Guidance for Industry
Antiretroviral Drugs Using Plasma HIV RNA Measurements —Clinical
Considerations for Accelerated and Traditional Approval
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. An alternative approach may be used if
such approach satisfies the requirements of the applicable
statutes and regulations.
This guidance is intended to assist
sponsors in the clinical development of drugs for the treatment
of human immunodeficiency virus (HIV) infection. Specifically,
this guidance addresses the Agency’s current thinking regarding
designs of clinical trials that use HIV ribonucleic acid (RNA)
measurements to support accelerated and traditional approvals of
antiretroviral drug products. It is also intended to serve as a
focus for continued discussions among the Division of Antiviral
Drug Products (DAVDP), pharmaceutical sponsors, the academic
community, and the public.
This guidance does not address specific
phase-1 and phase-2 development issues, development of alternate
dosing regimens, or the use of HIV-1 resistance testing. These
issues will be addressed separately in future guidance
documents.
In addition to consulting guidance
documents, sponsors are encouraged to contact the division to
discuss specific issues that arise during the development of an
antiretroviral drug product.
Accelerated approvals of antiretroviral
drugs have been based for years on changes in surrogate
endpoints, such as CD4 cell counts and plasma HIV RNA
levels. Traditional approvals were based on clinical endpoint
trials assessing the effects of a drug on mortality and/or HIV
disease progression. With the availability of potent
antiretroviral drug regimens and sensitive assays for assessing
plasma HIV RNA, the standards of clinical practice evolved to a
paradigm emphasizing maximal and durable HIV RNA suppression.
In addition, with the successes of combination therapy and the
subsequent decline of HIV-related illnesses (Palella et al.,
1998; Hogg et al., 1999), it became clear that a requirement for
clinical endpoint studies for every traditional approval was no
longer necessary nor feasible.
In July 1997, the Agency convened an
advisory committee meeting to consider the use of changes in HIV
RNA levels as endpoints in clinical trials supporting
traditional approval of antiretrovirals. To evaluate the
feasibility of using HIV RNA levels as a study endpoint, a
collaborative group of pharmaceutical, academic, and government
scientists investigated relationships between treatment-induced
changes in HIV RNA and clinical endpoints from ongoing and
completed antiretroviral trials (Murray et al., 1999; Hill et
al., 1998). In several analyses of more than 5000 patients in
multiple trials, a clear association was identified between
initial decreases in plasma
HIV RNA levels and reduction in the risk of clinical progression
and death. This relationship was observed across a range of
patient characteristics including pretreatment CD4 counts and
HIV RNA levels, prior drug experience, and treatment regimen.
Based on these data, the Division of Antiviral Drug Products
advisory committee concurred that treatment-induced decreases in
HIV RNA levels were highly predictive of meaningful
clinical benefit and that HIV RNA measurements could serve as
endpoints in trials designed to support both accelerated and
traditional approvals. The Division proposed that accelerated
approvals could be based on studies that show a drug’s
contribution toward shorter-term reductions in HIV RNA (e.g., 24
weeks) while traditional approvals could be based on trials that
show a drug’s contribution toward durability of HIV RNA
suppression (e.g., for at least 48 weeks). The committee agreed
with this proposal and also recommended that changes in CD4 cell
counts be consistent with observed HIV RNA changes when
considering approval of an antiretroviral drug.
According
to the regulations (21 CFR 314.500 - 314.510), three criteria
need to be addressed when considering the appropriateness of an
accelerated approval: (1) the disease studied must be serious
or life-threatening, (2) there must be an available surrogate
that is reasonably likely to predict clinical benefit, (3) there
must be demonstration of improved activity over approved drugs
or activity in a population in need of additional therapeutic
options. As stated in 21 CFR 314.500, accelerated approvals
apply to drugs that “have been studied for their safety and
efficacy in treating serious and or life-threatening illnesses
and that provide meaningful therapeutic benefit to patients over
existing treatments (e.g., ability to treat patients
unresponsive to, or intolerant of available therapy, or improved
patient response over available therapy).” An accelerated
approval can be based on a surrogate endpoint reasonably likely
to predict clinical benefit (but not necessarily a fully
established surrogate) or a clinical endpoint other than
irreversible morbidity or mortality, where the ultimate goal of
therapy is an effect on morbidity or mortality.
