Clinical Pharmacogenomics: Premarket Evaluation in Early-Phase Clinical Studies and Recommendations for Labeling

Administrative guidance, not enacted legislation. This captured publisher edition preserves the listed document and attachments; later developments are not independently consolidated.

10300.fnl.doc
Guidance for Industry
Clinical Pharmacogenomics: Premarket
Evaluation in Early-Phase Clinical
Studies and Recommendations for
Labeling
U.S. Department of Health and Human Services
Food and Drug Administration
Center for Drug Evaluation and Research (CDER)
Center for Biologics Evaluation and Research (CBER)
Center for Devices and Radiological Health (CDRH)
January 2013
Clinical Pharmacology
Clinical/Medical

Guidance for Industry
Clinical Pharmacogenomics: Premarket
Evaluation in Early-Phase Clinical
Studies and Recommendations for
Labeling
Additional copies are available from:
Office of Communications
Division of Drug Information, WO51, Room 2201
10903 New Hampshire Ave.
Silver Spring, MD 20993-0002
Phone: 301-796-3400; Fax 301-847-8714
http://www.fda.gov/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/default.htm
or
Office of Communication, Outreach, and Development (HFM-40)
Center for Biologics Evaluation and Research
Food and Drug Administration
1401 Rockville Pike, Rockville, MD 20852-1448
http://www.fda.gov/BiologicsBloodVaccines/GuidanceComplianceRegulatoryInformation/Guidances/default.htm
(Tel) 800-835-4709 or 301-827-1800
or
Office of Communication, Education, and Radiation Programs
Division of Small Manufacturers, International and Consumer Assistance
Center for Devices and Radiological Health
Food and Drug Administration
10903 New Hampshire Ave.
WO66, Room 4613
Silver Spring, MD 20993-0002
http://www.fda.gov/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/default.htm
Email: dsmica@cdrh.fda.gov
Fax: 301-827-8149
(Tel) Manufacturers Assistance: 800-638-2041 or 301-796-7100
(Tel) International Staff: 301-796-5708
U.S. Department of Health and Human Services
Food and Drug Administration
Center for Drug Evaluation and Research (CDER)
Center for Biologics Evaluation and Research (CBER)
Center for Devices and Radiological Health (CDRH)
January 2013
Clinical Pharmacology
Clinical/Medical

TABLE OF CONTENTS
I. 	INTRODUCTION................................................................................................................. 3
II. 	BACKGROUND ................................................................................................................... 4
A. 	Genetic Differences ........................................................................................................ 4
B. 	Pharmacogenomics Studies ............................................................................................ 5
III. PROSPECTIVE DNA SAMPLE COLLECTION ............................................................. 7
IV. CLINICAL EVALUATION OF PHARMACOGENOMICS ........................................... 9
A. 	General Considerations ................................................................................................... 9
B. 	Clinical Pharmacogenomic Studies .............................................................................. 10
1. 	PK and PD Studies in Healthy Volunteers ................................................................... 10
2. 	PK and PD Studies in Patients ...................................................................................... 12
3. 	Dose-Response (D/R) Studies....................................................................................... 13
C. 	Specific Considerations in Study Design ...................................................................... 13
1. 	Overview ....................................................................................................................... 13
2. 	Study Population ........................................................................................................... 14
3. 	Multiple Covariate Considerations ............................................................................... 15
4. 	Dose Selection .............................................................................................................. 15
5. 	Measurements of Interest .............................................................................................. 16
6. 	Statistical Considerations .............................................................................................. 16
V. 	INCLUDING PHARMACOGENOMIC INFORMATION IN LABELING ................ 19
REFERENCES ............................................................................................................................ 21
APPENDIX — Examples of the Value of Pharmacogenomics ................................................... 22

Contains Nonbinding Recommendations
3
Guidance for Industry1
Clinical Pharmacogenomics: Premarket Evaluation
in Early-Phase Clinical Studies and Recommendations for Labeling
This guidance represents the Food and Drug Administration's (FDA's) current thinking on this topic. It
does not create or confer any rights for or on any person and does not operate to bind FDA or the public.
You can use an alternative approach if the approach satisfies the requirements of the applicable statutes
and regulations. If you want to discuss an alternative approach, contact the FDA staff responsible for
implementing this guidance. If you cannot identify the appropriate FDA staff, call the appropriate
number listed on the title page of this guidance.
I. 	INTRODUCTION
This guidance is intended to assist the pharmaceutical industry and other investigators engaged
in new drug development in evaluating how variations in the human genome, specifically DNA
sequence variants, could affect a drug’s pharmacokinetics (PK), pharmacodynamics (PD),
efficacy, or safety.2 The guidance provides recommendations on when and how genomic
information should be considered to address questions arising during drug development and
regulatory review.
The application of pharmacogenomic approaches during drug development is an evolving
process that begins with discovery and continues through confirmation of clinical efficacy and
safety outcomes. The focus of this guidance, however, is to provide advice on general principles
of study design, data collection, and data analysis in early-phase trials. This guidance does not
address trial design or statistical analysis considerations for later-phase, randomized, controlled
clinical trials that are intended to draw definitive conclusions about treatment effects in a
genomic subgroup (e.g., enrichment designs, adaptive enrichment designs, simultaneous
hypothesis testing overall and within subgroups), or co-development of a drug and in vitro
diagnostic.
Rather, the considerations here are more relevant for exploratory and observational studies
intended to generate genomic hypotheses that may then be tested in prospectively designed phase
3 trials. For instance, early-phase data on genomic-dependent dosing or efficacy, even when not
definitive, can provide guidance on dosing or patient selection in later-phase trials, or inform the
strategy for further collection of genetic and related biomarker data in later controlled trials.
1 This guidance has been prepared by the Office of Clinical Pharmacology, Office of Translational Sciences, and an
Interdisciplinary Working Group with representatives from the Center for Drug Evaluation and Research (CDER),
Center for Biologics Evaluation and Research (CBER), and the Center for Devices and Radiological Health (CDRH)
at the Food and Drug Administration (FDA).
2 For the purposes of this guidance, the term drug includes both small molecule and biological products.

Contains Nonbinding Recommendations
4
FDA’s guidance documents, including this guidance, do not establish legally enforceable
responsibilities for industry. Instead, guidances describe the Agency’s current thinking on a
topic and should be viewed only as recommendations, unless specific regulatory or statutory
requirements are cited. The use of the word should in Agency guidances means that something
is suggested or recommended, but not required.
II. 	BACKGROUND
Pharmacogenomics (PGx) refers broadly to the study of drug exposure and/or response as
related to variations in DNA and RNA characteristics (see the International Conference on
Harmonization (ICH) E15 Definitions for Genomic Biomarkers, Pharmacogenomics,
Pharmacogenetics, Genomic Data and Sample Coding Categories).3 Drug exposure refers to the
PK profile following administration. Drug response refers to the PD response to the drug; that
is, all of the effects of the drug on any physiologic and pathologic processes, including those
related to effectiveness and those related to adverse reactions.
PD effects are generally characterized by an exposure-response (E/R) relationship of drugs, and
genetic differences can lead to changes in the steepness of the E/R curve, the location of the
curve (i.e., change in EC50 ), the maximum effect (Emax ), and other features of the E/R
relationship. The definition of PGx in this document focuses mainly on DNA sequence
variations and is not intended to include other related characteristics, such as proteomics and
metabolomics.4
A. 	Genetic Differences
Genetic differences between individuals can affect virtually all aspects of a disease and its
treatment, including the rate of disease occurrence; the risk of disease progression or recurrence;
the drug or drug class most likely to provide benefit; the therapeutic dose; the nature and extent
of beneficial responses to treatment; and the likelihood of drug toxicity. The genetic differences
likely to be of most relevance in drug development are those associated with genes in four broad
categories: (1) genes relevant to the drug’s PK (absorption, distribution, metabolism (including
formation of active metabolites), and excretion (ADME)); (2) genes that code for intended or
unintended drug targets and other pathways related to the drug’s pharmacologic effect; (3) genes
not directly related to a drug’s pharmacology that can predispose to toxicities such as immune
reactions; and (4) genes that influence disease susceptibility or progression. All of these genetic
factors can affect the benefit–risk profile of a drug product.
In general, the effects of genetic differences on PK are easier to characterize because they affect
a readily measured feature of the drug (its pharmacokinetics). There are far more examples of
3 FDA guidances, including ICH guidances, are available on FDA’s guidance Web page. We update guidances
periodically. To make sure you have the most recent version of a guidance document, check FDA’s guidance Web
site at FDA Basics for Industry > Guidances.
(http://www.fda.gov/ForIndustry/FDABasicsforIndustry/ucm234622.htm).
4 For the purposes of this guidance, the terms genetic and genomic are used interchangeably.

