|
Reviewer Guidance
Conducting a Clinical Safety Review of a New Product Application and
Preparing a Report on the Review
(PDF
format of this document)
U.S. Department of
Health and Human Services
Food and Drug Administration
Center for Drug Evaluation and Research (CDER)
February 2005
Good Review
Practices
Additional copies
are available from:
Office of Training and Communication
Division of Drug Information, HFD-240
Center for Drug Evaluation and Research
Food and Drug Administration
5600 Fishers Lane
Rockville, MD 20857
(Tel) 301-827-4573
http://www.fda.gov/cder/guidance/index.htm
U.S. Department of
Health and Human Services
Food and Drug Administration
Center for Drug Evaluation and Research (CDER)
February 2005
Good Review
Practices
TABLE OF
CONTENTS
I.
INTRODUCTION
II. GENERAL
GUIDANCE ON THE CLINICAL SAFETY REVIEW
III.
SPECIFIC GUIDANCE ON THE CONTENT OF THE SAFETY REVIEW
Reviewer
Guidance1
Conducting a
Clinical Safety Review of a New Product Application and Preparing a
Report on the Review
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 good review practice (GRP)
guidance is intended to assist reviewers conducting the clinical
safety reviews as part of the NDA and BLA review process, provide
standardization and consistency in the format and content of safety
reviews, and ensure that critical presentations and analyses will
not be inadvertently omitted. The standardized structure also
enables subsequent reviewers and other readers to readily locate
specific safety information.
This guidance is an expansion of
section 7 of the clinical review template and is entirely compatible
with that template. The structure of this guidance, as an annotated
outline, is meant to correlate exactly with the section headings of
the review template, providing the pertinent guidance under each
heading. The guidance also provides, as attachments, illustrations
of displays and graphs that have been used successfully in the past.
These are not requirements, but examples, and reviewers can
substitute for, or modify, them, or simply find them unnecessary in
particular cases. It is expected that new attachments will be added
as examples become available.
The commentary and suggestions under
each section of the guidance, together with appended examples,
provide suggested analyses, methods of presentations, and discussion
of special cases and potential difficulties. Some flexibility in
implementing the guidance will be needed, as different types of
applications and datasets may require modifications to the structure
outlined in this guidance. If sections are omitted, the review
should briefly explain the reason for the omission.
FDA's guidance documents, including
this guidance, do not establish legally enforceable
responsibilities. 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.
GENERAL GUIDANCE ON THE CLINICAL SAFETY REVIEW
This Good Review Practice (GRP)
Guidance provides an annotated outline of the safety component of a
clinical review of an application (NDA, BLA) and guidance
on how to conduct and organize the safety review.
2 It is usually most efficient and
informative to include all the safety findings, whatever the source,
in the safety section of the clinical review (i.e., apart from the
description of individual studies in the efficacy review). In some
cases, however, it may be more appropriate to discuss some or all
aspects of safety as part of the discussion of individual efficacy
studies and reference them in this section (e.g., studies with
mortality outcomes, development programs in which most of the safety
data come from one or two large multi-center studies, and when
evaluation and review of safety data may be more convenient or
informative study by study).
The safety review has two distinct
components: (1) identification and assessment of the significance of
the adverse events reported in clinical trials (controlled or
uncontrolled) and (2) evaluation of the adequacy of the applicant's
safety evaluation. This guidance describes an approach that
integrates safety findings across all studies and other clinical
experience. Consideration of the safety findings in individual
studies, without a thoughtful integration of the overall safety
experience, is not adequate for a safety review.3
Although much of the guidance in
this document is directed primarily toward the clinical reviewer and
toward the analysis of particular events, the evaluation of safety
data also involves analyses of event rates, estimation of risk over
time, exploration of possible subgroup differences, and
identification of risk factors associated with serious events, all
analyses that involve substantial knowledge of methods of validly
quantifying risk and providing measures of uncertainty. Clinical
reviewers should therefore collaborate with their biostatistical
colleagues when necessary in the preparation of reviews and consider
when it may be appropriate to conduct a joint statistical and
clinical review for particularly important safety issues.
The conceptual framework of this
guidance is similar to the framework used for advising manufacturers
on submitting safety data in FDA's Guideline for the Format and
Content of the Clinical and Statistical Section for New Drug
Applications (Clinical/Statistical guidance)4
as well as in the guidance M4: The Common Technical Document for
the Conduct of Human Clinical Trials for Pharmaceuticals - Efficacy.
Because several related terms are
used in this guidance that could cause some confusion, the following
explanations are intended as clarification.
For purposes of this guidance, the
term adverse reaction is used to refer to an
undesirable effect, reasonably associated with the use of a drug,
that may occur as part of the pharmacological action of the drug or
may be unpredictable in its occurrence. This term does not include
all adverse events observed during use of a drug, only those for
which there is some basis to believe there is a causal relationship
between the drug and the occurrence of the adverse event. The term
adverse event is used here to refer to
any untoward medical event associated with the use of a drug in
humans, whether or not it is considered drug-related. The phrases
serious adverse drug experience and serious adverse
event are used in this guidance to refer to any event
occurring at any dose, whether or not considered drug-related, that
results in any of the following outcomes:
- Death
- A life-threatening adverse
experience
- Inpatient hospitalization or
prolongation of existing hospitalization
- A persistent or significant
disability/incapacity
- A congenital anomaly or birth
defect.
Important medical events that may
not result in death, be life-threatening, or require hospitalization
may be considered serious adverse drug events when, based upon
appropriate medical judgment, they may jeopardize the patient or
subject and may require medical or surgical intervention to prevent
one of the outcomes listed in this definition. Examples of such
medical events include allergic bronchospasm requiring intensive
treatment in an emergency room or at home, blood dyscrasias or
convulsions that do not result in inpatient hospitalization, or the
development of drug dependency or drug abuse. Documents developed by
the International Conference on Harmonisation (e.g., E2) add to
serious events those that prolong hospitalization but do not include
cancer and overdose.
Finally, the term adverse
dropout is used in this guidance to refer to subjects who
did not complete the study because of an adverse event, whether or
not considered drug-related; adverse dropouts include subjects who
received the test drug, reference drugs, or placebo.
The safety review has four principal
tasks:
1) To identify and closely examine
serious adverse events that suggest, or could suggest, important
problems with a drug-- specifically, adverse reactions severe enough
to prevent its use altogether, to limit its use, or require special
risk management efforts
2) To identify and estimate the
frequency of the common (usually nonserious) adverse events that
are, or may be, causally related to the use of the drug;
3) To evaluate the adequacy of the
data available to support the safety analysis and to identify the
limitations of those data. At a minimum, this includes assessments
of whether the extent of exposure at relevant doses is adequate
4) To identify unresolved safety
concerns that will need attention prior to approval or that should
be assessed in the postmarketing period, including such concerns as
the absence of data from high-risk populations or potential
interactions
In addition, the safety review
should:
- Identify factors that predict the
occurrence of adverse reactions, including patient-related factors
(e.g., age, gender, ethnicity, race, target illness, abnormalities
of renal or hepatic function, co-morbid illnesses, genetic
characteristics, such as metabolic status, environment) and
drug-related factors (e.g., dose, plasma level, duration of
exposure, concomitant medication)
- Identify, where possible, ways to
avoid adverse reactions (dosing, monitoring) and ways to manage
them when they occur
- For a drug that is to be
approved, provide a comprehensive evaluation of risk information
adequate to support a factual and sufficient summary of the risk
information in labeling.
Approaches to evaluation of the
safety of a drug generally differ substantially from methods used to
evaluate effectiveness. Most of the studies in phases 2-3 of a drug
development program are directed toward establishing effectiveness.
In designing these trials, critical efficacy endpoints are
identified in advance, sample sizes are estimated to permit an
adequate assessment of effectiveness, and serious efforts are made,
in planning interim looks at data or in controlling multiplicity, to
preserve the type 1 error (alpha error) for the main end point. It
is also common to devote particular attention to examining critical
endpoints by defining them with great care and, in many cases, by
using blinded committees to adjudicate them. In contrast, with few
exceptions, phase 2-3 trials are not designed to test specified
hypotheses about safety nor to measure or identify adverse reactions
with any pre-specified level of sensitivity. The exceptions occur
when a particular concern related to the drug or drug class has
arisen and when there is a specific safety advantage being studied.
In these cases, there will often be safety studies with primary
safety endpoints that have all the features of hypothesis testing,
including blinding, control groups, and pre-specified statistical
plans.
In the usual case, however, any
apparent finding emerges from an assessment of dozens of potential
endpoints (adverse events) of interest, making description of the
statistical uncertainty of the finding using conventional
significance levels very difficult. The approach taken is therefore
best described as one of exploration and estimation of event rates,
with particular attention to comparing results of individual studies
and pooled data. It should be appreciated that exploratory
analyses (e.g., subset analyses, to which a great caution is
applied in a hypothesis testing setting) are a critical and
essential part of a safety evaluation. These analyses can, of
course, lead to false conclusions, but need to be carried out
nonetheless, with attention to consistency across studies and prior
knowledge. The approach typically followed is to screen broadly for
adverse events and to expect that this will reveal the common
adverse reaction profile of a new drug and will detect some of the
less common and more serious adverse reactions associated with drug
use.
With respect to assessment of
serious events, there are two distinct situations. First, there are
the events readily recognized as consequences, or at least potential
consequences, of the treatment (i.e., adverse reactions) because
they would be unusual in the population under study. Second, and
particularly critical, are serious events that are not so readily
attributed to the drug because they can occur even without the drug,
for example, because they are known to result from the underlying
disease or are relatively common in the population being studied
(e.g., heart attacks, strokes in an elderly population) and could
therefore represent intercurrent illness. Adverse events that do not
seem typical of what drugs do (that is, that are not hematologic,
hepatic, renal, dermatologic or pro-arrhythmic) can be especially
difficult to attribute to a drug. The history of the relatively late
recognition of the practolol syndrome (sclerosing peritonitis,
oculomucocutaneous syndrome), retroperitoneal fibrosis with
methylsergide (Sansert), pulmonary hypertension with aminorex and
other appetite suppressants, thromboembolic disease with oral
contraceptives, endometrial cancer with post-menopausal estrogens,
suicidal ideation with interferons, and more recently, cardiac
valvular disorders with fenfluramine, illustrates this problem.
Perhaps most difficult of all is the situation where the adverse
event is, or could be, a consequence of the disease being treated.
Thus, it was extremely difficult to discover that many drugs for
heart failure (beta agonists, phosphodiesterase inhibitor inotropes,
and a vasodilator, flosequinan) caused increased rates of the same
kinds of death seen with the underlying disease (i.e., due to
progressive heart failure or arrhythmias), that anti-arrhythmics
could provoke new arrhythmias, and that interferon could cause
depression in patients with cancer or multiple sclerosis, conditions
that are themselves associated with mood alteration. Distinguishing
the effects of a drug on the immune or other impaired systems in
patients with cancer or HIV infection can also be difficult. Many
years ago, a last resort drug for rheumatoid arthritis, azaribine (Triazure),
was approved despite a number of arterial thrombi seen during
development because those were thought to be more common in patients
with RA (the drug was removed from the market shortly after
approval, however, when unusual thrombotic events (e.g., thrombosis
of a digital artery) became apparent). Drugs for seizure disorders
and schizophrenia can be difficult to assess with respect to causing
sudden death because patients with the disorders they treat have a
relatively high rate of this event. Usually, the only way to
establish that these are adverse reactions is through controlled
trials of significant size. Sometimes, the controlled trials to
evaluate effectiveness will be large enough to address these issues,
but sometimes, where there is a significant concern, special, large
safety studies may be needed.
There is no simple answer to these
difficult assessments, but this guidance, similar to section H of
the Clinical/Statistical guideline (Guideline for the Format and
Content of the Clinical and Statistical Sections of New Drug
Applications)5and the
guidance M4 The CTD -
Efficacy, suggest an
approach, namely, close examination of all patients who die or who
leave a study prematurely because of any adverse event (whether or
not thought drug-related),6
with explicit consideration of the possibility that the event was
drug-related (the "prepared mind" approach). With respect to
discovering that a drug causes a modestly increased rate of serious
events that are relatively common in the population, only large
controlled trials can provide a satisfactory answer and the reviewer
needs to consider whether such trials are needed. In some cases,
there are reasonably well-established surrogate markers that can
predict severe injury. For example, an increased rate of
transaminase elevations accompanied by a small number of cases in
which bilirubin elevation accompanies the transaminase elevation can
predict the occurrence of more severe liver injuries in some
patients, and visual field defects may portend irreversible
peripheral vision loss. Similarly, substantial QT interval
prolongation on the electrocardiogram predicts the occurrence of
Torsade de Pointes-type ventricular tachycardia.
Before beginning the safety review,
the reviewer should identify and assemble (or locate electronically)
all available materials for the review. These materials include:
- The applicant's Integrated
Summary/Analysis of Safety (ISS)
- Adverse event tables in the NDA/BLA
submission7
- Case report forms (CRFs) for
patients who experienced serious adverse events or who dropped
out of a study because of an adverse event. The reviewer should
request these CRFs if the applicant does not include them in the
submission (although they are required under 21 CFR 314.50). If
the number of cases is very large (e.g., for dropouts) and many
of the events are similar, it may be reasonable to request only
a sample of CRFs.8 Note
that, in some cases, dropouts attributed to other reasons will
upon review be associated with an adverse event.
- Individual patient adverse
reaction data listings, laboratory listings, and baseline
listings, usually accessible electronically8
- The applicant's narrative
summaries of deaths, serious adverse events, and other events
that resulted in dropouts
- If available, displays of
individual patient safety data over time for patients who
experienced serious adverse events
- The safety sections of the
sponsor's proposed labeling
- Common Technical Document (CTD)
safety-related sections (module 2, Sections 2.5.5, 2.7.4), which
give an overview of the applicant's approach to the safety
evaluation and a detailed summary of the safety data
- Any other safety-related
documents, such as discussions of related drugs, descriptions of
use of adverse drug reactions (ADR) coding dictionaries to
combine data across studies, specific studies of safety
hypotheses
Although the review will assess the
data submitted, it may be useful to identify at the outset
particular concerns that will be explored because they are suggested
by the pharmacology of the drug or by safety concerns with
pharmacologically related drugs. Thus, the clearance pathway of a
drug will suggest certain potential drug-drug interactions or
certain effects of decreased renal or hepatic function. Similarly,
the pharmacologic class, and prior experience, could lead to focus
on particular laboratory or clinical abnormalities (e.g., muscle or
liver abnormalities with HMGCoA reductase inhibitors, QT
prolongation with fluoroquinolone anti-infectives, gastrointestinal,
renal, and cardiovascular effects of nonsteroidol anti-inflammatory
drugs, liver abnormalities with endothelin receptor antagonists,
cognitive impairment with sedating drugs, sexual dysfunction with
selective serotonin reuptake inhibitors). These concerns are
considered further in section 7.2.5.
Although there are no established
standards for auditing safety data in a submission, the review
should describe efforts to assess consistency of the data provided
(e.g., comparing information included in case report forms, case
report tabulations, and narrative summaries for individual
patients). For important adverse events, for example, it is
generally important to consider not only the applicant's narrative
description, but the associated CRF or hospital records and
submitted laboratory, radiology, or pathology results.
An important part of the safety
review is reviewing individual cases of death, serious adverse
events, adverse events leading to discontinuation (adverse
drop-outs), and discontinued patients who are lost to follow-up. One
reason to review the details of individual cases is to determine
whether the event was coded to the correct preferred term. The
assessment of causality for specific adverse events in NDAs/BLAs is
heavily dependent on comparisons of event rates between treatment
groups, and the numerator of these rate calculations includes events
coded to the same preferred term. A case might be incorrectly
included in the numerator of a rate calculation if the event is
incorrectly coded to a specific preferred term. Events may be
incorrectly coded to preferred term by the applicant when they
summarize the data or because an investigator used a verbatim term
incorrectly when recording the event in the case report form. An
example of incorrect coding would be if an investigator used the
verbatim term acute liver failure for a case of increased ALT
and the applicant coded the event to acute liver failure. One would
not want to include such a case in the numerator of a risk
calculation for acute liver failure. Similarly, a case could be
incorrectly excluded from a numerator. Inconsistent coding (e.g.,
peripheral edema coded as "heart failure" for one patient, but
"metabolic abnormality" for another) could result in an
inappropriately low numerator.
A second reason to conduct
individual case review of deaths, serious adverse events, and
adverse events leading to discontinuation is to determine whether
there is a likely explanation for the event other than the drug that
is the subject of the application, such as another drug or
concomitant illness (e.g., documented acetaminophen overdose in a
case of acute liver failure would argue against attribution to the
test drug; documented cholecystitis would argue against attribution
of cholestasis to the test drug). If there is no likely alternative
explanation for the event, the event must be considered at least
possibly drug-related, and should be included in a rate calculation.
A third reason for individual case
review of deaths, serious adverse events, and adverse events leading
to discontinuation is to look for other reasons that might exclude
the drug as a cause of the event. One example would be when an
adverse event occurred during a placebo washout period before
exposure to study drug occurred. Events that occur prior to exposure
would not be included in the numerator of risk calculations. Events
that begin long after discontinuation of the drug might also be
considered unlikely to be drug-related, but care must be taken in
excluding them, as there are examples of such late drug-caused
reactions (e.g., FIAU (fialuridine, a nucleoside analog) where liver
failure was seen well after the drug was stopped, probably because
it induced mitochondrial DNA damage that became a problem only when
mitochondria tried to replicate) and because some chronic reactions
might not be detected immediately.
A fourth reason for individual case
review of deaths, serious adverse events, and adverse events leading
to discontinuation is to look for results of rechallenge. A
potentially important source of information about causality is when
an individual is rechallenged with drug, accidentally or
deliberately. Recurrence with rechallenge is a potentially strong
indicator of causality, but interpretation of the results of
rechallenge is highly dependent on the natural course of the event
being considered. For noncyclical events that are exceedingly rare
in the background (e.g., acute liver failure, aplastic anemia)
recurrence of the event upon rechallenge (i.e., positive rechallenge)
provides strong evidence of causality. Positive rechallenges are
less definitive for diagnoses/events that can occur in cyclical or
recurrent fashion (e.g., worsening glucose control in a subject with
diabetes mellitus), but close observation of the patient's whole
course (i.e., both challenge periods and dechallenge periods) may be
helpful. Rechallenges that do not result in recurrence of the event
(i.e., negative rechallenge) suggest (but do not prove) that the
drug did not cause the event. One must consider such factors as
whether it was possible for the event to recur, the dose of drug and
duration of exposure at which the subject was rechallenged, and
whether the length of observation following rechallenge was
sufficient to allow recurrence of the event of interest.
It is important to distinguish the
processes described above from the causality analyses of
drug-related events often provided by investigators and applicants
in NDA/BLA submissions. The analyses of drug-related adverse events
presented by applicants are usually based on assessments made by
investigators at the time of an event, are highly dependent on
information about the side effect profile of the drug available at
the time of the study (e.g., what is in the investigator's
brochure), and are not informed by awareness of the entire safety
database. These analyses are generally not expected to provide much
useful information in assessing causality.
Assessment of the drug-relatedness
of an adverse event is fundamentally different for relatively
frequent and relatively rare events. For the former, a reviewer
would compare the incidence of adverse events occurring in the study
drug group to that in the placebo (or other control) group (in RCT).
For rare events, the expected rate in a clinical trial database
would be zero. Thus, if even a few cases (sometimes even a single
case) of a rare life-threatening event occurred when none was
expected, that would represent a serious safety problem for a drug
product that does not provide unique efficacy or some other
advantage over available treatments.
III.
SPECIFIC GUIDANCE ON THE CONTENT OF THE SAFETY REVIEW
The following sections bear the same
names and numbers of the section of the clinical review template
that contains the safety review (Section 7.0).
