies 109 medical judgement 109, 132, 138, 206
life-table approaches 163 Medical review and judgment 121
Limitations in the Identification and medical significance 113, 120
Evaluation of Safety Information 111 medication compliance 81
Limitations of Clinical Trials 7, 30 medication errors 30, 66, 90, 232
limitations of clinical trials 111 MEMO database 69
linear relationship 223 meta-analysis 47, 135, 143, 151, 157,
lines of evidence 276 158, 159, 209, 230
Line extension 58 Meta-analytical Techniques 10, 157
line listings 168, 170, 177, 178, 180, Metabolites 276
181, 182, 183, 210, 213, 214 milestones 8, 61, 117, 203
Listed 230, 241 minimum data elements 272
listedness 171, 172 missing data 134, 250
liver enzymes 81 modification of study protocols 111
local datasheet 171 monetary compensation 39
log rank test 155, 208, 209, 234 Morbidity and Mortality as Efficacy
lost to follow-up 153 Endpoints 9, 89
low power 145 morbidity and mortality rates 68
multi-arm trials 135
multidisciplinary teams 117
M multiple analyses 29, 138, 142, 205
“medically significant” cases 82 multiple hypotheses 144
managing risk 17, 57 multiple significance testing 145, 164
managing safety information 13, 72, 188 multiple unrelated causes 100
marked laboratory abnormalities 123, multiplicity 10, 29, 138, 139, 142, 144,
125, 126 205, 206, 230, 269
marketing approval 109
master database of clinical research
results 48 N
mechanism 42, 59, 60, 71, 92, 104, 116, “number needed to treat” 162
144, 173, 184, 194, 206, 277 narrative description of the case 87, 198
MedDRA® 87, 100, 101, 102, 104 Natural course of disease 64
MedDRA® codification principles 101 natural history of the disease 60, 67, 68,
MedDRA® Lowest Level Term 101 69, 114, 115, 121, 194, 202

296

group6_PH.indd 296 7.8.2007 12:21:03

 negative findings 38, 192 ongoing evaluation of the safety profile
negative results 47 109
neglected diseases 51 ongoing review of benefits 109
non-adherence to treatment 137 ongoing safety evaluation 110, 111, 201
non-clinical data 118 ongoing serious event 96
non-clinical research 88 ongoing signal detection 120
Non-clinical safety experience 65 open-label comparators 174, 211
non-clinical safety information 175 organizational structure 55, 103
non-parametric tests 234, 238 organ impaired patients 64
non-pharmacological treatments 116 orphan drugs 78
non-randomised and non-comparative outcome variable 224
studies 135, 205 outlying values 140, 207
non-serious adverse events 77, 83, 85, outside experts 59, 63
88, 94, 113, 120, 196, 197, 198, 202, outside investigator-sponsor 185, 215
269, 273 overall framework 55
non-serious adverse events of special overall periodic review 116
interest 85, 94, 197 overdose 276
non-significant differences 149
Non-specific “disorder” terms 101
non-specific signs and symptoms 86 P
nonmaleficence 37, 38 “possibly drug related” 179
notification of safety information 168 P-value 132, 142, 143, 144, 146, 148,
null hypothesis 133, 146, 155, 226, 231 149, 160, 164, 205, 207
null value 148, 149, 223 parallel-treatment groups 135
Number Needed to Harm (NNH) 162, parametric methods 156, 232
231 partially-blinded analysis 122
Number Needed to Treat (NNT) 231 Partnership for Human Research Protec-
number of subjects exposed 111, 112, tion, Inc, PHRP 42
123, 203 Patient Population Characteristics 9, 114
Patient privacy 40
payments 38, 192
O pediatrics 21, 29, 64
objective parameters 92 pediatric patients 66
observational cohort studies 156, 209 perceived benefits 70, 235
observational studies 31, 46, 78 perceived risks 235
observation period 95, 150, 207 peri-approval studies 78, 233
odds ratio 146, 147, 148, 149, 158, 161, periodic aggregate assessment 118, 202
231 Periodic and summary reviews 117
off-label (unapproved) uses 30, 49, 66 Periodic Communication of Safety
one-sided testing 144 Information 11, 180
One-sided vs Two-sided Testing 231 periodic listings 177