Contains Nonbinding Recommendations
5
genetic PK differences than examples of genetic differences on PD in most therapeutic areas. In
many cases, the mechanism that causes differences in PK is related to metabolizing enzymes or
transport proteins with well-established genetic polymorphisms, as is the case for CYP2C9,
CYP2C19, CYP2D6, or SLCO1B1, so that such differences can be anticipated. In these cases,
DNA sample collection, blood and/or urine drug concentration data, and well-characterized
phenotypic information are needed to determine the extent to which genetic polymorphisms in
metabolism and/or transporter genes influence exposure to drugs and/or active metabolites and
responses.
In contrast to genetic differences affecting PK, genetic differences affecting PD are more
difficult to detect because clinical effects are more complex traits, generally more variable
among individuals and complicated by variability over time within the same individual. Clinical
effects may also be influenced by many factors, including imprecision in measurement and
differences in drug exposure. Genomic effects on PD can, however, profoundly affect selection
of the optimal dose based on differences in dose- or exposure-response, safety, and efficacy. For
example, in addition to genetic variants in its metabolizing enzyme (CYP2C9), genetic variations
in warfarin’s target (VKORC1) significantly affect response and dose requirements (see warfarin
product labeling for details).
Drug product labeling has increasingly included information obtained during drug development
on the treatment effect or likelihood of treatment response in a subset of patients with a particular
genetic/genomic status, on altered risk–benefit balance in genetic subgroups, or on the need to
genotype to guide dosing. Drug product labeling has also been revised after approval, generally
based on postmarket experience, to include PGx information that can inform the benefit–risk
relationship or allow dosing of the drug to be adjusted for individuals (see Appendix for select
examples). It is hoped that ascertainment of genomic information throughout drug development
will enable earlier discovery of clinically important genomic differences (i.e., before marketing).
B. 	Pharmacogenomics Studies
PGx studies can contribute to a greater understanding of interindividual differences in the
efficacy and safety of investigational drugs. PGx research depends on the collection and use of
biological samples to generate data.
Across the drug development continuum, genomic data may be used for several purposes,
including (1) identifying the basis for PK outliers and intersubject variability in clinical response;
(2) ruling out the role of polymorphic pathways as clinically significant contributors to variable
PK, PD, efficacy, or safety; (3) estimating the magnitude of potential drug–drug interactions; (4)
investigating the molecular or mechanistic basis for lack of efficacy or occurrence of adverse
reactions; and (5) designing clinical trials to test for greater effects in specific subgroups (i.e., use
in study enrichment strategies) (Zineh and Pacanowski, 2010).
Drug development is commonly described in phases (21 CFR 312.21). The first two phases
provide initial information about safety and efficacy and ideally examine a broad range of doses,
so that the larger, later adequate and well-controlled trials (phase 3) that are generally needed to
support marketing approval can be efficiently designed (e.g., with a narrower and reasonable

Contains Nonbinding Recommendations
6
range of doses, good patient selection criteria, and appropriate endpoints). In addition,
information about PK and pertinent PD effects (generally effects on biomarkers considered
indicative of activity and/or effectiveness) can provide proof-of-concept, which may improve
success of later trials. Information on PD often includes evidence of E/R (usually dose-response
but sometimes supplemented by concentration-response modeling) and, when possible, pertinent
subset information (e.g., demographic, disease severity) that may improve the design of phase 3
trials by refining dose selection and identifying patients with potentially greater responses.
PGx assessment in early-phase clinical studies may:
1. 	Identify populations that should receive lower or higher doses of a drug, or longer
titration intervals, based on genetic effects on drug exposure, dose-response, early
effectiveness, and/or common adverse reactions. Often, these differences are related to
differences in drug absorption, distribution, excretion, or metabolism. The latter are
generally identified by genetic variations in polymorphic enzymes that define metabolic
status. However, genetic factors that affect dose- or exposure-response relationships also
could help define the dose range for later trials.
2. 	Identify responder populations based on phenotypic, receptor, or genetic characteristics, a
critical element in treatment individualization that has been used primarily in the
oncologic setting. Predicted differences in response can lead to enrichment strategies
based on such predictive markers.
3. 	Identify high-risk groups. Drugs that cause serious adverse effects generally will not be
acceptable in most settings, unless the adverse effects can be predicted. There are
examples of genetic markers (typically in immunology-related genes like those in the
HLA family) that effectively predict susceptibility to serious adverse reactions (see
Appendix). It is possible to link adverse drug reactions to genetic factors (e.g.,
metabolic, immunologic) and manage them in later trials, supporting approval of drugs
with particular value, such as those that address an unmet medical need. To date, the
most likely use of such information would be to identify poor metabolizers or ultra-rapid
metabolizers (e.g., CYP2D6) whose blood levels of parent or relevant metabolites could
be markedly affected — in trials, they could be excluded or their doses modified to
account for genetic variations. Genetic information in early-phase studies can also be
useful in characterizing the maximum tolerated doses.
The phase 1 and 2 studies considered in this guidance are often described as exploratory in that
they are not intended to provide the definitive evidence of safety and effectiveness needed to
support drug approval. Nonetheless, PGx studies can provide mechanistic support and be
applied in the design or analysis of later trials, potentially improving their efficiency or
likelihood of success if the genetic factor can help predict the likelihood and magnitude of
response.
Genomic analyses are also common in phase 3 trials (although often not prespecified), and
responses can be examined in relation to genotype. In most cases, such post hoc analyses of
efficacy will be considered exploratory, but in certain cases, strong subset findings of

Contains Nonbinding Recommendations
7
significantly altered benefit–risk balance have been included in labeling (e.g., KRAS mutations
with cetuximab and panitumumab indicated for colon cancer). PGx studies in phase 3 trials can
also provide valuable information on the efficacy and safety impact of genetic effects identified
in early-phase studies.
Phase 2 studies that suggest genomic influences can lead to phase 3 trials that incorporate
findings into prespecified hypotheses. Examples might include enriching the study with
genomically defined individuals, determining a dose based on demonstrated variability in earlier
studies, or defining a priori hypothesis testing of a primary endpoint in a genomic subset.
Decisions to pursue prospective strategies in phase 3 trials will depend on a variety of factors,
such as the clinical context, availability of alternative treatments, and the risk–benefit profile of
the drug. When prospective strategies to apply genetic information to the use of a drug are
planned, early consultation with the appropriate centers (i.e., CDER, CBER, and/or CDRH) is
highly recommended.
It is FDA’s position that if a companion diagnostic is required for therapeutic selection, an FDA-
approved or -cleared test will be required at the same time that the drug is approved.5 An in vitro
PGx test would be considered a companion diagnostic device if it will provide information that is
essential for the safe and effective use of a therapeutic product as directed in labeling.
III. 	PROSPECTIVE DNA SAMPLE COLLECTION
This section of the guidance specifically pertains to collection of DNA samples from appropriate
tissue sources (e.g., blood, buccal) for analysis of germline (i.e., inherited) gene variations as
opposed to acquired somatic mutations (e.g., tumor mutations). This section focuses on DNA
specimens; nevertheless, many of the principles discussed may apply to other genetic materials
obtained from human tissues, such as RNA or proteins, for which pre-treatment expression
patterns may be used to characterize a disease and predict patient’s therapeutic response or
prognosis.
An important prerequisite to successful use of genetic information in drug development is the
appropriate collection and storage of DNA samples from a large enough number of participants
in clinical trials, both exploratory studies and the adequate and well-controlled trials intended to
demonstrate effectiveness and safety. Ideally, baseline DNA samples should be collected from
all patients in all arms of clinical trials in all phases of drug development. The voluntary and
incomplete nature of many exploratory genetic studies often raises concerns about potential bias
and statistical power, which could compromise the scientific rigor of such studies. Because
premarket clinical trials represent the largest and richest source of controlled clinical data for
drugs, the drug development and public health value of DNA banking and importance of
complete data should be recognized and efforts made to collect baseline genetic information
when appropriate.
5 FDA has issued a draft guidance for industry and FDA staff In Vitro Companion Diagnostic Devices. Once
finalized, that guidance will reflect the Agency’s thinking on this topic.