This guidance organizes the safety
review into three main sections:
- Methods and Findings (Section
7.1)
This section contains 17
subsections. Overall, this section should describe the relevant
data sources, the safety assessments that were carried out, and
the major findings of the detailed safety review. Section 7.1
should use a systematic approach to describing available data.
It should focus first on the serious and potentially serious
reactions, the kind that can affect the approval decision or
severely limit the use of the drug (see Subsections 7.1.1 to
7.1.4). Focus should then move to the more common reactions that
rarely influence approval but are often critical to patient and
physician acceptance of the drug. Section 7.1 should then
consider less common events, laboratory findings, vital signs,
electrocardiograms (ECGs), immunogenicity, human
carcinogenicity, human reproductive toxicity, withdrawal
phenomena, abuse potential, and overdose (Subsections 7.1.5 to
7.1.17).
- Adequacy of Patient Exposure
and Safety Assessments (Section 7.2)
This section should address the
adequacy of exposure (e.g., overall patient numbers and numbers
for specific demographic subsets, duration of exposure, the dose
levels at which exposure took place,9
the quality and completeness of safety evaluations, whether all
necessary evaluations were conducted (e.g., animal tests, in
vitro tests, long-term safety testing, specific assessments of
ECG effects), and whether any additional safety assessments are
needed (either pre- or postapproval). This section should also
include a subsection on additional submissions of safety data,
including safety update(s).
- Summary of Selected
Drug-Related Adverse Events, Important Limitations of the Data,
and Conclusions (Section 7.3)
This guidance also discusses general
analytical methods that may be useful for multiple aspects of the
safety assessment in section 7.4 (see page 45). This section
discusses pooling of data, explorations for adverse reaction
predictive factors, dose-dependency evaluations, time dependency
evaluations, duration of adverse reactions, drug-demographic
interactions, drug-disease interactions, and drug-drug interactions.
The annotated outline of the review
begins here:
7.0 Integrated
Review of Safety
This section consists of 17
subsections (e.g., death, other serious adverse events, laboratory
findings). Each of these subsections is organized somewhat
differently, depending on the content. In presenting analyses in the
safety review, it is important to clearly distinguish between the
applicant's analyses and conclusions and those of the reviewer.
In discussing serious adverse events
and dropouts (Sections 7.1.1, 7.1.2, and 7.1.3), it is critical that
the reviewer identify individual patients in a way that enables
subsequent readers to readily access data and supporting information
if needed (e.g., study #, investigator #, patient ID#).
Identifying Deaths Relevant to
the Safety Review
Deaths occurring during the
following time periods or under the following conditions should be
assessed:
- Deaths occurring during
participation in any study, or during any other period of drug
exposure
- Deaths occurring after a patient
leaves a study, or otherwise discontinues study drug, whether or
not the patient completes the study to the nominal endpoint, if
the death:
- is the result of a
process initiated during the study or other drug exposure,
regardless of when it actually occurs; or
- occurs within a time period
that might reflect drug toxicity for a patient leaving a study
or otherwise discontinuing drug. For drugs with prompt action
and relatively short elimination half-lives, 4 weeks is a
reasonable time period. For drugs with particularly long
elimination half-lives or drug classes with recognized potential
to cause late occurring effects (e.g., nucleoside analogs, gene
therapies, or cell transplants), deaths occurring at longer
times after drug discontinuation should be evaluated.
The reviewer should consider all
deaths that occurred in a drug's development program and any other
reports of deaths from secondary sources (e.g., postmarketing or
literature reports), without regard to investigator or applicant
judgment about causality. It is also important to consider deaths on
control treatments for comparison, even though they are obviously
not related to the drug in the application. Individual deaths should
be listed in a table (see Table 7.1.1.1), unless they are an
effectiveness study outcome.
Applicants will provide line
listings of all patients who died in studies, together with a brief
narrative. (See E3 Structure and Contents of Clinical Reports,
Section 12.3.10 The narratives
may also be placed in the Integrated Summary of Safety.)
Distinguishing Expected from
Unexpected Deaths
Certain causes of death are
sufficiently unusual in the absence of drug therapy, even in large
databases, that they would almost always be considered unexpected
(e.g., aplastic anemia or acute hepatic necrosis) and deserve
detailed individual discussion. Other fatal events occur at such
frequency in the general population that they would be expected to
occur in any large database absent drug therapy (e.g., fatal strokes
and heart attacks), especially in the elderly.11
In most cases, these events need to be examined for frequency but
discussion of individual cases is not helpful. Expected deaths would
include:
- Deaths in studies in which
mortality is an endpoint and the cause of death is expected for
the disease or condition
- Deaths in studies in diseases
where high mortality rates are expected and the cause of death is
expected (e.g., cancers). Note, however, that early deaths in
cancer studies are a concern as patients are usually selected for
clinical trials because they were not expected to die soon.
- Coincidental deaths resulting
from progression of underlying disease present at enrollment in a
study (e.g., a patient who dies from progression of cancer or
Alzheimer's Disease or an acute myocardial infarction attributed
to underlying coronary artery disease present prior to study entry
- Deaths from intercurrent
long-term illness. These include the wide variety of fatal events
that can be seen in any population, especially a relatively
elderly population, such as sudden death (presumably representing
an arrhythmia), fatal infections, surgical emergencies, or
intracranial hemorrhage).
Even though fatal events may be
expected in a population, the reviewer should not without further
consideration readily accept the conclusion that a fatal event is
due to the underlying disease or an intercurrent illness and not the
drug. For each fatal event, the reviewer should specifically
consider the possibility that the event represents an as yet
unsuspected adverse reaction. Even if there is nothing about these
deaths to suggest a drug cause, it is critical to assess whether the
rate of these events is increased. The best way to do this, of
course, is by comparison with a control group (a single trial or
pooled), but if no control group is available, it may be of value to
look at databases of other drugs used in the same population.
When distinguishing between
unexpected and expected deaths, the reviewer should make clear the
bases for the distinctions (e.g., early deaths in cancer patients
are unexpected if the entered patients were chosen because they were
not expected to die soon; hematologic deaths are unexpected in a
postinfarction study). For unexpected deaths, the individual medical
events associated with the death should be carefully evaluated and
discussed in detail in the review. Expected deaths should be
classified as to type of death, but it is usually not necessary to
discuss in detail the individual medical events associated with
those deaths. What is critical is to consider whether there is a
suggestion that their rate is increased, the adequacy of the data to
evaluate this, and the need to know more (e.g., because of
experience with related drugs).
Pooling of Relevant Data
Before conducting any mortality
analyses, the reviewer must consider the poolability of the data
pertinent to deaths. If data are not poolable, analyses should be
conducted for separate databases, then examined together. See
section 7.4.1. for discussion of pooling.
Overall Mortality Analysis
The review should include an
analysis of overall mortality for all phase 2 and 3 exposures across
treatment groups as well as cause-specific mortality to the extent
possible. The fineness of classification depends on the
quality of data and the number of events (cardiovascular (broad) vs.
AMI, sudden death, CHF (more specific)), recognizing that assessing
cause-specific mortality is very difficult even in the best
circumstances, such as a study in which there is an attempt to
describe such endpoints prospectively.12
Death from an acute myocardial infarction can be indistinguishable
from death resulting from an arrhythmia, for example. Analyses
should be corrected for differences in drug exposure using
person-time in the denominator to calculate mortality rates.13
If person-time exposure is not included in the submission (ideally,
it should be requested at the pre-NDA/pre-BLA meeting), it should be
requested as soon as the need is recognized. This correction can be
done only for those deaths for which person-time data are available.
It may be useful to present both crude mortality and mortality
expressed in person-time in an appendix table (see Table 7.1.1.2 for
sample display). Life table approaches may be helpful in cases when
there are more than a few deaths, and when the direction of
different studies varies significantly. Ideally, one would have
mortality data from other databases for comparison (e.g., from other
drugs in the same class).
Discussion of Applicant's
Assessment of Deaths
The reviewer should describe and
evaluate the applicant's assessment of deaths, including the
following:
- The applicant's criteria for
including deaths in the NDA/BLA (e.g., whether the criteria were
reasonable, whether the criteria were met)
- The methods used by the applicant
to detect and classify deaths
- The applicant's method of
analyzing overall mortality and cause-specific mortality
- The applicant's judgments on the
drug-relatedness of events associated with deaths
Reviewer's Assessment of Deaths
The reviewer's assessment of deaths,
reflecting both the applicant's and the reviewer's analyses, should
include the following:
- Listing of information upon which
reviewer assessment is based (e.g., CRFs, narrative summaries,
consultant reports, autopsy reports)
- Tabular Summary of Deaths. Deaths
should be summarized in an appendix table, as illustrated in Table
7.1.1.1. It may be useful to distinguish between those deaths for
which exposure data are available and those for which such data
are unavailable (e.g., for postmarketing deaths, exposure data may
never be available). In the table and subsequent discussion, there
should be a clear identifier so that subsequent reviewers can
identify the particular patient.
- Analysis of overall mortality for
Phase 2 and 3 drug exposures across treatment groups (see Table
7.1.1.2)
- Analysis of cause-specific
mortality across treatment groups (this could use a table similar
to Table 7.1.1.2)
- The reviewer's overall judgment
about the drug-relatedness of medical events associated with death
(i.e., which deaths were probably explained by factors other than
the study drug (e.g., another drug, underlying illness, another
illness common in the population) and which could not reasonably
be explained by such factors). Differences from the applicant's
evaluation should be noted and discussed.
- Further discussion of the
individual events associated with death and believed to be
potentially drug-related, either because they are increased in
rate compared to control or because of the nature of the event
(e.g., events typically drug-related, such as aplastic anemia or
acute hepatic necrosis, or events that would not be expected in
the population studied such as sclerosing abdominal or pulmonary
conditions or rapidly progressive unexplained renal failure). Any
uncertainty about drug-relatedness should lead to inclusion of the
event. For each of these individual events, brief narratives
should be included in the review or (if numerous) attached in an
appendix.
- Other relevant analyses, such as
analysis of dose response (administered dose, body weight and
surface area adjusted dose, cumulative dose, schedule (including
duration of infusion for IV drugs) analysis of mortality within
critical subgroups (e.g., demographic, disease severity, excretory
function, concomitant therapy), drug-demographic, drug-disease,
and drug-drug interactions (see section 7.4)
- When deaths occur in uncontrolled
studies, best available estimates of mortality in the population
studied, in the absence of the treatment (see Section 7.4.3)
- When deaths are relatively
frequent, the reviewer should consider some of the approaches
described for Common Adverse Events (see section 7.1.5)
Identification of Nonfatal,
Serious Adverse Events
The reviewer should identify,
without regard to the applicant's causality judgment, all serious
adverse events that occurred in the drug's development program or
were reported from secondary sources (e.g., postmarketing or
literature reports). Serious adverse events may, in addition to
signs, symptoms, and diagnosable events, include changes in
laboratory parameters, vital signs, ECG, or other parameters of
sufficient magnitude to meet the regulatory definition of a serious
adverse drug experience.14
Applicants generally provide a line
listing of all patients in phase 2 and 3 of the development program
who had an event meeting FDA's criteria for a serious adverse event.
For each such event, the applicant should also provide a brief
narrative (see ICH E3 Structure and Content of Clinical Study
Reports, Section 12.3.2). Because the definition of serious
is subject to some interpretation, the reviewer should make clear
how the applicant created the list. For example, applicants may
include events considered serious by investigators, even if they do
not technically meet the FDA or ICH definition for a serious event.
If such events are included, the inclusion parameters should be
noted.
Discussion of Nonfatal Serious
Adverse Events
This section of the review should
contain the following:
- A brief description of data
sources used in the review of individual cases (e.g., case report
forms, applicant's narrative summaries, hospital records).
- The Tabular Summary of serious
adverse events (see sample listing, Table 7.1.2.1)
- An analysis of overall rate of
serious events and rate of specific serious events, for each
treatment group in critical subgroups (e.g., demographic, disease
severity, excretory function, concomitant therapy), and by dose.
The median duration of exposure should be examined across
treatment groups. If there is a substantial difference in exposure
across treatment groups, incidence rates should be calculated
using person-time exposure in the denominator, rather than number
of patients in the denominator, similar to the presentation for
deaths in Table 7.1.1.2.
- Reviewer's overall assessment of
which serious adverse events were probably explained by factors
other than the study drug (e.g., another drug, underlying illness,
another illness common in the population) and which could not
reasonably be explained by such factors) including any pertinent
information from non-serious events (e.g., seizures
not leading to hospitalization, all syncope) that may be related
to the serious event.
- Further discussion of each
individual serious adverse event judged to be drug related (i.e.,
each adverse reaction), as needed, including any relationship of
the reaction to death. For each of these reactions, brief
narratives should be included in the review or (if numerous)
attached in an Appendix.
- A discussion or listing of
serious events considered unlikely to be drug-related (may be
identified in the Tabular Summary, illustrated in Table 7.1.1.1).
If serious nonfatal adverse events
are relatively frequent, the reviewer should consider some of the
approaches described for Common Adverse Events (see section 7.1.5).
FDA regulations require that the
CRFs from patients who discontinue treatment in association with
an adverse event (adverse drop-out) be submitted with the
application (21 CFR 314.50(f)(2)) and their analysis constitutes a
critical part of the safety evaluation (see Section 7.1.3.2).
ICH E3 (Guideline on
Structure and Content of Clinical Study Reports) defines a new
category of other significant adverse events. It includes:
- Marked hematological or other lab
abnormalities not meeting the definition of serious. This will
need to be an individual judgment, probably depending on the drug
(e.g., CPK elevation could have a different implication for a
statin and a different drug)
- Any events that led to an adverse
dropout or any other intervention such as dose reduction or
significant additional concomitant therapy (an expansion of the
adverse dropout concept that appears in 21 CFR 314.50(f)(2),
and in the Clinical/Statistical guideline, and M4: The CTD -
Efficacy)
- Potentially important
abnormalities not meeting the above definition of serious and not
leading to death or modification of therapy (e.g., a single
seizure, syncopal episode, orthostatic symptoms)
If the applicant has included
listings for other significant adverse events these may be
described here, under a subsection separate from the discussion of
dropouts (see Section 7.1.3.3). Alternatively, marked laboratory
changes may be described under Laboratory Findings (Section 7.1.6).
The review should contain an overall
profile of dropouts from clinical trials. The profile should
classify dropouts from the overall phase 2 and 3 study pool by
reason for dropping out (e.g., adverse event, treatment failure,
lost to follow-up). Where there are clinically relevant differences
in dropout rates for certain subsets (e.g., dropouts in
placebo-controlled trials vs. dropouts in other studies; dropouts in
certain demographic or disease-related subgroups), the profile
should also classify dropouts for those subsets. The reviewer should
explain the basis for selecting identified subsets and provide
mutually exclusive tabulations in which individual patients are
counted only once.15
Ordinarily, the dropouts should be categorized in a table or tables
appended to the safety review (see Table 7.1.3.1.1). It can be
useful to display, graphically or in tables, the cumulative dropout
rates for each treatment group within each study, especially for
cause-specific reasons, when this information is available and to
assess patient baseline risk factors that contribute to differential
cumulative dropout patterns. When pooling data, consideration of
dropout patterns over all studies may reveal information that is
useful to the overall safety evaluation.
When classifying dropouts, the
reviewer should carefully examine the reasons identified by the
applicant for subjects dropping out. Heightened scrutiny is
warranted for:
- Dropouts classified as
administrative, lost to follow-up, or a similar term
- Dropouts for which the applicant
changed the investigator's determination of the reason for the
drop out.
Discontinuations attributed to
adverse events require submission of the corresponding CRF but CRFs
may not be submitted for dropouts classified as administrative or
lost to follow-up. If such CRFs are not available, the reviewer may
need to request at least a sample of them to determine whether these
dropouts may have occurred in association with an adverse event.
Where dropouts are reclassified by an applicant (i.e., assigned a
reason for dropping out other than the one given in the CRF), the
review should indicate how and by whom such reclassifications were
made and comment on the appropriateness of decisions.
Ordinarily, the reviewer should
combine patients categorized as dropping out for intercurrent
illness and patients categorized as dropping out for adverse drug
reactions (if the applicant makes that distinction) under the
general category of dropouts for adverse clinical events (CRFs
should be provided for both of these categories). This
categorization is neutral from the standpoint of causality judgment,
recognizes the great difficulty in making distinctions between
adverse drug reactions and intercurrent illness, and encourages the
reviewer to consider the possibility that what seemed to be another
illness or a consequence of the underlying illness was, in fact, an
adverse drug reaction.
The reviewer should examine the
number and distribution of dropouts to identify potential problems
with study conduct or analyses (e.g., a substantial number of
dropouts due to lost to follow-up and sites with
disproportionately high dropout rates should be a sign of concern).
For example, early dropouts generally, and differential (drug group
vs placebo group) early dropouts in particular, often present
difficulties in conducting and interpreting the effectiveness
analysis and may suggest breakdown of blinding. The review should
discuss any concerns about dropouts and the methods employed by the
reviewer to address them.
The analysis of adverse events
associated with dropouts is important for two distinct reasons.
First, it identifies the type and frequency of adverse events that
patients were unable to tolerate even in a clinical trial setting,
where there is arguably more support for enduring adverse events
than in a clinical practice setting. This provides important
prescribing information that can contribute to dose selection, and
in some cases, to choosing a method of titration. In most cases,
there will be little doubt about which of these events are
attributable to the drug because the events will be of relatively
high frequency, even if withdrawals because of them are not, and the
main issue will be their frequency and importance. It is usually not
necessary to review these events case by case.
Second, and the reason CRFs for
dropouts due to adverse events are provided automatically to
reviewers, these adverse dropouts may provide a clue to unexpected,
but important, adverse reactions, (e.g., fibrosing intra-abdominal
or pulmonary illnesses, progressive liver or kidney diseases,
cardiac valve damage, neurological diseases, arteritis,
thromboembolic diseases, all of which have been caused by drugs)
that can easily be dismissed as intercurrent illness. The frequency
of these events is likely to be very low, and the review should
contain an analysis of each such adverse event that resulted in
withdrawal from the study, whether or not the event was attributed
to the drug. The reviewer should avoid dismissing such events as
intercurrent illness and specifically consider the possibility that
each dropout not due to a known effect of the drug might reflect an
unexpected effect of the drug. The applicant will usually provide a
line listing of adverse dropouts (Table 7.1.3.2.1) and this listing
(which need not be attached to the review) can serve to identify
events needing further scrutiny. Review of the CRFs can often
provide critical insights. The reviewer should describe how she or
he analyzed these events.
With respect to the more common
adverse events leading to discontinuation of treatment, the review
should present:
- The incidence of adverse events
associated with dropouts. Ideally, incidence would be presented in
a table or tables appended to the safety review with separate
tables for subsets of the overall clinical data pool in which
there were clinically meaningful differences in dropout incidence
(Table 7.1.3.2.2). Tables displaying incidence of events should
include each event that led to a dropout even if a single patient
had more than one such event.
- Whether the event can reasonably
be considered drug-related; this conclusion will be based on
comparisons between treatment groups in controlled trials and can
be informed by the overall rate of the adverse event (Section
7.1.5), and the known pharmacology of the drug.
- The dose response and time
dependency of the drop-outs and drug-demographic, drug-disease,
and drug-drug interactions (see section 7.4 General Methodology).
For the rarer events that could
suggest an important adverse reaction, the critical review
determination is whether any of these events suggest drug-induced
injury. These events need to be considered individually, with
narratives and reference to other databases as appropriate.
Where the review contains
applicant-generated tables, it is important for the reviewer to
determine and describe how the tables were created. A table may
identify one or more adverse events as having caused a particular
patient to withdraw, in which case it would represent the actual
incidence of specific adverse events that led to dropout. This
approach is preferred. Alternatively, a table may list the adverse
events that a subject experienced at the time of dropout and not
identify any event (or events) as causing the dropout. This approach
does not provide the actual incidence of adverse events associated
with dropouts and is of less value. The reviewer should make clear
in the review which of these approaches was used, or whether an
alternative approach was used.
