Clinical Therapeutics/Volume 38, Number 7, 2016

Review Article
Current Challenges in Cancer Treatment
Jon Zugazagoitia, MD, PhD1,2; Cristiano Guedes, MD1; Santiago Ponce, MD1,2;
Irene Ferrer, PhD2; Sonia Molina-Pinelo, PhD2; and Luis Paz-Ares, MD, PhD1,2
1
Medical Oncology Department, Hospital Universitario 12 de Octubre, Madrid, Spain; and 2Instituto de
Investigación Iþ12. Lung Cancer Clinical Research Unit CNIO, Iþ12, Madrid, Spain

ABSTRACT death cell protein-1/programmed death cell ligand-1

Purpose: In this review, we highlight the current [PD-1/L1] and anticytotoxic T lymphocyte antigen-4
concepts and discuss some of the current challenges monoclonal antibodies) are restricted to a minority of
and future prospects in cancer therapy. We frequently patients, and no predictive markers are yet robustly
use the example of lung cancer. validated that could help us recognize these subsets
Methods: We conducted a nonsystematic PubMed and optimize treatment delivery and selection. To
search, selecting the most comprehensive and relevant achieve long-term survival benefits, drug combina-
research articles, clinical trials, translational papers, and tions targeting several molecular alterations or cancer
review articles on precision oncology and immuno- hallmarks might be needed. This will probably be one
oncology. Papers were prioritized and selected based on of the most challenging but promising precision cancer
their originality and potential clinical applicability. treatment strategies in the future.
Findings: Two major revolutions have changed Implications: Targeting single molecular abnormal-
cancer treatment paradigms in the past few years: ities or cancer pathways has achieved good clinical
targeting actionable alterations in oncogene-driven responses that have modestly affected survival in some
cancers and immuno-oncology. Important challenges cancers. However, this approach to cancer treatment
are still ongoing in both fields of cancer therapy. On is still reductionist, and many challenges need to be
the one hand, druggable genomic alterations are met to improve treatment outcomes with our patients.
diverse and represent only small subsets of patients (Clin Ther. 2016;38:1551–1566) & 2016 Elsevier HS
in certain tumor types, which limits testing their Journals, Inc. All rights reserved.
clinical impact in biomarker-driven clinical trials. Key words: cancer therapy, checkpoint inhibitors,
Next-generation sequencing technologies are increas- drug development, immunotherapy, lung cancer, next-
ingly being implemented for molecular prescreening in generation sequencing, precision oncology, targeted
clinical research, but issues regarding clinical interpre- therapy.
tation of large genomic data make their wide clinical
use difficult. Further, dealing with tumor heterogene-
ity and acquired resistance is probably the main
limitation for the success of precision oncology. On INTRODUCTION
the other hand, long-term survival benefits with Cancer is a major public health problem worldwide.
immune checkpoint inhibitors (antiprogrammed Global demographic characteristics predict an

The basis for this article has been presented as a Keynote Lecture at
the 12th Congress of the European Association of Clinical Pharmacol-
ogy and Therapeutics in Madrid, Spain. Scan the QR Code with your phone to obtain
FREE ACCESS to the articles featured in the
Accepted for publication March 17, 2016.
http://dx.doi.org/10.1016/j.clinthera.2016.03.026 Clinical Therapeutics topical updates or text
0149-2918/$ - see front matter GS2C65 to 64842. To scan QR Codes your
phone must have a QR Code reader installed.
& 2016 Elsevier HS Journals, Inc. All rights reserved.

July 2016 1551

 Clinical Therapeutics

increasing cancer incidence in the next decades, with trials), immunotherapy OR immuno-oncology (255,507
420 million new cancer cases annually expected by results; 14,081 clinical trials), immune checkpoint inhib-
2025. According to GLOBOCAN data, 14.1 million new itors OR PD-1/L1 blockade (769 results; 17 clinical trials),
cases and 8.2 million deaths from cancer were estimated and nonsmall-cell lung cancer. Articles were selected
in 2012.1 Cancers of the female breast, colorectal, mainly on the basis of their clinical applicability, and we
prostate, and lung are the most frequently diagnosed prioritized for practice-changing clinical studies, some
cancers in Europe.2 Lung cancer remains the leading translational papers, and selected comprehensive reviews
cause of cancer incidence and mortality worldwide.1 published in the last 5 years. Relevant articles were also
The increasing knowledge of molecular and tumor identified through searches of the authors’ files and when
biology has notably changed cancer treatment para- reviewing other papers and their respective bibliographies.
digms during the past 15 years. Formerly, cancer was Unpublished reports from scientific conferences were
classified and treated solely according to organs of identified across meeting libraries and abstract books.
origin or simplistic histomorphologic features. In a Only articles published in English were included. All of the
seminal paper published by Schiller et al3 in 2002, references cited in this article were reviewed. The final
completely overlapping survival curves were found in reference list was generated on the basis of originality and
advanced nonsmall-cell lung cancer (NSCLC) relevance to the broad scope of this review.
patients after use of 4 different platinum-based che-
motherapy doublets with third-generation drugs. Even
though the trial was limited to lung cancer, it found TARGETING ACTIONABLE ALTERATIONS IN
that cancer treatment based on a broad use of ONCOGENE-DRIVEN CANCERS
cytotoxic chemotherapies in unselected patients had The essential premise of genotype-based precision
reached its therapeutic plateau. In addition, it became oncology is that tumor-specific molecular abnormal-
clear that the development of molecularly targeted thera- ities can be targeted with accurate, effective, and
pies and treatment selection based on particular molecular potentially less-toxic therapies. Extensive preclinical
alterations was needed. Since then, 2 pillars have driven work and primary discoveries of somatic, single-gene
the subsequent evolution of cancer treatment: new tech- genomic abnormalities that could be pharmacologi-
nology acquisition for tumor molecular profiling and the cally targeted opened the first gateways for genomic
discovery of predictive molecular targets. Together, these precision oncology. More recently, comprehensive
efforts have materialized the 2 recent revolutions in cancer and integrative characterization of many cancers using
treatment. First, genotype-directed precision oncology, high-throughput technologies under the auspices of
that is, tailoring personalized therapies to subsets harbor- national (eg, The Cancer Genome Atlas, funded by the
ing specific genomic abnormalities across different tumor National Cancer Institute and National Human Ge-
types. Second, targeting components of the tumor micro- nome Research Institute in the United States) or
environment, in particular the immune system and the international (eg, International Cancer Genome Con-
antitumor immunity. In this review, we will succinctly sortium) efforts, has led both to a new era of genomic
describe the fundamental premises of these 2 anticancer or molecular taxonomy of cancer and to the discovery
strategies. We will also highlight some of the major of cancer genes and biomarkers for therapy.4
challenges ahead in both fields of cancer treatment, There are 3 crucial issues for successful clinical
frequently using the example of lung cancer. biomarker development: biologic plausibility (the
identified genomic alteration is responsible for malig-
nant transformation and tumor progression), analyt-
METHODS ical validity (it can be detected with robust, reliable,
We did a nonsystematic review of current concepts in and clinically applicable genomic tests), and clinical
precision oncology. References for this review were validity (the prognostic or predictive utility of the
identified through searches of PubMed using the terms biomarker has been validated in clinical trials and
precision oncology (8301 results; 313 clinical trials), community-based clinical cohorts). At the same time,
oncogene addiction OR targeted therapies (102,601 it must be emphasized that clinical biomarkers might
results; 4883 clinical trials), next-generation sequencing have diagnostic, prognostic, predictive, or pharmaco-
OR early drug development (69,901 results; 2201 clinical genomic utilities.5 Predictive biomarkers are the most