Under
accelerated approval, continued marketing of a drug is subject
to certain conditions outlined in the regulations, principally,
an agreement to conduct further studies to establish clinical
benefit. Since traditional approvals of antiretroviral drugs may
now be based on changes in HIV RNA to establish benefit, shorter
term reductions in HIV RNA levels supporting an accelerated
approval can be considered surrogate endpoints for
longer, durable suppression of HIV RNA levels.
Because continuous
treatment with multiple antiretrovirals is necessary to achieve
HIV suppression and because a substantial number of patients may
no longer respond to available treatment options because of poor
tolerability or the emergence of resistance, new antiretroviral
agents are needed. Therefore, the Division strongly encourages
sponsors to study patients who have few remaining approved
treatment options. Demonstrating safety and drug activity (using
a surrogate endpoint) in populations in need of new therapeutic
options is consistent with the intended goals of the accelerated
approval regulations.
For
patients who have no remaining treatment options and are at
imminent risk of disease progression secondary to severe
immunodeficiency, participation in randomized controlled
registrational trials may not always be desirable. For such
patients, participation in open-label, compassionate-use safety
studies or expanded access studies may be more appropriate. This
was the consensus of the advisory committee panel held on
January 2001. The Division encourages sponsors with promising
new drugs to develop access protocols as soon as safely feasible
in the drug development plan for patients in need. Trial design
considerations for studies of heavily treatment experienced
patients are addressed further in section III. C., Efficacy
Considerations for Accelerated Approval.
The
Division also recognizes that some investigational drugs may not
be appropriate for patients with substantial amounts of prior
antiretroviral drug exposure, particularly when an
investigational drug has demonstrated broad cross-resistance to
other drugs in its class. Therefore, other characteristics may
also support a drug as a candidate for accelerated approval
including improved efficacy or improved safety or tolerability
over existing drugs. In addition, a more convenient dosing
schedule (e.g., tid to bid or qd), novel mechanism of action,
different clinical cross-resistance profile, favorable drug
interaction profile, or utility in specific populations in need
of therapies (e.g., pregnant women, children) may also support a
drug as a candidate for accelerated approval. Such advantages
should be demonstrable with data.
The majority of
antiretroviral accelerated approvals to date have been supported
by safety data from at least 400 to 500 patients who received
the proposed dose for marketing (or higher doses) for
approximately 6 months. Although HIV is considered
life-threatening, the numbers of patients studied for previous
antiretroviral accelerated approvals have approximated, or
exceeded, the International Committee on Harmonisation (ICH)
guidance for drugs intended for long-term treatment of
non-life-threatening conditions.
The ICH guidance recommends the collection of safety data on at
least 300 to 600 patients receiving the proposed dose for 6
months with safety data on a total of 1,500 patients when
including patients with shorter-term drug exposures. The
recommended safety database of 300-600 patients for 6 months was
chosen to allow a reasonable chance to identify adverse events
occurring at a frequency as low as 1:100. The ICH guidance also
states that additional safety data on longer term use in a
smaller cohort, than stated above, is advisable.
Applicants are
encouraged to discuss their proposed safety database with the
Division prior to submitting an NDA. On occasion, specific
findings in preclinical or phase 1-2 development may indicate
the need for a database that is larger or longer in duration to
adequately evaluate potential drug toxicity.
Controlled and comparative safety data are
preferred. Safety data from uncontrolled or expanded access
protocols may be useful, but often lack the degree of detailed
reporting obtained in controlled clinical trials. In addition,
the assessment of causal relationships between a drug and an
adverse event is more difficult when relying on uncontrolled
safety data.
Every
attempt should be made to design randomized, blinded (or
partially blinded), controlled trials that permit one to clearly
assess an investigational drug’s contribution, as part of a
combination regimen, toward decreases in HIV RNA. Studies in a
broad range of patient populations (gender, age, and race) and a
range of pretreatment characteristics (e.g., advanced and early
disease, heavily pretreated and treatment naïve, as appropriate)
are recommended to characterize the activity of the drug. The
Division recommends that NDAs include at least two adequate and
well-controlled studies of a minimum of 24 weeks duration to
support accelerated approval. However, given that some
patients will have longer term follow-up, submissions should
include some data past 24 weeks when possible. In treatment
naïve patients, analyses at earlier time points (e.g., 16 weeks)
have proven to be less discriminating for detecting important
differences between treatment regimens. In addition, prior to
24 weeks, some patients may have HIV RNA levels that are still
declining, especially when measured with sensitive assays.
