Study Design

1. Types of epidemiological studies

Research in population health centres on questions related to: Magnitude (How common is the problem?) Causality (Does this risk factor cause the problem?) Evaluation (Does this intervention work?).

There are a number of different types of studies used to investigate these research issues. For a particular causal relationship, the process of evidence accumulation generally involves several studies to answer a series of questions (shown below). The primary aim of the research process in animal health being to identify the cause of disease, or the factors that predispose to disease, so that appropriate preventative and control measures can be implemented. In observational studies, the researcher observes what is happening or has happened within a population, without attempting to interfere in the natural progression of disease events being studied. In intervention (or experimental) studies, the researcher deliberately implements a treatment or prevention protocol and measures its impact on the disease in the study population. Further for observational studies, the researcher observes what is happening or has happened within a population in order to answer a question related to magnitude or causality. There are 2 categories depending on whether the research is purely descriptive (eg a survey to answer questions such as what is the level of disease or of production?) descriptive studies or whether it will involve a study that seeks to formally compare (or analyse) the difference in disease level between groups in order to address a causality question - analytical studies.

2. Using epidemiological studies to identify causes of disease

Some studies are better than others for determining whether there is a cause-and-effect relationship between a potential risk factor and a disease process. Several study designs are discussed below, in order from those providing weak through to those providing strong evidence for such a relationship. Descriptive studies a. Case Reports Case reports are detailed presentations of a single case or a small number of cases (usually less than 10). Good Things: New and unfamiliar diseases or manifestations or associations of disease are Highlighted Often are the only means of describing rare or unusual clinical events and therefore are rich sources of hypotheses about disease presentation, risk, prognosis and treatment Low cost

Epi/PG/mpg

Epi/PG/mpg

Descriptive studies
Observational studies Case report Case series Survey MAGNITUDE Level of disease occurrence?

CAUSALITY Link between risk general factor and disease general occurrence?

Analytical studies
Cross-sectional Case-control with ill-defined population (hospital-based) Case-control with well-defined population (population based) Cohort

Intervention studies
Randomised controlled trial

EVALUATION Response to intervention?

Epi/PG/mpg

Bad Things: Particularly susceptible to bias as they involve a small and highly selected group of patients There is no control group

b. Case Series A case series is a study of a larger number of cases (more than 10) with a particular disease Good Things: A common way of delineating the clinical picture of a disease and larger numbers means that descriptive statistics can be used As with case reports, there is no comparison (control) group Only tentative conclusions can be drawn

Bad Things:

c. Survey A survey evaluates the level of disease or of production in a group of animals. If all the animals in the population are examined (a census approach) or a representative sample of the animals in a population is examined at one point in time to determine disease status then a measure of prevalence is obtained. However often surveys do not examine all or a representative sample of a population and therefore do not provide a prevalence figure that is an estimate of the prevalence in the target population. Good Things: Surveys can be repeated to determine if changes are occurring over time If all animals or a representative samples are included then internal validity will be maximised, that is, the results reflect the true picture in the target population Co-operation from livestock owners is easy to obtain. May be based on farmer recording. Will not provide a measure of prevalence if a representative sample is not used. Not very useful for monitoring rare diseases or diseases of short duration or diseases that are under control at the time of data collection.

Bad Things:

Analytical studies d. Cross-sectional Studies A cross-sectional study is a structured investigation in which information is collected on each animal in a group, some of whom are diseased while others are healthy, on a single occasion. The proportion of the group that is diseased constitutes the prevalence of disease. Good Things:

Epi/PG/mpg

These studies can be repeated to determine if changes are occurring over time Many different methods can be used to collect information including questionnaires (by mail, phone or visit), abattoir sampling and farm visits Formal sampling methods can be used to ensure internal validity, that is, the results reflect the true picture in the target population Moderate cost and relatively quick and easy to do Co-operation from livestock owners is easy to obtain Because disease and the possible factors responsible for the disease are measured simultaneously, it is often unclear which came first. That is, the time sequence is obscured and causal and non-causal associations cannot be differentiated. Not very useful for monitoring rare diseases or diseases of short duration or diseases that are under control at the time of data collection Control of extraneous variables may be incomplete Cannot estimate incidence of disease

Bad Things:

e. Case-Control Studies Case control studies are designed to find out whether a risk factor is more common in animals with a given disease (cases) than animals without the disease (controls). After selection of a group case animals and a group of control animals, each group is examined for the presence or absence of particular risk factors.