1552 Volume 38 Number 7

 J. Zugazagoitia et al.

useful markers in daily practice, as they simulta- and NTRK1, 2, 3 rearrangements, FGFR1, 2, 3
neously enable both selection of subsets that will obtain amplifications or mutations, and DDR2 mutations
the greatest benefits from a certain treatment and (the last 2 predominantly in squamous cell carcino-
exclusion of those who will not benefit from therapy. mas)13 (Table I). Importantly, using multiplexed
Prognostic markers, however, are informative of patient genomic testing, druggable genomic targets (inclu-
outcomes irrespective of treatment, and are therefore less ding EGFR, ALK, ROS-1, BRAF, HER2, MET,
frequently used in the clinic for treatment decisions. RET, PIK3CA genomic alterations) were identified in
NSCLC is one example that illustrates the para- up to 64% of the samples within a large multicenter
digms of genomics precision oncology. From the cohort of lung adenocarcinoma patients. Those patients
initial one size fits all described in the study by Schiller receiving appropriate matched therapies achieved longer
et al,3 shortly after 3 research groups found that the survival compared with those with driver alterations but
presence of mutations in the tyrosine kinase domain not receiving targeted therapies or those without
of EGFR gene (EGFR-activating mutations in exons oncogenic aberrations susceptible of specific
1821) conferred sensitivity to EGFR tyrosine kinase treatment.14 Finally, integrative and comprehensive
inhibitors (TKIs).6,7,8 Subsequently, a landmark molecular characterizations of both adenocarcinomas
trial was initiated enrolling metastatic pulmonary and squamous cell carcinomas have already been
adenocarcinoma patients enriched for the existence published, and it is thought that a genomics-based
of EGFR mutations (Asiatic, never or light smokers) classification of human lung tumors will become a reality
who were randomly assigned to receive first-line in the near future.15 Of course, oncogene-targeted
chemotherapy or an EGFR TKI (gefitinib). A signifi- therapies have been successful for treating other subsets
cant treatment interaction (P o 0.0001) was found of solid tumors apart from NSCLC4,16 (Table I).
between EGFR mutation status and treatment arm, so Three important issues must be pointed out regard-
that EGFR mutationpositive patients selectively ing oncogene-targeted therapies. First, most of these
benefited from gefitinib in terms of progression-free examples illustrate the key concept of oncogene
survival (PFS) (n ¼ 261; hazard ratio [HR] ¼ 0.48; addiction. In these tumor models, a single driver
95% CI, 0.360.64; P o 0.0001) compared with mutation confers distinct biologic properties and is
EGFR mutationnegative patient (n ¼ 176; capable of driving the main oncogenic capabilities so
HR ¼ 2.85; 95% CI, 2.053.98; P o 0.0001), tumor cells become strongly dependent on that specific
underscoring the predictive effect of EGFR-activating genomic alteration for survival.17 Second, these
mutations for the benefit of EGFR TKIs.9 genomic drivers represent small subsets of patients
This predictive benefit in terms of response rates across different solid tumors.4,16 And third, the vast
(RRs) and PFS has been confirmed in at least majority of genomic alterations that are clinically
6 molecularly selected (patients with EGFR mutation validated as predictive markers are genes encoding
positive tumors) randomized trials, including Cauca- for molecules involved in pathways related to
sian populations.10 ALK gene fusions have undergone sustained proliferation or apoptosis inhibition18
similar preclinical and clinical validation and, together (Table I). One possible exception is the tumor
with ROS-1 rearrangements, are both well-established carrying trunk mutations in genes involved in DNA
predictors for the benefit of ALK-ROS-1 TKIs.11,12 repair mechanisms. For instance, germ-line BRCA1, 2
Other targetable oncogenes, whose clinical data come mutations have been successfully validated as predic-
from smaller cohorts or early clinical trials, include tive markers for the benefit of poly ADP ribose
B-RAF and HER2 mutations, and MET amplifications, polymerase inhibitors (which are synthetic lethal in
which, with an approximate 40% RR with tailored the presence of homologous repair deficiency) in
therapies, are formally recognized as predictive targets advanced, high-grade, serous ovarian carcinomas.19,20
in NSCLC. Finally, there are other oncogenic drivers
whose targeted inhibition has shown encouraging
results in preclinical models and clinical responses in CANCER IMMUNOTHERAPY
few case series. These targets are currently being more The longstanding hypothesis of cancer immunoediting
widely studied in clinical trials, including, among is now recognized as a core process of tumorigenesis.
others, MET mutations (exon 14 skipping), RET, It is well known that many solid tumors are

July 2016 1553

Clinical Therapeutics

Table I. Oncogene addicted tumors and matched targeted therapies.

Tumor Type Prevalence, Selected Drugs with Available Clinical Response

and Gene Genomic Alteration % Data in Molecularly Selected Patients Rate, %

NSCLC
EGFR Activating mutations 1015 Gefitinib, erlotinib, afatinib, 5070*
dacomitinib, NG EGFR TKI†
ALK Gene rearrangements 25 Crizotinib, NG ALK TKI‡ 5070*
ROS-1 Gene rearrangements 12* Crizotinib, NG ROS1 TKI 5070*
HER2 Activating mutations 12 Afatinib, dacomitinib, neratinib, 2040
trastuzumab§
BRAF Activating mutations 12 Vemurafenib, dabrafenib§ 40
MET Amplification and point 15|| Crizotinib, cabozantinib, INC280§ 3040
mutation
RET Gene rearrangements 13 Cabozantinib, vandetanib, 
sunitinib, sorafenib§
NTRK1, 2, 3 Gene rearrangements 13 Entrectinib, LOXO-101§ 
FGFR1, 2, 3¶ Amplification and 520 BGJ398, AZD4547§ 
activating mutations
DDR2 ¶
Activating mutations 13 Dasatinib§ 
Melanoma 5065
BRAF Activating mutation (V600) 60 Vemurafenib, dabrafenibþtrametinib#
Breast cancer
HR Overexpression 60 Hormonal therapies** 5070
HER2 Amplification 20 Trastuzumab, pertuzumab, lapatinib 80
Prostate cancer 70 Hormonal therapies††
AR Overexpression, Abiraterone,‡‡ enzalutamide‡‡ 5070§§
Amplification
Gastric cancer
HER2 Amplification 20 Trastuzumab 4050
GIST
c-kit and Activating mutations 80 Imatinib 5085
PDGFR

AR ¼ androgen receptor; GIST ¼ gastrointestinal stromal tumor; HR ¼ hormonal therapies; Inh ¼ inhibitors; NG ¼ next
generation; TKI ¼ tyrosine kinase inhibitors.
*
Data correspond to Caucasian populations.
||
De novo. MET amplification might occur in up to 10% of EGFR-acquired resistance.
¶
Predominantly mutated in squamous-cell lung carcinomas.
†
Including osimertinib and rociletinib.
‡
Including alectinib and ceritinib.
§
Not approved drugs.
#
Trametinib is a MEK inhibitor.
**
Including LHRH agonists, tamoxifen, aromatase inhibitors.
††
Including LHRH agonists, bicalutamide.
‡‡
Castration-resistant prostate cancer.
§§
Lower responses for castration-resistant prostate cancers.

1554 Volume 38 Number 7

 J. Zugazagoitia et al.

immunogenic.21 During malignant transformation, non- Immune-suppressive inhibitory checkpoint mole-