If a submission separates out
information on adverse events that led to dose reduction or
significant additional concomitant therapy,16
but not to discontinuation of treatment, those findings should be
described using an approach similar to that proposed above for
adverse dropouts.
In addition to reviewing deaths,
serious adverse events, and adverse events associated with dropouts,
it may be useful to construct algorithms involving combinations of
clinical findings that may be a marker for a particular toxicity
(e.g. serotonin syndrome, cough, chest congestion and shortness of
breath that may constitute drug-related bronchospasm, or
drug-induced Parkinsonism). When such algorithms are used, the
algorithm and results of the search using the algorithm should be
described in the review. Generally, and where possible, such
searches should be done while the reviewer is blinded to treatment,
as this will minimize bias when identifying cases.
It should be noted that the causal
relation of a drug to uncommon serious adverse events may be
supported by less serious events that are more common. For example,
the likelihood that a drug caused a small number of cases of serious
liver toxicity may be supported by a higher rate of transaminase
elevation.
This section of the review focuses
on establishing the common adverse reaction profile for the drug and
determining the content of the adverse reaction table(s) to be
included in labeling. NDAs typically contain numerous tables and
analyses of adverse event incidence (e.g., by study, by various
pools of studies, and for the overall database). In general, what
are included are TESS, treatment emergent signs and symptoms
(i.e., signs and symptoms not present at baseline, or not present at
the severity seen on treatment). To approach these data, the
reviewer should generally go through the steps outlined in sections
7.1.5.1-7.1.5.5.
Adverse events can be elicited by
open-ended questions or checklists with varying degrees of
specification. Each approach has advantages and disadvantages, but
results can differ greatly and may lead to marked differences in
reported adverse event rates across studies (it would not usually be
appropriate to pool results obtained using both methods). The
reviewer should describe the applicant's method or methods of
eliciting adverse event data in clinical trials, including whether
checklists were used, the frequency with which patients were
assessed, and whether the approaches differed among studies.
Identification of signs (abnormal findings observed by a clinician)
would seem to be less of a problem, as these are elicited by
physical examination, but use of a physician exam checklist could
lead to a different result from a more general requirement for
physical exam. If different approaches were used (e.g., checklists
in studies conducted in the United States, open-ended inquiries in
European studies), the reviewer should consider and discuss in the
review the effect, if any, on the adequacy of adverse event
information collected.
Although investigator adverse
reaction terms are provided as part of study reports and are listed
in case report tabulations, the integrated analysis of the ISS
requires the applicant to use some way of grouping closely related
events to obtain an overall rate for a category of events. This is
accomplished by using a so-called dictionary of preferred
terms, such as COSTART,17
MedDRA,18 the latter a more
granular listing developed under the auspices of ICH. These
dictionaries are in fact lists of preferred terms and leave
(especially COSTART and other older dictionaries) considerable
discretion to the classifier to choose the term that best reflects
the verbatim term reported by the investigator. The categorization
of such systems, however, may not capture, or can dilute, the true
meaning of certain events. In addition, terms used in COSTART may
not be informative (e.g., pain, tooth disorder). It is expected that
as MedDRA becomes more widespread, this will no longer be such a
problem. It is critical that the reviewer assess the appropriateness
of the applicant's categories and the coding of adverse event
verbatim terms to preferred terms and understand how the verbatim
terms (including terms in languages other than English) were
classified.
In assessing the applicant's coding
of events, the reviewer should compare the applicant's preferred
terms to the verbatim terms used by investigators and patients,
focusing on the events leading to dropouts or other changes in
treatment as well as to serious adverse events. The applicant will
usually provide (ideally this will have been agreed to at the pre-NDA/BLA
meeting, but if not, it should be sought early in the review) the
following tables and listings for assistance with this assessment;
they should be provided in a form the reviewer can manipulate, such
as a SAS transport file, not just in PDF format:
- Adverse event tables in
individual study reports based on investigator terms for events
- A comprehensive line listing of
all adverse events in phase 2 and 3 studies with a column
containing investigator terms coded under a preferred term (see
Table 7.1.5.2.1. This table is for reference only; it would not be
included in the review)
- Listing of preferred terms and
the investigator and patient terms that were subsumed under the
preferred term. This table is for reference only; it would not be
included in the review, although parts of it might be).
The reviewer should consider the
following:
- Whether terms are too narrow (splitting),
resulting in an underestimation of the true incidence for a
particular event or syndrome (e.g., somnolence, drowsiness,
sedation, and sleepiness probably all refer to the same event)
- Whether the terms are too broad
or over-inclusive (lumping), so that important events that
should be examined separately are diluted by less important events
(e.g., loss of consciousness and syncope subsumed under
hypotensive events or hypotension)
- Whether terms used lack a
commonly understood meaning (e.g., mouth disorder, tooth disorder,
GI disorder) and, if so, whether the incidence of individual
events subsumed under these terms should be expressed separately
or mapped to a different preferred term
- Whether terms exaggerate a
finding (acute liver failure for a transaminase elevation) or
minimize the importance of an event (hypotension for a syncope
episode)
- Whether the coding of adverse
events is similar across treatment groups
In any of these cases, the reviewer
(or the applicant at the request of the reviewer) may have to
recalculate rates using alternative terms or different groups of
terms.
Usually it will be impossible to
evaluate all or even most adverse event terms in this much detail.
However, certain preferred terms should always have the subsumed
verbatim terms examined because they may conceal important events.
For example, accidental injury often includes fractures
and/or lacerations related to falls. The fall, itself, however, may
not have been captured as an adverse event. Additionally, edema
may sometimes include facial edema. Since facial edema often
represents an allergic reaction, one would not want allergic events
lumped together with peripheral edema events. In general, adverse
event terms associated with discontinuation or serious consequences
deserve the closest scrutiny, but other classifications should be at
least spot-checked. The review should comment on how this issue was
addressed.
7.1.5.3
Incidence of Common Adverse Events - Assessment of Various
Databases
Applicants typically prepare a
wide variety of tables of adverse event rates for individual
studies and pools of various studies. Those tables generally
include investigator causality assessments and severity
ratings. The tables the reviewer considers useful should be
appended to the review. Incidence rates for common adverse
events may be estimated from the relatively small portion of
the overall database that is contained in the controlled
(especially placebo-controlled) trials. For these more common
events, the ability to compare rates on drug with a control
outweighs the disadvantage of basing the rate estimates on
fewer subjects. In determining incidence rates for common
adverse events, the reviewer should identify the subset of
trials in the phase 2 and 3 database that will provide the
best estimate of rates and develop tables of event rates based
on that judgment.
- If possible, the reviewer should
rely on pooled data from studies using the same comparator group
(e.g., only placebo-controlled trials) and of roughly similar
duration. If some of the trials also had an active control, rates
for that group (pooled) can also be included (see Table
7.1.5.3.1). If different doses were used, both a pooled all doses
group and individual dose groups can be shown. The best comparison
is of the groups included in all studies (drug at a particular
dose and placebo), but the others (active controls, individual
dose groups) may also be useful (also see Section 7.4 for broader
discussion of pooling).
- If there are not adequate numbers
of patients in such trials to give meaningful rate estimates, the
reviewer should consider pooling placebo-controlled trials, active
control trials, and three arm trials (i.e., trials that do not all
have the same control group). Even when this approach is needed
overall, smaller subsets of studies, or even individual studies,
can be used to examine high-frequency events.
- Most applicants will construct
adverse event tables by compiling and presenting the numbers
and/or percentages of patients experiencing an adverse event in a
study (or the absolute number of adverse events experienced in a
group), without regard to the duration of treatment received. This
is often satisfactory for relatively short-term studies. If
studies of significantly different durations are pooled, however,
or if there is a different discontinuation rate in the treatment
arms and the risk of the adverse reaction persists over time, one
must consider these durations to understand the real occurrence
rate that patients will experience. One way to deal with the
problem of different durations is to use the total person-time
exposure for each treatment group and calculate the rate of the
adverse event per period of exposure (# of patients with adverse
event total person-time exposure), rather than the risk (# of
patients with adverse event total number of patients). This is
particularly useful for the more important adverse reactions and
reactions that occur at a fairly constant rate over time, but the
person-time approach can also be used when the hazard rate changes
over time. In this case, however, the observation period must be
broken into component periods (e.g., evaluating person-time rates
for each treatment for month 1, month 2, ....).
- If concurrently controlled data
are unavailable, overall rates from well-monitored, single-arm
databases can be used to provide some indication of rates that
were observed in treated patients, but there is little ability to
establish causality except insofar as reactions are predicted by
the known pharmacology of the drug.
For the most part, attributions of
causality by the investigators should be discounted, and adverse
events should be assessed without regard to attribution. Also, in
general, tables should give rates for all severities of a given
effect, although in some cases (notably cytotoxic drugs), it is
important to distinguish more and less severe reactions, as the
former may be therapy-limiting or may affect the overall
benefit-risk conclusion for the drug. For events with high
background rates (e.g., headache, fatigue, and other events that
occur frequently independent of drug therapy), however, display of
all reported events can result in a high event rate that obscures
drug-relatedness. This can be a particular problem when time on drug
is prolonged. For example, it is common for studies of 4 to 6 weeks
duration to report headache at a high (20 to 25 percent) rate. In
that case, considering the severity or causality assessment of such
events may allow a better assessment (e.g., if severe headaches are
found only in the drug-treated group). Events that are more severe
and for which subjects have multiple occurrences while on drug
therapy are more likely drug-related. In determining incidence,
however, both single occurrence and multiple occurrence events
should be counted as one event.
Some categories of adverse events
(e.g., decreased cognitive or sexual function) are notoriously
difficult to detect without special efforts, such as targeted
questionnaires. If the database includes special studies intended to
identify these events, they should generally be given more credence
than nontargeted studies, which tend to substantially underestimate
rates (See Section 7.1.9). Incidence rates should be based on
findings from the targeted studies.
The review should contain a table
(or tables) that presents the best overall display of commonly
occurring adverse events, generally those occurring at a rate of 1
percent or more (but lower rates can be presented for very large
databases). The table, or tables, will form the basis for the
adverse reaction table in labeling. The table may use a higher cut
off than 1 percent if doing this does not lose important
information. Adverse events that are equally common on drug and
placebo, or more common on placebo, are usually omitted. The
frequency cut-off for inclusion of adverse events in the table
(e.g., > 1%) is inherently arbitrary. If one is used, the review
should explain how the threshold was determined. It may also be
informative to include tables that distinguish between common
adverse events on the basis of severity. It is most common to group
adverse events within body systems, but a display by descending
frequency may also be useful.
For common adverse events, the
reviewer should attempt to identify those events that can reasonably
be considered drug related. Although it is tempting to use
hypothesis-testing methods, any reasonable correction for
multiplicity would make a finding almost impossible, and
studies are almost invariably underpowered for statistically valid
detection of small differences. The most persuasive evidence for
causality is a consistent difference from control across studies,
and evidence of dose response. The reviewer may also consider
specifying criteria for the minimum rate and the difference between
drug and placebo rate that would be considered sufficient to
establish that an event is drug related (e.g., for a given dataset,
events occurring at an incidence of at least 5 percent and for which
the incidence is at least twice, or some other percentage greater
than, the placebo incidence would be considered common and drug
related). The reviewer should be mindful that such criteria are
inevitably arbitrary and sensitive to sample size.
For adverse events that seem drug
related (the analyses suggested can have no value for unrelated
events), the reviewer should perform the following additional
analyses (see Section 7.4, General Methodology, for discussion of
methods for the explorations and analyses identified below), as
appropriate:
- Explorations for dose dependency.
These are important. The reviewer should ordinarily rely on fixed
dose studies, as titration studies tend to show that those who
tolerate higher doses have lower adverse reaction rates, but in
some cases titration studies may show a clearly increased rate of
adverse reactions with dose. It may also be useful to evaluate
safety as a function of weight-adjusted dose, body
surface-adjusted dose, or cumulative dose. Dose increases may be
associated with adverse reactions or the severity of adverse
reactions.
- For events that occur commonly,
explorations of time to onset
- For common, troublesome events
(e.g., somnolence, nausea) explorations of adaptation to develop
information on the time course of, and tolerance for, such events
- Explorations for demographic
interactions (rates and comparisons with control for demographic
and other subsets) for at least the more common and important
adverse events. The applicant will have provided such analyses
under 21 CFR 314.50(d)(5)(vi). Note that this analysis may require
use of less optimal tables of pooled results (see section
7.1.5.3).
- Explorations for drug disease and
drug-drug interactions if there is a strong signal for an
interaction or good rationale for expecting an interaction
- Selective exploration of certain
adverse events in an attempt to better characterize them. For
example, if rash appears to be drug related, the reviewer may want
to look more closely at individual cases of rash. The applicant's
line listing of all adverse events across the entire phase 2 to 3
databases would be a good source for identifying individual cases
of rash. If a subject dropped out because of rash, the applicant
should have provided a narrative discussion of the event, which
would also be a good source for attempting to better characterize
the event. Although the data collected on nonserious adverse
events is usually sparse, the reviewer could still request
additional information from the applicant on commonly occurring
adverse events that require further characterization.
- When adverse events of a given
type vary markedly in severity, separate analyses of each severity
may be useful.
A description of the methods used in
such additional explorations should be provided, with all results,
interpretations, and pertinent discussion. Where an applicant's
analysis is considered inadequate, this should be noted and an
alternate developed by the reviewer or requested from the applicant.
In general, a fairly large database
is needed to evaluate less common adverse events. To identify
relatively rare events of significant concern, the reviewer has to
examine the occurrence of adverse events over the entire phase 2 to
3 database, including data for which there is no useful concurrent
control. The overall database is typically heterogeneous, including
uncontrolled exposure for varying durations and at varying doses,
and is unlikely to lend itself to meaningful estimates of rates or
assessment of causality (except where there has been rechallenge).
Thus, it may be sufficient for the reviewer to group these data in
gross categories of incidence and by body system. For example, it
may be useful to categorize less common events in order of
decreasing frequency within the following incidence ranges:
- Adverse events occurring at rates
less than or equal to 1/100
- Adverse events estimated to occur
at rates between 1/100 and 1/1000
- Adverse events estimated to occur
at rates less than 1/1000
The reviewer should then develop a
condensed list of reactions to be included in the Adverse Reactions
section of labeling.19 This
list should eliminate events that are common in the general
population and not likely to be drug related and adverse events
characterized by terms that are too vague to be helpful, unless the
reviewer is able to identify a more meaningful term that was
subsumed into the vague term when the adverse event was coded by the
applicant (see Section 7.1.5.2 above).
Some of the reactions in the
condensed list may be of particular concern, but insufficiently
clear as to whether they are caused by the drug to lead to a
Warning/Precaution in the labeling. In that case, it is useful to
notify the safety evaluator in the Office of Drug Safety who will be
monitoring the drug after marketing.
The approach to review of laboratory
findings (chemistry, hematology, and urinalysis) is generally
similar to that suggested for the other categories of safety data.
As considered in greater detail below, the review should identify
laboratory tests performed in the clinical studies, describe the
dataset from which laboratory findings information is obtained,
describe the methods used to assess findings, discuss pertinent
findings, and review the more important findings in depth.
Laboratory findings discussed in detail in other sections of the
review (e.g., Section 7.1.2 Other Serious Adverse Events, Section
7.1.3 Dropouts and Other Significant Adverse Events) need not be
discussed in detail in this section. This section should refer to
the more detailed discussions of such findings elsewhere in the
review.
The review should provide an
overview of what laboratory testing (chemistry, hematology, and
urinalysis) was carried out. It is preferable to summarize the
overall approach, rather than provide detailed comments about
laboratory testing for each study. The review should contain the
following to the extent relevant to the data:
- Discussion of any discrepancies
between planned analyses and analyses that were done (e.g., tests
omitted or added, changes in planned frequency of testing)
- Discussion of procedures used to
evaluate abnormal values (e.g., whether patients were followed
until their values normalized, whether any patients were
rechallenged, the procedures used for sample analysis (i.e.,
central or local labs, windows of time in which lab values
were considered20 )
- A summary table identifying the
numbers of patients exposed to test drug who had baseline
laboratory values and follow-up assessments
- Whether results of unscheduled
lab tests were included in the principal analyses and tables
The reviewer should note that
laboratory tests obtained at unscheduled visits (e.g., when a
patient is hospitalized for an adverse event) are often not included
in the NDA/BLA laboratory database. In those cases, the only place a
reviewer would learn of an abnormal laboratory value might be a
narrative summary (or occasionally a CRF). Too often, however, the
narrative summary includes only a preferred or verbatim term (e.g.,
acute renal failure) and does not include the laboratory
value of interest (e.g., BUN/creatinine). In such cases, the
laboratory data of interest should be requested from the applicant.
Controlled comparisons generally
provide the best data for deciding whether there is a signal of an
effect of a drug on a laboratory test. Placebo-controlled trials are
generally short term, however, and therefore unsuitable for
assessing late-developing abnormalities, so that longer term data
also need to be examined. If there is no concomitant control,
comparison may need to be made with similar populations outside the
NDA (e.g., in other applications). In identifying the sample
population for comparison of laboratory values, the reviewer should
pool relevant studies. The review should explain how the studies to
be pooled were selected. In comparing laboratory values, there are
additional considerations when using pooled data (in addition to
those discussed in Section 7.4.1 Methodology, Pooling), including:
- The methods of sample collection
and handling in different studies
- The assay methods used in
different studies
- The reference ranges used in
different studies
Several analyses may be needed.
Separate analyses should be performed for patients with normal
values at baseline, for patients with abnormal values at baseline,
and for patients without baseline values. In general, there will
need to be at least one analysis that includes all data (data from
planned or unplanned visits, values collected as follow-up to
abnormal findings).
This review should generally include
three standard approaches to the analysis of laboratory data. The
first two analyses are based on comparative trial data. The third
analysis should focus on all patients in the phase 2 to 3
experience. Analyses are intended to be descriptive and should not
be thought of as hypothesis testing. P-values or confidence
intervals can provide some evidence of the strength of the finding,
but unless the trials are designed for hypothesis testing (rarely
the case), these should be thought of as descriptive. Generally, the
magnitude of change is more important than the p-value for the
difference.
The central tendency analysis
generally compares mean or median changes from baseline across
treatment groups, and the review should contain the results of these
analyses for all laboratory measurements. Although marked outliers
are typically of greatest interest from a safety standpoint (see
below), at times a potentially important effect may be revealed only
in analyses looking at differences in mean change from baseline. For
example, several drugs that cause modest decreases in uric acid
because of a uricosuric effect have caused acute renal failure (ticrynafen,
suprofen) in inadequately hydrated patients. Suprofen was withdrawn
from the market for this reason. Mean changes in electrolyte levels
can also signal risks.
It is generally useful to include as
appendices tables providing data on central tendency (see Table
7.1.7.3.1.1). The reviewer should note and discuss signals that
emerge from these tables and indicate those for which further study
is needed, if any.
The review should focus on patients
whose laboratory values deviate substantially from the reference
range. Applicants usually include displays and analyses designed to
detect such outliers. The relevant data would come from shift
tables, scatter plots, box plots, cumulative distribution displays,
and tables providing incidence of patients across treatment groups
who had a potentially clinically important deviation from normal on
one or more laboratory parameters while on treatment (see Tables
7.1.7.3.2.1). In analyzing outliers, the reviewer should be aware of
the following:
- Regression to the mean (and an
apparent upward shift) can be expected if patients are screened
for normality, giving a shift even if there is no drug effect;
comparison with control groups is critical.
- If there are more measurements
performed during treatment than baseline and abnormal values are
randomly occurring, there is more opportunity for outliers during
treatment. Again, comparison with a control group is critical.
- For important laboratory
parameters, the reviewer should carefully consider the cut-points
used by the applicant to define normal and abnormal.