Blinded comparisons with controls are preferred
because they reduce biases resulting from differences in
management, treatment, or assessment of patients arising from
investigator or subject knowledge of the randomized treatment.
The Division acknowledges that there are situations in which
blinding drugs or regimens may not be feasible; however, in most
cases the difficulties associated with blinding a study are not
insurmountable. For example, blinding may be difficult when
drugs require dose adjustments based on other components of a
regimen; however, this could be accomplished by similarly dose
adjusting the placebo. In combination regimens, the number of
placebos could be large or unmanageable; however, in most cases
blinding of one component of a regimen is all that is
necessary. The Division encourages sponsors to provide access
to their products’ placebo formulations for use in clinical
studies conducted outside their own drug development.
Sponsors designing studies in which blinding may be
difficult or infeasible should discuss the proposal with the
Division in advance, to review potential modifications that
might facilitate blinding and to discuss the potential impact of
open-label therapy on interpretation of results. When blinding
is impossible, open-label protocols should have detailed
procedures for treatment switches and toxicity management, since
differential implementation of protocol procedures among
treatment arms in open-label studies may impair interpretability
of study results. For example, the validity of the results of
open-label studies may be questioned if there are large
differences between treatment arms with respect to nonprotocol-specified
treatment discontinuations. In such instances DAVDP expects
additional sensitivity analyses using different methods of
handling treatment discontinuations or missing data, including
those that treat the control and study arms asymmetrically (see
Appendix B).
Conducting
controlled, comparative studies in patients who have exhausted
many treatment regimens may call for innovations in study
design, including the use of multiple investigational agents,
factorial comparisons, and collaboration among two or more
sponsors. Collaboration and the use of multiple investigational
agents are strongly encouraged; however, phase-3 studies
intended to support registration should be designed such that
the treatment effect of each drug of interest can be isolated.
In addition, the potential for drug-drug interactions,
particularly those that may require dosing changes, should be
considered in advance.
For heavily
treatment experienced patients, the sponsor may consider
studying doses that are higher than those studied in treatment
naïve patients. Higher concentrations may be necessary to treat
less susceptible HIV isolates. However, this approach may be
limited by the amount of supporting preclinical data or by
previously observed dose-limiting toxicities in the clinic.
As previously stated, patients with no approved
treatment options and who are at imminent risk of progression
may be better suited for expanded access or compassionate use
protocols. Inclusion criteria for registrational studies should
be such that patients are likely to complete the intended study
period.
·
Superiority trial designs
Phase-3 superiority trials can include add-on
or substitution comparisons. In the first case, the
investigational drug plus a standard combination regimen is
compared to placebo plus the same standard regimen. In some
cases, an experimental drug could be added to a background
regimen of drugs that the participant or investigator chooses
from a list of possibilities.
For substitution comparisons, the investigational
drug is substituted for a component of a standard regimen. This
regimen is then compared to the standard regimen.
·
Noninferiority trial designs
Noninferiority
trials use substitution comparisons as described above.
For noninferiority comparisons, it is important that the
contribution of the substituted drug to a regimen’s overall
activity be previously characterized in the population of
interest. This is often referred to as a study’s assay
sensitivity. This information should be used to support a
noninferiority comparison and to calculate an appropriate sample
size. This will be discussed further under section III. C. 3.,
Choice of Control Arms.
·
Dose comparison trial designs
Phase 2 dose ranging studies that have demonstrated
a significant dose response may provide supportive data for an
accelerated approval of an antiretroviral drug. Generally, dose
comparison studies should include a large enough range of doses
to demonstrate a response slope. Given that it would not be
desirable to design a large phase 3 protocol using doses of
antiretrovirals that are known or predicted to be suboptimal
based on preclinical or clinical data, phase 3 efficacy studies
that primarily rely on dose comparisons may be difficult to
design. Sponsors should discuss proposals for dose comparison
studies with the Division in advance.