self, tumor-associated antigens or neoepitopes resulting cules generated upon T-cell activation that regulate
from gene mutations are created, which can be recog- the immunologic synapse between T cells and den-
nized by the immune system. This process is called dritic cells in lymph nodes (eg, cytotoxic T lymphocyte
immune surveillance. At least initially, adaptive, tumor antigen-4 and B7.1), modulating T-cell activation; or
antigenspecific T-cell responses are generated, leading between T cells and tumor cells in the tumor bed (eg,
to cancer-cell elimination.21,22,23 Several steps are PD-1/L12), modulating immune rejection or the
recognized in this anti-tumor immune response that have effector phase, are among the most relevant targets
been comprehensively depicted in a seminal review by for immunotherapy.26-28 Monoclonal antibodies
Chen and Mellman,24 commonly known as the cancer blocking cytotoxic T lymphocyte antigen-4 and PD-
immunity cycle. However, it is clear that tumors finally 1/L1 receptors have yielded clinical benefits in several
escape from immune attack, despite functional immune tumor types and, contrary to classic cytotoxic chemo-
systems. Incomplete tumor elimination is followed by an therapy or even targeted therapy, they provide dura-
equilibrium phase, in which cancer cells shape their ble, long-term survival benefits. Particularly appealing
microenvironment and initiate complex mechanisms of clinical results have been reported with PD-1/L1
immune evasion that will finally lead to immune escape blockers and treatment paradigms have changed in
and tumor progression.21,25 Actually, immune evasion is an increasing list of solid tumors, particularly mela-
considered one of the core hallmarks of cancer18 and, noma, NSCLC, or kidney cancer27-29 (Table II). The
importantly, targeting these immune-suppressive mecha- long-term overall survival (OS) benefit achieved with
nisms is revolutionizing cancer treatment. Cancer cells nivolumab compared with standard second-line che-
can generate immune-suppressive networks in every step motherapy in advanced NSCLC is an example of this
of the cancer immunity cycle. Thus, tumors can disrupt encouraging clinical efficacy. In the CheckMate 017
the generation of tumor-reactive T cells (defective den- trial, 272 advanced, previously treated, squamous-cell
dritic cell maturation and activation, defective T-cell lung cancer patients were randomly assigned to
activation), T-cell trafficking (immune-suppressive che- receive nivolumab (n ¼ 135) or docetaxel (n ¼ 137).
mokine milieu) or T-cell cross of the tumor vasculature The patients who received nivolumab had a 3-month
(tumor endothelium is both a physical barrier for T cells improvement in OS (HR ¼ 0.59; 95% CI, 0.440.79;
and an active immune suppressor by generation of P ¼ 0.00025), and 1-year survival rates were 42%
angiogenic (eg, vascular endothelial growth factor) and compared with 24%. There was also a treatment benefit
immunosuppressive factors (eg, prostaglandin E2, indole- in terms of RR and PFS in favor of nivolumab.30 With a
amine 2,3-dioxygenase, TIM-3, and PD/L1-2). For those similar study design, the CheckMate 057 trial confirmed
T cells that manage to “home” the tumor, they reach an the OS improvement in the nonsquamous-cell lung
immune-suppressive local tumor microenvironment do- cancer subsets (12.2 months vs 9.7 months, respectively;
minated by tumor-associated immune-suppressive leuko- HR ¼ 0.73; 95% CI, 0.590.89; P ¼ 0.002).
cytes (regulatory T cells, myeloid-derived suppressor Intriguingly, median PFS did not favor nivolumab (2.3
cells), and a number of soluble immunosuppressive months vs 4.3 months, respectively),31 which underscores
molecules (generated mainly by these cells, which include the paradigm shift of immunotherapy compared with
transforming growth factorβ, interleukin-10, adeno- other anticancer drugs relative to its long-term survival
sine, and indoleamine 2,3-dioxygenase, among many benefits.
more), that prevent the encounter with their tumor- Remarkably, PD-1/L1 blockers have a better tox-
antigen target or suppress their immune functions. icity profile than conventional chemotherapy.29 While
Finally, cancer cells can evade the final step of immune immune-related adverse events might be relatively
attack and immune rejection by avoiding antigen recog- frequent, they are mostly mild and can be well
nition (eg, downregulation of major histocompatibility managed with adequate training. Apart from immune
complex class I molecules) or, for instance, by upregulat- checkpoint inhibitors, many other important immune-
ing a number of membrane receptors that induce based therapies are being developed, which can be
apoptotic signals in T cells (FasL and tumor necrosis practically classified into active or passive immuno-
factorrelated apoptosis-inducing ligand) or immune therapies according to the implication of host’s
tolerance (PD/L12).24,25 immunity when eliciting anti-tumor responses.

July 2016 1555

1556

Clinical Therapeutics
Table II. Practice changing Phase IIIII randomized trials with anticytotoxic T lymphocyte antigen-4 and anti PD-1/L1 agents in solid tumors.

Treatment Study First

Agent Tumor Type Scenario Phase Control Arm Response Rate Median OS Author

PD-1/L1 Nivolumab Melanoma First line III Dacarbazine 40% vs 13.9% NRe vs 10.8 mo Robert64
73% vs 42% 1-y OS
NSCLC Pretreated* III IC-CT 32% vs 11% NR Weber63
Squamous Pretreated III Docetaxel 20% vs 9% 9.2 mo vs 7.3 mo Brahmer30
Non-squamous Pretreated III Docetaxel 19% vs 12% 12.2 mo vs 9.4 mo Borghaei31
RCC Pretreated III Everolimus 25% vs 5% 25 mo vs 19.6 mo Motzer61
Pembrolizumab Melanoma Pretreated* II IC-CT 25% vs 4% NR Ribas68
First line III Ipilimumab 33.7% vs 11.9% NR in any group Robert69
74.1% vs 58.2% 1-y
OS
NSCLC Pretreated III Docetaxel 18% vs 9% 12.7 mo vs 8.5 mo Herbst70,†
CTLA4 Ipilimumab Melanoma Pretreated III Gp100 10.9 mo vs 1.5 10 mo vs 6.4 mo Hodi71
mo
First line‡ III Dacarbazine 15.2% vs 10.3% 11.2 mo vs 9.1 mo Robert72
Nivolumabþ Melanoma First line III Ipilimumab 61% vs 11% NR Postow73
Ipilimumab First line III Monotherapy§ 57.6% vs 43.7% vs NR Larkin74
19 %

IC-CT ¼ investigator’s choice chemotherapy; OS ¼ overall survival; NSCLC ¼ nonsmall-cell lung cancer; NR ¼ not reported; NRe ¼ not reached; RCC ¼ renal-cell
carcinoma.
*
Progressed after ipilimumab, or ipilimumab and a BRAF inhibitor if they were BRAF (V 600) mutation-positive.
†
Large international Phase I trial leading to pembrolizumab approval in NSCLC. Among PD-L1 þ patients (Z50%), response rate ¼ 45.2% and median OS not
Volume 38 Number 7

reached.
‡
The experimental arm consisted in ipilimumabþdacarbazine.
§
Nivolumab and ipilimumab, respectively.
 J. Zugazagoitia et al.

Active immunotherapies (which include immune panels are thought to detect multiple potentially
checkpoint modulators) depend on the host’s capa- relevant mutations across several cancer genes
bility of mounting a T-cellbased antitumor attack. (n ¼ 20300) at once, with sufficient scalability for
Contrary, passive immunotherapies, such as adoptive clinical requirements. NGS gene panels can be either
T-cell therapy, are those strategies that contain in- polymerase chain reaction (PCR) amplicon-based
trinsic anti-tumor properties.32 Immunostimulatory or hybrid capture-based. Each assay has its own
antibodies against activatory checkpoint receptors advantages and disadvantages for routine clinical
(eg, anti-CD137 or anti-CD40, among others) are application. Because PCR amplicon-based panels rely
promising targets for anticancer immunotherapy and on PCR amplification of commonly mutated exons of
are being tested in clinical trials.33 candidate genes before massively parallel sequencing,
Cancer vaccines are by far the most studied active they are subject to limitations inherent to PCR DNA
immunotherapies in the clinic across a wide variety of amplification (eg, false-positive results due to
tumor types, with mostly discouraging results.34,35 polymerase-induced artifacts in formalin-fixed paraf-
Polyvalent, cell-based cancer vaccines (eg, dendritic fin-embedded tissue samples or false-negative results
cell or tumor-cell vaccines) containing a wide range of in samples with low tumor DNA input). In addition,
tumor-associated antigens are promising, but are not their predominant applications include the identifica-
without technical difficulties for clinical applicability. tion of point mutations and deletions, so they need to
There are comprehensive reviews in the field, which be commonly combined with other techniques, such
we recommend for interested readers.36 As far as us immunochemistry of fluorescence in situ hybrid-
passive immunotherapies are concerned, adoptive ization in order to cover a wider spectrum of pre-
T-cell therapy is among the most exciting anticancer dictive genomic abnormalities (eg, copy number
immunotherapy. Early experiences in acute lympho- variations or gene rearrangements).5,39
blastic leukemia have shown encouraging clinical On the contrary, hybrid capture-based panels do not
efficacy results that merit further study in larger trials require amplicon generation and can reliably detect copy
and other solid tumors.37,38 number variations and gene rearrangements, but demand
more complex computational infrastructure.39 Several
customized gene panels are commercially available for
CHALLENGES AHEAD IN PRECISION Clinical Laboratory Improvement Amendmentcertified
ONCOLOGY laboratories and are increasingly being used in many
Clinical Implementation of Next-generation academic centers for routine genotyping and patient care.
Sequencing Technologies These early clinical experiences have found that
As stated previously, predictive genomic abnormal- customized gene panel genotyping is not only feasible
ities are uncommon, diverse in nature, and distributed and adaptable to the limitations and challenges inherent to
across many tumor types. These premises, together clinical settings (frequent limited biopsy tissue available,
with the increasing number of predictive mutations in low tumor cellularity, formalin-fixed paraffin-embedded
each tumor type, force the requirement of compre- tissue samples, or rapid turnaround time required for
hensive, multiplexed, and highly sensitive sequencing therapeutic decisions, among others),40 but is also clinically
techniques for routine clinical care.4,5 useful for routine cancer care or patient prompt selection
Classic Sanger sequencing does not fulfill these for biomarker-driven early clinical trials.41,42 However, at
criteria and is definitely not cost-effective, as it lacks least for the moment, these experiences have been limited
enough sensitivity, cannot test multiple genomic mu- to large academic centers and are not yet widely available
tations simultaneously, and is unable to detect ge- for the entire oncology community.
nomic abnormalities other than point mutations or Whole exome and genome or RNA sequencing plat-
small insertions or deletions.39 To meet these forms are rarely used for routine clinical genotyping and
requirements, high-throughput next-generation se- are mostly used for research purposes. Several issues limit
quencing (NGS) technologies have been developed.5 their clinical applicability. The first issue is the manage-
Two next-generation molecular diagnostics can be ment of massive amounts of genomic datasets that
distinguished: customized gene panels and whole require complex bioinformatics for proper data analysis,
exome and genome or transcriptome panels. Gene which, at least for the moment, is not compatible with