- If values used to identify
outliers are too extreme, important findings may not be
identified.
- If values used to identify
outliers are not large enough, important findings may be obscured
by grouping important outliers and trivial findings (e.g., values
greater than two times upper limit of normal for transaminase are
common in many datasets and may not distinguish hepatotoxic from
non-hepatotoxic agents; 3-fold and higher elevations appear to be
more discriminating).
Decisions about what criteria to use
to identify outliers should, if possible, be made at the pre-NDA
meeting. Because it is not possible to know in advance what criteria
will be optimal for detecting between-group differences, it may be
useful to conduct analyses using cut points other than those chosen
by the applicant. In addition, it may be useful to consider
between-group comparisons of the following:
- Cumulative or other distributions
of data, rather than solely proportions of patients meeting some
arbitrary criterion
- Patients with large shifts within
the normal reference range
- Patients who meet outlier
criteria for more than 1 variable simultaneously (e.g.,
transaminase and bilirubin)
- Patients having persistent
abnormalities (more likely to be real deviations)
Analyses of outliers should serve as
a source of signals for events to explore in more depth. The
reviewer should discuss signals that emerge and indicate those for
which further exploration is needed. The details of the explorations
carried out and the results should be provided in subsection 7.1.6.4
as described below.
The reviewer should carefully
analyze individual patients with large changes in laboratory values.
These changes are much more likely to identify significant problems
than mean or median changes from baseline. Applicants typically
provide a list that identifies patients with extreme changes,
usually specified in advance. Individual patient data displays
should be available to the reviewer for all such patients. Even for
relatively uncommon events, it is helpful to compare rates in
treatment and control groups.
Discontinuation of treatment for a
laboratory abnormality may be considered a marker of perceived
clinical importance of the finding. It is again useful to compare
treatment groups, taking into account duration of treatments, for
rates of discontinuation for particular laboratory abnormalities.
Because of the importance of looking at dropouts for laboratory
changes (even a small number of marked abnormalities, such as liver
function or WBC count, may signal major problems), all such dropouts
in the phase 2 to 3 population should be identified. The reviewer
should generally analyze and comment on each individual patient
identified as dropping out for any significant laboratory
abnormalities. In some cases, it is critical to note whether
appropriate testing has been carried out to rule out
non-drug-related mechanisms (e.g., viral hepatitis serological
testing in patients with transaminase elevation or more severe liver
injury) and whether appropriate additional tests have been performed
(e.g., bilirubin in patients with transaminase elevation).
Additional analyses may be
appropriate for certain laboratory findings, including analyses for
dose dependency, time dependency, and also drug-demographic,
drug-disease, and drug-drug interactions (see Section 7.4
Methodology). The review should discuss the rationale for additional
explorations, the methods used, and the results and interpretations.
Certain laboratory assessments are
so critical to the safety assessment that they deserve special
attention in any review. For example, hepatotoxicity has been an
important cause of drug marketing withdrawals from the 1950s (iproniazid)
to the present (ticrynafen, benoxaprofen, troglitizone, bromfenac)
and has led to important limitations on the use of many more drugs (isoniazid,
labetalol, trovafloxacin, tolcapone, nefazodone, felbamate). At
present, it appears that a potential for severe hepatotoxicity may
be signaled by a set of findings sometimes called Hy's Law,
based on the observation by Hy Zimmerman, a major scholar of
drug-induced liver injury, that a pure hepatocellular injury leading
to jaundice had serious implications, a 10 to 50 percent mortality.
Over the years, this observation has led to the following
proposition:
In a drug development database, a
potential for severe hepatotoxicity is signaled by the following set
of findings:
1. An increased rate of
transaminase elevations (3x ULN, 5x ULN, 10x ULN, etc.) in
treated patients compared to control
2. No significant evidence
of obstruction (elevated AP), although some elevation may
follow severe hepatocellular injury
3. A very small number of
cases (two, perhaps even one) of transaminase elevation
accompanied by a rise in bilirubin to 2x ULN
The explanation for the usefulness
of this signal is the high capacity of the liver for bilirubin
excretion; it takes a good deal of damage to the liver to impair
bilirubin excretion (in the absence of obstruction). This signal has
been present for troglitizone, bromfenac, and dilevalol (never
approved in the United States, but hepatotoxic in Portugal).
Table 7.1.7.5.1 is an outline of a
comprehensive assessment of available data pertinent to potential
hepatotoxicity. A similar outline will be developed for assessment
of electrocardiographic QT abnormalities, a risk factor for
potentially fatal arrhythmias (see section 7.1.9).
Vital signs can be analyzed and
reported using an approach essentially identical to that taken for
laboratory data. This section should be organized in a similar
manner to the laboratory section.
7.1.8.1
Extent of Vital
Signs Testing in the Development Program
7.1.8.2
Selection of
Studies and Analyses for Overall Drug-Control Comparisons
7.1.8.3
Standard Analyses
and Explorations of Vital Signs Data
7.1.8.3.1
Analyses Focused
on Measures of Central Tendency
7.1.8.3.2
Analyses Focused
on Outliers or Shifts from Normal to Abnormal
7.1.8.3.3
Marked Outliers
and Dropouts for Vital Signs Abnormalities
7.1.8.4
Additional Analyses
and Explorations
7.1.9 Electrocardiograms (ECGs)
ECG data can be analyzed and
reported using an essentially identical approach to that taken for
laboratory data. The adequacy of the assessment (see 7.2) may be
especially important in this case, given recent experience with
drugs that prolong the QT interval and cause the ventricular
tachycardia known as Torsade de Pointes (TdP). A guidance document
on the design, conduct and interpretation of clinical studies
assessing the effects of drugs on the QT interval is under
development as a part of the ICH effort. The current version, a step
2 guidance (E14: Clinical Evaluation of QT/QTc Interval Prolongation
and Proarrhythmic Potential for Non-Antiarrhythmic Drugs) can be
found at http://www.ich.org. The safety review should provide in
this section an overview of effects on the QT interval, organized in
a similar manner to the laboratory section. This section of the
safety review should summarize the results of any studies designed
specifically to assess the effects of the drug on the QT interval.
This section should describe the
number of baseline and on-study ECGs obtained, who read the ECGs,
and what methodology was used (e.g., automatic, blinded
cardiologists).
7.1.9.2
Selection of
Studies and Analyses for Overall Drug-Control Comparisons
7.1.9.3
Standard Analyses
and Explorations of ECG Data
7.1.9.3.1
Analyses Focused
on Measures of Central Tendency
7.1.9.3.2
Analyses Focused
on Outliers or Shifts from Normal to Abnormal
7.1.9.3.3
Marked Outliers
and Dropouts for ECG Abnormalities
7.1.9.4
Additional Analyses
and Explorations
7.1.10
Immunogenicity
Data on the impact of immunogenicity
(if applicable) on safety, efficacy, and/or clinical pharmacology
and pharmacokinetics may be summarized in this section and
referenced throughout the review.
All therapeutic proteins have the
potential to elicit antibody responses. An antibody response to a
protein may have no consequences or, in some cases, can lead to
potentially serious sequelae. Adverse immune responses to a protein
drug could result in one or more of the following outcomes:
- For a product that is intended as
replacement for a missing endogenous substance, antibodies could
neutralize the replacement product and generate a clinical
deficiency syndrome.
- Neutralization of a protein
product by blocking antibodies could reduce the efficacy of
a life-saving product.
- Antibody development could result
in a life-threatening hypersensitivity response.
Factors that tend to increase the
likelihood of an immune response include whether the protein is
highly conserved in nature (less likely if it is), whether the
protein product is administered via the subcutaneous route (more
likely if it is), and whether the protein intended for chronic use.
This section of the review should assess the adequacy of the
immunogenicity data provided to address these issues.
Although formal studies in humans of
the carcinogenic effects of drugs and biologics are uncommon,
reflecting the expectation that induction of cancer would occur over
a very long period of exposure, a systematic assessment of human
tumors reported during drug development can provide useful safety
information in some cases. Such an assessment would be appropriate
where controlled studies are of long duration (e.g., more than a
year), especially for drugs or biologics that have positive
genotoxicity or animal carcinogenicity findings or are known immune
modulators.
The review should describe and
discuss results of any studies designed to evaluate a specific
safety concern or concerns. These studies may include:
- Studies to assess whether a drug
has safety concerns common to its pharmacologic class (e.g., a
study to assess effects of a benzodiazepine hypnotic on driving,
respiration, memory, or next day psychomotor functioning)
- Studies in topical products
(including systemic products delivered by a patch) to assess
cumulative irritancy, contact sensitizing potential,
photosensitivity, and photoallergenicity
- Studies to characterize a drug's
effect on QT interval, part of most modern development efforts
- Studies intended to demonstrate a
safety advantage over therapeutic alternatives (less
extrapyramidal effect for an antipsychotic, less sedation for an
anti-histamine, less cough from an angiotensin II blocker than an
ACE inhibitor). Such studies must include the comparator agent (a
failure to see the side effect in a placebo-controlled study is
usually not informative without the active control to demonstrate
assay sensitivity).
- Studies in special populations
thought to be at increased risk and likely to use the drug.
In labeling, the results of these
studies should, as appropriate, supersede data from less targeted
studies (e.g., observational safety data collected from efficacy
trials).
The review should contain a
discussion of abuse potential and any apparent withdrawal symptoms.
For therapeutic classes with a history of abuse potential and
withdrawal phenomena (e.g., sedative/hypnotics and anxiolytics),
studies are usually performed to assess these issues. The review
should comment on the adequacy and findings of these studies. For
other drugs, adverse events that emerge after discontinuation of the
drug should be assessed to determine whether they may indicate a
withdrawal phenomenon. If the applicant evaluated the potential for
withdrawal phenomena, the review should indicate whether there was a
prospective or post-hoc assessment of withdrawal emergent signs and
symptoms (during drug taper or following discontinuation) and
discuss the implications of the approach used on the reliability of
the findings.
Although formal studies in humans of
the effects of drugs on reproduction, pregnancy, or lactation are
uncommon, the review should summarize any drug exposure in pregnant
or nursing women, including any inadvertent exposure during the
drug's development and exposure identified from secondary sources
(e.g., postmarketing surveillance). If there is no information on
drug exposure in pregnant or lactating women, the review should
acknowledge that fact. The review should discuss positive and
negative findings.
Increasingly, clinical reviewers are
presented with analyses of height and weight data collected during
studies of pediatric subjects. These data are generally inadequate
to allow for definitive conclusions about an effect of drug on
growth for several reasons. Assessment of the effect of drug on
growth requires accurate measurements, particularly for height, and
in most studies, height is not measured accurately. Growth is a
process that occurs over long periods of time, and controlled trials
of several weeks duration may not provide a sufficient period of
observation to assess the effect of drug on growth. Open label
studies can offer longer periods of time to observe effects on
growth, but the lack of a control group limits the ability to
separate the effect of drug and underlying disease on growth. Review
of height and weight data for possible effects on growth makes use,
in part, on approaches described above in the laboratory data
section. Analysis of changes in central tendency and outlier
analysis, for example, apply to the evaluation of the effect of a
drug on growth. There are, however, some distinctive issues that
must be considered.
First, the sponsor should describe
how weight and height were measured. The manner in which these
measurements were made will bear on how much confidence the reviewer
can have in the data provided. For example, a development program in
which the measurement schedule and methodology were standardized and
in which the study staff were trained in measurement, will result in
more reliable data than a development program that did not
standardize procedures. The review should therefore include a
description of the measurement methodology.
Second, growth is not constant
throughout childhood and varies by age and sex. Without
consideration of these factors at baseline, absolute mean changes in
weight and height can give misleading results. Adjustment of growth
for age and sex can be done by conversion of a child's height and
weight to a z-score, which is the number of standard deviations that
an individual's measurement is from the mean for age and sex matched
children in the general population. A decrease in mean z-score for a
group is interpreted as evidence of a lag in growth compared to what
would be predicted using general population data. In a controlled
trial, differences in mean z-score changes from baseline between
treatment groups may provide evidence of an effect of drug on
growth. Declines in mean z-scores in open label studies are less
easily interpreted because these could result from the effect of
drug or could be caused by the disease for which the treatment is
being studied.
Sponsors should provide analyses of
height and weight data that assess measures of central tendency and
outlier analyses using height and weight z-scores. Although results
from these analyses will not provide definitive proof of drug
related effects on growth in most cases, they may help identify
candidates for prospective studies of the effect of drug on growth
in children. The review team should request such analyses at the
pre-NDA meeting.
The review should summarize all
overdose experience with a drug in humans (including both
information provided by the applicant and information obtained from
secondary sources) and describe the constellation of signs,
symptoms, and other abnormalities one might expect to see in
association with overdose. Phase 1 data should be reviewed to
identify subjects who may have received higher doses than those used
in later phases of study. In addition, patients with certain
physiological differences that would compromise their ability to
clear a drug (e.g., renal impairment, hepatic impairment, limited
CYP4502D6 activity for a drug cleared by this isozyme) may provide
data relevant to the clinical implications of overdose.
Relevant findings from postmarketing
experience, if any, should be described briefly here and referenced
in the summary section (Section 7.3).
Section 7.1 is an assessment of the
adverse events seen during the development program. Section 7.2
should provide the reviewer's comments on the adequacy of drug
exposure and the safety evaluations performed as part of the
development program. This section addresses the regulatory question
of whether or not all tests reasonably applicable were
conducted to assess the safety of the new drug. Was there adequate
experience with the drug in terms of overall numbers of patients and
in appropriate demographic subsets of patients? Were doses and
durations of exposure appropriate? Were all (or not all) appropriate
tests performed in the exposed patients? Were all necessary and
appropriate animal tests performed? Were all the appropriate
clinical tests carried out (e.g., electrocardiographic assessment of
effects on QT interval)? Was the drug adequately worked up
metabolically? Were appropriate in vitro studies of drug-drug
interaction carried out according to current guidelines? Were all
potentially important findings adequately explored: for example, to
what extent was psychomotor impairment specifically assessed in a
drug that is sedating?
If important data are missing, this
could influence the regulatory action on the drug. A critical task
of the reviewer in this section is identification of specific
concerns that need to be addressed by the applicant, either before
approval or postapproval. Even more than for most other parts of the
review, the reviewer needs to be conscious of recent developments
and discuss issues broadly. Finally, this section is the place for
detailed comments on the quality and completeness of the data
provided.
The review should clearly describe
the studies and overall extent of the data supporting the evaluation
of safety. The reviewer should then make a judgment about the
adequacy of the clinical experience with the new drug for assessing
safety.
In this section the reviewer should
identify and characterize the primary safety data sources used in
conducting the review. If these are described elsewhere in the
review, this section can reference those sections. The primary
source is generally the database derived from the applicant's
development program. Studies in this program will generally have
full study reports related to safety, or studies that are grouped
for analysis of safety in an Integrated Summary of Safety; case
report forms will be available. These studies usually will have been
closely monitored. Secondary sources may also be available and may
be of critical importance (e.g., for a drug already available in
other countries), and there may be some parts of the database that
have had limited analyses (i.e., only for deaths and adverse
dropouts); these are described in section 7.2.2.
Tables and graphs are useful in
describing the data sources for the safety review. Generally, the
reviewer should use the tables and graphs in this section to
characterize the overall database. The detailed tables and other
displays for this subsection may be included in an appendix to the
review, but summary tables and narrative statements should be
included here. The reviewer should also characterize the per patient
data (narratives, CRFs, CRTs and electronically accessible databases
for baseline information. See section 7.4 for discussion of ability
to link databases.
The reviewer should include in an
appendix a table, such as that illustrated in Table 7.2.1.1.1,
enumerating all subjects and patients across the entire development
program, phases 1 to 3. This is a critical table that identifies the
important patient pools and denominators for subsequent analyses and
incidence estimates.
The reviewer should also include an
appendix table that provides brief descriptive information for all
individual studies, including study design (fixed dose vs. flexible
dose, parallel vs. crossover), dosing schedule, study location
(foreign vs. domestic), treatment groups and doses, sample sizes,
patient population (elderly). Studies that were designed to assess a
particular aspect of safety (e.g., ECG, ophthalmic) should be noted.
Most NDAs/BLAs will include a table of all studies, as such a table
is called for in the Clinical/Statistical guidline and in the
Common Technical Document.
Applicants sometimes segregate
certain clinical trials from their primary source data (see 7.2.2,
secondary source data), especially foreign data. This may be
appropriate, especially if there is a basis for believing that these
data differ substantially in quality and/or completeness or in
critical aspects of investigator practice from the data included in
the primary source database. This is a matter of judgment, however,
and cannot be assumed to be valid. An explanation should be provided
in the review describing the basis for decisions about what data
were included and what excluded from the primary source data.
An NDA/BLA generally includes data
from patient samples that are at different levels of completeness in
terms of data entry, information collected, and validation. Table
7.2.1.1.1 should include patient counts (or estimates) from all
studies contributing data, regardless of these factors. Data cutoff
dates or database lock dates for the various databases
comprising the NDA/BLA should be identified at this point in the
review. For example, the cutoff date for the overall safety database
derived from completed studies might be more distant, while the
cutoff date for submitting serious adverse events from all studies
may generally be more recent. These dates may likely need updating
during the course of NDA/BLA review as more data become available.
The reviewer should include appendix
tables in a format similar to that illustrated in Table 7.2.1.2.1
(showing percent distribution within treatments of patients by age,
gender, and race as well as weight in various groups), providing
overall demographic information for phase 1 and phase 2 to 3 study
pools separately. It may be appropriate to provide demographic
displays for subsets within these larger pools at other points in
the review.
There are many ways to summarize the
dose and duration experience with a new drug. Either can be
expressed as mean, median, maximum, with histograms or other
displays that give the numbers exposed at various doses or for
various durations. A particularly useful approach is to provide
combined dose and duration information. It is suggested that the
review contain tables in the format illustrated in Table 7.2.1.3,
enumerating patients on the basis of mean daily dose of the NDA/BLA
drug and duration of administration for phase 1 and phase 2 to 3
study pools separately. If the study used a titration design, the
modal dose (if 2 different doses were used for the same duration,
the larger, or maximal modal dose) may be the more useful summary
statistic. It is particularly important to examine the subgroup of
patients who received a dose at least as large as the dose intended
for marketing.
It may also be useful to provide
similar tables based on maximum dose, modal dose, dose expressed as
mg/kg or mg/m,2 or even plasma concentrations, if such
data are available.
It may also be useful to provide
similar tables for various subgroups (e.g., males and females
separately, various age groups separately, and patients with various
comorbid illnesses of interest separately). There should be similar
displays for active control drugs if any were included in trials for
the new drug.
Finally, it may be useful for the
review to include an appendix table providing total person time
exposure data for the NDA/BLA drug, active control, and placebo, for
the phase 2 to 3 database.
Secondary source data are (1) data
derived from studies not conducted under the applicant's IND and for
which CRFs and full study reports are not available,21
or studies so poorly conducted (e.g., poor ascertainment for adverse
events), that they cannot be reasonably included in the primary
source database, (2) postmarketing data, and (3) literature reports
on studies not conducted under the IND. Often the applicant may have
made the distinction between the data considered primary source data
and other data, and the reviewer needs to examine the rationale for
this distinction.
The secondary sources should be
briefly described. It is worth emphasizing that secondary source
data may be a critical source of information for review, despite the
generally lower quality of these data, because they often provide
the larger database needed to look for less common serious adverse
events and may be reliable with respect to deaths and serious
adverse events.
The NDA/BLA should be clear in
describing exactly what other studies provided data and what the
basis was for not integrating such data with the primary source data
(e.g., no CRFs, no study reports, not adequately monitored). Lack of
clarity in this should be noted by the reviewer.
If postmarketing data are available,
this section should describe briefly the type of information
available for review. An example of such a description would be a
comment that a line listing for (a specified number of) spontaneous
reports from marketing in (country) was provided, along with
narrative summaries for the serious adverse events among the reports
and an estimate of product use in (country) during that time period.