·
Modified factorial comparison trial designs
This design is
actually a variation of an add-on superiority design, except
that multiple investigational agents may be tested
simultaneously. This design may be useful when studying
patients who are unable to construct a viable antiretroviral
regimen from approved drugs.
A hypothetical
example of a modified factorial is shown below. It is considered
modified because not all possible cells of the factorial
are studied. Excluding cells (e.g., background therapy alone)
allows participants a greater possibility of being randomized to
a treatment arm with more potentially active drugs. In the
example below, the study tests superiority of arm 1 to arms 2
and 3 for two investigational drugs, X and Y. For drug Y,
superiority of Arm 1 vs. 2 should be demonstrated. For Drug X,
superiority of Arm 1 vs. 3 should be demonstrated to support
efficacy. This or alternative study designs using multiple
investigational agents may be appropriate but should be
discussed with the Division in advance.
·
Other Designs
As stated above,
the Division acknowledges that designing adequate studies in
heavily treatment experienced patients is difficult and may
require modifications of typical study designs and procedures.
Such issues were discussed during the January 2001 advisory
committee meeting. In general, consensus was reached that use
of external or historical controls would not be advisable
because a large degree of heterogeneity in the treatment
experienced patient population would impair the interpretability
of such a comparison.
However, the committee did offer support for
several similar proposals of a hybrid type design in which
randomized comparisons were continued for short periods (days to
weeks) followed by longer term assessments of the activity of a
new drug as part of a combination regimen. There were several
similar proposals. One example of such a hybrid design is shown
below. Part 1 allows discrimination of the short-term antiviral
effects of components of the regimen. Part 2 evaluates the
durability of the antiviral effect produced by a combination
regimen including investigational drug X. Further support for
the investigational drug’s (X) contribution to a durable
antiviral effect could be provided by a correlation between
baseline phenotypic susceptibility and antiviral response.
Example B: Two-Part Hybrid
Design
Part 1
(Duration approximately 2 weeks)
Arm 1: Investigational drug X
added to old regimen
Arm 2: Continue old regimen
Arm 3: Switch to new
background regimen
Part 2: (22 weeks for a
total of 24 weeks)
All Arms: Receive new
background + investigational drug X
The duration of
part 1 would depend on what is known regarding the rate of
emergence of resistant strains under selective pressure with
drug X and/or other drugs of the same class. Part 1 would allow
a clean, albeit short, discrimination of the regimen’s
components toward the overall activity of the regimen.
A 24-week
antiviral response rate that varied in accordance with baseline
phenotypic susceptibility to drug X could offer supportive
evidence that drug X was contributing to the regimen’s overall
efficacy at 24 weeks. For example, such a study might show that
subgroups of patients with baseline phenotypic susceptibility to
drug X of less than 5 fold of wild type, 5 to 10 fold of wild
type, and greater than 10 fold of wild type had 24 week
virologic response rates of 90 percent, 60 percent and 30
percent, respectively. This would be analogous to, though not
as convincing as, a dose response. In this case patients are
not randomized to dose but evaluated prospectively according to
their baseline differences in susceptibility. However, this
interpretation could be confounded if patients with better
baseline phenotypic susceptibility to drug X also had better
baseline susceptibility to other drugs in the regimen. In this
case, the argument for using baseline susceptibility to support
a particular drug’s contribution toward activity would be
weakened.
Because of the
limitations of this design, such studies should not provide the
principal support for an NDA package. However, such a study may
complement a well-controlled study(ies) by providing important
data for the use of the investigational drug in a population of
need.
Every attempt
should be made to include study treatment regimens consistent
with standards of clinical practice while the trial is being
conducted. In general, current HIV treatment guidelines
emphasize the importance of using at least 3 potentially active
drugs (if possible) when constructing a regimen. Proposals for
control arms that deviate from current standards of care should
be supported by convincing scientific rationale and/or data and
discussed with the Division before implementation. Because of
the evolving nature of accepted standards of HIV treatment,
appropriate comparison regimens can be expected to change over
time.