July 2016 1557

 Clinical Therapeutics

the short timeframe needed for therapeutic decision basket trial for BRAF V600 mutant nonmelanoma
making. The second issue is the difficulty clinically cancers was published recently.45 Notably, one of the
interpreting the high quantity of noninformative, un- key findings of this trial is that the oncogenic
known significance genomic data that they generate. The properties and subsequent clinical targetability of
third issue is the economic cost.39,40 genomic drivers are dependent on specific tumor
Whichever NGS platform is used, clinical interpretation types. As such, in this trial, encouraging activity in
of the genomic results is definitely one of the most NSCLC subsets receiving vemurafenib (RR ¼ 43%)
important challenges for the wide clinical applicability of was found, but not in other tumor types, such as
NGS technologies. We highlight 3 characteristics that colorectal cancers, anaplastic thyroid cancers,
affect the clinical relevance of a particular genomic cholangiocarcinomas, or ovarian cancers.45 On the
abnormality: functional impact (biological relevance), other hand, histology-dependent “umbrella trials”
clinical effects, and targetability. Mutations affecting restrict the inclusion criteria to a particular tumor
hotspot regions of biologically relevant cancer genes are, type, and eligible patients are genotyped for multiple
in general, functionally important, clinically relevant, and mutations that match with different drugs. These trials
mostly targetable (“driver mutations”), but this function- frequently incorporate randomization strategies (ei-
ality might be inherent to particular tumor types. How- ther biomarker-positive patients only or all
ever, it might be difficult to determine whether non- biomarker-positive and negative patients) to include
hotspot, previously undescribed or unknown significant a control arm where patients receive nontargeted
mutations in biologically relevant genes are true drivers therapy in order to assess the prognostic or predict-
that can match with effective drugs, or are instead ability of the biomarker in question.44 Examples
functionally or clinically irrelevant “passenger muta- include the I-SPY 2 (Investigation of Serial Studies to
tions.”39,43 This situation might be frequent in carcinogen- Predict Your Therapeutic Response With Imaging and
induced tumors carrying high mutational loads, such as Molecular Analysis 2; Clinicaltrials.gov ID:
melanoma or lung cancer. In addition, clinical diagnostic NCT01042379) and SAFIR-01 (High Throughput
NGS adds further complexity because matched germline Technologies to Drive Breast Cancer Patients to
DNA is usually unavailable and the risk of falsely Specific Phase III Trials of Targeted Agents; Clin-
considering a rare polymorphic variant as a potential icalTrials.gov ID: NCT01414933) trials in breast
candidate driver cannot be excluded.39 Several publically cancer or Master Protocol (A Biomarker-Driven
available databases and levels of evidence have been Master Protocol for Previously Treated Squamous
proposed in order to determine the clinical significance Cell Lung Cancer [Lung-MAP]; ClinicalTrials.gov
and actionability of genetic abnormalities.43 What is ID: NCT02154490), and ALCHEMIST (Adjuvant
evident is that expert multidisciplinary teams, including Lung Cancer Enrichment Marker Identification
pathologists, molecular biologists, geneticists, bioinfor- and Sequencing Trial; ClinicalTrials.gov ID:
matics, and oncologists, are needed for adequate data NCT02194738) trials in lung cancer.43 Another
analysis, interpretation, and therapeutic decision guiding strategy, which can be applied to basket or umbrella
in the clinic. trials, is to randomly assign genotype-selected patients
to molecularly tailored therapies versus conventional
Conducting Biomarker-driven Clinical Trials therapies instead of to specific drugs.44 The
The low mutation frequency of some genomic recently reported, histology-agnostic, French SHIVA
abnormalities, together with their wide distribution (A Randomized Proof-of-Concept Phase II Trial Com-
across many tumor types, raise major challenges when paring Therapy Based on Tumor Molecular Profiling
testing the clinical impact of precision cancer medi- Versus Conventional Therapy in Patients With Re-
cine. Designs of genotype-enriched clinical trials can fractory Cancer) study is one example of these trials.46
be classified first according to whether they are Finally, another different strategy to investigate the
histology-based or not. “Basket trials” are histology- clinical value of personalized therapy is the “N-of-1”
independent, as they include patients with different trial design, which compares, within each patient, the
tumors harboring a common genomic abnormality to clinical outcome of a mutation-tailored therapy
receive a matched therapy within the framework of a against the most recent nontargeted regimen.44 The
Phase III trial.44 As an example, an important Phase II international WINTHER trial (A Study to Select