Contains Nonbinding Recommendations
8
When known genetic factors are likely to influence the efficacy, safety, or dosing of the
investigational drug, the comparator, or the background therapy, DNA should be collected from
all subjects to specifically test those genetic factors for subgroup analysis. Specifically, genomic
objectives should be stated in the protocol and described in the informed consent form as
applicable in light of the known properties (e.g., metabolism, mass balance, E/R relationship) of
the investigational drug or the comparator, including genomic objectives related to drug
metabolizing enzyme variants; clinically relevant transporter variants; drug target or signaling
protein variants that clearly affect activity or expression; or biomarkers for disease prognosis or
response to background therapy.
Potential PGx differences in efficacy and/or safety can arise from genetic variants that are not yet
as well characterized as the metabolism or transporter genes and, thus, not anticipated in
advance. Therefore, general DNA sample collection for exploratory analyses should be a routine
consideration, and is strongly encouraged (or plans are). It then becomes possible to seek
explanations for differences in PK, PD, efficacy, tolerability, or safety that were not anticipated
prior to beginning the study, noting, of course, potential multiplicity and bias issues.
In some situations, DNA collection may ultimately have limited utility and, therefore, may not
be needed. This may be the case, for example, for trials evaluating topical drugs for which
systemic exposure is expected to be low. However, collecting DNA to characterize the role of
genetic factors is particularly important for drugs with high intersubject variability in PK or PD;
bimodal or trimodal distributions for measured PK or PD parameters; observed PK or PD
differences between racial or ethnic groups; narrow therapeutic ranges; or potential safety issues
(e.g., structural safety alerts, QT prolongation, liver enzyme abnormalities). Because many of
these signals are not evident until a large number of subjects have been exposed, plans to collect
DNA should be considered at the outset of a drug development program. When possible under
applicable laws, regulations, and ethics committee polices, consent for DNA collection for
exploratory analyses should be obtained from all participants in all arms of clinical trials. FDA
recognizes that clinical trials are often conducted globally, and international regulations and
ethics committee policies vary with respect to collection and storage of genetic material (Warner
et al., 2011; Ricci et al., 2011).
When complete sample acquisition is not possible, efforts should be made to obtain as high a
sample acquisition rate as possible, with additional target collection of DNA samples from those
subjects who are identified as PK or PD outliers, who experience a clinical endpoint of interest,
or who experience a severe or treatment-limiting adverse event. In cases of incomplete sample
acquisition, the specific reasons should be described (e.g., in the applicant’s study report) and
any potential bias estimated when possible.
DNA collection and storage plans should be specified before initiation of a study to minimize the
potential for sample selection bias, even if these samples are studied only at a later time during or
after a study. An effort should be made to collect genetic samples at enrollment and/or at
baseline to avoid potential bias associated with delayed collection (e.g., selection of tolerant or
responsive subjects). This is particularly important when many patients do not complete the
study, do not comply with the protocol, withdraw from the trials before experiencing a clinical
outcome, or die during the trial. For drugs used in acute care settings, it may not be feasible to

Contains Nonbinding Recommendations
9
collect DNA on a voluntary basis (i.e., using a separate consent), nor would a genomic
biomarker be easily applied in practice. DNA collection in this scenario should still be
considered for supportive analyses if indicated, but the timing of consent and sample collection
may occur later in the trial as appropriate to the clinical scenario.
Because historically many pharmacogenomic relationships have been identified several years
after the drug has been marketed, DNA should be retained in the event that new genomic issues
arise after the completion of the studies when possible under applicable laws, regulations, and
ethics committee polices. Samples should be retained for a time period that will permit
postmarket analysis should the need arise (e.g., at least 15 years). Alternatively, when data are
generated from higher-throughput technologies (e.g., genome-wide arrays, ADME chips, and
high-throughput sequencing), data appropriate to the study objectives may be retained instead of
samples, when possible under applicable laws, regulations, and ethics committee polices. As a
general matter, sample retention is preferred in the event the need arises to re-assay the samples.
Routine collection of DNA samples in phase 2 or 3 trials could provide applicants with an
opportunity to investigate the causes of lack of efficacy or the occurrence of toxicity in different
individuals, using such approaches as exploratory genome-wide association investigations and
candidate gene or targeted pathway analyses. So, although the need for genomic tests and
possible relationship to dosing, efficacy, or safety may not be suspected at study initiation, they
may become of interest at a later time. This is especially relevant when such samples can be
analyzed along with samples from other similar studies. Therefore, whenever possible, informed
consent procedures should anticipate this possibility. Attention should be paid to the appropriate
sample collection, handling, storage, and retention to enable possible exploration of genotype-
phenotype relationships after study completion.
Samples that can be used for DNA analysis include a range of biological materials such as blood
or buccal cells. In addition to germline mutations, there are also somatic or acquired mutations
to be considered, for example, in biopsies from tumors. Like germline mutations, not only can
somatic changes be related to drug response, but they may also predict the severity of a disease
and disease prognosis (e.g., likelihood of metastasis) and can be used to identify subgroups of
patients most suited for outcome trials because of a relatively large rate of events (prognostic
enrichment). There are, for example, a variety of breast cancer markers that can predict
likelihood of recurrence, a potentially critical selection factor for an adjuvant trial. Information
to support the quality and integrity of DNA during sample collection and storage, along with
information to show that the DNA material can be used for consistent and reproducible analysis,
should be provided in an applicant’s study report.
IV. 	CLINICAL EVALUATION OF PHARMACOGENOMICS
A. 	General Considerations
The value of DNA sample collection and the information that analysis of these samples can
provide will vary for different drugs and indications. Known PGx factors identified during
nonclinical assessments that affect PK or PD should be considered, particularly when the