As is the case for most spontaneous reports, these reports are
likely to be difficult to interpret. Important events will be
described in appropriate sections (e.g., 7.1.1 and 7.1.2, Deaths and
Other Serious Events).
Relevant literature may be
incorporated in various sections of the NDA, but is ordinarily
included in section 5.4 of the Common Technical Document (M4: The
CTD -
Efficacy).
The NDA/BLA may include a separate literature section or the
literature may be provided or referenced as called for in various
places in the Clinical/Statistical guidline in section II F, Other
Studies and Information. The applicant should have provided a
description of the search strategy to assess the world literature
(e.g., databases used, key search words), the personnel who carried
it out (their credentials) and whether the search relied on
abstracts or full texts (including translations) of articles. A
cutoff date for the literature search should also have been
provided. A copy (translated as required) should have been submitted
for any report or finding judged by the applicant to be potentially
important.
This section of the review should
describe what information from the literature search was provided
for review, the extent to which the above description of an ideal
presentation was met, and whether any missing information is
important (and/or was obtained by the reviewer). Independent
literature reviews conducted by the reviewer should be described
here as well.
Actual safety findings should be
described in appropriate sections of the safety review to present
from the literature reports in this section of the review.
In evaluating the adequacy of
clinical experience with the drug, the reviewer should refer to
current ICH guidance on extent and duration of exposure needed to
assess safety22 as well as the
draft guidance on Pre-Marketing Risk Assessment.23
The review should specifically address the following:
- Whether an adequate number of
subjects were exposed to the drug, including adequate numbers of
various demographic subsets and people with pertinent risk factors
- Whether doses and durations of
exposure were adequate to assess safety for the intended use
- Whether the design of studies
(open, active-control, placebo-control) was adequate to answer
critical questions
- Whether potential class effects
were evaluated (e.g., for anti-arrhythmic effects, evaluation of
the potential for pro-arrhythmic effects) and whether problems
suggested by pre-clinical data were assessed
- Whether patients excluded from
the study limit the relevance of safety assessments (e.g.,
diabetics, people over 75, people with recent myocardial
infarction, people with renal or hepatic functional impairment, or
people on other therapy). This may depend on the signals of
toxicity that were observed in the patients who were studied.
The clinical reviewer should not
attempt a general assessment of the preclinical program, but rather,
comment on whether preclinical testing was adequate to explore
certain potential adverse reactions, using preclinical models based
either on a drug's pharmacology or on clinical findings that emerged
early in clinical development. For example, for a drug anticipated
to cause QT prolongation because of its drug class or because QT
prolongation was seen in phase 1 studies, there are in vitro models
to evaluate this potential. The reviewer should note whether such
studies were done. If such studies were performed, the results would
be summarized in the Pharmacology Review.
The reviewer should comment on the
adequacy of routine clinical testing of study subjects, including
efforts to elicit adverse event data and monitor laboratory
parameters, vital signs, and ECGs. In assessing the adequacy of
clinical testing, the reviewer should consider the adequacy of the
methods and tests used and the frequency of testing. The adequacy of
specific testing intended to assess certain expected or observed
reactions should be discussed under subsection 7.2.7.
The reviewer should be alert to the
absence of data in an NDA laboratory database for analytes that are
typically included in routine laboratory monitoring. For example, it
was discovered after approval that the NDA laboratory database for
the anti-epileptic drug zonisamide did not have data on serum
bicarbonate. It was later determined that this drug is associated
with a non-anion gap metabolic acidosis. The serum bicarbonate data
would have been helpful in identifying this adverse reaction
earlier.
Knowledge of how a drug is
metabolized and excreted is critical to anticipating safety problems
in patients with impaired excretory or metabolic function and
problems resulting from drug-drug interactions.
Drug-drug interaction assessment is
a critical part of a modern drug development program and should
evaluate the drug both as a substrate for interactions (interference
with its clearance) and as an inducer or inhibitor of the clearance
of other drugs. The reviewer should comment on the adequacy of in
vitro and in vivo testing carried out by the applicant to identify
the following:
- The enzymatic pathways
responsible for clearance of the drug and the effects of
inhibition of those pathways, notably CYP450 enzymes and p-glycoproteins
- The effect of the drug on CYP450
enzymes (inhibition, induction) and the effects of the drug on the
PK of model compounds
- The major potential safety
consequences of drug-drug interactions
Details of these assessments will be
found in the Clinical Pharmacology Review and in the summary of that
evaluation in the Medical Officer's Review.
The reviewer should discuss the
adequacy of the applicant's efforts to detect specific adverse
reactions that are potentially problematic and might be expected
with a drug of any class (e.g., QT prolongation or hepatotoxicity)
or that are predicted on the basis of the drug class (e.g., sexual
dysfunction with SSRI antidepressants). The reviewer should also
discuss whether the applicant should have made efforts to assess
certain events that it did not assess. The reviewer should also
discuss pertinent negative findings (absence of findings) for a drug
in this section of the review (see examples below).
The adverse events that warrant
specific attention will vary depending on the characteristics of the
drug and the drug class. The known pharmacology of the drug would
suggest some evaluations (e.g., first dose effects for peripheral
alpha blockers, tolerance and withdrawal effects for central alpha
agonists, urinary retention with anti-cholinergics, QT prolongation
with type III anti-arrhythmics, extrapyramidal effects with
antipsychotics, muscle pain with statins), while experience with
other members of the class would suggest others (e.g.,
hepatotoxicity with thiazolidinedione PPAR gamma agonists (glitizones),
tendon problems with fluoroquinolones). There should be a subheading
for each adverse reaction that warrants special consideration (even
if not observed) and, under each subheading, a discussion of what
was done to detect the reaction and the adequacy of the approach.
The following list of potential adverse reactions, and some of the
drug and therapeutic classes that might trigger higher interest in
them, may be a useful starting point in assembling a list (it is
also important to examine labeling for other members of the drug's
pharmacologic class):
Hepatotoxicity (NSAIDs,
thiazolidinedione PPAR gamma agonists)
Pancreatic toxicity
QT prolongation (any
antiarrhythmic, antipsychotic, antihistamine, fluoroquinolone)
Vasodilator effects, such as
hypotension (alpha blockers) or edema (dihydropyridine calcium
channel blockers)
Withdrawal effects (beta
blockers, central alpha agonists, SSRIs, narcotics)
Orthostatic hypotension (any
antihypertensive, antipsychotics)
Hypertension (any
sympathomimetic or phosphodiesterase inhibitor)
Tachycardia
Neutropenia (drugs related to
ticlopidine, procainamide, clozapine)
Bleeding (drugs inhibiting any
aspect of clotting or platelet function, NSAIDs)
Aplastic anemia
Increased coagulation times
Muscle injury (any HMG CoA
Reductase Inhibitor (statin) or other lipid-lowering drug)
Sedation (any psychotropic
drug)
CNS stimulation
Anticholinergic activity
Allergic reactions
Sexual dysfunction (any
antidepressant, sedating drug)
Elevated intraocular pressure
Cataracts
Retinopathy
Worsening glucose
tolerance/diabetes (diuretics, atypical antipsychotics)
Pro-arrhythmic effects and
increased mortality (most nonbeta blocker anti-arrhythmics)
Increased CHF and SD mortality
(any inotrope, some negative inotropes such as calcium channel
blockers)
Nephropathy (NSAIDs)
Example 1: If orthostatic
hypotension was an expected adverse reaction, but was not observed,
the reviewer should determine whether the applicant made efforts to
detect it and, if so, whether the applicant's approach (e.g., timing
and frequency of vital signs testing) was adequate to detect it.
Example 2: If QT prolongation
was observed in phase 1 studies, the reviewer should ascertain
whether the applicant made efforts, beyond routine ECG testing, in
phases 2 and 3 to explore the consequences in patients of the
observed QT prolongation and, if so, whether those efforts were
adequate, including adequate exposure to higher doses. For example,
how did the applicant follow-up patients who experienced clinical
events that may be manifestations of torsade de pointes (e.g.,
syncope, dizziness, or palpitations)? Holter monitoring, for
example, might have been appropriate in such patients.
The reviewer should provide general
overall assessments of the quality and completeness of the data
available for conducting the safety review and describe the bases
for these assessments. More than that, attention to completeness and
quality of assessment is important throughout the review. The
reviewer should recognize that quality may differ from the
primary source data and for data over which the applicant had less
control. The following examples illustrate some of the ways in which
applicants can differ in the quality and completeness of data they
provide:
- Applicants may differ in what
they include in a CRF. For example, if additional laboratory data
are collected at unscheduled visits or after the normal end of a
trial, some do not include these data. Such data may be stored in
some other place (a correspondence file). Sometimes
additional information is attached to the front of the CRF as
queries. If CRFs do not indicate any additional testing beyond
the routine assessments, the reviewer should ascertain whether
additional testing was done to reassess abnormal values before the
next routine visit (e.g., at an unscheduled visit). If the CRFs do
not indicate that additional testing was performed, the reviewer
should ask the applicant if additional laboratory data are
available.
- If it is apparent that the CRF
contains insufficient information about an adverse event (e.g., if
a patient was hospitalized for an adverse event), the reviewer
needs to determine whether there is additional information
available. Such an observation also raises the general question of
whether all pertinent data have been included in CRFs.
- The reviewer should be concerned
about patients with abnormal clinical or laboratory findings who
are lost to follow up, particularly if there are significant
numbers of such patients. In these situations, the reviewer may
consider asking the applicant to attempt to obtain the needed
follow up information. If the information cannot be obtained, it
may be appropriate to perform sensitivity analyses to assess the
possible impact of missing data, assuming a worst-case outcome.
- The reviewer should be
particularly alert to situations in which applicants make changes
in CRFs to reclassify adverse events or reasons for subjects
dropping out without the investigator's agreement. There is
greater concern where serious adverse events are reclassified and
reclassifications are done without blinding. The reviewer should
ask the applicants about procedures used (if unclear) and attempt
to assess the impact of multiple changes on the safety evaluation.
- For electronic data, the reviewer
should clarify what information is, and is not, included in the
electronic files. For example, if a reviewer is relying on
electronic files from the case report forms, it is important to
know what, if any, information from the CRFs was not included. A
separate file may be needed for any missing data.
7.2.9 Additional
Submissions, Including Safety Update
The initial NDA/BLA submission
may not contain all information pertinent to the safety
evaluation. Further data submissions may be planned at the time
of initial submission and filing (e.g., results of additional
long-term follow-up), may represent responses to specific
questions or discipline review letters, or may be part of the
safety update required under regulations (21 CFR 314.50(d)(5)(vi)(b)).
It is critical to review these data to determine whether safety
conclusions are affected, particularly with respect to serious
or fatal events.
This section should:
- Describe safety submissions,
noting whether the results have been incorporated into the rest of
the review or are considered in this section
- For those safety matters not
incorporated into the rest of the review, discuss any data with
important implications for safety. In general, this will involve
deaths, adverse dropouts and other serious events, and these
should be considered as in sections 7.1.1, 7.1.2, 7.1.3, as
appropriate to the (usually) small numbers. Only if these events
alter the overall safety picture will a more detailed discussion
of the entire area (e.g., deaths, liver injury) be needed.
Any reports of important changes in
foreign labeling or new studies that give insight into more common
events should also be noted.
This section of the review should
briefly summarize each of the adverse reactions that the reviewer
considers important and drug-related (i.e., this should constitute a
problem list for the drug). For each adverse reaction, there
should be a separate subheading followed by a brief summary of the
reaction and references to sections of the review (e.g., other parts
of the safety section, Clinical Pharmacology, studies described in
the Efficacy section) containing more detailed information about the
adverse reaction generally, or specific aspects of the reaction. The
review should integrate by reference all relevant details about the
reaction, including patient identifying numbers for certain patients
(e.g., for deaths). Below is a sample summary section entry for QT
prolongation:
QT
Prolongation
Dose-related QT prolongation
compared to control was seen in all controlled trials, with
a mean change of 20 msec at 100 mg/day (peak), the
recommended maximum dose, and smaller changes at lower
doses; 5 percent of patients had QTc values over 500 msec at
some point, compared with ___ percent on placebo. The drug's
metabolism is predominately via CYP4503A4, so that moderate
inhibitors of this enzyme could lead to greater QTc
prolongation. The QTc effects of doses greater than 100 mg
have not been studied.
- See Section 7.1.1
(Deaths) at page __ for discussion of deaths that may be
related to QT prolongation and detailed discussion of the
finding
- See Section 7.1.9 (ECGs)
at page __ for discussion of ECG changes
- See Section 7.1.10
(Special Studies) at page __ for dose response study of QT
prolongation (doses of 10, 40 and 100 mg)
- See Section 7.2.4
(Metabolic and Interaction Workup) at page __ for
discussion of the adequacy of the applicant's in vivo and
in vitro assessments of the metabolism of (Drug) and
potential relation of drug-drug interactions to QT
prolongation
- See Section 3.2 (Animal
Pharmacology/Toxicology) at page __ for discussion of the
animal models used to evaluate effects on K channels, and
QT prolongation
- See Section 7.2.1.1.2.3
(Literature) at page __ for published articles about
similar products and methodological suggestions.
- See Section 7.1.13
(Overdose) at page __.
- Patient ID numbers for
possibly relevant deaths: ______, ______, ______.
|
As the QT prolongation example
shows, it is useful to identify the various sections of the clinical
review that can be referenced for additional details about an
identified adverse event. If the review is converted into a PDF
file, bookmarking can be used to electronically link the text in the
problem list to earlier sections of the review.
In addition, in this section the
reviewer should provide summary recommendations for further studies,
with a reference to section 7.2 for more details.
The review should also provide
overall conclusions about the safety of the drug, including:
- Overall assessment of the
available safety information, referring both to what it has shown
and its adequacy
- The limitations of the available
data
- Additional information needed,
including both further analyses and additional studies.
- Comparison, to the extent
possible, of the safety of the drug under review to the safety of
other available products, and the basis for that comparison
(direct comparative data vs. clinical opinion)
- Whether a risk management program
(beyond labeling) is needed and why
- Analysis of likely uses beyond
labeling, (e.g., in more severe patients, in other diseases, in
children)
- Whether there is a need for
postmarketing safety studies
This section of the guidance
describes analytical methods that have general application to the
safety review and provides a location in the review for any general
discussion of methodological issues not discussed elsewhere,
organized by the subsections listed here, with additional sections
as needed. It is important to consider early in the review whether
the available patient level data will allow the analyses the
reviewer intends. For example, in examining whether particular
baseline risk factors are related to an adverse event, the reviewer
will either need to extract the baseline characteristics from case
report tabulations or be sure the information is available in a
retrievable form. Similarly, it may be important to link individual
safety observations with other on therapy data, such as dose,
duration of treatment, concomitant therapy, other adverse events,
lab data or effectiveness results (it is obviously best if such
issues are considered at pre-NDA/BLA meetings).
Before estimating the incidence of
adverse events, the reviewer must select the patient sample of
interest. Pooling data from different studies can improve the
precision of an incidence estimate (i.e., narrow the confidence
intervals by enlarging the sample size). Better precision is
particularly important for lower frequency events, which can be
difficult to detect and may not occur in some studies. Pooling can
also provide the larger database that will permit explorations of
possible drug-demographic or drug-disease interactions in subgroups
of the population. Pooling can also, however, obscure real
potentially meaningful differences between studies. The review
should explain why any pooling used in the review was chosen. When
making decisions about pooling, the reviewer should consider the
following:
- It is most appropriate to combine
data from studies that are of similar design, that is, similar in
dose, duration, choice of control, methods of ascertainment, and
population (checklist vs. general inquiries vs. no prompt at all;
in psychiatric drug trials it is typical for obsessive compulsive
patients to spontaneously report adverse events more frequently
than schizophrenic patients. It is also possible that different
populations may have different vulnerabilities to a drug, and
therefore, different risk profiles.) When the studies are similar
in design but differ in duration, it may be critical to account
for exposure duration and to look for time-dependent events.
- Even when the pooled analysis is
the primary one, it is important to explore the range of
incidences across the studies being pooled. For a specific adverse
event, if the incidence differs substantially across the
individual studies in a pool, the pooled value should not be used,
as it is probably not meaningful and, in some cases, could obscure
important information about predictors for that event. (In one
case, for example, several studies were combined and a
reassuringly low estimate of phototoxicity was obtained.
Subsequent examination of individual study results found one study
with a substantial rate of phototoxicity. The study was the only
outpatient study done (i.e., the only one in which patients had an
opportunity to be exposed to sunlight).) In some situations, the
incidence may be best described by the range in the various
studies. For the photoxicity example above, however, the most
relevant data are those from the outpatient study, the only study
that was conducted under conditions pertinent to intended use.
- In some cases, observed
differences in rates in various studies can be explained (e.g.,
better ascertainment, different populations), so that a consistent
rate can be determined from a subset of studies.
- Formal tests for extreme values
may be useful to assess appropriateness of assay pooled data
(e.g., test of heterogeneity such as the Breslow-Day Chi-Square
test could be used). Alternatively, the reviewer might use a more
subjective approach, such as determining if the direction of the
difference is always the same across studies, or use a graphic
display of incidence by study to informally consider the extent of
variability and to identify outliers; outliers may be important in
identifying subgroups of patients who are at particular risk for
certain adverse reactions.
In pooling data, usually the
numerator events and denominators for the selected studies are
simply combined. Other more formal weighting methods can be used
(e.g., weighting studies on the basis of study size or inversely to
their variance). The review should describe how the pooling was
performed, as well as the rationale for selection of the method
used.
Adverse reaction rates may differ
considerably from one patient population to another and may change
over time. Factors that may affect the safety profile of a drug
should be explored during the review. Explorations for common
predictive factors, such as dose, plasma level, duration of
treatment and concomitant medications, and patient-predictive
factors such as age, sex, race, concomitant illnesses, are
considered below. In general, these explorations are meaningful only
for adverse events that appear to be drug-related (see Section
7.4.3).
If data from randomized, parallel,
fixed-dose studies (or data from studies in which patients were
randomized to fixed dose ranges), are available, they should be
analyzed for evidence of dose dependency for any adverse reactions.
If plasma concentration data are available, it may be useful to
explore plasma concentration effect relationships as well. It may
also be useful to reconfigure dose as mg/kg or mg/m2, to
decrease the effect of size or weight differences on drug exposure.
Dose-response relationships should also be examined in demographic
subgroups (e.g., females, blacks, elderly patients). Dose-dependency
analyses are usually performed by simple inspection of incidence
rates across different doses or different weight or body surface
area-adjusted doses. Formal statistical testing can also be used. If
formal statistical tests are performed for a study that includes
placebo control as well as different doses, and a drug-placebo
difference is apparent, it may be desirable to focus on between-dose
group differences.
Flexible Dose Titration Studies
Although it is tempting to try to
extract dose-response or plasma level-response data from flexible
dose (titration) studies, and the ICH dose-response guideline24
encourages this, there are many potential problems with such
analyses. In particular, many adverse reactions show considerable
time dependency, some occurring early, some late. It is easy to
confound dose (or plasma concentration) with duration when dose is
increased over time. In some cases, such as anticancer drugs or
drugs that are known to produce anti-cholinergic or sedating
reactions, the drug is dose-adjusted to toxicity, which will often
obscure any dose-response relationship. In addition, if dose is
increased only in patients without adverse effects (i.e., subjects
who are resistant to them), the higher doses will be associated with
lower adverse effect rates. On the other hand, if dose is titrated
to clinical effect, and adverse reactions occur late (so that they
do not affect the dose given), analysis of the rate with respect to
dose may be useful. For example, erythropoietin, used to treat
anemia in patients with chronic renal failure or cancer, is titrated
to maintain hemoglobin within a specific range. Given the delayed
therapeutic response (erythropoesis), analysis of adverse events by
dose or cumulative dose prior to a reaction can give insight into
dose-related toxicity.