1558 Volume 38 Number 7

 J. Zugazagoitia et al.

Rational Therapeutics Based on the Analysis of strong clinical efficacy.44 Actually, the need for
Matched Tumor and Normal Biopsies in Subjects randomized controlled trials in this setting is a
With Advanced Malignancies; ClinicalTrials.gov ID: matter of unresolved debate in the oncology
NCT01856296) is genomically and transcriptomically community.
characterizing different tumor types across 6 countries to
compare PFS rates between targeted versus most recent Tumor Heterogeneity and Resistance
untargeted therapy with a modified N-of-1 design.47 During cancer development, tumors acquire so-
All of these trial designs mentioned are mostly matic mutations in an evolutionary Darwininan
thought to investigate “early signals” of activity of a model. Cells that acquire certain mutations gain
matched targeted therapy. New adaptive statistical survival advantage and dominate localized tumor
designs or end points (eg, disease control rate at areas by displacing those lacking these genomic alter-
8 weeks or early tumor shrinkage) are being increasingly ations. This process is enhanced by consecutive clonal
used in order to optimize this proof-of-concept proc- expansions. According to this model, all cells within a
ess.44 Importantly, multicenter collaborative networks tumor would be biologically similar, and thus equally
are essential to facilitate molecular prescreening susceptible to acquiring mutations and spawning new
(comprehensive molecular profiling) and subsequent subclones. However, several lines of evidence indicate
patient selection for these clinical trials.4 Besides the that tumor initiation and progression could rely on a
challenge, there are 2 other main issues that may limit relatively minor population of self-renewing cancer
new molecularly guided drug development. The first is stem cells. These cancer stem cells would be the
target prioritization for trial eligibility.43 As stated ancestors of a much larger population of more differ-
previously, misclassification of “driver and passenger” entiated cells with limited proliferative capacity. To
mutations can result in false-negative results, with the further complicate this picture, the following cancer
consequent drug rejection or exclusion of potential stem cell models for tumor initiation and progression
candidates that could truly benefit from therapy. The have been proposed: the strictly hierarchical model, in
second issue is tumor molecular heterogeneity. Cur- which cancer stem cells are a biologically distinct
rent practice assumes that a one-site, single-tumor population within the tumor and the only ones with
biopsy is representative of the whole tumor genomic self-renewing and tumorigenic potential; and the non-
burden of the patient, underestimating the existence of hierarchical model, where potentially every tumor cell
intratumor heterogeneity. While driver mutations (particularly the transit-amplifying or progenitor cells,
usually dominate all metastatic sites and this hetero- ie, the daughters of cancer stem cells) have plasticity
geneity especially affects subclonal, probably passen- and the potential ability to de-differentiate and re-
ger mutations, increasing knowledge of tumor clonal enter the stem cell state in response to intrinsic or
evolution highlights the fact that subclonal popula- microenvironmental factors.49 Darwinian and cancer
tions might affect the growth of clonally dominant stem cell models are not mutually exclusive, as they
cells, and are also responsible for acquired oncogene still rely on linear clonal successions. However, they
resistance.43 Therefore, next-generation clinical trials are still oversimplistic because, if true, homogeneous
that take into account the premise of tumor hetero- clones would dominate tumor masses. Instead, tumor
geneity are being developed.48 Finally, when positive progression is accelerated by the progressive genomic
signals from the early clinical trials mentioned are instability inherent in cancer cells or exogenous
apparent, they usually need confirmation in larger, carcinogens (eg, sunburn and smoking), which
preferably Phase III randomized trials in order to get increase the generation of new mutations, probably
the drug approved for clinical use. Conducting large, exceeding Darwinian selection.18
Phase III randomized studies in molecularly selected In turn, the course of tumor progression, rather
cohorts with mutations that hardly represent 1% to than being linear, occurs as a branched model, with
5% of overall cases might take several years, or even tumor masses composed of increasing numbers
be unfeasible in some rare tumor types. Accelerated of genetically distinct subclones. Actually, several
drug approvals are sometimes granted by the high-throughput whole-genome sequencing studies
regulatory authorities when there is a strong across different tumor types have reported this, reveal-
biomarker or drug matching that translates into ing high genomic variability within primary tumors

July 2016 1559

 Clinical Therapeutics

and metastatic regions.50 In a breakthrough article, divided in 3 groups: those that imply second genomic
integrative genomic characterization of multiple alterations in the drug target (eg, secondary mutations
primary and metastatic lesions of 2 renal-cell carcino- in the TKI domain of EGFR or ALK), those
mas found that there was genomic, transcriptomic, and that reactivate the same pathway (eg, MEK mutation
functional heterogeneity within separate tumor sites, in BRAF-mutant melanoma treated with BRAF
resulting in different tumor subclones. In addition, inhibitors) or those involving activation of bypass
25% to 50% of the detected mutations were private track signaling pathways.53 Nongenomic alterations
(tumor regions not detected in order).51 (transcriptional and epigenetic changes) and
This high level of intratumor heterogeneity has also intratumoral immunity have been described as other
been confirmed in other tumor types, including lung sources of acquired resistance to targeted therapy, but
cancer. Multiregion whole-exome sequencing in 11 these mechanisms are less well established.56 Among
localized adenocarcinomas did reveal high levels of the first group, EGFR T790M mutations (exon 20)
intratumor heterogeneity but, importantly, the major- are detected in up to 60% of patients with acquired
ity of mutations (76%) were present in all tumors resistance to EGFR TKIs.57 Targeted therapy against
regions, and most of the driver mutations tended to be this secondary genomic event with third-generation
truncal, occurring early in tumorigenesis.52 This was EGFR TKIs (osimertinib or rociletinib) has changed
also found in the study by Gerlinger et al51 and similar the natural history of resistant EGFR disease, with
reports in other solid tumors that suggest that despite 60% RR and preliminary 10-month median PFS in
high levels of tumor heterogeneity, recurrent targets T790Mþ subsets.58,59 As far as the second group is
and driver mutations are dominant and present in concerned, functional activation of RAS-RAF-MEK-
every tumor clone and tumor region, representing ERK pathway represents probably the most important
robust therapeutic targets.50 This means that a bypass track leading to acquired resistance, both
single-site biopsy would be sufficient to detect those in EGFR and ALK-mutant lung cancers.57,60
driver mutations that can match with their respective Combination strategies targeting bypass tracks have
targeted therapies or, in other words, there is no need shown encouraging results in experimental models,55
to perform multiple biopsies and check for hetero- and are currently being investigated in clinical trials.
geneity for therapeutic decisions in this setting. How- Importantly, using these combinations as upfront
ever, increasing evidence suggests that some subclonal treatments may delay the emergence of acquired
populations may support the growth and survival of resistance or turn tumors less aggressive or
neighboring, clonally dominant cells, turning some heterogeneous upon disease progression.
regions more aggressive than others. In addition, With all of this in mind, it is evident that repeated
subclonal populations and branched clonal evolution biopsies at progression are needed to determine
are partly responsible for drug resistance.43 resistant mechanisms and their potential targeted
Tumor heterogeneity, together the emergence of inhibition. This can be a challenge in routine clinical
resistant clones on drug pressure (clonal evolution) are practice, as it might result in increased morbidity or be
definitely the main challenges for the success of unfeasible in certain cases.16 Genomic analysis of
precision oncology, as they are both closely related circulating tumor cells or circulating-free DNA
and explain the therapeutic failures in clinical practice. (“liquid biopsies”) could potentially minimize these
In the case of solid tumors, lung cancer has served as issues, and these platforms are being developed in
the main source to study acquired resistance to TKIs ongoing studies in the resistance setting and next-
in oncogene-addicted tumors. Actually, the genetic generation biomarker-driven clinical trials.43 Other
basis and some clinically relevant mechanisms of advantages include the possibility of optimal
acquired resistance in EGFR- or ALK-mutant lung genotyping in the absence of adequate tumor tissue
cancers have been characterized, enabling the develop- and monitoring the molecular evolution while patients
ment of selective drugs to overcome resistant clones.53 are on treatment.61 Finally, in order to avoid or delay
This has been done either by studying clinical samples the emergence of more aggressive mechanisms of
(re-biopsies taken from patients on TKIs at disease resistance, next-generation targeted therapies (third-
progression)54 or in experimental models.55 In generation TKI monotherapies and potentially combi-
general, mechanisms of acquired resistance can be nations) are being investigated as first-line treatments

1560 Volume 38 Number 7

 J. Zugazagoitia et al.

rather than therapies upon disease progression in treatment benefits and thus optimize treatment selection
EGFR- and ALK-mutant lung cancers. Liquid biopsies is probably the main challenge in the field of cancer
could be helpful to characterize the level of tumor immunotherapy currently. Expression of PD-L1 as-
heterogeneity in these patients, thus enabling the sessed by immunochemistry has been the most studied
selection of potentially personalized upfront therapies marker in this regard, and its potential predictive
in combination. It remains to be seen whether these capacity has been analyzed in most of the randomized
blood-based genomic profiling or other approaches to controlled trials investigating the role of PD-1/L1 block-
face heterogeneity and monitor resistance will be ers. However, the predictive role of PD-L1 expression is
scalable for clinical purposes. Table III details the far from being clarified at the moment.
most important challenges of genomics precision Evidence across different tumor types suggests that
cancer medicine. although the higher tumor PD-L1 expression the
better RR and survival rate, treatment benefits are
Predictive Markers for Immunotherapy and not necessarily restricted to PD-L1positive subsets.62
Precision Immuno-oncology In metastatic melanoma, 2 nivolumab pivotal trials
PD-1/L1 blockade has represented a real break- reported that approximately 20% to 30% of PD-
through in cancer treatment. However, long-term sur- L1negative patients responded, as compared with
vival benefits are achieved only in a small fraction of 50% of PD-L1positive subsets.63,64 Importantly,
patients (10%20%). Deciphering what characterizes improved OS was reported with nivolumab
these subsets and developing molecular biomarkers that compared with dacarbazine for PD-L1negative
can help us identify which patients will obtain larger patients in the CheckMate 066 trial (1-year OS

Table III. Premises and challenges for genomics precision oncology.