Contains Nonbinding Recommendations
10
threshold between activity and toxicity is narrow. Exploratory human PGx investigations
generally begin with in vitro studies, followed by clinical pharmacology studies in humans to
assess the PK and PD properties of the drug that might be associated with gene variants in
metabolizing enzymes, transporters, or drug target receptors. For example, if in vitro studies
show that a molecule’s metabolism in human cell systems relies on a well-established
polymorphic gene, such as CYP2C19, and metabolism is a major route of elimination in humans,
it would almost always be important to determine the contribution of genomic factors to
variability in PK. These data inform decisions about whether subsequent clinical studies need to
take PGx differences into account (e.g., in dose-response studies). When exposures are variable
and exposure is correlated with response, these findings may subsequently translate to dose
selection.
Various technology platforms are available to rapidly characterize the contribution of established
and evolving allelic variations of hundreds of metabolism and transporter genes simultaneously
in clinical pharmacology studies to generate hypotheses and better understand variability in PK
and/or PD related to genomic factors. Candidate gene approaches are reasonable when a drug
interacts with receptors and other targets that have known functional polymorphisms, or when
the drug is subject to metabolism or transport by a protein that has known functional
polymorphisms. When the pharmacology of the drug (i.e., mechanism of action or
metabolism/transport) is not well-characterized, high-throughput platforms (such as ADME gene
or genome-wide arrays or high-throughput sequencing) should be considered. Because higher-
throughput technologies may generate false positives, it is important to confirm the role of a
newly identified marker with experimental evidence, such as in vitro enzyme or transport studies.
To design informative studies and interpret study results appropriately, careful attention should
be paid in clinical pharmacology studies to differences, if known, in the prevalence of ADME-
related gene variants among racial or ethnically distinct groups. The genetic markers selected for
analysis should be appropriate to the population being studied. In smaller studies, it is prudent to
include rarer functional alleles for evaluation so subjects are not misclassified because a
particular (rare) allele was not tested. Drug-metabolizing phenotype assignment based on
genotypes should be carefully considered in clinical pharmacology studies.
B. 	Clinical Pharmacogenomic Studies
In vitro studies of metabolism, transport, or drug targets could help identify the need for human
PGx studies and contribute to the design and analysis of those studies. The following types of
clinical pharmacology studies provide opportunities to prospectively integrate PGx factors for
assessing interindividual variability and its implications for subsequent clinical studies.
Additionally, retrospective analysis of individual or pooled studies (e.g., of similar design,
independent subjects) can be performed to characterize genetic effects on PK or PD. In later-
stage phase 3 clinical trials intended to support efficacy and safety conclusions in genomic
subsets, stratified randomization or planned subset assessment would generally be expected.
1. 	PK and PD Studies in Healthy Volunteers

Contains Nonbinding Recommendations
11
Single- and multiple-dose PK studies provide important initial information on drug PK
and can suggest the level of interindividual variability in PK that can be expected in later
trials. These studies can provide information on common gene variants affecting ADME,
and collection of DNA samples from all participants is recommended so that analysis can
be performed on individual subjects to retrospectively evaluate the causes of PK outliers
and to help explore the PK parameter distribution.
When there are serious concerns about the toxicity of an investigational drug due to
excess exposure at higher doses in individuals with genetically mediated alteration in
metabolism (e.g., for drugs with anticipated dose-limiting toxicity and
polymorphically mediated metabolism), prospectively genotyping subjects will
identify those subjects who are at risk, so that they can (1) receive lower doses or (2)
be excluded from PK studies. Subjects with variant genotypes can then be enrolled
after there is a better understanding of the in vivo relevance of the metabolic pathway
and the relationship between drug exposure and safety. For example, in phase 1
oncology studies designed to characterize the maximum tolerated dose, it would be
important to understand whether excess exposures are restricted to subjects with
certain genotypes so that that these subjects do not inappropriately cause a generally
suboptimal dose to be selected as the maximum tolerated dose — a critical
consideration for determining which doses to carry forward to subsequent trials.
When in vitro and/or mass balance studies suggest that an investigational drug is
metabolized to a large degree by a polymorphic pathway (e.g., CYP2D6), PGx analyses
should be conducted in single- and/or multiple-dose PK studies in healthy subjects to
evaluate common gene variants with known phenotypic effects to determine the extent of
variability and the maximal differences in systemic exposure between genotypes. In
most cases, genotyping can be performed retrospectively (after completion of the studies)
to evaluate observed variability in PK and PD if relevant PD measurements are available.
Because many clinical pharmacology studies may be individually underpowered to
definitively characterize the effects of gene variants on PK or PD, even those that are
common, a pooled analysis of independent subjects from multiple studies with similar
protocols may be necessary.
In some situations, it may be advisable to conduct dedicated clinical pharmacology
studies with balanced, prospective, genotype-based enrollment, or to target enrollment of
subjects with variant genotypes to permit meaningful retrospective analyses. Such
strategies may be appropriate when exposures are strongly correlated with response or
safety and the genetic factors expected to affect metabolism or transport are not common
enough in the general population to be evaluated in an individual clinical pharmacology
study. Also, it may be important to quantify the PK or PD differences between genotypes
in a controlled manner after clinically relevant pharmacogenetic effects have been
observed. When considered in the context of population PK and other analyses, such
studies may help define genotype-specific dosing strategies.
Special consideration should be given to a drug for which conversion to an active
metabolite from an inactive parent compound occurs through a polymorphic metabolism

Contains Nonbinding Recommendations
12
pathway. It is advisable to characterize the metabolism of the drug and study the
biological activity of the relevant metabolites early in drug development. In these cases,
differences in metabolite exposure among individuals may have implications for dosing,
efficacy, and safety. When drug metabolites have important clinical effects (e.g.,
clopidogrel and codeine), failure to form the active metabolite (e.g., in poor metabolizers)
may have profound effects on response while overproduction (e.g., in ultrarapid
metabolizers) may result in untoward toxicities (e.g., codeine). It is particularly critical
from the beginning of development in these cases to include subgroups of subjects with
genetic variants in metabolic or transporter pathways.
Strategies other than a targeted candidate gene approach can also be useful for probing
the causes of variability early in drug development, even before there is understanding of
the influence of genetic factors on drug response. Strategies include routine retrospective
genotyping of subjects in early-phase clinical trials using a gene chip that includes a large
number of possible candidate metabolism and transporter genes, some of which may
influence PK and/or PD. Such studies may identify the gene variants that contribute to
PK or PD variability that were not anticipated based on prior in vitro or clinical data.
Exploration of other genetic contributors to PK is indicated when the exposure-response
relationship is well-defined, observed drug exposures are variable across subjects, and in
vitro data do not implicate specific metabolism pathways or transporters. Because of the
multiple hypotheses being tested with this approach, associations can be confirmed by in
vitro studies or staged analysis designs (e.g., discovery cohort followed by replication
cohort) before considering additional prospective clinical studies.
For a drug that is a substrate of a polymorphic enzyme or transporter, the evaluation of
comparative PK in the subgroups genetically defined as UM (ultra-rapid metabolizers),
EM (extensive metabolizers), IM (intermediate metabolizers), and PM (poor
metabolizers) often provides essential information on potential drug–drug interactions.
This is helpful in prioritizing subsequent drug interaction studies and in estimating the
extent of interactions by that specific pathway. For example, the difference in drug
exposures between EM and PM subgroups would generally represent the most extreme
change that could be caused by a strong inhibitor of that pathway. Similarly, an
alternative to a genotype-specific PK study in the case of a polymorphic metabolizing
enzyme is to administer the investigational drug to extensive metabolizers with and
without concomitant administration of a known strong inhibitor of the metabolic
pathway. For example, an individual who is a CYP2D6 EM can be converted to a de
facto CYP2D6 PM by concomitant administration of a strong CYP2D6 inhibitor. For
drug interaction study design and selection of strong inhibitors of metabolic pathways,
refer to the FDA guidance for industry on Drug Interaction Studies — Study Design,
Data Analysis, and Implications for Dosing and Labeling.
2. 	PK and PD Studies in Patients
If important variability in PK of active species (i.e., parent drug and/or its active
metabolite) is observed in healthy volunteers, the significance of this finding should be
considered in the design of subsequent studies in patients (e.g., in dose/response studies