297

group6_PH.indd 297 7.8.2007 12:21:04

 periodic reports to investigators 182, Pooling of control groups 161
183, 184, 213, 214 Pooling of Data 10, 123
Periodic Review and Evaluation 122 pooling of data 164
periodic review of aggregate data 122 population-based data 161, 210
periodic summary reports 26, 257 population trends 125
period of observation 95 possible safety signal 126
person-time 150, 151, 152, 156, 208, post-approval pharmacovigilance 13
209
post-approval studies 48, 83, 113
personal data 45
post-authorization study requirements 78
pharmacogenetics 32, 45
post-marketing reporting 167, 172
pharmacogenetic data 81
post-marketing risk management plan
pharmacokinetic data 81, 128
63, 193
pharmacology studies 32, 65, 68, 71,
124, 233 post-study monitoring 96
Pharmacovigilance 13, 15, 17, 22, 27, post-study observation period 95
43, 53, 56, 91, 100, 117, 171, 176, post-treatment events 150, 207
218, 224, 225, 230, 232, 246, 271 post-treatment follow-up 150
pharmacovigilance in children 29 post-treatment value 140, 206
pharmacovigilance plan 24, 242 postmenopausal hormone therapy 70
PHARMO Record Linkage System 69 potential for abuse 66
PHASES OF CLINICAL STUDIES (I potential human toxicity 88
– IV) 232 potential risk 60, 62, 123
Phase I (Human Pharmacology) 80, 82, potential risk factors 123
196, 232
potential safety issues 61, 67
Phase II (Therapeutic Exploratory) 232
power 10, 29, 31, 133, 138, 140, 141,
Phase III (Therapeutic Confirmatory)
142, 143, 145, 157, 205, 234, 236,
232
238
Phase IIIb 78
powerful analysis 234
Phase IV (Therapeutic Use) 232
pre-approval periodic reports 61
Phase IV investigators 183, 213
pre-study training 82
Phase IV trials 32, 78
precursors (prodromes) 88, 198
physical examination findings 120, 203
physiological or potentially toxic effects Preferred Terms 101, 102
124 pregnancy 81, 90, 101, 273
placebo controls 115 pregnancy testing 81
placebo patients 68 premature termination 23, 109
point estimate 148, 156, 207, 233 Prescription Event Monitoring (PEM)
Poisson distribution 156, 209, 233 69, 287
Poisson regression 156, 209 preventive therapies 67
pooled control groups 143, 161, 210 preventive vaccines 82
pooled placebo groups 128 PRIM&R 43
pooling datasets 104 primary analyses of AE data 97, 99, 200

298

group6_PH.indd 298 7.8.2007 12:21:04

 privacy 8, 27, 40, 41, 45, 90, 260, 262, rare adverse drug reaction 113
279 rare adverse events 143, 205
probable diagnosis 99 Rates per person-time 151, 208
Problems in Safety Analyses 10, 137 rate ratio 151, 152, 208
product-specific procedures 57, 193 re-consenting 46, 47
product profile 63 reactive metabolite formation 8, 72, 194
project management 8, 59, 64, 194 reasonable possibility of a causal rela-
prompt communications 58, 193 tionship 170, 171, 211
prompt notification 176, 184, 212, 214 reasons for discontinuation 121, 123,
Prompt Reporting Other than Case Re- 180
ports 11, 175 rechallenge 119, 276
proof of concept studies 233 reference population 234
prophylactic treatments 82, 196 reference risk 234
proportional hazards regression 155, 209 reference standards for laboratory values
proprietary information 48 94
protecting trial participants 101, 111 registry of ongoing trials 48
protocol amendments 58, 221 Regression 234
Protocol design 133 regression coefficients 223
Protocol for Biomedical Research 37 regulatory reporting status 120, 174
protocol procedure-related adverse event regulatory submission of data 105
176 relative risk 133, 146, 147, 148, 151,
PSURs 15, 61, 81, 172, 181, 184, 213, 156, 161, 162, 223, 231, 233, 234,
214 235
publication bias 47, 157 relevant product information 117
publication policy for trials 47 renewing informed consent 186
reportability of a case 172
reporting clock 176
Q Reporting Odds Ratio 147
QT prolongation 66, 70, 103 reporting of events to investigators and
quality (manufacturing) issues 32 ethics committees 177
quality of life 81, 88 reporting ratio 147
quarterly line listings 177, 178, 181 resource-poor and developing countries
quarterly review of safety data 61 51
questionnaires based on diagnostic results of all studies 48
standards 98 result summaries 49
Review of Individual Cases 10, 121
risk 21, 66, 68, 71, 109, 111, 113, 116,
R 121, 131, 132, 146, 147, 149, 150,
“reasonable possibility” 85, 197, 220 151, 154, 156, 158, 161, 162, 168,
random-effects models 159 231, 233, 234
Rank 234, 238 risk-management strategies 113
rapid onset 139, 206 risk and harm 21, 191