Premise Clinical requirements Main Challenges Ahead

Diverse and increasing number of Implementing next-generation Technical issues

predictive genomic abnormalities sequencing technologies for Managing large amounts
across tumor types routine care of genomic datasets
Clinical interpretation of
the genomic data
Economic costs and
accessibility
Low prevalence of some genomic Biomarker-driven trial designs to Large, multi-institutional
abnormalities and widely test their clinical impact efforts required
distributed across many tumor Long recruitment period;
types time and costs
Optimizing molecular
prescreening methods and
timing for patient selection
Target prioritization
Tumor heterogeneity
Tumor heterogeneity Liquid biopsies and next-generation Robust assay development
clinical trials Clinical translation for
routine care
Emergence of acquired Next-generation drugs targeting resistant Emerging toxicities and
resistance clones and combinatorial strategies tolerability
(upfront or at disease progression)

July 2016 1561

 Clinical Therapeutics

67.8% vs 37.4%, respectively).64 In patients with notably higher for NSCLCs with Z50% PD-L1
squamous-cell NSCLC and renal-cell carcinoma expression treated with pembrolizumab in the KEY-
treated with nivolumab, RRs and OS improved NOTE 001 and 010 trials.66,70 Several issues limit the
irrespective of PD-L1 expression.30,65 However, in interpretability of these results. The first ones are
non-squamous NSCLC, a strong predictive effect for assay-related, as different antibodies or cutoff values
PD-L1 expression was suggested. For the PD- have been used across clinical trials.62 Second, RR
L1positive subgroup, OS was doubled in favor of might not be the ideal clinical end point to measure
nivolumab (17.2 months compared with 9 months the predictive capacity of antiPD-1/L1 drugs. And
with docetaxel; HR ¼ 0.59; 95% CI, 0.430.82). third, both PD-1/PD-L1 are upregulated upon
However, no differences were observed between nivo- interferon (IFN)-γ release and are therefore inducible
lumab and docetaxel in PD-L1negative tumors (10.4 rather that tight molecular markers.28,62
months vs 10.1 months, respectively; HR ¼ 0.90; 95 In this sense, PD-L1 might be overexpressed in 2
CI, 0.661.24; P value for interaction o0.001).31 In different tumor contextures. First, aberrant oncogenic
line with these data, RRs and OS outcomes were pathways may intrinsically lead to PD-L1 upregulation,

Table IV. Potential predictive markers for PD-1/L1 blockade.

Study First
Biomarkers Premise Author

PD-L1 expression High PD-L1 expression might be correlated with better response rate Fusi62
and longer survival that absence of PD-L1 expression. However, some
PD-L1negative patients might still have durable responses and survival
benefits. Assay-related inconsistencies and the fact that PD-L1 is an
inducible marker that may vary according to tumor contextures limit the
interpretability of the results across clinical trials.
Genomic High mutational burden has been shown to correlate with longer Rizvi75
complexity overall survival as compared with low mutational burden. This has been Le76
shown in carcinogen-induced tumors (melanoma, NSCLC) and in tumors Snyder77*
developed in a context of DNA repair deficiency (colorectal
carcinoma with microsatellite instability). Of note, mismatched deficient
colorectal carcinomas show higher tumor-infiltrating lymphocytes.
Importantly, in the case of NSCLC, smokers tend to respond better that
nonsmokers, which is a paradigm shift as compared with oncogene-addicted
lung adenocarcinomas, which are almost exclusive to nonsmokers.
Transcriptomic Type I interferon-based transcriptomic signatures might be associated Ribas78
signatures with treatment benefits. This has been recently shown in metastatic
melanoma but might be in principle applicable to other tumor types.
Tumor infiltrating Pre-existing CD8þ T cells distinctly located at the invasive tumor margin Tumeh79
lymphocytes are associated with tumor regression upon PD-1/L1 blockade in metastatic
melanoma. Importantly, using next-generation sequencing
for T-cell antigen receptors, infiltrating T cells were found to be clonal,
suggesting a tumor-antigenspecific T-cell responses, which, in turn,
were inhibited by adaptive PD-L1 expression by tumor cells

NSCLC ¼ nonsmall-cell lung cancer.

*
Data for anti-CTLA4 blockade.

1562 Volume 38 Number 7

 J. Zugazagoitia et al.

resulting in constitutive PD-L1 expression. And second, bed or adjacent to immune infiltrates in the invasive
PD-L1 might be upregulated upon IFN-γ release as a margins), or in what contexture PD-L1 is overexpressed
mechanism of adaptive tumor resistance against pre- would improve its utility as a predictive marker. Beyond
existing anti-tumor immunity, reflecting immune rejec- PD-L1 expression, other promising biomarkers, includ-
tion.28,67 One could argue that adaptive, rather than ing the pre-existence of tumor-infiltrating lymphocytes,
constitutive, PD-L1 expression might be a surrogate high tumor mutational loads, or type I IFN-based
marker for better outcomes with PD-1/L1 blockers. transcriptomic signatures, have been reported to be
However, there is no clear explanation as to why some potential positive predictive markers for the benefit of
minority of PD-L1negative patients still obtain treat- these drugs (Table IV).28 Essentially, all of them are
ment responses. The complex interaction networks of surrogates of a common denominator, namely,
the immune system could explain that, even in the immunogenicity (mutational burden) or anti-tumor im-
absence of PD-L1 expression, some antitumor immunity munity (tumor-infiltrating lymphocytes, IFN signature)
could be reinstated upon PD-1/L1 blockade by the in a context of adaptive immune resistance (Figure 1). In
activation of other pathways that could lead to tumor this sense, probably one of the most promising
immune rejection.62 Therefore, PD-L1 expression as- biomarkers for immunotherapy is measuring the
sessed by immunochemistry does not seem to satisfy all genomic contexture of the tumor. Because somatic
of the methodologic and biologic properties required to mutations have the potential to encode immunogenic
become the ideal biomarker. It is likely that establishing neoantigens, tumor genomic profiling might determine
common testing criteria and better defining which cells the sensitivity not only to immune checkpoint inhibitors,
(tumor cells or immune infiltrates), localization (tumor but also for other promising immunotherapies. Actually,

CARCINOGENS DNA REPAIR DEFICIENCY

Genomic instability and high

mutational load
(NEOANTIGENS)

CD8+TILs

Type I IFN
dependent
genes
ADAPTIVE IMMUNE
RESISTANCE

PD-L1 EXPRESSION

Figure 1. Surrogates of pre-existing immune activation and adaptive immune resistance as markers for the
benefit of immune chelpoint inhibitors.