Contains Nonbinding Recommendations
13
in genotype-defined subgroups). When specific genotypes are shown to reliably predict
blood levels and drug effect, this knowledge can be used in the subsequent design of
other clinical trials, for example, by using genotypes to (1) select patients for trials (e.g.,
enrichment with potential responders and elimination of patients likely to experience
toxicity); (2) stratify groups within trials; and (3) adjust doses in trials. These steps can
increase the average effect, decrease toxicity, and improve the chances of overall success
of the study. The trials can also reveal exposure differences that are not clinically
critical, in which case additional PGx studies may be given lower priority.
3. 	Dose-Response (D/R) Studies
D/R studies are usually conducted in phase 2 using biomarkers or clinical endpoints that
are relevant to clinical efficacy and safety to (1) provide proof of concept, (2) identify
doses for phase 3 trials, and (3) establish dose-response for relatively common adverse
effects. Both PK differences (i.e., metabolism and transport) and PD differences (e.g.,
shift in concentration-response curves) can lead to differences in D/R in individuals. If
previous PK and/or PD studies suggest that a genotype or phenotype is important in
influencing systemic exposure-response or efficacy and safety responses, D/R studies that
stratify dose groups by genotype or specific genotype-guided D/R studies (PK adjusted
D/R or even a concentration controlled study) should be considered. In the latter studies,
doses are defined by expected blood levels in individuals rather than by administered
dose.
Drug blood level evaluation in D/R studies, even if the study was not planned to assign
patients to groups by blood levels, can help interpret results when there are major
differences in blood levels resulting from genomic factors as well as apparent variability
in D/R relationships. Genetic information can be included in D/R or
concentration/response (C/R) models much like any other clinical covariate (see section
C.3, Multiple Covariate Considerations). Explanations related to genomic factors can
sometimes be persuasive on their own or can lead to hypotheses to be tested in further
studies, where patients would be stratified by genotype.
C. 	Specific Considerations in Study Design
1. 	Overview
The choice of study design depends on prior knowledge and the purpose of the study.
The study is straightforward when the goal is to compare PK in genomically defined
subgroups of healthy volunteers or patients. Often, PK can be assessed in the relevant
subgroups in the same study. This design is similar to studies in people with hepatic or
renal impairment. These studies provide information on exposure in genomically defined
subgroups and, depending on an understanding of the PD consequences of blood levels
(e.g., concentration/response relationships), could influence dosing in later randomized
controlled trials. The information and results would generally be included in product
labeling.

Contains Nonbinding Recommendations
14
Less well developed than such PK studies are study designs used to identify genomically
distinguished PD responses. When a particular genomic influence is reasonably well
understood, patients should be stratified and responses analyzed by subset. This would
be possible, for example, for oncology settings when tumor markers (e.g., estrogen
receptor status, EGFR mutation status, and KRAS mutation status) are thought to predict
response, and historically when other markers have had similar potential (e.g., high vs.
low renin, systolic or diastolic dysfunction in heart failure). In a more exploratory
setting, PK-PD studies or D/R studies measuring a biomarker can be examined for
genomic predictors of PD effects. Earlier exploratory studies in drug development will
generally look for effects on such a marker, while studies of clinical endpoints would be
assessed later in drug development. In some cases, of course, there is no marker, and
only clinical outcomes can be studied. In later trials, such relationships, if not
anticipated, would in most cases be considered exploratory (i.e., needing prospectively
defined confirmation).
Analytical validation of genotyping and phenotyping methods should be established
before initiating a clinical PGx study. Appropriate quality control materials, standards,
and calibrators (where applicable), as well as validated protocols, should be established to
provide assurance of continuing analytical performance over time and across testing sites.
Analytical validity is critical if the genomic biomarker is intended to select patients for
entry into pivotal efficacy studies; if the results of the genomic test are intended to
determine whether a patient is to receive the drug, the assay may require approval under
an Investigational Device Exemption by CDRH. For use as an in vitro device companion
diagnostic, the test must be approved or cleared, except in specified circumstances, at the
same time the drug is approved for marketing (see footnote 5). For specific advice on
analytical and clinical validity requirements for FDA approval or clearance of a genomic
assay, sponsors should consult the Office of In Vitro Diagnostics and Radiological
Devices (OIR) in CDRH, especially if co-development of a drug and diagnostic is
planned.
2. 	Study Population
Clinical PGx studies focused on pharmacokinetics are usually performed in phase 1 using
healthy volunteers, with additional attention to the effects of sex, age, and race/ethnicity.
Safety considerations may preclude the use of healthy volunteers for certain drug classes
(e.g., cytotoxic anticancer drugs). Studies of patients for whom the investigational drug
is intended provide the opportunity to explore PD or clinical endpoints not measurable in
healthy volunteers.
The exclusion of patients with certain genotypes from a clinical trial should be
considered on a case-by-case basis, but may be appropriate when the
concentration/response relationship is reasonably well known and it is clear that subjects
with certain genotypes would experience toxicities or not respond to the low exposure of
active drug that would be achieved. Similarly, the clear absence of a drug target (e.g., the
gene for a cell surface receptor needed for anticancer activity) might lead to exclusion of
such patients. A potential problem, however, may be lack of information as to what

Contains Nonbinding Recommendations
15
receptor level is needed for response. Many quantitative genomic markers do not lead to
“all or none” responses; therefore, inclusion of patients with a range of target levels in
early-phase trials is advisable to enable exploration and refinement of optimal diagnostic
cut-offs for prospective testing in phase 3 trials.
When a drug–drug interaction study is intended to evaluate the impact of an
investigational drug as an inhibitor of enzyme metabolism, it would not be appropriate to
enroll poor metabolizers of that enzyme if it is polymorphic. In some instances, an
evaluation of the extent of drug interactions in subjects with various genotypes may be
helpful (refer to the FDA’s guidance for industry on Drug Interaction Studies — Study
Design, Data Analysis, and Implications for Dosing and Labeling.
3. 	Multiple Covariate Considerations
Many observable phenotypes of drug response in humans result from the interactions of
multiple factors or covariates, including genetic, demographic, and environmental factors.
The understanding of specific covariates (e.g., age, sex, and race) and gene-covariate
interactions on variability in drug response could be useful in understanding the relative
impact of genetics, versus other nongenetic factors, on the PK, PD, dosing, efficacy, and
safety of a drug. For example, some clinical studies have found that genetic variants in
CYP2C9 and VKORC1 increase the risk of bleeding in patients taking standard doses of
warfarin. However, genetic variations are not the only factors that increase the risk of
bleeding. The dose of warfarin in the context of a patient’s body surface area (BSA) and
age may also influence the bleeding risk. In most cases, therefore, it is important to
understand the risk associated with multiple factors — both genetic and nongenetic.
Mathematical simulations using population-based, physiological PK models (i.e.,
physiologically based pharmacokinetic (PBPK) models) that simultaneously integrate
various patient-intrinsic and -extrinsic factors can provide an understanding of the
potential complex changes in E/R relationships in patients when multiple covariates are
present. Some applications of these models, including the design of clinical trials to
evaluate the effects of drug-metabolizing enzyme polymorphisms on PK and PD, are
increasingly being seen in regulatory submissions (Zhao et al., 2011).
4. 	Dose Selection
A clinical PGx study should be conducted at relevant clinical doses. A lower dose or
different titration interval can be used in subjects with certain genotypes that could cause
high and unsafe exposure or excessive pharmacological response to the drug.
Interpretation of findings in a clinical PGx study, such as changes in exposure in specific
genotypes, may be aided by a good understanding of dose- or concentration-response
relationships for both desirable and undesirable drug effects in the general population and
in subpopulations with different genetic variations. FDA’s guidance for industry on
Exposure-Response Relationships — Study Design, Data Analysis, and Regulatory
Applications discusses considerations in the evaluation of D/R and
concentration/response (C/R) relationships.