Cumulative Dose Dependency
For certain adverse reactions, it
may be possible to demonstrate a relationship between cumulative
dose and the occurrence of the reaction (e.g., liver fibrosis and
cirrhosis with methotrexate, cardiotoxicity with doxorubicin, renal
toxicity with Amphotericin B). For drugs that are used chronically,
the reviewer should consider the possibility that cumulative dose
may predict toxicity and discuss this in the review.
The reviewer should explore time
dependency of adverse reactions in two ways - time to onset of the
finding and duration of the finding:
Time of Onset
Although most adverse reactions
occur early in treatment and may be best characterized by a crude
incidence rate (number with the reaction divided by number exposed),
others may occur only after some delay of weeks, months, or longer.
A crude incidence rate, based on a patient population exposed
predominantly for short periods, will understate the importance of
such adverse reactions for chronically used drugs. For important
adverse reactions that occur later in treatment, there should be
explorations of the time dependency of the reaction. Possible
methods include:
- A life table (Kaplan-Meier graph)
describing risk as a function of duration of exposure (i.e.,
cumulative incidence)
- Plotting risk for discrete time
intervals over the observation period (i.e., a hazard rate curve)
reveals how risk changes over time.
- Adjusting for duration of
exposure by expressing the adverse reaction rate in terms of
person-time (person-time is duration of exposure summed across all
patients, e.g., 2 patients each exposed for 6 months = 1
patient-exposure-year). This approach is useful only when one can
safely assume that the hazard rate is constant over time.
Duration of Adverse Event
Certain adverse events that occur at
initiation of treatment may appear to diminish in
frequency with continued use. Possible explanations for this
phenomenon include adaptation or tolerance, decreased reporting of
the event even by patients though it is still occurring at the same
rate, and reduced dose or dropping out in patients with the event.
For drugs used chronically and for which there was an adverse event
that seemed to diminish in frequency over time, it may be useful to
characterize and quantify the change. It would be important, for
adverse events of interest, to determine whether the decreased rate
simply reflected discontinuation by affected patients or real
adaptation. One way to make this distinction is to identify a cohort
that experienced an event of interest during a specified period of a
trial, but nonetheless completed the trial, and observe the rate of
the event in that cohort over time. This cohort of survivors could
be compared to a similar cohort of placebo recipients who
experienced the same event at baseline. The same approach could be
used for adverse events occurring later in treatment. It is usually
sufficient to do such analyses for those adverse reactions that are
relatively common and likely to be drug related (see Section 7.1.5.4
for methods to identify drug-related events).
Numerous methods can be used to
analyze age, gender, and race implications for safety, and
applicants must present analyses of safety information for these
population subsets. In most cases, there will be pharmacokinetic
information available for some or all of these subsets, which may
help in interpreting adverse event rates. In some cases, it may be
useful to construct subgroups based on more than one factor. For
example, bleeding is the principal risk associated with use of
thrombolytic agents in patients with acute myocardial infarction.
Women tend to have more bleeding than men, and risk is inversely
related to weight. Thus, an analysis by gender weight subgroups can
identify the group at greatest risk of bleeding (thin women). It may
also be useful to consider age-gender or race-gender subgroups.
Formal analysis should be limited to events considered common (e.g.,
occurring at an incidence of at least 2 percent) and that occur at a
clearly greater rate on drug than placebo. In small studies or for
low frequency events, there will usually not be sufficient power to
detect differences between groups, so that these analyses will
usually be based on pooled data. In general, these analyses are
descriptive, comparing risk of an event in one subset with the risk
in another (men vs. women, old vs. young, black vs. white); as these
comparisons obviously do not reflect randomization to the subset
(baseline characteristic) of interest, formal statistical
comparisons are usually not warranted. For these descriptive
comparisons, two approaches deserve consideration; when the control
rates of adverse events differ for population subsets these
approaches can provide quite different results: (1) evaluation of
relative risk (RR) (cumulative risk on drug/cumulative risk on
comparison drug or placebo) and (2) evaluation of attributable risk
(AR) (cumulative risk on drug - cumulative risk on comparison drug
or placebo).
When background event rates differ
by demographic subgroup, relative risk analysis will provide a
quantitative estimate of the difference in effect of the drug, but
the attributable risk may be a better estimate of the importance of
the risk in the subsets. To illustrate, consider a comparison of
drug-induced nausea for males versus females. Suppose the rate of
nausea on placebo is 1 percent for men and 3 percent for women and
that on drug it is 3 percent for men and 9 percent for women. The
risk ratios (RR) for both sexes are 3 and the relative risk for men
and women (RRf/RRm) is one (no difference), yet the attributable
risk is much greater for women than men (6 percent vs. 2 percent), a
finding of possible importance in treatment. Such a difference has
been observed for several adverse reactions of amlodipine, a calcium
channel blocker, and is described in labeling as a gender
difference, even though the RR's are the same.
The reviewer should be alert to the
possibility that co-morbidity will affect the adverse reaction
profile of the drug (i.e., a drug-disease interaction). Such
interactions can arise from abnormalities of excretory function
(renal or hepatic disease), and typically, the applicant will have
carried out formal pharmacokinetic studies in patients with hepatic
and renal disease to indicate the potential for such reactions. The
reviewer needs to consider, in that case, whether PK differences are
manifested as differences in adverse reaction rates. Apart from
differences in adverse reaction rates related to PK differences,
differences in rates can also reflect true differences in
susceptibility to adverse reactions (i.e., real pharmacodynamic
differences). In general, the same methods described for exploring
drug-demographic interactions can be applied here.
The clinical reviewer should be
alert to the potential of drug-drug interactions to affect the
safety profile of the drug. Again, these interactions could be
either pharmacokinetic (affecting elimination of the drug) or
pharmacodynamic, in either case leading to observed differences in
adverse reaction rates for the subgroups receiving or not receiving
co-administered drugs. Typically, there will be formal interaction
studies to evaluate potential pharmacokinetic effects of concomitant
therapy on drugs metabolized by CYP450 enzymes, but PK interactions
can also occur through effects on renal excretion and transport
(P-glycoprotein) proteins. True pharmacodynamic interactions are
less frequently recognized but can be important (e.g., marked
hypotension when sildenafil is given with organic nitrates). In
general, the same methods described for exploring drug demographic
interactions can be applied here.
In assessing the critical question
of whether an adverse event is caused by a drug, whether the drug is
capable of causing that adverse event in the population is usually
of greater interest than whether the drug caused the event in each
patient who reported the event, but the approach to causality is
distinctly different for relatively common events and relatively
rare, serious events.
Common Events
Where events are common and occur in
multiple patients in controlled trials, it is usually not necessary
or helpful to consider each case individually. Rather, all reported
cases can be considered potentially drug-related, and causality is
assessed by comparing the rates of reports in patients treated with
test drug and in control groups. If an event is clearly more
frequent with test drug than the control, it can be attributed to
treatment with the test drug.
Uncommon, Serious Events
Causality judgments are much more
difficult for uncommon (e.g., < 1/1000) serious events where there
are, in most cases, no useful comparisons to control groups. The
reviewer therefore must form a judgment as to the plausibility of
drug-relatedness for the individual cases.
- The following questions should be
considered:
1. Was the patient in fact
exposed to drug and did the adverse event occur after drug
exposure?
2. Did the patient have a
clinical experience that meets the criteria for the adverse
event of interest? (Establishing a standard case definition
may be helpful here.)
3. Is there a reasonably
compelling alternative explanation for the event? (For
example, recent benzene exposure for a case of aplastic
anemia; the event is a well-recognized consequence of the
patient's underlying illness.)
4. Is the adverse event of a
type commonly associated with drug exposure, such as
hematologic, hepatic, renal, dermatologic or pro-arrhythmic
events? (But also see below caution about discarding events
that do not seem plausibly drug related.)
- After assessing individual cases
to identify events that could be drug-related and for which there
are no compelling alternative explanations, the reviewer should
compare the observed rate of occurrence of the event in the
database with a best estimate about the background rate for the
event for the population being studied. For an event like aplastic
anemia, with a background rate of perhaps 1 per million person
years, finding even one case suggests a causal relationship. For
events that occur more frequently in the absence of drug therapy
(e.g., MI, stroke, sudden death, seizure, which could occur at
rates of 0.1 to 1 percent, depending on the population), the
finding of one or two cases may be very difficult to interpret in
the absence of a substantial controlled trial database.
- The reviewer should also evaluate
any other information about the drug that bears on causality
including:
1. Whether the drug is a
member of a class of drugs known to be causally associated
with the event of interest
2. Presence of other adverse
events in the database that may be associated with the event
of interest (e.g., a general finding of drug associated
transaminitis or animal findings suggestive of hepatotoxicity
would substantially strengthen the signal generated by the
finding of a single case of hepatic failure)
3. Positive re-challenge with
the drug (although it would be unusual to deliberately
re-challenge for a serious event, there may occasionally be
inadvertent re-exposures that are informative)
- Caution Concerning Relative
Plausibility of Uncommon, Serious Events
The reviewer should be cautious
about dismissing uncommon, serious events that
don't seem
plausibly drug-related and should consider differences in
common less serious adverse reactions that might predict the
uncommon serious reactions with longer use. There are numerous
examples of uncommon, serious adverse reactions that are uniquely
associated with a drug or drug class:
- tendon rupture associated
with the quinolone antibiotics
- heart valve lesions
associated with fenfluramine
- practolol syndrome
- retroperitoneal fibrosis
with Sansert
- pulmonary
hypertension with Aminorex (a European weight loss drug),
and various other drugs
- suicidal ideation with
interferons, Accutane
- intussusception with
rotovirus vaccine
- pulmonary fibrosis with
amiodarone
List of
Tables
TABLE 7.0.1 INDEX FOR
LINKING IDENTIFIED PATIENTS WITH SUPPLEMENTARY PATIENT
INFORMATION IN THE NDA (CRFS, NARRATIVE SUMMARIES, AND
PATIENT DATA LISINGS)
TABLE 7.1.1.1 DEATHS
LISTING
TABLE 7.1.1.2 MORTALITY BY
TREATMENT GROUP FOR POOL OF PHASE 2-3 STUDIES WITH NEW
DRUG
TABLE 7.1.2.1 SERIOUS
ADVERSE EVENT LISTING
TABLE 7.1.3.1.1 DROPOUT
PROFILE: INCIDENCE OF DROPOUT BY TREATMENT GROUP AND
REASON FOR PHASE 2 TO 3 STUDIES WITH NEW DRUG
TABLE 7.1.3.2.1 ADVERSE
EVENT DROPOUT LISTING
TABLE 7.1.5.2.1 TREATMENT
EMERGENT ADVERSE EVENT LISTING
TABLE 7.1.5.3.1 TREATMENT
EMERGENT ADVERSE EVENT INCIDENCE FOR POOL OF 6-WEEK
PLACEBO-CONTROLLED TRIALS
TABLE 7.1.7.3.1.1 MEAN
CHANGE FROM BASELINE FOR SERUM CHEMISTRY PARAMETERS IN
POOL OF PLACEBO-CONTROLLED STUDIES
TABLE 7.1.7.3.2.1
INCIDENCE OF POTENTIALLY CLINICALLY SIGNIFICANT CHANGES
IN SERUM CHEMISTRY PARAMETERS FOR POOL OF PLACEBO
CONTROLLED STUDIES FOR NEW DRUG
TABLE 7.1.7.5.1 HEPATOTOXICITY
EVALUATION
TABLE 7.2.1.1.1
ENUMERATION OF SUBJECTS/PATIENTS FOR NEW DRUG
DEVELOPMENT PROGRAM
TABLE 7.2.1.2.1
DEMOGRAPHIC PROFILE FOR PHASE 2-3 STUDIES WITH NEW DRUG
TABLE 7.2.1.3.1 NUMBER
(PERCENT) OF PATIENTS RECEIVING NEW DRUG ACCORDING TO
MEAN DAILY DOSE AND DURATION OF THERAPY IN PHASE 2-3
STUDIES (N=2500)
|
|
|
|
|
|
|
|
Table 7.0.1
Index for
Linking Identified Patients with Supplementary Patient
Information in the NDA (CRFs, Narrative Summaries, and Patient
Data Listings1 |
Study Number2 |
Patient Number3 |
Case Report
Forms |
Narrative
Summaries |
Patient Data
Listings |
|
|
Volume4 |
Pages5 |
Volume |
Pages |
Volume |
Pages |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1Separate indices
should be provided for patients exposed to new drug, active
control drugs, and placebo.
2Study numbers should
be numerically ordered and tabbed as separate sections within the
index.
3Patient numbers should
be numerically ordered within each study section.
4The volume number
provided in this index should be the unique volume number assigned
to the volume as part of the complete NDA, and not a separate
volume number assigned to the volume as part of a section of the
NDA.
5The page numbers
provided in this index should be the unique page numbers assigned
for the entire volume, and not separate page numbers assigned to
the separate sections that might be included in any particular
volume.
|
|
|
|
|
|
|
|
|
|
Table 7.1.1.1
Deaths Listing1,2,3
Treatment =
New Drug4
Cutoff Date5 |
Trial |
Center |
Patient |
Age
(yrs) |
Sex |
Dose6
(mg) |
Time7
(Days) |
Source8 |
Person
Time9 |
Description10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1A footnote should
describe the rule for including deaths in the table (e.g., all
deaths that occurred during a period of drug exposure or
within a period of up to 30 days following discontinuation
from drug and also those occurring later but resulting from
adverse events that had an onset during drug exposure or
during the 30-day follow up period). Other rules may be
equally appropriate.
2Deaths occurring
outside the time window for this table should be listed
elsewhere.
3This table should
be provided by the sponsor in electronic format. The exact
design of the table and the preferred electronic format should
be established in discussions between the sponsor and the
reviewing division.
4Similar lists
should be provided for patients exposed to placebo and active
control drugs.
5This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available
for entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
6Dose at time of
death, or if death occurred after discontinuation, note that,
as well as last dose before discontinuation.
7Days on drug at
time of death; or if death occurred after discontinuation,
note how many days on drug before discontinuation and also how
many days off drug at time of death.
8This listing
should include all deaths meeting the inclusion rule, whether
arising from a clinical trial or from any secondary source
(e.g., postmarketing experience). The source should be
identified in this column (i.e., 10 for deaths
arising from primary source clinical trials and 20
for those arising from secondary sources).
9This column should
identify patients (yes/no) for whom person-time data are
available, so the reviewer can know which patients were
included in the mortality rate calculations.
10Since narrative
summaries should be available for all deaths, the description
can be very brief (e.g., myocardial infarction, stroke,
pancreatic cancer, suicide by drowning).
|
|
|
|
|
|
|
Table 7.1.1.2
Mortality by
Treatment Group
for Pool of Phase
2-3 Studies with New Drug1,2,3
Cutoff Date4 |
Treatment
Group5 |
Total Number
of Patients6 |
Total Number
of Deaths7 |
Crude
Mortality8 |
Patient Exposure
Years (PEY)9 |
Total Deaths with
Person-Time10 |
Mortality per
100 PEY11 |
New
Drug |
|
|
|
|
|
|
Active
Control |
|
|
|
|
|
|
Placebo |
|
|
|
|
|
|
1This table
provides data comparing overall mortality across treatment
groups for the pool of all phase 2 to 3 studies in the
development program. Similar tables may be appropriate for
other subgroups within the phase 2 to 3 program (e.g., a table
may be provided for a pool of all similarly designed
short-term placebo controlled trials). Similar tables may be
appropriate for certain individual trials of interest. All
deaths should be counted, regardless of the investigator's or
the sponsor's judgment about causality, including (1) any
deaths occurring during participation in any of the studies in
the target pool, (2) any deaths occurring after a patient
leaves any of the targeted studies, whether prematurely or
after completion to the nominal endpoint, if the death is (a)
the result of a process initiated during the study, regardless
of when it actually occurs, or (b) occurs within 4 weeks of a
patient leaving a study, or longer for drugs with particularly
long elimination half-lives or from drug classes with known
late occurring effects. The actual rule used for including
deaths should be provided in a footnote to the table. In case
there are substantial deaths of specific causes, it may be
appropriate to provide data for cause specific mortality as
well.
2Patients
participating in crossover trials should be enumerated for
each of the pertinent columns of the table (e.g., a patient
receiving treatment in each of the three arms of a 3-way
crossover study comparing new drug, active control, and
placebo would be included in all three columns.
3This table should
be provided by the sponsor in electronic format. The exact
design of the table and the preferred electronic format should
be established in discussions between the sponsor and the
reviewing division.
4This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available
for entry. Generally, this date should, be no more than
several months prior to the submission date for an NDA. This
date as well as this table may likely need to be updated
during the course of NDA review as more data become available.
5In the sample
table, only 1 row is provided for an active control
group. One such category may suffice for certain NDAs, but may
not for others, and the decision regarding how to categorize
active control patients should be made in consultation with
the reviewing division. Similarly, for this table, only 1 row
is provided for new drug, with the implication that all new
drug patients, regardless of dose, should be included in the
calculations for that column. Other approaches (e.g.,
distinguishing patients on the basis of dose) may be equally
appropriate.
6The Ns in these
rows should match the N's in Table 5.1.1.1., and if not, an
explanation should be provided in a footnote.
7This is the total
number of deaths for each group.
8This is simply the
total number of deaths divided by the total number of patients
exposed in each group.
9This column should
provide person-time in patient exposure years (PEY). This
table assumes a constant hazard rate; however, in certain
situations, it may be appropriate to stratify by increments of
exposure.
10This is the
subset of total deaths for which person-time is available.
11This is the
number of deaths for whom person-time is available divided by
PEY for each group, and multiplied by 100.
|
|
|
|
|
|
|
|
|
|
|
Table 7.1.2.1
Serious Adverse Event Listing1,2,3
New Drug Clinical Trials
Source: Phase 2-3 Trials4
Sorting A: Randomized Treatment, Trial #, Investigator/Center
#, Patient #5
Treatment = New Drug6
Cutoff Date7 |
Trial |
Center |
Patient |
Age
(yrs) |
Sex |
Dose8
(mg) |
Time9
(days) |
Body System |
Prefrrd Term |
Adverse Event10 |
W/D11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 7.1.3.1.1
Dropout Profile: Incidence of Dropout by Treatment Group and
Reason
for
Phase 2 to 3 Studies with New Drug1,2,3
Cutoff Date4:
|
Reasons for
Dropout5 |
Treatment
Groups6 |
|
New Drug
N =
|
Placebo
N =
|
Active Control
N =
|
Lack of Efficacy |
%7 |
% |
% |
Adverse Event |
% |
% |
% |
Lost to Follow up |
% |
% |
% |
Other |
% |
% |
% |
Total Dropouts |
% |
% |
% |
1This sample table
should be based on a pool of all trials in the phase 2 to 3
development program. Similar tables may be appropriate for other
subgroups within the phase 2 to 3 program (e.g., a table should
be provided for a pool of all similarly designed short-term
placebo controlled trials). Similar tables may be appropriate
for certain individual trials of interest.
2Patients
participating in crossover trials should be enumerated for each
of the pertinent columns of the table (e.g., a patient receiving
treatment in each of the three arms of a 3-way crossover study
comparing new drug, active control, and placebo would be
included in all three columns).
3This table should be
provided by the sponsor in electronic format. The exact design
of the table and the preferred electronic format should be
established in discussions between the sponsor and the reviewing
division.
4This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
5This sample table
includes 4 categories for dropout, but a more detailed breakdown
may be of interest as well.
- The adverse event category
here would include all patients identified as dropping out
for adverse events, regardless of whether or not the events
were judged by the investigator or sponsor to be drug
related and regardless of what other reasons may have been
identified in association with dropout. Patients identified
as dropping out for intercurrent illness would ordinarily be
included under this adverse event category. Similarly, a
patient identified as dropping out for an adverse event and
lack of efficacy would also ordinarily be included under
this adverse event category.