July 2016 1563

 Clinical Therapeutics

engineered, tumor-antigenspecific adoptive T-cell REFERENCES

therapies (T cells that express cloned T-cell receptors 1. Ferlay J, Soerjomataram I, Dikshit R, et al. Cancer
or chimeric antigen receptors for specific tumor antigens incidence and mortality worldwide: sources, methods
or epitopes) based on the specific patient’s mutanome, is and major patterns in GLOBOCAN 2012. Int J Cancer.
probably one of the most promising strategy for 2015;136:E359–E386.
2. Ferlay J, Steliarova-Foucher E, Lortet-Tieulent J, et al. Cancer
precision immunoncology.37,38
incidence and mortality patterns in Europe: estimates for 40
countries in 2012. Eur J Cancer. 2013;49:1374–1403.
3. Schiller JH, Harrington D, Belani CP, et al. Comparison of
CONCLUSIONS four chemotherapy regimens for advanced non-small-cell
Despite major advances, the current approach to face lung cancer. N Engl J Med. 2002;346:92–98.
cancer treatment is still reductionist. Targeting single 4. Roychowdhury S, Chinnaiyan AM. Translating genomics
molecular abnormalities or cancer pathways has for precision cancer medicine. Annu Rev Genomics Hum
achieved good clinical responses that have modestly Genet. 2014;15:395–415.
affected survival in some cancers. However, targeting 5. Simon R, Roychowdhury S. Implementing personalized
a single hallmark or pathway with a single drug cancer genomics in clinical trials. Nat Rev Drug Discov.
(“magic bullet”) will not likely lead to cancer cure. 2013;12:358–369.
6. Paez JG, Jänne PA, Lee JC, et al. EGFR mutations in lung
We predict that drug combinations against several
cancer: correlation with clinical response to gefitinib
molecular alterations or cancer hallmarks, in a way
therapy. Science. 2004;304:1497–1500.
that is similar to what we have done with HIV
7. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations
treatment, might be a promising therapeutic strategy in the epidermal growth factor receptor underlying re-
to treat cancer in the near future. Given the encourag- sponsiveness of non-small-cell lung cancer to gefitinib.
ing success of immunotherapy (particularly check- N Engl J Med. 2004;35:2129–2139.
point inhibitors), PD-1/L1 blockade may constitute a 8. Pao W, Miller V, Zakowski M, et al. EGF receptor gene
basic pillar for these combinations. Immunotherapies mutations are common in lung cancers from “never smokers”
could be combined either among each other and are associated with sensitivity of tumors to gefitinib and
(2 checkpoint inhibitors or a checkpoint inhibitor plus erlotinib. Proc Natl Acad Sci U S A. 2004;101:13306–13311.
an immunostimulatory antibody) or with other anti- 9. Mok TS, Wu Y-L, Thongprasert S, et al. Gefitinib or
cancer agents (including targeted agents in oncogene- carboplatin-paclitaxel in pulmonary adenocarcinoma.
N Engl J Med. 2009;361:947–957.
driven cancers). Toxicity will be one of the key
10. Mok T, Yang J-J, Lam K-C. Treating patients with EGFR-
limiting factors when implementing these combination
sensitizing mutations: first line or second line–is there a
strategies in the clinic, and early recognition and
difference? J Clin Oncol. 2013;31:1081–1088.
management of adverse events will surely be a core 11. Solomon BJ, Mok T, Kim D-W, et al. First-line crizotinib
component for treatment success. Ultimately, we versus chemotherapy in ALK-positive lung cancer. N Engl J
evoke that personalized combination strategies ac- Med. 2014;371:2167–2177.
cording to pathways or hallmarks that specifically 12. Shaw AT, Ou S-HI, Bang Y-J, et al. Crizotinib in ROS1-
drive each patients’ tumor biology within next- rearranged non-small-cell lung cancer. N Engl J Med. 2014;
generation precision oncology initiatives will be one 371:1963–1971.
of the main challenges and most promising strategies 13. Morgensztern D, Campo MJ, Dahlberg SE, et al. Molec-
for cancer treatment in the future. ularly Targeted Therapies in Non–Small-Cell Lung Cancer
Annual Update 2014. J Thorac Oncol. 2015;10:S1–S63.
14. Kris MG, Johnson BE, Berry LD, et al. Using multiplexed
assays of oncogenic drivers in lung cancers to select
CONFLICTS OF INTEREST
targeted drugs. JAMA. 2014;311:1998–2006.
The authors have indicated that they have no conflicts
15. Clinical Lung Cancer Genome Project (CLCGP), Network
of interest regarding the content of this article. Genomic Medicine (NGM). A genomics-based classification
of human lung tumors. Sci Transl Med. 2013;5. 209ra153.
16. Arnedos M, Soria J-C, André F, Tursz T. Personalized
ACKNOWLEDGMENTS treatments of cancer patients: a reality in daily practice, a
All the authors contributed in the writing and review- costly dream or a shared vision of the future from the
ing of the article. oncology community? Cancer Treat Rev. 2014;40:1192–1198.

1564 Volume 38 Number 7

 J. Zugazagoitia et al.

17. Sordella R, Bell DW, Haber DA, cancers. Br J Cancer. 2015;112: diagnostics for routine clinical use:
Settleman J. Gefitinib-sensitizing 1421–1427. overcoming the roadblocks to per-
EGFR mutations in lung cancer 30. Brahmer J, Reckamp KL, Baas P, sonalized cancer medicine. Ann On-
activate anti-apoptotic pathways. et al. Nivolumab versus Docetaxel col. 2015;26:1830–1837.
Science. 2004;305:1163–1167. in Advanced Squamous-Cell Non- 40. Gagan J, Van Allen EM. Next-
18. Hanahan D, Weinberg RA. Small-Cell Lung Cancer. N Engl J generation sequencing to guide can-
Hallmarks of cancer: the next gen- Med. 2015;373:123–135. cer therapy. Genome Med. 2015;7:80.
eration. Cell. 2011;144:646–674. 31. Borghaei H, Paz-Ares L, Horn L, 41. Rodon J, Saura C, Dienstmann R,
19. Jayson GC, Kohn EC, Kitchener HC, et al. Nivolumab versus Docetaxel et al. Molecular prescreening to
Ledermann JA. Ovarian cancer. Lan- in Advanced Nonsquamous Non- select patient population in early
cet. 2014;384:1376–1388. Small-Cell Lung Cancer. N Engl J clinical trials. Nat Rev Clin Oncol.
20. Ledermann J, Harter P, Gourley C, Med. 2015;373:1627–1639. 2012;;9:359–366.
et al. Olaparib maintenance ther- 32. Galluzzi L, Vacchelli E, Bravo-San 42. Hollebecque A, Massard C, Soria
apy in patients with platinum- Pedro J-M, et al. Classification of J-C. Implementing precision medi-
sensitive relapsed serous ovarian current anticancer immunotherapies. cine initiatives in the clinic: a new
cancer: a preplanned retrospective Oncotarget. 2014;5:12472–12508. paradigm in drug development.
analysis of outcomes by BRCA sta- 33. Melero I, Grimaldi AM, Perez- Curr Opin Oncol. 2014;26:340–346.
tus in a randomised phase 2 trial. Gracia JL, et al. Clinical develop- 43. Andre F, Mardis E, Salm M, et al.
Lancet Oncol. 2014;15:852–861. ment of immunostimulatory mono- Prioritizing targets for precision
21. Schreiber RD, Old LJ, Smyth MJ. clonal antibodies and opportunities cancer medicine. Ann Oncol. 2014;
Cancer immunoediting: integrating for combination. Clin Cancer Res. 25:2295–2303.
immunity's roles in cancer suppres- 2013;19:997–1008. 44. Dienstmann R, Rodon J, Tabernero
sion and promotion. Science. 2011; 34. Vansteenkiste JF, Cho B, Vanakesa J. Optimal design of trials to dem-
331:1565–1570. T, et al. MAGRIT, a double-blind, onstrate the utility of genomically-
22. Tartour E, Zitvogel L. Lung cancer: randomized, placebo-controlled guided therapy: Putting Precision
potential targets for immunotherapy. phase III study to assess the efficacy Cancer Medicine to the test. Mol
Lancet Respir Med. 2013;1:551–563. of the RECMAGE-A3 þ AS15 can- Oncol. 2015;9:940–950.
23. Galon J, Angell HK, Bedognetti D, cer immunotherapeutic as adjuvant 45. Hyman DM, Puzanov I, Subbiah V,
Marincola FM. The continuum of therapy in patients with resected et al. Vemurafenib in Multiple Non-
cancer immunosurveillance: prog- MAGE-A3-positive Non-Small Cell melanoma Cancers with BRAF V600
nostic, predictive, and mechanistic Lung Cancer (NSCLC) [Abstract]. Mutations. N Engl J Med. 2015;373:
signatures. Immunity. 2013;39:11–26. Ann Oncol. 2014;25(suppl 4):iv409. 726–736.
24. Chen DS, Mellman I. Oncology meets 35. Butts C, Socinski MA, Mitchell PL, 46. Le Tourneau C, Delord J-P,
immunology: the cancer-immunity et al. Tecemotide (L-BLP25) versus Gonçalves A, et al. Molecularly
cycle. Immunity. 2013;39:1–10. placebo after chemoradiotherapy targeted therapy based on tumour
25. Motz GT, Coukos G. Deciphering for stage III non-small-cell lung molecular profiling versus conven-
and reversing tumor immune sup- cancer (START): a randomised, tional therapy for advanced cancer
pression. Immunity. 2013;39:61–73. double-blind, phase 3 trial. Lancet (SHIVA): a multicentre, open-
26. Pardoll DM. The blockade of im- Oncol. 2014;15:59–68. label, proof-of-concept, rando-
mune checkpoints in cancer immu- 36. Melero I, Gaudernack G, Gerritsen mised, controlled phase 2 trial.
notherapy. Nat Rev Cancer. 2012; W, et al. Therapeutic vaccines for Lancet Oncol. 2015;16:1324–1334.
12:252–264. cancer: an overview of clinical trials. 47. Rodon J, Soria J-C, Berger R, et al.
27. Ribas A. Releasing the Brakes on Nat Rev Clin Oncol. 2014;11:509–524. Challenges in initiating and con-
Cancer Immunotherapy. N Engl J 37. June CH, Riddell SR, Schumacher ducting personalized cancer therapy
Med. 2015;373:1490–1492. TN. Adoptive cellular therapy: a trials: perspectives from WINTHER,
28. Topalian SL, Drake CG, Pardoll race to the finish line. Sci Transl a Worldwide Innovative Network
DM. Immune checkpoint blockade: Med. 2015;7. 280ps7. (WIN) Consortium trial. Ann Oncol.
a common denominator approach 38. Rosenberg SA, Restifo NP. Adoptive 2015;26:1791–1798.
to cancer therapy. Cancer Cell. 2015; cell transfer as personalized immu- 48. Catenacci DVT. Next-generation
27:450–461. notherapy for human cancer. clinical trials: Novel strategies to
29. Homet Moreno B, Ribas A. Anti- Science. 2015;348:62–68. address the challenge of tumor
programmed cell death protein-1/ 39. Heuckmann JM, Thomas RK. A new molecular heterogeneity. Mol Oncol.
ligand-1 therapy in different generation of cancer genome 2015;9:967–996.