Contains Nonbinding Recommendations
16
5. 	Measurements of Interest
 	PK Parameters
PK measurements and parameters that should be useful for consideration of genotypic
effects on drug exposure include AUC, Cmax , and time-to-Cmax (Tmax ), as well as other PK
parameters such as clearance, volumes of distribution, and half-lives. Additional
measures, such as trough drug concentrations in multiple-dose PK studies (if associated
with a PD measurement, an efficacy endpoint, or an adverse effect) can help determine
appropriate dosing strategies to achieve similar exposure across different subsets of the
population.
 	Biomarkers of Drug Response (PD)
Biomarkers of drug response related to a drug’s intended pharmacological effect,
suspected off-target effects, and/or safety, when available, should be incorporated into
clinical PGx studies to measure whether or not genetic factors influencing exposure or
target response will have an impact on clinical outcomes.
6. 	Statistical Considerations
Clinical studies evaluating the effect of PGx factors during early drug development are
intended to address questions concerning variability in PK and short-term PD endpoints
in healthy volunteers and patient groups. Although there are well-established variants in
metabolism (e.g., CYP2D6 and CYP2C19 poor metabolizers), most pharmacogenomic
research is exploratory at this stage and is often intended to discover relationships for
which no prior hypotheses exist. Even when genomic factors are expected to be
important (e.g., genetic determinants of receptor characteristics), there may be
considerable uncertainty as to clinical consequences.
Statistical considerations in PGx studies are important. The hypotheses and conclusions
arising from early-phase clinical studies (e.g., need for different doses for different CYP
genotypes) should be sufficiently supported with credible data. When not definitive,
those data should define a hypothesis to be pursued later in drug development with more
rigorous study designs. Considering the limited number of subjects within or across
early-phase clinical pharmacology studies, such studies may be able to identify only large
differences in PK resulting from genomic differences (e.g., CYP2D6 poor metabolizers)
that are likely to be clinically relevant; genomic PD differences will generally need
further study.
For exploratory and observational studies, which generally do not involve randomization
of subjects to treatment, the statistical concepts that are most relevant to clinical
pharmacogenomics include the following:
1. 	Avoid confounding by balancing the testing of samples on all controllable
confounding factors, which include technical variations (e.g., day, operator,

Contains Nonbinding Recommendations
17
instrument, reagent lot, sampling handling/storage/preparation/dilution) and
patient variations (e.g., outcome, disease stage, treatments, co-morbidities, other
medical history, demographics, clinical covariates). This can be accomplished by
employing experimental design principles such as replication, blocking,
randomization (e.g., of sample test order), and masking/blinding (e.g., of assay
operator to outcome, and, conversely, of outcome evaluator to assay result).
2. 	Control for the multiplicity and the risk of incorrectly identifying associations in
genomic data when many searches are performed (e.g., using Bonferroni
correction, false discovery rate, or other method).
3. 	Quantitatively characterize the preliminary marker classifier performance (e.g.,
sensitivity, specificity, predictive values, reclassification, or other metrics) and
explore the prognostic and predictive attributes of the marker when appropriate.
4. 	Address the potential for population stratification due to admixture and other
confounding factors (i.e., factors that are responsible for the effect and that are
associated with a genetic difference).
5. 	Assess the reproducibility of a genetic finding (strength of association, magnitude
of association, repeatability in multiple datasets, subgroup response) so that it can
be used reliably and evaluated in follow-up development in later-phase clinical
trials.
Highly multiplexed genotyping methods are available, for instance, to interrogate
hundreds or thousands of sequence variations in ADME-related genes. These methods
may be useful when exploring causes for PK variability in the absence of in vitro data
that suggest a single causal pathway. Early-phase studies are often not adequately
powered to define PGx relationships when only a few subjects with a variant genotype
may have been enrolled. Consequently, pooled analysis of clinical pharmacology studies
with similar protocols enables more comprehensive and reliable investigations of
multiple genetic factors while retaining the ability to control for false positive results.
The correlative analysis methods employed in such studies are similar to the
pharmacometric methods used to evaluate the effects of age, race, and other clinical
covariates on drug exposure. Considering the exploratory nature of these investigations,
replication of the findings or experimental studies (e.g., follow-up in vitro studies) may
be recommended to confirm clinical relationships between a genetic factor and drug
exposure. Typical methods to filter genetic markers for association testing should be
applied (e.g., Hardy-Weinberg equilibrium deviations).
In PGx substudies that are not randomized, it is possible for the substudy population to be
different from the overall trial population in a variety of ways. If a trial is randomized
and the substudy is selected based on a baseline feature, the groups in the randomized
sample, if they are a reasonably large fraction of the sample, should be sufficiently
similar to allow for meaningful assessment. In all cases, however, a between-group
difference could be the result of the genetic difference or of some characteristic related to
the genetic difference. The statistical analysis of genetic substudies should include an
assessment of differences in clinical or demographic characteristics and an assessment of
heterogeneity in treatment effects between the substudy and overall trial populations.
When incomplete sampling is related to geographic or site differences in participation,

Contains Nonbinding Recommendations
18
treatment effects at the participating sites should also be estimated. Power analysis
should be performed to determine what effect sizes can feasibly be identified in a subset
of the trial, including the power to detect genotype by treatment interactions.
Statistical issues for early pharmacogenomic assessments can be divided roughly into the
following three categories according to the use of the PGx information:
1. 	To define differences in metabolism or clearance that will affect the PK of a drug
When a genetically polymorphic pathway has an important role in the drug’s PK, the
magnitude of the variability in PK parameters according to genotype should be examined.
In some cases, the effect of the altered PK will confer an important effect on PD or
clinical endpoints, depending on the C/R relationship. Studies will initially simply
compare PK in groups with normal and gene-variant metabolism, and often the
differences will be large and readily described. If the clinical consequences of the
genetic differences need to be investigated, studies in patients with normal and genetic-
variant profiles should be studied (in separate studies, or preferably as strata within a
single study), and the differences in response between the two groups should be noted.
2. 	To define differences in the magnitude or presence of a favorable response to a
treatment
When genetic differences do not lead to differences in blood levels of active drug or
metabolite, but rather to changes in the C/R relationship (e.g., slope in C/R curve, or at
the extreme, lack of any PD effect), validation and precise definition of the genomic
difference should be evaluated in a trial in which genetic strata can be studied, preferably
in the same trial. Thus, cancer patients whose tumors bear a particular genetic marker
and patients without the marker would be stratified in a controlled trial and differences in
response between the groups studied.
A critical question is how such markers would be identified initially as an a priori
hypothesis. If not understood as a mechanism of drug action initially, they should be
identified later, after studying the exposure of an unselected population to the drug.
Response could then be explored in relation to a wide range of markers, possibly with
samples from related studies. Considering the small sample size available in premarket
clinical databases, exploratory studies such as genome-wide association studies can be
powered to detect only large genetic effects. Even if a strong association were seen, it
would generally be necessary to conduct a focused examination to replicate the
relationship in recognition of the possibility of false findings from multiple testing and
bias within the initial dataset and the typically small sample sizes at this stage. A plan for
an interim evaluation, perhaps of a critical biomarker, could provide early confirmation if
the effect were in fact large.
In addition, if use of the test result for the purpose of selecting patients or dosing
regimens for clinical trials leads to labeling that directs treatment to a genomically
defined patient subset, the test would ordinarily require CDRH approval.