- Lost-to-follow up is an
important outcome to track, since it reflects on the overall
conduct of the studies.
- The other category
is intended to include all other reasons that may generally
be considered nontreatment related. This category is often
identified as administrative, and includes such
reasons as patient refused further participation, patient
moved away, patient improved, patient not eligible, protocol
violation, unknown.
- Decisions about what
categories to include should be made in consultation with
the reviewing division.
6In the sample table,
only 1 column is provided for an active control group.
One such category may suffice for certain NDAs, but may not for
others, and the decision regarding how to categorize active
control patients should be made in consultation with the
reviewing division. Similarly, for this table, only 1 column is
provided for new drug, with the implication that all new drug
patients, regardless of dose, should be included in the
calculations for that column. Other approaches (e.g.,
distinguishing patients on the basis of dose) may be equally
appropriate. The N's in these column headings should match the
N's in Table 5.1.1.1., and if not, an explanation should be
provided in a footnote.
7Numbers for this
table should be rounded to the nearest integer.
|
|
|
|
|
|
|
|
|
|
|
|
Table 7.1.3.2.1
Adverse Event Dropout Listing1,2,3
New
Drug Clinical Trials
Source: Phase 2-3 Database4
Sorting A: Randomized Treatment, Trial #, Investigator/Center #,
Patient #5
Treatment = New Drug6
Cutoff Date7: |
Trial |
Center |
Patient |
Age
(yrs) |
Sex |
Dose8
(mg) |
Time9
(days) |
Body
System |
Prefrrd Term |
Adverse Event10 |
Serious11 |
Outcome12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1This is a line
listing of all reported adverse events identified as leading to
discontinuation, regardless of whether or not they were
considered drug related, for all patients participating in
trials identified as sources for this listing. Thus, all events
categorized as intercurrent illness leading to
discontinuation would, nevertheless, be included in this
listing, and any judgments about attribution can be included in
the narrative summary. This listing is a critical component of
the integrated safety summary.
2The variables
included in this listing include:
- Trial #
- Center #
- Patient # (a unique number
that identifies this patient in the NDA database)
- Age
- Sex
- Dose (in mg) at time of
event onset
- Time (i.e., duration, of
exposure (in days) at time of event onset)
- Body system category for
event (using COSTART or other thesaurus)
- Preferred term for event
- Adverse event as reported
by investigator and/or patient
- An indication of whether or
not the event met definition for serious
- Outcome
- Race
- Weight
- Height
- Dose expressed as mg/kg,
mg/mm2, or even plasma concentration, if
available
- Other drug treatment
- Severity of adverse event
(mild, moderate, severe)
- Action taken (e.g., none;
decrease dose, discontinue treatment)
- Causality assessment by
investigator (related, not related)
- Location in NDA of CRF,
patient narrative summary)
3The exact design of
the table and whether or not it needs to be provided in
electronic format should be established in discussions between
the sponsor and the reviewing division.
4Similar listings may
be provided for individual studies as part of full reports for
such studies and, possibly, for other pools that are subsets of
this larger pool.
5It is essential to
provide this listing in two different forms (i.e., sorting A (by
patient) and sorting B (by adverse event)). This listing is for
sorting A, by patient, and permits the reviewer to explore all
the adverse events reported as leading to discontinuation for
each individual patient. Sorting B (by adverse event) should be
as follows: Randomized Treatment, Body System, Preferred Term,
Adverse Event, Trial, Center, Patient #, Age, Sex, Dose, Time,
Serious. Sorting B permits the reviewer to explore all the
adverse events of a similar type reported as leading to
discontinuation.
6This sample listing
is for all new drug patients across all studies in the phase 2
to 3 development program. Similar listings should be provided
for active control and placebo patients.
7This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
8This column should
include the dose being administered (in mg/day) at the time the
event occurred.
9This column should
include the time (i.e., duration of exposure (in days)), at the
time the event occurred.
10This column should
include the adverse event in the language reported by the
investigator and/or patient, i.e., before coding.
11This column should
include an indication of whether or not the adverse event met
the criteria for serious as defined for the development
program overall.
12This column should
categorize the outcome upon follow up evaluation for the adverse
event leading to discontinuation, as follows:
(R) Resolved
(P) Persisting
(U) Unknown
|
|
|
|
|
|
|
|
|
|
|
|
Table 7.1.5.2.1
Treatment Emergent Adverse Event Listing1,2,3
New
Drug Clinical Trials
Source: Phase 2-3 Database4
Sorting A: Randomized Treatment, Trial #, Investigator/Center #,
Patient #5
Treatment = New Drug6
Cutoff Date7: |
Trial |
Center |
Patient |
Age
(yrs) |
Sex |
Dose8
(mg) |
Time9
(days) |
Body
System |
Preferred Term |
Adverse Event10 |
Serious11 |
W/D12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1This is a line
listing of all reported treatment emergent adverse events,
regardless of whether or not they were considered drug related,
for all patients participating in trials identified as sources
for this listing. A footnote should identify all studies
contributing to this pool.
2The variables
included in this listing include:
- Trial #
- Center #
- Patient # (a unique number
that identifies this patient in the NDA database)
- Age
- Sex
- Dose (in mg) at time of
event onset
- Time (i.e., duration, of
exposure (in days) at time of event onset)
- Body system category for
event (using COSTART or other thesaurus)
- Preferred term for event
- Adverse event as reported
by investigator and/or patient
- An indication of whether or
not the event met definition for serious
- An indication of whether or
not the event led to withdrawal
- Race
- Weight
- Height
- Dose expressed as mg/kg,
mg/mm2, or even plasma concentration, if
available
- Other drug treatment
- Duration of adverse event
- Timing of adverse event
relative to last dose
- Severity of adverse event
(mild, moderate, severe)
- Action taken (none,
decrease dose, discontinue treatment)
- Outcome
- Causality assessment by
investigator (definitely, probably, possibly, or unlikely
related)
- Location in NDA of CRF,
patient narrative summary
3The exact design of
the table and whether or not it needs to be provided in
electronic format should be established in discussions between
the sponsor and the reviewing division.
4Similar listings may
be provided for individual studies as part of full reports for
such studies, and possibly for other pools that are subsets of
this larger pool.
5It is essential to
provide this listing in two different forms (i.e., sorting A (by
patient) and sorting B (by adverse event)). This listing is for
sorting A, by patient, and permits the reviewer to explore all
the adverse events reported for each individual patient. Sorting
B (by adverse event (i.e., 1 row for each occurrence of each
adverse event)) should be as follows: Randomized Treatment, Body
System, Preferred Term, Adverse Event, Trial, Center, Patient #,
Age, Sex, Dose, Time, Serious, W/D. Sorting B permits the
reviewer to explore all the reported adverse events of a similar
type.
6This sample listing
is for new drug patients (i.e., for all patients exposed to New
Drug in the phase 2 to 3 studies that are part of the Integrated
Primary Database). Similar listings should be provided for
active control and placebo patients.
7This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
8This column should
include the dose being administered (in mg/day) at the time the
event occurred.
9This column should
include the time (i.e., duration of exposure (in days)), at the
time the event occurred.
10This column should
include the adverse event in the language reported by the
investigator and/or patient (i.e., before coding).
11This column should
include an indication of whether or not the adverse event met
the criteria for serious as defined for the development
program overall. 12This column should include an
indication of whether or not the adverse event led to
discontinuation of the assigned treatment.
|
|
|
|
Table
7.1.5.3.1
Treatment-Emergent Adverse Event Incidence
for Pool of
6-Week Placebo-Controlled Trials1-10
Cutoff Date11: |
Body System/
Adverse Event12-14 |
Percentage of
Patients Reporting Event15 |
|
New Drug
N16= |
Active Control
N= |
Placebo
N= |
Body as a Whole |
|
|
|
Headache |
|
|
|
Etc. |
|
|
|
Cardiovascular System |
|
|
|
Postural Hypotension |
|
|
|
Etc. |
|
|
|
Gastrointestinal System |
|
|
|
Constipation |
|
|
|
Etc. |
|
|
|
. |
|
|
|
. |
|
|
|
Urogenital System
Impotence17 |
|
|
|
Etc. |
|
|
|
. |
|
|
|
. |
|
|
|
1This table compares
the incidence of treatment emergent adverse events across
treatment groups for a pool of similarly designed
placebo-controlled trials of new drug. Generally, an arbitrary
threshold incidence for new drug patients is used as a criterion
for selecting adverse events to include; > 1% for new
drug is a commonly used rule, but others may be equally
appropriate. The criterion used should be noted in the table
title or in a footnote.
2Study pools other
than those described for this sample table may be equally
appropriate, and similar tables useful for individual trials may
also be of interest.
3In the sample table,
only 1 column is provided for an active control group.
One such category may suffice for certain NDAs, but may not for
others, and the decision regarding how to categorize active
control patients should be made in consultation with the
reviewing division.
4Similarly, for this
table, only 1 column is provided for new drug, with the
implication that all new drug patients, regardless of dose,
should be included in the calculations for that column. Other
approaches (e.g., dividing patients on the basis of dose), may
be equally appropriate. If the studies used were fixed-dose
studies, it is generally most informative to preserve the dose
categories in constructing this table. However, dose categories
that are not relevant to the doses that are being recommended
for use may reasonably be omitted from this table. It is
generally not useful to try to artificially construct dose
categories from dose titration studies, since there is often
confounding of dose and time.
5Data are often
available on the investigator's opinion regarding whether or not
any particular adverse event was in fact related to the drug
being taken. Some reviewers consider this useful information and
may construct tables that include only those events considered
possibly, probably, or definitely drug-related by the
investigator. Others ignore such judgments and include all
reported adverse events, with the view that the control groups,
especially placebo if present, should permit one to make
causality decisions, regardless of the investigators' judgments
about drug-relatedness. Either approach can be acceptable, but
it is critical that a footnote indicate clearly when adverse
events are not included due to investigators' judgments that
they were not drug-related, since this approach may reduce the
adverse event rates that appear in the table.
6Data are also often
available on the intensity of the reported adverse events,
generally including categories of mild, moderate, or severe.
Adverse event tables may ignore such classifications and pool
all events together, or some attempt may be made to focus only
on a subset of reported events (e.g., only those classified as
severe). Again, either approach is acceptable, but it is
important to describe in a footnote what approach was taken.
7Not uncommonly, a
new drug is developed for more than one indication. If adverse
event rates appear to be to occur at similar rates across the
indications, it may be reasonable to pool the data in creating
an adverse events table, possibly one providing greater
precision. However, it is not inconceivable that adverse event
rates may vary depending on the population studied, and if this
appears to be the case, pooling may not be appropriate.
8Adverse events that
occur at a rate for placebo that is > the rate for new
drug should be removed from the table and noted only as a
footnote.
9Patients
participating in crossover trials should be included in the
calculations for each of the pertinent columns of the table
(e.g., a patient receiving treatment in each of the three arms
of a 3-way crossover study comparing new drug, active control,
and placebo would be included in the calculations for all three
columns).
10This table should
be provided by the sponsor in electronic format. The exact
design of the table and the preferred electronic format should
be established in discussions between the sponsor and the
reviewing division.
11This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
12Adverse events
should be organized under body system categories.
13Within each body
system category, adverse events should be ordered according to
decreasing frequency.
14Adverse events
during exposure are generally obtained by spontaneous report and
recorded by clinical investigators using terminology of their
own choosing. Consequently, it is not possible to provide a
meaningful estimate of the proportion of individuals
experiencing adverse events without first grouping similar types
of events into a smaller number of standardized event
categories. Generally a table of this type should use these
preferred adverse event terms, and a footnote should identify
the system used for coding investigator terms. Adverse event
terms that convey no useful information (e.g., joint disorder),
should be replaced by more clinically useful terms or deleted.
15Percentages should
be rounded to the nearest integer. Although not strictly
hypothesis testing, p-values give some feeling for the strength
of the finding and should be produced for all new drug/placebo
pairwise comparisons and any p-values meeting a p < 0.05 level
of significance should be noted by an asterisk (*) as a
superscript to the %.
16The N for each
column should be provided at the column heading, so that only
the percentage of patients having that adverse event need be
included in the table, and not the actual number.
17The rates for
gender specific adverse events (e.g., impotence) should be
determined using the appropriate gender specific denominator,
and this fact should be indicated with a footnote.
|
|
|
|
|
|
|
|
|
|
Table
7.1.7.3.1.1
Mean Change
from Baseline for Serum Chemistry Parameters1
in Pool of
Placebo-Controlled Studies2,3,4
Cutoff Date5: |
Serum
Chemistry Parameters
and Units of
Measure6 |
Treatment
Groups7,8 |
|
New Drug |
Placebo
|
Active Control |
|
N9 |
BL10 |
Change from BL11 |
N |
BL |
Change from BL |
N |
BL |
Change from BL |
Albumin (g/dl) |
|
|
|
|
|
|
|
|
|
Alkaline Phosphatase (U/L) |
|
|
|
|
|
|
|
|
|
Bilirubin, total (mg/dl) |
|
|
|
|
|
|
|
|
|
BUN
(mg/dl) |
|
|
|
|
|
|
|
|
|
CK
(U/L) |
|
|
|
|
|
|
|
|
|
Calcium (mg/dl) |
|
|
|
|
|
|
|
|
|
Cholesterol (mg/dl) |
|
|
|
|
|
|
|
|
|
Creatinine (mg/dl) |
|
|
|
|
|
|
|
|
|
GGT
(U/L) |
|
|
|
|
|
|
|
|
|
Glucose (mg/dl) |
|
|
|
|
|
|
|
|
|
LDH
(U/L) |
|
|
|
|
|
|
|
|
|
Phosphorus (mg/dl) |
|
|
|
|
|
|
|
|
|
Potassium (mmol/L) |
|
|
|
|
|
|
|
|
|
Sodium (mmol/L) |
|
|
|
|
|
|
|
|
|
Triglycerides (mg/dl) |
|
|
|
|
|
|
|
|
|
Uric Acid (mg/dl) |
|
|
|
|
|
|
|
|
|
1This table provides
data comparing the mean change from baseline across treatment
groups for serum chemistry parameters. An acceptable alternative
would be to provide median change from baseline. The
postmeasurement is generally the worst value during treatment.
2This sample table is
based on a pool of similarly designed placebo controlled trials.
Other pools, as well as individual trials may also be of
interest.
3Patients
participating in crossover trials should be enumerated for each
of the pertinent columns of the table (e.g., a patient receiving
treatment in each of the three arms of a 3-way crossover study
comparing New Drug, active control, and placebo would be
included in all three columns).
4This table should be
provided by the sponsor in electronic format. The exact design
of the table and the preferred electronic format should be
established in discussions between the sponsor and the reviewing
division.
5This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
6The parameters
included in this list are for illustration. In general, the list
should include all those serum chemistry parameters measured in
whatever pool of studies is the focus of the table. Similarly,
the units of measure are for illustration, and these details
should be worked out in consultation with the reviewing
division.
7In the sample table,
only 1 column is provided for an active control group.
One such category may suffice for certain NDAs, but may not for
others, and the decision regarding how to categorize active
control patients should be made in consultation with the
reviewing division.
8Similarly, for this
table, only 1 column is provided for new drug, with the
implication that all new drug patients, regardless of dose,
should be included in the calculations for that column. Other
approaches (e.g., dividing patients on the basis of dose), may
be equally appropriate. If the studies used were fixed-dose
studies, it is generally most informative to preserve the dose
categories in constructing this table. However, dose categories
that are not relevant to the doses that are being recommended
for use may reasonably be omitted from this table. It is
generally not useful to try to artificially construct dose
categories from dose titration studies, since there is often
confounding of dose and time.
9N represents the
number of patients who had the serum chemistry parameter of
interest assessed at baseline and at least one follow up time.
10This column should
provide the baseline means for all the serum chemistry
parameters of interest.
11This column should
provide the mean change from baseline to patient's worst on drug
value for each of the serum chemistry parameters of interest.
While not hypothesis testing, p-values provide some measures of
the strength of the finding and should be produced for all new
drug/placebo pairwise comparisons and any p-values meeting a p <
0.05 level of significance criterion should be noted by an
asterisk (*) as a superscript to the mean change from baseline.
|
|
|
|
|
|
|
|
|
|
Table 7.1.7.3.2.1
Incidence of Potentially Clinically Significant Changes in Serum
Chemistry Parameters1
for
Pool of
Placebo Controlled Studies for New Drug2,3,4
Cutoff Date5: |
Serum
Chemistry Parameters and PCS Criteria
7
L=Low; H=High;
ULN=Upper Limits of
Normal |
Treatment
Groups6 |
|
New Drug |
Placebo |
Active Control |
|
Total
Pts8 |
Abnormal |
Total
Pts |
Abnormal |
Total
Pts |
Abnormal |
|
|
Nbr9 |
%10 |
|
Nbr |
% |
|
Nbr |
% |
Albumin-L (< 2.5 g/dl) |
|
|
|
|
|
|
|
|
|
Alkaline P'tase-H (> 400 U/L) |
|
|
|
|
|
|
|
|
|
Bilirubin, total-H (> 2 mg/dl) |
|
|
|
|
|
|
|
|
|
BUN-H (> 30 mg/dl) |
|
|
|
|
|
|
|
|
|
CK-H (> 3XULN) |
|
|
|
|
|
|
|
|
|
Calcium-L (< 7 mg/dl) |
|
|
|
|
|
|
|
|
|
Calcium-H (> 12 mg/dl) |
|
|
|
|
|
|
|
|
|
Cholesterol-H (> 300 mg/dl) |
|
|
|
|
|
|
|
|
|
Creatinine-H (> 2 mg/dl) |
|
|
|
|
|
|
|
|
|
GGT-H
(> 3XULN) |
|
|
|
|
|
|
|
|
|
Glucose-L (< 50 mg/dl) |
|
|
|
|
|
|
|
|
|
Glucose-H (> 250 mg/dl) |
|
|
|
|
|
|
|
|
|
LDH-H
(> 3XULN) |
|
|
|
|
|
|
|
|
|
Phosphorus-L (< 2.0 mg/dl) |
|
|
|
|
|
|
|
|
|
Phosphorus-H (> 5.0 mg/dl) |
|
|
|
|
|
|
|
|
|
Potassium-L (< 3.0 mmol/L) |
|
|
|
|
|
|
|
|
|
Potassium-H (> 5.5 mmol/L) |
|
|
|
|
|
|
|
|
|
SGOT/AST-H (> 3XULN) |
|
|
|
|
|
|
|
|
|
SGPT/ALT-H (> 3XULN) |
|
|
|
|
|
|
|
|
|
Sodium-L (< 130 mmol/L) |
|
|
|
|
|
|
|
|
|
Sodium-H (> 150 mmol/L) |
|
|
|
|
|
|
|
|
|
Triglycerides-H (> 300 mg/dl) |
|
|
|
|
|
|
|
|
|
Uric Acid (F)-H (> 8.0 mg/dl) |
|
|
|
|
|
|
|
|
|
Uric Acid (M)-H (> 10.0 mg/dl) |
|
|
|
|
|
|
|
|
|
1This table provides
data comparing the incidence across treatment groups of patients
who were normal at baseline meeting criteria of having had a
change on any of the listed serum chemistry parameters of
potential clinical significance (PCS). Separate listings should
be provided for patients who were abnormal at baseline and met
these PCS criteria.
2This sample table is
based on a pool of similarly designed placebo controlled trials.
Other pools, as well as individual trials may also be of
interest.
3Patients
participating in crossover trials should be enumerated for each
of the pertinent columns of the table (e.g., a patient receiving
treatment in each of the three arms of a 3-way crossover study
comparing New Drug, active control, and placebo would be
included in all three columns).
4This table should be
provided by the sponsor in electronic format. The exact design
of the table and the preferred electronic format should be
established in discussions between the sponsor and the reviewing
division.