July 2016 1565

 Clinical Therapeutics

49. Chaffer CL, Weinberg RA. How does 62. Fusi A, Festino L, Botti G, et al. 71. Hodi FS, O'Day SJ, McDermott DF,
multistep tumorigenesis really pro- PD-L1 expression as a potential et al. Improved survival with ipili-
ceed? Cancer Discov. 2015;5:22–24. predictive biomarker. Lancet Oncol. mumab in patients with metastatic
50. Swanton C. Intratumor heterogeneity: 2015;16:1285–1287. melanoma. N Engl J Med. 2010;363:
evolution through space and time. 63. Weber JS, D'Angelo SP, Minor D, 711–723.
Cancer Res. 2012;72:4875–4882. et al. Nivolumab versus chemother- 72. Robert C, Thomas L, Bondarenko I,
51. Gerlinger M, Rowan AJ, Horswell S, apy in patients with advanced et al. Ipilimumab plus dacarbazine for
et al. Intratumor heterogeneity and melanoma who progressed after previously untreated metastatic mela-
branched evolution revealed by anti-CTLA-4 treatment (CheckMate noma. N Engl J Med. 2011;364:2517–
multiregion sequencing. N Engl J 037): a randomised, controlled, 2526.
Med. 2012;366:883–892. open-label, phase 3 trial. Lancet 73. Postow MA, Chesney J, Pavlick AC,
52. Zhang J, Fujimoto J, Zhang J, et al. Oncol. 2015;16:375–384. et al. Nivolumab and ipilimumab ver-
Intratumor heterogeneity in localized 64. Robert C, Long GV, Brady B, et al. sus ipilimumab in untreated mela-
lung adenocarcinomas delineated Nivolumab in previously untreated noma. N Engl J Med. 2015;372:
by multiregion sequencing. Science. melanoma without BRAF mutation. 2006–2017.
2014;346:256–259. N Engl J Med. 2015;372:320–330. 74. Larkin J, Chiarion-Sileni V, Gonzalez
53. Camidge DR, Pao W, Sequist LV. 65. Motzer RJ, Escudier B, McDermott R, et al. Combined Nivolumab and
Acquired resistance to TKIs in solid DF, et al. Nivolumab versus Everolimus Ipilimumab or Monotherapy in Un-
tumours: learning from lung cancer. in Advanced Renal-Cell Carcinoma. treated Melanoma. N Engl J Med. 2015;
Nat Rev Clin Oncol. 2014;11:473–481. N Engl J Med. 2015;373:1803–1813. 373:23–34.
54. Katayama R, Shaw AT, Khan TM, et al. 66. Garon EB, Rizvi NA, Hui R, et al. 75. Rizvi NA, Hellmann MD, Snyder A,
Mechanisms of acquired crizotinib re- Pembrolizumab for the treatment et al. Cancer immunology. Mutational
sistance in ALK-rearranged lung Can- of non-small-cell lung cancer. landscape determines sensitivity to
cers. Sci Transl Med. 2012;4. 120ra17. N Engl J Med. 2015;372:2018–2028. PD-1 blockade in non-small cell lung
55. Crystal AS, Shaw AT, Sequist LV, 67. Ribas A. Adaptive Immune Resistance: cancer. Science. 2015;348:124–128.
et al. Patient-derived models of How Cancer Protects from Immune 76. Le DT, Uram JN, Wang H, et al.
acquired resistance can identify ef- Attack. Cancer Discov. 2015;5:915–919. PD-1 Blockade in Tumors with
fective drug combinations for can- 68. Ribas A, Puzanov I, Dummer R, Mismatch-Repair Deficiency. N Engl
cer. Science. 2014;346:1480–1486. et al. Pembrolizumab versus J Med. 2015;372:2509–2520.
56. Hugo W, Shi H, Sun L, et al. Non- investigator-choice chemotherapy 77. Snyder A, Makarov V, Merghoub T,
genomic and Immune Evolution of for ipilimumabrefractory mela- et al. Genetic basis for clinical response
Melanoma Acquiring MAPKi Resist- noma (KEYNOTE-002): a rando- to CTLA-4 blockade in melanoma.
ance. Cell. 2015;162:1271–1285. mised, controlled, phase 2 trial. N Engl J Med. 2014;371:2189–2199.
57. Chong CR, Jänne PA. The quest to Lancet Oncol. 2015;16:908–918. 78. Ribas A, Robert C, Hodi FS, et al.
overcome resistance to EGFR- 69. Robert C, Schachter J, Long GV, Association of response to pro-
targeted therapies in cancer. Nat et al. Pembrolizumab versus Ipili- grammed death receptor 1 (PD-1)
Med. 2013;19:1389–1400. mumab in Advanced Melanoma. N blockade with pembrolizumab (MK-
58. Jänne PA, Yang JC-H, Kim D-W, Engl J Med. 2015;372:2521–2532. 3475) with an interferon-inflammatory
et al. AZD9291 in EGFR inhibitor- 70. Herbst RS, Baas P, Kim DW, et al. immune gene signature [Abstract].
resistant non-small-cell lung cancer. Pembrolizumab versus docetaxel for J Clin Oncol. 2015;33(suppl):3001.
N Engl J Med. 2015;372:1689–1699. previously treated, PD-L1-positive, 79. Tumeh PC, Harview CL, Yearley JH,
59. Sequist LV, Soria J-C, Goldman JW, advanced non-small-cell lung can- et al. PD- 1 blockade induces re-
et al. Rociletinib in EGFR-mutated cer (KEYNOTE-010): a randomised sponses by inhibiting adaptive im-
non-small-cell lung cancer. N Engl J controlled trial. Lancet. Dec 18, 2015 mune resistance. Nature. 2014;515:
Med. 2015;372:1700–1709. [Epub Ahead Of Print]. 568–571.
60. Wilson FH, Johannessen CM, Piccioni
F, et al. A functional landscape of
resistance to ALK inhibition in lung
cancer. Cancer Cell. 2015;27:397–408.
61. Haber DA, Velculescu VE. Blood-
based analyses of cancer: circulating Address correspondence to: Luis Paz-Ares, MD, PhD, Medical Oncology
tumor cells and circulating tumor Department, Hospital Universitario 12 de Octubre, Avenida de Córdoba,
DNA. Cancer Discov. 2014;4:650–661. s/n, 28041, Madrid, Spain. E-mail: [email protected]

1566 Volume 38 Number 7