From a patient
management perspective, use of control regimens that have been
determined to be clearly suboptimal, as based on clinical
studies or consensus of expert panels reviewing pertinent data,
would jeopardize the viability of a study. From a regulatory
perspective, controls used for efficacy comparisons should
generally be those that have demonstrated durable viral
suppression. Although one particular control regimen cannot be
recommended as the most appropriate, some regimens are clearly
inappropriate from either a regulatory or patient management
perspective. For example, regimens consisting of only two of
the currently approved nucleoside reverse transcriptase
inhibitors (NRTI) in treatment naïve patients would not be
considered appropriate control regimens from a patient
management perspective.
Most antiretroviral drugs have not demonstrated
activity in all situations or populations. In fact, some drugs
are known to lack significant activity in some treatment
experienced patients due to cross-resistance. However, for
noninferiority trials, an active control must have
reproducible and well-defined clinical activity in the context
of the regimen and population to be studied. This is referred
to as assay sensitivity for a clinical trial. In other
words, the quantitative contribution of the control toward a
regimen’s overall activity should be previously demonstrated.
Otherwise, a study drug may appear to be noninferior to an
active control when in reality neither drug had contributed
toward the activity of a regimen. Readers should refer to the guidance prepared by
the ICH entitled, E10 Choice of Control Group in Clinical
Trials. This document addresses key issues relating to the
design of noninferiority studies.
Protocols
should include procedures for clinical management based on
changes in HIV RNA. However, to facilitate interpretation of
study results, it is critical that management decisions be made
in a uniform manner. This is particularly important for
open-label studies. Protocol procedures that allow treatment
switches for patients who never achieve HIV RNA levels below an
assay limit should be applied consistently across treatment
arms. For example, some protocols allow treatment naïve
patients who have not achieved an HIV RNA reduction of 1 log10
by 8 weeks to switch their antiviral regimen. These criteria
may vary depending on the population studied and the response
that is expected or desired.
It is anticipated
that plasma HIV RNA measurements will be used as the primary
assessment of drug efficacy. Sponsors are encouraged to use
sensitive approved HIV RNA assays, particularly for studies in
treatment naïve patients. More sensitive assays may allow
greater discrimination of differences in treatment effects. In
addition, the Division recommends the use of FDA approved HIV
RNA assays to ensure that assay performance characteristics are
understood. Currently, approved assays are the Roche Amplicor
HIV-1 Monitor standard and ultrasensitive tests and
the NucliSens HIV-1
QT.
The Division prefers that approved assays be used. However, the
Division also recognizes that new assays may offer potential
improvements over existing technologies, perhaps with regard to
sensitivity or detection of additional viral clades. In such
cases, sponsors can choose to use unapproved
assays but should be prepared to provide the Division with
information supporting the limits and performance
characteristics of the investigational assay (See Appendix A).
This will permit an independent evaluation of an assay’s
limitations. Since the formal review of HIV RNA assays falls
under the purview of the Center for Biologics, the Division’s
review of an unapproved assay’s performance will focus on the
interpretability of data with respect to the particular clinical
trials in the NDA. Thus, the Division’s review of assay
performance data does not imply that the given assay is
validated or FDA-approved for patient prognosis and/or
monitoring. Furthermore, this review does not imply that the
given assay is automatically acceptable for future clinical
trials.
When feasible, the portion of subjects achieving
HIV RNA levels below the assay limit at 24 weeks should be the
primary endpoint for accelerated approval. Although such
analyses are consistent with the current goals of clinical
practice, it should be emphasized that the Division views this
endpoint as a stringent tool for assessing drug activity and not
necessarily a recommended threshold for patient management. The
advantage of an endpoint that assesses response below an assay
detection limit, compared to one assessing mean changes, is that
it is more likely to be protected from treatment changes
resulting from a perceived lack of efficacy. In addition,
handling missing data is simplified when using a response
endpoint compared to one assessing mean changes. However, the
Division acknowledges that less stringent criteria or
thresholds, other than confirmation of HIV RNA levels above an
assay limit, may be useful for managing patients or for altering
drug regimens.