From Fletcher et al. (1996) Case-control studies need the following: A series of cases and a series of controls or a comparison group that does not have the disease. Researchers then look backward in time (ie. retrospectively) to determine the frequency of the exposure factor in the two groups. There must be enough animals in the study so that chance is less likely to play a part in the observed results. Controls must be representative of the study population from which the cases came and must have an equal opportunity to receive the risk factor as cases. A well-defined population in which the cases are obtained from all the cases arising in a herd, and the controls are randomly selected from that herd. Epi/PG/mpg

Good Things: Well-suited to the study of rare diseases or of those with long incubation periods Relatively quick and easy to do and low cost Can detect an effect using far fewer subjects compared to cohort studies (useful for rare diseases). Can explore the relationship between several risk factors and the disease at one time Risk of disease can be estimated using the odds ratio Relies on recall or accurate records for information on past risk factor exposure so these studies are particularly prone to selection and measurement bias The results are not representative of the general population and prevalence and incidence cannot be calculated Information may be incomplete Selection of controls can be difficult (and must not break certain selection rules) Statistical associations discovered do not prove causal relationships

Bad Things:

f. Cohort Studies In a cohort study, a cohort of healthy animals capable of developing the disease of interest, some exposed to a specific risk factor and some not, are observed for a period of time and the incidence of disease in each group compared. This type of study examines the relationship between exposure to a risk factor and subsequent disease incidence.

From Fletcher et al. (1996) Good Things: Incidence can be calculated Logical process from exposure to outcome and provides very strong evidence of the effect of the exposure factor under consideration Exposure can be measured without bias because at that time the outcomes are not known Other exposure factors (potential confounders) can be measured and accounted for during analyses Provides estimates of relative risk Relatively expensive and large numbers of subjects are required to study rare diseases Epi/PG/mpg Bad Things:

No assurance that cases will occur Follow-up and maintenance of reliable records is difficult Must be careful to ensure minimal selection, measurement and confounding bias

Intervention studies g. Clinical and Field Trials As stated in the section on trials, effective field and clinical trials are randomised controlled trials in which there are adequate numbers of animals, there is blinding of both the animal owners and investigators, and there are carefully standardised methods of measurement and analysis. Good Things: The most powerful way to establish a cause and effect relationship The study design guards against differences between groups of animals under investigation Not suited to diseases with long latent periods

Bad Things:

3. Measuring the strength of association

Data from case-control and cohort studies are usually entered into a 2 2 contingency table: Disease Cases A C A+C Noncases B D B+D

Risk Factor

Exposed Unexposed

A+B C+D

Two measures of the strength of association between the risk factor and the disease can be easily calculated from this table relative risk and odds ratio. Relative Risk RR = A / (A + B) C / (C + D) COHORT STUDY

As stated in an earlier section, the relative risk indicates the chance of an event occurring in the exposed group relative to the chance of the event occurring in the unexposed group. It can be calculated in cohort studies but not case-control studies. The higher the relative risk, the greater the association between the risk factor and the occurrence of disease. A RR of 1 indicates no association, whereas a RR of less than 1 indicates a sparing or protective effect. Usually RR is presented a 95% confidence interval. The 95% confidence interval is the range of values that have a 95% probability of including the true value of the statistic, in this case the RR. If the 95% confidence interval includes the value 1, the RR is considered not significant. Odds Ratio OR = [A / (A + C)] / [C / (A + C)] [B / (B + D)] / [D / (B + D)] = A / C = AD CASE-CONTROL B/D BC STUDY

Epi/PG/mpg

The odds ratio is another measure of association used particularly in case-control studies. In these studies relative risk cannot be calculated because the rates of disease in the exposed and unexposed groups are unknown. However, when disease in a population is relatively infrequent (less than 5%), the OR is very close in magnitude to what the RR would be if it could be calculated (Martin et al., 1987). The odds ratio measures the odds (or chance) of disease being present when a factor is present compared to the odds of disease being present when the factor is absent. The statistical significance of the OR is determined by calculating 95% confidence intervals; if these intervals include the value 1, the OR is considered not to be significant. Odds ratio (OR) measures the odds (or chance) of disease being present among exposed individuals compared to the odds of disease being present among unexposed individuals. It can be calculated in cohort, case-control and cross-sectional studies. In case-control studies of relatively rare diseases (less than 5% prevalence), the OR closely approximates the expected magnitude of the relative risk (if the RR could be calculated). The statistical significance of the OR is determined by calculating the 95% confidence interval; if this includes the value 1, the OR is considered not significant.

Epi/PG/mpg