299

group6_PH.indd 299 7.8.2007 12:21:05

 risk assessment 23, 56, 60, 113, 114, sample size 112, 133, 134, 149, 160,
119, 192, 235, 266 204, 227
risk difference 146, 149, 158, 162, 235 Saskatchewan Health Plan 69
Risk estimation 235 scatter plots 140, 207, 235
risk evaluation 16, 66, 73, 127, 235 senior safety committee 62
risk management 7, 24, 25, 27, 30, 55, Sensitivity 236
56, 63, 109, 113, 114, 186, 192, 193, serious adverse event 77, 85, 86, 87, 89,
235, 246 94, 102, 104, 121, 122, 173, 175,
Risk Management Plan 8, 55, 56, 63, 73, 197, 198, 200, 202, 211, 212, 248,
193, 224 249, 269
risk minimization 56, 57, 60, 63, 66, 73, serious adverse events 25, 28, 77, 82, 83,
111, 113, 192, 201, 235 84, 85, 88, 89, 90, 92, 94, 104, 113,
risk perception 235 117, 120, 121, 122, 163, 175, 196,
Risk perception and acceptance 127 197, 198, 202, 261, 269
risk ratio 146, 149, 151, 234 serious adverse event form 77, 86, 92,
197, 271, 279
roles and responsibilities 27, 28, 31, 39,
41, 43, 55, 57, 59, 80, 99, 184, 193, Serious Adverse Event or Reaction 236
194, 227, 255 severity 94, 98, 102, 114, 115, 158, 175,
rule of three 112, 149, 160 200, 209, 225, 226, 273, 276
shared responsibility 116
Signal 10, 117, 237
S signal detection 97, 99, 117, 118, 120,
“shift” tables 164 147, 199, 202
“study treatment-emergent” findings 95 signal strengthening 237
“support” by a company 81 Significance 230, 237
SADR 218 Significant 125, 180, 212, 237
SAE 122, 173, 211, 218, 247, 256, 281 significant new safety information 178,
Safety-Signal Detection 117 181, 213
safety database 61, 67, 89, 94, 103, 104, significant safety signals 184, 214
201, 270 signs and symptoms 9, 25, 65, 79, 82,
Safety Data Management 9, 16, 17, 77, 86, 87, 92, 93, 98, 99, 100, 164, 198,
97, 167, 169, 170, 172, 201 199, 247, 248, 282
Safety Management Process 11, 184 Simes 145, 238
Safety Management Team 58, 62, 184, site personnel 80, 92
193, 214 special interest cases 124, 271
safety monitoring 13, 23, 28, 41, 57, 77, special monitoring procedures 60
78, 91, 95, 110, 134, 168, 191, 271, special populations 21, 64, 65, 66, 68
272 special study data 81, 94
safety signal 68, 121, 125, 126 specific list of possible ADRs 93
safety surveillance/risk management specific milestones 117, 203
program 27 specific report forms 122
safe and effective 21, 40 Sponsor 79, 238

300

group6_PH.indd 300 7.8.2007 12:21:05

 sponsor’s terms 99 Submission Data Standard (SDS) 105
spontaneous reports 11, 17, 89, 94, 100, summary of the emerging safety profile
117, 118, 147, 172, 175, 183, 200, 183, 213, 214
211, 214 summary reviews 49, 117
SSAR 219 supporting information 87, 198
Stakeholders 7, 39 Surrogate variables 139
standardization of terminology 99 Survey on Pharmaceutical Company
standardization of clinical trial protocol Safety-Management Practices 245
elements 106 survival analysis 29, 137, 152, 153, 155,
standardized data management tech- 156, 159, 164, 208, 224, 238
niques 105 survival curves 154, 208
standardized form 94 SUSAR 219, 238
Standardized MedDRA® Queries Suspected Unexpected Serious Adverse
(SMQs) 102 Reaction (SUSAR) 238
standard data elements 169 syndrome 86, 98, 99, 100, 138, 206
standard global practices 29 Systematic Error 238
standard of care 63 System Organ Classes (SOCs) 101, 102
standard template 78
standard terms and definitions 99
standard therapy as a comparator 51 T
standard treatment 116 “treatment emergent signs and symp-
start of safety data collection 95, 199 toms” 164
statistical and clinical significance 132, targeted laboratory investigations 88
204 targeted study 66
statistical association 132 target population 22, 57, 60, 61, 67, 68,
statistical cut-off level 145 115, 127
statistical inference 148 template on safety and efficacy review
statistical methods to detect signals 117 110
statistical power 29, 140, 141, 142, 157, Temporary hold 67
205 Termination of program 67, 109
statistical software programs 149 terms and definitions 7, 27, 99, 132, 218
statistical terms 27, 132, 218, 234, 236 Threshold Considerations 12, 275
statisticians 24, 41, 44, 105, 106, 230 threshold for adding a new serious AE
stopping or modifying the trial 134 186
stopping rules 29, 122, 268 time-dependent phenomena 134, 135, 205
study manuals 94 timeframe for periodic reports to investi-
study monitors 78, 80 gators 183, 213
Study protocols 78 time at risk 150, 208
sub-populations 29, 122, 268 time dependency 29, 138, 206
subgroup analyses 123, 134, 203 time of dosing 124
subjective component 92 time on drug 29, 138, 206
subjective data 93 Time on or off Treatment 150