5This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry. Generally this date should be no more than several months
prior to the submission date for an NDA. This date as well as
this table may likely need to be updated during the course of
NDA review as more data become available.
6In the sample table,
only 1 column is provided for an active control group.
One such category may suffice for certain NDAs, but may not for
others, and the decision regarding how to categorize active
control patients should be made in consultation with the
reviewing division. Similarly, for this table, only 1 column is
provided for new drug, with the implication that all new drug
patients, regardless of dose, should be included in the
calculations for that column. Other approaches (e.g.,
distinguishing patients on the basis of dose), may be equally
appropriate.
7The parameters
included in this list are for illustration. In general, the list
should include all those serum chemistry parameters measured in
whatever pool of studies is the focus of the table. Similarly,
the proposed criteria for potentially clinically significant
are for illustration, and these details should be worked out in
consultation with the reviewing division.
8The total number of
patients for each parameter should represent the number of
patients for the treatment group who (1) had that parameter
assessed at baseline and at least one follow up time and (2) for
whom the baseline assessment was normal.
9The number abnormal
represents the subset of the total number who met the criterion
in question at least once during treatment. A separate listing
should provide patient identification for those patients meeting
the criterion.
10Percentage of the
total number meeting the criterion should be rounded to the
nearest integer. While not strictly hypothesis testing, p-values
should be produced for all new drug/placebo pairwise comparisons
and any p-values meeting a p < 0.05 level of significance should
be noted by an asterisk (*) as a superscript to the %.
Table 7.1.7.5.1
Hepatotoxicity
Evaluation
I. Data Collection
A. Overview of liver
chemistry data (tests performed, frequency, specific
follow-up plans for abnormal values)
B. Specific follow-up plan if
chemistry is elevated at end of treatment
C. Re-challenge plan, if any
D. Exclusions from studies because
of liver chemistry abnormalities, if any
II. Observations
A. Abnormal liver
chemistries seen in controlled trials (separate for pooled
placebo controlled, active controlled) with greater than two
week exposure. Rates can be given as events/exposed;
positive findings can be also analyzed as events per patient
year and examined for rates over time.
1. Rates of 3x, 5x, 10x,
20x ULN elevations of AST (SGOT), ALT (SGPT), and either
ALT or AST
2. Rates of any elevations of
bilirubin; rate of elevated bilirubin to >1.5x ULN
3. Rates of alkaline phosphatase
(AP) > 1.5x ULN
4. Rates of elevated transaminase
accompanied by elevated bilirubin.
All rates should be given for both
drug and control group.
B. For total database with
exposure > two weeks (i.e., including uncontrolled)
Same as for controlled database
(1-4)
C. Individual events
1. Listing of patients
with any elevated transaminase (>3x ULN), without more
than slight AP elevation, associated with increase in
bilirubin to > ULN.
2. Show time course of enzyme and
bilirubin elevations
3. For such patients, review
clinical situation
a. Ethanol history
b. Evidence viral hepatitis
d. Special studies, notably Bx
e. Possible
confounding, including concomitant illness,
concomitant medications (known hepatotoxins, including
acetaminophen)
III. Possible problems/signals
A. Any patient with elevated
transaminase (to at least 3x ULN, generally higher), no
evidence of obstruction (elevated AP) and even modestly (2X
ULN) elevated bilirubin. Greater elevation of bilirubin is
stronger signal.
B. Greater rate than control
of 3x, 5x, 10x, etc. elevations of transamisne.
|
|
|
|
Table
7.2.1.1.1
Enumeration of
Subjects/Patients for
New Drug
Development Program1,2,3,4
Cutoff Date5:
|
Study
Groups |
Treatment
Groups |
|
New Drug |
Active Control6 |
Placebo |
Completed Phase 1 (Clinical
Pharmacology) |
Single Dose |
120 |
30 |
30 |
Multiple Dose |
60 |
30 |
30 |
Ph
1 Subtotal |
180 |
60 |
60 |
Completed Phase 2-3 (Studies of
Proposed Indication) |
Placebo Control7 |
|
|
|
Fixed Dose |
500 |
150 |
150 |
Flexible Dose |
100 |
100 |
100 |
Active Control |
|
|
|
Fixed Dose |
200 |
100 |
0 |
Flexible Dose |
100 |
100 |
0 |
Uncontrolled |
|
|
|
Short Term |
100 |
0 |
0 |
Long Term |
700 |
0 |
0 |
Ph
2-3 Subtotal |
12008 |
450 |
250 |
Ongoing Phase 2-3 Studies (Studies
of Proposed Indication) |
Placebo Control |
|
|
|
Flexible Dose |
1509 |
0 |
1509 |
SD
Subtotal |
120 |
30 |
30 |
MD
Subtotal |
1410 |
480 |
430 |
Grand Total |
1530 |
510 |
460 |
1This table provides
a count by study type of the subjects/patients exposed to new
drug, active control, and placebo across the entire set of
studies in the development program that contributed safety and
efficacy data for new drug. It should include all
subjects/patients known or assumed to have received even a
single dose of assigned treatment. It should exclude
subjects/patients who are known not to have received any of the
assigned treatments or for whom no follow up information is
available subsequent to the assumed receipt of assigned
treatment. A separate listing of all such patients should be
provided. [Note: If this list includes more than a few patients,
this may indicate a potentially important problem in the conduct
of studies.]
In creating this table, it is
necessary to classify and group studies on the basis of several
characteristics. For the purposes of this table, the following
characteristics and distinctions were deemed important:
- Phase 1 vs Phases 2 to 3
- Completed vs Ongoing and
Blinded
- Single Dose vs Multiple
Dose
- Controlled vs Uncontrolled
- Short-Term vs Long-Term
- Placebo-Controlled vs
Active-Controlled
- Fixed Dose vs Flexible Dose
Obviously, there are other
features that may be important as well, and that could lead to
additional breakdowns within the table or to separate tables
(e.g., different indications, inpatient vs outpatient status,
differences in the quality and completeness of data collected
across different studies, foreign vs domestic). The
characteristics to be used in classifying studies for the
purpose of this table should be decided in consultation with the
designated reviewing division at FDA.
In addition to this table that
enumerates patients by category of study, it would be useful to
have a table that enumerates patients by each individual study
in the development program. This would be an expanded version of
the above table that enumerates patients for each study (i.e.,
each of the categories in the above table would identify and
provide data for the individual studies comprising that
category). Sponsors ordinarily provide such a table.
2Patients
participating in crossover trials should be counted in each of
the pertinent columns of the table (e.g., a patient receiving
treatment in each of the three arms of a 3-way crossover study
comparing new drug, active control, and placebo would be counted
in all three columns).
3Footnotes to this
table should identify by study number all those studies
comprising the various study groupings for this table. For
example, in the sample table, the fixed dose placebo controlled
trials contributing to the counts for that category should be
listed in a footnote, and similarly for all other categories.
4This table should be
provided by the sponsor in electronic format. The exact design
of the table and the preferred electronic format should be
established in discussions between the sponsor and the reviewing
division.
5This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table likely need to be updated during the course
of NDA review as more data become available.
6In the sample table,
only 1 column is provided for an active control group.
One such category may suffice for certain NDAs, but may not for
others, and the decision regarding how to categorize active
control patients should be made in consultation with the
reviewing division.
7In this table, a
decision was made to pool all studies having a placebo arm,
whether or not an active control arm was also included. Thus,
the active control category includes only those active control
studies that did not have a placebo control arm. Other
approaches to grouping studies may be equally appropriate.
8The intent of this
table is to provide a count of unique subjects/patients exposed
to new drug, etc. in the development program. Since patients
often participate in more than 1 study in a development program,
it is necessary to have an approach to avoid counting patients
more than once for the subtotals and grand totals. The approach
used in this table is to include in parentheses in the pertinent
cells of the table a count of the patients in that cell total
who have already been counted by virtue of having participated
in a previous study (e.g., a patient in an open extension trial
should have been previously counted in an acute, controlled
phase). The subtotals of unique individuals exposed to the
assigned treatment can then be calculated by subtracting the sum
of all numbers in parentheses from the sum of all the cell
totals for each column (e.g., in this table, the completed phase
2 to 3 subtotal for new drug is 1700 less the 500 patients
already counted in short-term controlled trials, or 1200).
9Frequently, some
studies may be ongoing and blinded at the time of NDA
submission, even though some individual patients having
experienced serious adverse events may have been unblinded. In
these instances, the table should include estimates of the
numbers of patients exposed to new drug, etc. from these
studies, since exact counts may not be available. Footnotes
should indicate when the table entries are based on estimates
rather than exact counts.
|
|
|
|
Table
7.2.1.2.1
Demographic
Profile for Phase 2-3 Studies with New Drug1,2,3,4,5
Cutoff Date6:
|
Demographic
Parameters |
Treatment
Groups7,8 |
|
New Drug
N =
|
Placebo
N =
|
Active Control
N =
|
Age
(years) Mean
Range
Groups9
< 40
40-64
> 65 |
%
%
% |
%
%
% |
%
%
% |
Sex
Female
Male |
%
% |
%
% |
%
% |
Race10
Caucasian
Non-Caucasian |
%
% |
%
% |
%
% |
Weight (kg) Mean
Range |
|
|
|
1This table should be
based on a pool of all trials in the phase 2 to 3 development
program. Similar tables may be appropriate for other subgroups
within the phase 2 to 3 program and also for certain individual
trials of interest. The specific trials included should be
listed.
2Patients
participating in crossover trials should be included in the
calculations for each of the pertinent columns of the table
(e.g., a patient receiving treatment in each of the three arms
of a 3-way crossover study comparing New Drug, active control,
and placebo would be included in the calculations for all three
columns).
3Numbers for this
table should be rounded to the nearest integer.
4This sample table
includes 4 demographic categories of obvious interest, however,
others may be of interest as well (e.g., height, severity on
baseline measures of disease severity). It may also be of
interest to look at combinations of characteristics, such as
gender and age (e.g., women under 50).
5This table should be
provided by the sponsor in electronic format. The exact design
of the table and the preferred electronic format should be
established in discussions between the sponsor and the reviewing
division.
6This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
7In the sample table,
only 1 column is provided for an active control group.
One such category may suffice for certain NDAs, but may not for
others, and the decision regarding how to categorize active
control patients should be made in consultation with the
reviewing division. Similarly, for this table, only 1 column is
provided for new drug, with the implication that all new drug
patients, regardless of dose, should be included in the
calculations for that column. Other approaches (e.g.,
distinguishing patients on the basis of dose), may be equally
appropriate.
8If, as is often the
case, the Ns available for calculating any particular
demographic parameter are less than the Ns in the column
headings, these Ns should be provided, along with an
explanation, in footnotes.
9If there are
pediatric exposures, these should be broken out as well.
10Other approaches to
racial categorization may be substituted for that proposed in
this sample table.
|
|
|
|
|
|
|
|
|
Table
7.2.1.3.1
Number
(Percent) of Patients Receiving New Drug According to Mean1,2,3,4,5,6,7
Daily Dose and
Duration of
Therapy in Phase 2-3 Studies (N=2500) Cutoff Date8: |
Duration
(Weeks) |
Dose9
(mg) |
|
0<Dos<5 |
5<Dos<10 |
10<Dos<20 |
20<Dos<30 |
30<Dos<50 |
50<Dos |
Total
(AnyDos) |
(%) |
0<Dur<1 |
6 |
19 |
31 |
31 |
25 |
13 |
125 |
(5%) |
1<Dur<2 |
6 |
19 |
31
|
31 |
25 |
13 |
125 |
(5%) |
2<Dur<4 |
13 |
37 |
62 |
63 |
50 |
25 |
250 |
(10%) |
4<Dur<12 |
31 |
94 |
156 |
156 |
125 |
63 |
625 |
(25%) |
12<Dur<24 |
25 |
75 |
125 |
125 |
100 |
50 |
500 |
(20%) |
24<Dur<48 |
25 |
75 |
125 |
125 |
100 |
50 |
500 |
(20%) |
48<Dur<96 |
13 |
37 |
62 |
63 |
50 |
25 |
250 |
(10%) |
96<Dur |
6 |
19 |
31 |
31 |
25 |
13 |
125 |
(5%) |
Total
(AnyDur) |
125 |
375 |
623 |
625 |
500 |
252 |
2500 |
(100%) |
(%) |
(5%) |
(15%) |
(25%) |
(25%) |
(20%) |
(10%) |
(100%) |
|
1This table is
calculated by first categorizing patients on the basis of the
interval of exposure for each (e.g., a patient exposed for 6
weeks would be counted in the 4<Dur<12 row). The mean
daily dose is then calculated for each patient for dose
categorization (e.g., a 6-week patient with a mean daily dose of
15 mg would be counted in the 10<Dos<20 column). Patients
are enumerated in only 1 cell of the matrix (i.e., this is a
mutually exclusive display). The dose and duration intervals
need to be designed specifically for the drug of interest. The
specific trials included should be listed. As with any table
summarizing data from disparate sources, it does not address all
information needs, and it should be interpreted with caution
(e.g., mean doses in the 4-12 row refer to mean doses over 0-12
weeks, not 4-12 as one might think). Nevertheless, the
information provided provides useful information.
2Similar tables can
be prepared for median, for modal, and for maximum dose.
3The same table can
be generated for any individual study or for any pool of
studies.
4The same table can
be generated for any subgroup of interest (e.g., on the basis of
age, sex, race, comorbid condition, concomitant medications, or
any combination of these factors).
5Similar tables
should be provided for active control drugs and placebo.
6If the total N for
this table does not match the total N from Table 5.1.1.1, as may
be the case (e.g., if dose or duration data are not available
for all exposed patients counted in Table 5.1.1.1, a footnote
should provide an explanation for the discrepancy).
7This table should be
provided by the sponsor in electronic format. The exact design
of the table and the preferred electronic format should be
established in discussions between the sponsor and the reviewing
division.
8This is the data
lock date for entering data into this table (i.e., the date
beyond which additional exposed patients were not available for
entry). Generally this date should be no more than several
months prior to the submission date for an NDA. This date as
well as this table may likely need to be updated during the
course of NDA review as more data become available.
9Dose may also be
expressed as mg/kg, mg/m2, or in terms of plasma
concentration if such data are available.
1 This guidance has been prepared by
the Integrated Summary of Safety group, a subcommittee of Good
Review Practices Track 8. The Track 8 Committee has been charged
with developing a guidance for the clinical review of a marketing
application under the Good Review Practices (GRP) initiative.
2 It is recognized that no drug is
safe in the sense of being entirely free of adverse effects.
Reference in the Food, Drug and Cosmetic Act to the 93safety94 of a
drug for the uses recommended in labeling has been interpreted as
meaning that the benefits of a drug outweigh its risks for those
uses. The safety review, however, is not a risk benefit analysis,
but rather, is the part of the NDA review that assesses and
describes the risks of the drug.
3 It is important to distinguish
between the concept of performing an integrated safety review and
the separate question of whether or not to pool data across studies
in the conduct of that review. For the purpose of this document, an
integrated safety review refers to the principle of bringing
together in one place in the review all data and analyses pertinent
to a particular safety issue (e.g., liver toxicity). Whether one
looks primarily at data from individual studies or at datasets
resulting from pooling of certain studies to address a particular
safety concern is not critical to the concept of an integrated
review. Either approach, or both approaches, will usually be used by
a reviewer in carrying out an integrated review.
4 See
http://www.fda.gov/cder/guidance/statnda.pdf
5 See
http://www.fda.gov/cder/guidance/statnda.pdf
6 21 CFR 314.50(f)(2) requires that
CRFs be submitted for each patient who died during a clinical study
or who did not complete the study because of an adverse event,
whether believed to be drug related or not. It should be clear from
the application that the ISS and other safety reports include all
adverse events that were seen during development, not just those
judged by investigators or the applicant to have been potentially
drug-related. This is also a useful point to make at a pre-NDA/BLA
meeting.
7 If the reviewer determines that
adverse event tables provided by the applicant are accurate and
fairly represent the data they purport to display, the tables may be
included in the safety review as appendices. If applicant-generated
tables are used, the review should identify the applicant as the
source.
8 The reviewer should be able to
easily access individual patient information. The reviewer may want
to clarify formatting and accessibility concerns at the pre-NDA
meeting. For hardcopy submissions, an index that directs the
reviewer to the exact location (volume and page number) of the CRF,
the narrative summary, and the individual patient safety data
display is essential (for sample index see Table 8.0.1). For
electronic submissions, the PDF files should have sufficiently
detailed bookmarks to offer easy navigation by the reviewer.
For example, narratives should be bookmarked by patient ID number,
not just by study treatment or treatment assignment.
9 The proportion of patients exposed
to the dose range that is effective should be considered. A total
exposure consistent with ICH recommendations (1500 total with
300-600 for 6 months and 100 for one year), may, on examination,
reveal far fewer who received an effective dose (i.e., the dose that
would be used). ICH E-1 ([The Extent of Population Exposure to
Assess Clinical Safety: For Drugs Intended for Long-Term Treatment
of Non-Life-Threatening Conditions, March 1995 (http://www.fda.gov/cder/guidance/iche1a.pdf)])
is clear in its expectation that the 6-month and one year exposures
should be at dosage levels intended for clinical use. Although it is
silent with respect to the 1500 figure, exposure at lower doses
would not be expected to be informative about the safety of the
clinically useful dose.
10 The International Conference on
Harmonisation of Technical Requirements for Registration of
Pharmaceuticals for Human Use (ICH) E3 Structure and Contents of
Clinical Study Reports.
11 Note that unexpected is
used differently from its use in 21 CFR 312.32, where it refers to
adverse events not identified in the investigator's brochure and
therefore reportable in an IND Safety Report.
12 Temple R, G Pledger, The FDA's
Critique of the Anturane Reinfarction Trial, N Engl J Med
303:1488-1492, 1980.
13 Since placebo and active control
patients can generally have had shorter durations of exposures than
patients given the new drug, they may have had less opportunity for
serious events to have occurred.
14 See 21 CFR 312.32(a); 314.80(a);
600.80(a).
15 Mutually exclusive
refers to the reason for dropping out. Patients should be identified
with only one of the reasons. However, patients may be represented
in more than one column (treatment group) of a table (e.g., patients
in a crossover study may have survived several treatment arms and
then dropped out).
16 This is recommended in the ICH
E3 guidance.
17 Food and Drug Administration,
Coding Symbols for Thesaurus of Adverse Reaction Terms, 5th
ed., FDA, Rockville, MD, 1995.
18 MedDRA (Medical Dictionary for
Regulatory Activities),
http://www.meddramsso.com/NewWeb2003/index.htm
19 A draft guidance for industry and
reviewers, Content and Format of the Adverse Reactions Section of
Labeling for Human Prescription Drugs and Biologics, was issued
in June 2000. Once finalized, it will represent the Agency's
thinking on this topic.
20 Applicants may consider only lab
values obtained within a certain window around the
protocol-specified date for collection. In some cases, the
laboratory data obtained outside the window may be available, but
the applicant may choose not to include it.
21 If CRFs are available from any
such studies and the data quality is comparable to that of data from
studies conducted under the applicant's IND, these data would
ordinarily be included in the primary source database.
22 ICH E1A The Extent of
Population Exposure to Assess Clinical Safety: For Drugs Intended
for Long-Term Treatment of Non-Life-Threatening Conditions
(http://www.fda.gov/cder/guidance/iche1a.pdf ) recognizes possible
differences in expected exposure (e.g., more patient exposure would
be expected for drugs with small effects, or drugs that are used
prophylactically in well populations, where only a small fraction of
patients will benefit).
23 A draft guidance Premarketing
Risk Assessment was issued in May 2004. Once finalized, it will
represent the Agency's thinking on this topic.
24 See ICH E4Dose-Response
Information to Support Drug Registration (http://www.fda.gov/cder/guidance/iche4.pdf).
Back
to Top
Back to Guidance Page
PDF requires the free Adobe
Acrobat Reader
Date created: November 30, 2005 |
|