Alternatively, for studies in patients who are
unlikely to maximally suppress virus, responses other than those
below an assay limit may be considered, such as a sustained 1
log10 decline in HIV RNA. Also, assessing mean
changes in HIV RNA from baseline, including those that are
averaged over time, may be useful for treatment experienced
patients in which HIV RNA reductions below the assay limit are
not expected to be frequent. Sustained HIV RNA reductions of at
least 0.5 log have been shown to correlate with reductions in
the risk of disease progression. Assessing mean changes from
baseline are not considered as useful in patient populations
that are able to achieve HIV RNA reductions below an assay
limit. Sponsors are encouraged to discuss with the Division in
advance which endpoint may be most appropriate for a particular
protocol.
Analyses evaluating changes over time in CD4
cell counts should accompany the analyses of HIV RNA. Clinical
endpoint data (CDC class C events) should also be collected,
analyzed, and submitted with the NDA. However, the frequency of
such events is likely to be low particularly over a 24-week time
period.
The choice of
delta for noninferiority testing should be discussed with the
Division prior to study initiation because one delta is not
appropriate for all study designs. In the past, many
noninferiority studies have been powered based on a delta of 10
percent to 12 percent. In most cases, this allowed sponsors to
plan studies that would be reasonably powered and feasible to
conduct. Such a delta has been useful when comparing the most
potent component of a three-drug regimen in treatment naïve
individuals. However, the sponsor should ultimately attempt to
choose a delta based on prior knowledge of the quantitative
contribution of the active control (substituted part of the drug
regimen) to the regimen as a whole. This contribution should be
determined in a similar population with a similar length of
follow-up of the proposed study. For noninferiority testing,
sponsors should employ two-sided 95 percent confidence intervals
adjusted for multiple comparisons. If one-sided confidence
intervals are used, the alpha should be 0.025.
Both noninferiority and superiority can be assessed
in the same study provided that the noninferiority comparison
and choice of delta have been specified prior to study
initiation and/or provided that the choice of delta can be
clearly justified based on previous clinical data.
Analyses that include all randomized patients
should be included in all NDAs. Sometimes patients who were
randomized but never returned to receive drug, can be removed
from the analyses if the study is blinded or if both arms are
similar in this regard. In such analyses, patients who have
introduced a new antiretroviral drug to the regimen (except for
prespecified changes in the background regimen in which the
reason for change is due to the background drug and not due to
the study drug or control) have discontinued study, have been
lost to follow-up, or for whatever reason have missing HIV RNA
data should be considered to have HIV RNA levels above the assay
limit. This is sometimes referred to as a noncompleter
equals failure (NC=F) analysis. In general, missing HIV RNA
data between study visits with values below the assay limit does
not constitute treatment failure.
In addition to a NC=F analysis, an analysis
comparing documented virologic failures in evaluable patients
should also be submitted. Any inconsistencies in treatment
outcomes for these analyses should be explained. Sensitivity
analyses that use different methods of handling treatment
discontinuations and missing data should also be provided in
support of efficacy.
The Division prefers that study results be reported
in a manner in which the reason for treatment failure can be
easily discerned. For accelerated approvals, simple proportions
may be calculated for the possible treatment outcomes at 24
weeks as shown in the example in Table 1. Refer to Appendix B
for computing treatment outcomes and displaying 48-week data for
traditional approval.
Table 1.
Example of Data Display of Treatment Outcomes at 24 weeks
(accelerated approval)
Outcome at 24 weeks |
Experimental
N (%) |
Control
N (%) |
Below
assay limit |
xx (xx%) |
xx (xx%) |
Above
assay limit |
xx (xx%) |
xx (xx%) |
Death |
xx (xx%) |
xx (xx%) |
Drug
change or discontinuation due to adverse events |
xx (xx%) |
xx (xx%) |
Drug
change or discontinuation due to other reasons
Consent withdrawn
Loss
to follow up
Nonadherence
Pregnancy
Protocol violation
Other |
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
|
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
xx (xx%)
|
In this guidance, the term traditional
approval refers to the usual marketing clearance mechanism
for the majority of drugs that have demonstrated clinical
efficacy as shown by a treatment effect on a clinically
meaningful endpoint. In most cases, results from at least two
well-controlled studies should be included in an NDA to support
traditional approval. The guidance entitled Providing
Clinical Evidence of Effectiveness for Human Drug and Biological
Products (May 14, 1998) should be consulted for discussion
of circumstances in which approval may be considered on the
basis of a single trial.