Contains Nonbinding Recommendations
19
3. 	To identify genomic predictors of an increased likelihood of an adverse effect
Relating a PGx characteristic to adverse drug reactions (ADRs) requires, first, analysis of
patients who experience the ADR to seek a common PGx characteristic. This would
generally involve a case-control approach, comparing the frequency of the putative PGx
predictor in patients with and without the adverse reaction. Given the generally small
sample sizes and substantial multiplicity in such searches, a very large genetic effect on
risk for the adverse event and clear statistical significance for the association would
generally be needed to be persuasive. In many cases, it may not be possible to
characterize PGx relationships for relatively rare or severe adverse reactions in the
premarket setting. High baseline sample acquisition rates and comprehensive follow-up
data may, however, allow for better characterization of the relationship between a genetic
variant and risk for adverse reactions. As a general matter, in addition to high sample
acquisition rates, a clinically meaningful definition of cases and clearly described
controls would be important. Alternative definitions of cases and controls may be useful
as sensitivity analyses (e.g., using different thresholds for biochemical laboratory values
to define drug-induced organ injury). The ultimate usefulness of the data will depend on
aspects of the results that include, but might not be limited to, the strength of the
association, consistency of the findings across subgroups (e.g., race groups/countries),
and relevance to the U.S. population, as well as the intention for the data's use (e.g.,
removal of a clinical hold, modified labeling).
V. 	INCLUDING PHARMACOGENOMIC INFORMATION IN LABELING
In general, labeling should include information on PGx only if it is useful to inform prescribers
about the impact (or lack of impact) of genotype on phenotype; or to indicate whether a genomic
test is available, and, if so, to indicate whether testing should be considered, is recommended, or
is necessary. If applicable, a “Pharmacogenomics” subsection should be included in the
CLINICAL PHARMACOLOGY section (e.g., as “12.5 Pharmacogenomics”) of the prescribing
information (PI) and should include clinically relevant data or information on the effect of genetic
variations affecting drug therapy.
In labeling, pharmacogenomic information can include, but is not limited to, the following:
 	Information on the frequencies of alleles, genotypes, haplotypes, or other genomic markers
of relevance
 	Description of the functional effects of genomic variants (e.g., genetically based differences
in enzyme activity such as reduced cytochrome P450 enzyme activity attributable to
polymorphisms in a CYP gene)
 	Effect of genotype on important PK parameters or PD endpoints
 	Description of pharmacogenomic studies that provided evidence of genetically-based
differences in drug benefit or risk
 	Dosing and patient selection recommendations based on genotype

Contains Nonbinding Recommendations
20
When the information has important implications for the safe and effective use of a drug and the
consequences of the genetic variations result in recommendations for restricted use, dosage
adjustments, contraindications, or warnings, this information will be summarized in other
sections of the labeling, as appropriate (e.g., BOXED WARNING, INDICATIONS AND
USAGE, DOSAGE AND ADMINISTRATION, CONTRAINDICATIONS, WARNINGS AND
PRECAUTIONS, DRUG INTERACTIONS). A cross-reference to the more detailed data
supporting the clinical recommendation (typically in CLINICAL PHARMACOLOGY or
CLINICAL STUDIES) should be included. If there is PGx information that is important to
convey to patients, it should be summarized in the PATIENT COUNSELING INFORMATION
section (17).
Detailed information about clinically relevant genetic information should be consolidated into the
most appropriate labeling section. The following chart shows types of PGx information that
could appear in various sections of labeling. 6
Section of Label 	Types of Information
INDICATIONS AND USAGE 	PGx information related to proper patient
selection (e.g., the need for PGx testing)
DOSAGE AND ADMINISTRATION 	Dosing recommendations for subgroups of
patients based on genetic makeup
BOXED WARNING,
CONTRAINDICATIONS, WARNINGS AND
PRECAUTIONS, and/or ADVERSE
REACTIONS
PGx information affecting drug safety
WARNINGS AND PRECAUTIONS and USE
IN SPECIFIC POPULATIONS
Genotype(s) that are known to be associated
with an adverse reaction in a specific
population
DRUG INTERACTIONS 	Relevant information concerning the role of
genetic variations in drug-drug interactions,
and the clinical consequences of the
combination of genetic polymorphisms in
protein(s) in the context of the drug’s
metabolism, transport, and action
CLINICAL PHARMACOLOGY 	PGx impact on PK or PD
CLINICAL STUDIES (if studied and the
evidence is substantial; or if observed neutral
findings (i.e., lack of a pharmacogenetic effect)
would be pertinent clinical information)
Efficacy differences related to PGx
6 FDA has issued the draft guidance for industry Labeling for Human Prescription Drug and Biological Products —
Implementing the New Content and Format Requirements. Once finalized, this guidance will represent the
Agency’s thinking on this topic.

Contains Nonbinding Recommendations
21
REFERENCES
FDA Guidances 7
Drug Interaction Studies — Study Design, Data Analysis, and Implications for Dosing and
Labeling (Draft)
E15 Definitions for Genomic Biomarkers, Pharmacogenomics, Pharmacogenetics, Genomic
Data and Sample Coding Categories (ICH guideline)
Exposure-Response Relationships — Study Design, Data Analysis, and Regulatory Applications.
In Vitro Companion Diagnostic Devices (Draft)
Labeling for Human Prescription Drug and Biological Products — Implementing the New
Content and Format Requirements (Draft)
Publications
Mallal S, et al. HLA-B*5701 screening for hypersensitivity to abacavir. N Engl J Med.
2008;358(6):568-79.
Mega JL, et al. Cytochrome p-450 polymorphisms and response to clopidogrel. N Engl J Med.
2009;360(4):354-62.
Mega JL, et al. Genetic variants in ABCB1 and CYP2C19 and cardiovascular outcomes after treatment
with clopidogrel and prasugrel in the TRITON-TIMI 38 trial: a pharmacogenetic analysis. Lancet.
2010;376(9749):1312-9.
Ricci DS, et al. Global requirements for DNA sample collections: results of a survey of 204 ethics
committees in 40 countries. Clin Pharmacol Ther. 2011;89(4):554-61.
Warner AW, et al. Challenges in obtaining adequate genetic sample sets in clinical trials: the perspective
of the industry pharmacogenomics working group. Clin Pharmacol Ther. 2011;89(4):529-36.
Zhao P, et al. Application of physiologically based pharmacokinetic (PBPK) modeling and
simulation in regulatory review. Clin Pharmacol Ther. 2011 Feb;89(2):259-67.
Zineh I, Pacanowski MA. Pharmacogenomics in the assessment of therapeutic risks versus
benefits: inside the United States Food and Drug Administration. Pharmacotherapy.
2011;31(8):729-35.
7 FDA guidances, including ICH guidances, are available on FDA’s guidance Web page. We update guidances
periodically, and draft guidances will be finalized. To make sure you have the most recent version of a guidance
document, check FDA’s guidance Web site at FDA Basics for Industry > Guidances.
(http://www.fda.gov/ForIndustry/FDABasicsforIndustry/ucm234622.htm).