301

group6_PH.indd 301 7.8.2007 12:21:06

 time profiles 139 underreporting of clinical research 47
Time Since Start of Treatment 155 unequal randomisation 159
time to occurrence 224 UNESCO 37
time to study withdrawal 121 unexpected ADR 226
Timing of Safety Evaluation 9, 116 unexpected and unusual results 134
torsade de pointes 70 Universal Instrument on Bioethics 37
Toxicology studies 88 Unlabeled 230
training 43, 80, 82, 86, 87, 93, 195, 198, Unlisted 230
265 unusual adverse event pattern 123
translated version of the original term(s) update of important safety information
99 177
transparency 38, 48, 192, 200 update to the DCSI 187, 215
treatment discontinuation 120 updating informed consent 187, 215,
Trends in average values 141, 207 255
trial report 134, 150, 204, 207 urgent public health needs 31
true rate 143, 149, 150, 160 urgent safety issues 184, 214
true value 148, 207, 223 US IND Rules 45
Two Sided Testing 143 utility theory 33
Types of Analyses 10, 120, 135
Type I error 133, 144, 206, 239
Type II error 145, 239 V
vaccines 21, 32, 79, 82, 127, 195, 196,
217, 248
U variation and uncertainty 133
“undercode” 102, 200 ventricular arrhythmias 70
UKECA (United Kingdom Ethics Com- verbatim AE terms 94, 97, 199
mittee Authority) 45 verbatim term 99, 101, 273
unapproved uses 49 vital signs 117, 120, 203
unblinded analyses 123 volume of reports to regulators and
unblinded clinical trial cases 182, 213 investigators 178
unblinded comparator cases 174, 211
Unblinded Data 9, 103
unblinded expedited reports 183, 214 W
unblinded placebo cases 174 waiver from obtaining informed consent
unblinded reports 103 186
unblinded studies 118 wash-out period 95, 135
Unblinding 11, 172 Weibull Distribution 156, 239
unblinding expedited single case reports weight of evidence 128, 275
172 well established safety profiles 77
uncertainty 27, 132, 133, 134, 143, 148, Western Pacific Regional Forum for
149, 154, 158, 161, 204, 207, 209, Harmonization of Herbal Medicines
210, 223, 228, 232 (FHH) 84

302

group6_PH.indd 302 7.8.2007 12:21:06

 WHO ADR database 69
WHO definitions 217
Within-patient changes 134
Women’s Health Initiative (WHI) 70,
147, 151, 152, 153
Working Group members 25, 85
worsening of the disease 115

Y
Yate’s Correction 148, 239

303

group6_PH.indd 303 7.8.2007 12:21:07

group6_PH.indd 304 7.8.2007 12:21:07
 CIOMS
Management of Safety Information from Clinical Trials
CIOMS publications may be obtained directly from CIOMS,
c/o World Health Organization, Avenue Appia, 1211 Geneva 27,
Switzerland or by e-mail to [email protected]
Management of
Both CIOMS and WHO publications are distributed by the
Safety Information
World Health Organization, Marketing and Dissemination,
Avenue Appia, 1211 Geneva 27, Switzerland and are available
from booksellers through the network of WHO sales agents.
from Clinical Trials
A list of these agents may be obtained from WHO by writing
to the above address.
Report of CIOMS Working Group VI

Price: CHF 40.– Geneva 2005

GROUP6_COVER DEF.indd 1 7.8.2007 12:17:50