Contains Nonbinding Recommendations
22
APPENDIX — Examples of the Value of Pharmacogenomics
PGx information obtained from genomic investigations during the course of drug development
can inform drug development and enhance the effectiveness and safety of drugs. Several
postapproval examples have illustrated the value of pharmacogenomic discoveries in
understanding variable dose requirements (e.g., warfarin (VKORC1/CYP2C9), tetrabenazine
(CYP2D6), pimozide (CYP2D6)), failure of intended pharmacological effect (e.g., clopidogrel
(CYP2C19), cetuximab and panitumumab (KRAS)), and adverse reaction predisposition (e.g.,
abacavir (HLA-B), carbamazepine (HLA-B)). 8
These examples illustrate principles that, if considered early in drug discovery and development,
can (1) provide experimental evidence for a pharmacogenomic interaction (or, importantly,
evidence that genetic variation is actually not clinically relevant); (2) define the relevance of
polymorphic pathways to drug metabolism and pharmacokinetics; (3) help determine whether
entry criteria for dose-finding studies should be modified to include or exclude certain subsets of
patients; (4) identify important stratification and enrichment factors for clinical trials across
phases of drug development; and (5) aid in development of instructive labeling for therapeutic
products under review for regulatory approval.
Abacavir (Ziagen): Improving the Benefit–Risk (B/R) Relationship by Identifying Patients at
High Risk for a Serious Adverse Reaction
Abacavir, which is used alone or in combination with other drugs, is an antiretroviral
drug used in the treatment of HIV-1 infection. An abacavir hypersensitivity reaction
(HSR) was observed in about 5 to 8% of clinical trial patients, so that hypersensitivity
was a well-recognized problem at the time of approval. The clinical manifestations of the
HSR included fever and/or rash and, to a lesser degree, gastrointestinal symptoms
(nausea, vomiting, diarrhea, and stomach pain) and/or respiratory symptoms (cough,
shortness of breath, and sore throat) that emerged within the first 6 weeks of treatment in
more than 90% of patients with HSR. Symptomatology worsened with continued therapy
and could be life-threatening, but usually resolved upon discontinuation of the drug.
Clinical diagnosis was imprecise because of the patients’ concurrent illness or drug
treatments, and there was an HSR rate of 2 to 3% in the standard of care arm without
abacavir in blinded clinical trials. The hypersensitivity reactions were an important
limitation to the use of abacavir.
Approximately 3 to 4 years after marketing approval of abacavir, new PGx research
identified an allele (HLA-B*5701) that appeared to be associated with the
hypersensitivity reactions, but the sensitivity and specificity of this predictor of HSR
varied between studies and racial populations. A 6-week, randomized controlled trial
(called PREDICT-1) (Mallal S et al., 2008) was undertaken to assess the clinical utility of
8 The examples provided illustrate historical approaches to updating the labels of drugs. Currently, FDA expects
that if a diagnostic test is essential for the safe and effective use of a therapeutic product (i.e., a companion
diagnostic), there should be a cleared/approved test with the appropriate intended use available concurrent with the
drug label change.

Contains Nonbinding Recommendations
23
HLA-B*5701 screening before beginning abacavir treatment. Abacavir-naïve patients (n
= 1956) were randomized 1:1 to an abacavir-containing regimen with HSR monitoring
according to standard of care (control arm), or to an abacavir-containing regimen with
HSR monitoring preceded by prospective HLA-B*5701 screening (PGx arm). In the PGx
arm, patients who tested positive for HLA-B*5701 were excluded, and only HLA-B*5701
negative patients were enrolled.
The trial had two co-primary endpoints: the rate of clinically suspected HSR, and the rate
of immunologically confirmed HSR, defined as HSR with a positive patch test reaction.
The incidence of clinically suspected HSR was 7.8% and 3.4% in the control and PGx
arms respectively (P<0.001). The positive predictive value of the HLA-B*5701 test for
clinically suspected HSR was 61.2% and the negative predictive value was 95.5%. For
immunologically confirmed HSR, the rate was 2.7% for the control arm and 0% for the
PGx arm (P<0.001). The positive predictive value of the HLA-B*5701 test for
immunologically confirmed HSR was 47.9% and the negative predictive value was
100%, that is, about half of the patients with a positive HLA-B*5701 test developed
confirmed HSR while no patient with a negative HLA-B*5701 test did. These findings
illustrate the importance of a precise definition of the event that is to be predicted by the
genetic marker (i.e., phenotype definition).
The impact of the PREDICT-1 results was substantial because the study was a
prospectively planned randomized trial with essentially full ascertainment of genomic
status of every randomized subject. The study provided demonstration of clinical
usefulness (i.e., near total ability to avoid abacavir-induced HSR with an acceptable false
positives rate in the screening). The results of this study influenced the inclusion of
strong recommendations for HLA-B*5701 screening in professional guidelines and in the
approved labeling (updated July 2008).
Clopidogrel (Plavix): Identifying Patients with Reduced Response to a Drug
Clopidogrel is a platelet adenosine diphosphate (ADP)-receptor antagonist that is
indicated for reduction of atherothrombotic events in patients with recent myocardial
infarction, recent stroke, peripheral artery disease, and acute coronary syndrome.
Clopidogrel is a prodrug with no antiplatelet activity, but about 15% of the dose is
metabolized to an active metabolite in a two-step process involving multiple CYP
enzymes, one of which is the polymorphic CYP2C19. CYP2C19 has four different
metabolizer phenotypes: ultrarapid, extensive, intermediate, and poor.
Analyses of data from several PK/PD studies demonstrated that carriers of at least one
loss-of-function (LOF) allele of CYP2C19 showed reduced exposure to the active
metabolite and less inhibition of platelet aggregation (Mega et al., 2009). These
observations have been independently replicated in numerous clinical pharmacology
studies. The relationship between the CYP2C19 genotype and PK/PD was further
extended to clinical outcomes in several population- or clinical trial-based cohort studies
(Mega et al., 2010). For example, CYP2C19 genotypes were determined in patients
enrolled in the TRITON-TIMI 38 trial who voluntarily consented to provide their blood

Contains Nonbinding Recommendations
24
samples for DNA analysis. In TRITON-TIMI 38, carriers of LOF CYP2C19 alleles had a
higher rate of death, nonfatal myocardial infarction, or nonfatal stroke than non-carriers
(no LOF allele) following percutaneous coronary intervention. In addition, the rate of
stent thrombosis over the same time period was approximately three times higher in
carriers.
Based on the results from multiple clinical pharmacology and outcome studies, the label
of clopidogrel was updated in May 2009 and again in March 2010 to include PGx
information related to the diminished antiplatelet responses and the increased risk of
cardiovascular events in patients with reduced CYP2C19 function. Many publications
have replicated the association between CYP2C19 variants and clopidogrel
responsiveness since the last labeling update. Collection of DNA in phase 3 trials helped
substantiate the findings of several clinical pharmacology studies.
Warfarin (Coumadin): Selecting Optimal Doses Based on Genotype-Based Differences in PK
and/or PD
Warfarin is a coumarin-based anticoagulant that is widely used for the short- and long-
term management of thromboembolic disorders, such as deep-vein thrombosis, and to
prevent stroke and systemic embolic events in patients with atrial fibrillation and those
undergoing orthopedic surgeries. A relatively large number of patients experience life-
threatening bleeding complications from warfarin. It has been consistently a top ten-
ranked cause of drug-induced serious adverse reactions. Major bleeding episodes are
reported relatively frequently. However, it is also essential to achieve adequate
anticoagulation to prevent thromboembolic events that warfarin is intended to prevent.
Warfarin has a narrow therapeutic range, with wide variation in dose requirements for
individual patients, and dose is modified by testing of INR (International Normalized
Ratio), a measure of coagulation inhibition. Titrating warfarin-naïve patients to a stable
INR range (e.g., 2-3, sufficient but not excessive anticoagulation) in a reasonable time is
a significant challenge for health care professionals because of the many genetic (e.g.,
CYP2C9 and VKORC1) and nongenetic (e.g., sex, body size, drug-drug interactions, diet)
factors affecting the PK and PD of warfarin. Underlying genetic factors have been
shown to account for approximately 35-40% of the variation in the maintenance dose.
CYP2C9 is the hepatic enzyme responsible for metabolizing S-warfarin, which is 3-5
times more potent than the R-enantiomer. Genetic polymorphisms affecting CYP2C9
(e.g., the *2 and *3 alleles) are common in the general population, resulting in decreased
clearance and higher blood levels of S-warfarin.
Warfarin works by inhibiting vitamin K epoxide reductase (VKOR), which is encoded by
the VKORC1 gene. Polymorphisms in this gene affect an individual’s response to
warfarin. Studies have shown that CYP2C9 and/or VKORC1 genotypes play an
important role in determining the dose of warfarin. In August 2007, FDA updated the
warfarin labeling to provide general information about genetic testing and to encourage
health care professionals to use this information for initial dosing to reduce the risk of
both bleeding and undertreatment. Subsequently, in January 2010, the warfarin labeling

Contains Nonbinding Recommendations
25
was updated to include a dosing table to be used for dosing when CYP2C9 and VKORC1
genotype information for a patient is known.