It is expected
that traditional approvals will be supported by safety data from
a minimum of 500 patients who have received a drug for
approximately 48 weeks. In addition, sponsors should provide
any safety data from earlier study enrollees who have been
followed for periods longer than 48 weeks.
Because multiple
adverse events have been observed with chronic administration of
antiretroviral therapy, mechanisms should be used for
systematically evaluating adverse events over prolonged periods
following traditional approval. Controlled comparisons and
prospectively evaluated cohorts may be helpful in characterizing
and defining drug associations for late-occurring adverse
events. Therefore, after traditional approval, the Division
strongly encourages sponsors to continue to collect safety data
in key randomized studies or other treatment cohorts for
prolonged periods (3-5 years). For long-term follow-up, study
procedures and data collection may be targeted or streamlined to
relieve some of the burden of long-term safety assessments.
Controlled and
comparative safety data are preferred. Safety data from
uncontrolled compassionate use protocols may be supportive, but
often lack the degree of detailed reporting obtained in
controlled clinical trials. Uncontrolled safety data rarely
allow one to assess causal relationships between a drug and an
adverse event.
Since a
risk-benefit ratio is always considered in any approval, careful
attention should be paid to treatment discontinuations for
intolerance or toxicity. Sponsors should make every attempt to
ascertain reasons for treatment and/or study discontinuations.
In addition, patients who did not have a protocol-defined
dose-limiting toxicity, but nonetheless had an unresolved
intolerance or adverse event at the time of discontinuation,
should be classified as discontinuing treatment secondary to
drug intolerance and not due to patient’s choice
or other. Analyses should be performed to evaluate
reasons for treatment discontinuations, possible baseline risks
for treatment intolerance, and time until a dose-limiting
adverse reaction. Such analyses are particularly crucial in
studies with a substantial proportion of treatment
discontinuations (greater than 20-25 percent), or with different
rates of discontinuations among treatment arms.
The same studies that are evaluated at 24 weeks for
accelerated approval may be continued for 48 weeks and longer to
support traditional approval. The duration of these studies
should permit the last patient randomized and still continuing
therapy to have passed the 48-week time point. The final study
report should include all available data at the time of
analysis, including that beyond 48 weeks. Some studies should
be continued beyond traditional approval, as feasible, as part
of a post-marketing commitment to assess safety and efficacy of
chronic administration. Data regarding longer term virologic
assessments in principal studies can be displayed in drug
labeling.
Issues relating to choice of control arms,
comparisons, and study procedures are discussed in previous
sections.
Traditional
approvals can be based on study results that show the drug’s
contribution toward sustained suppression of plasma HIV RNA
levels. However, some applications can contain a combination of
clinical endpoint studies and HIV RNA studies. Thus, the types
of studies included will affect the wording of the indications
granted.
The sponsor should
also include supportive analyses of CD4 count
responses and clinical endpoints. In virologic studies, the
investigational drug should also show no deleterious effect on
clinical progression (CDC-C AIDS-defining illnesses and death);
however because the study is unlikely to have sufficient power
to detect treatment differences in clinical events,
statistically significant differences between treatment arms are
not expected. In addition, studies generally should show
favorable CD4 responses; however, there
may be clinical situations in which significant CD4
increases may not be demonstrated, such as in patients who
already have normal range CD4 counts.
Study results
showing discordant CD4 responses will require close
examination and if not explained could call the primary endpoint
into question. Drugs that produce major discordance between the
desired HIV RNA and CD4 cell count responses (i.e., a
drug that produces decreases in both HIV RNA and CD4
cell counts instead of decreases in HIV RNA with increases in CD4
cell counts) should probably be evaluated using clinical
endpoint studies.
The proportion of
patients with HIV RNA levels below the assay limit at 48 weeks
(or longer) and time-to-loss-of-virologic-response will be
considered primary endpoints for trials supporting traditional
approval. These virologic response endpoints are discussed
below.
·
Time-to-loss-of-virologic-response
Studies designed
to assess time-to-loss-of-virologic-response allow participants
who have lost a virologic response to switch or seek new therapy
without compromising major study analyses. One definition for
time-to-loss-of-virologic-response is the time between
randomization (or start of treatment) and the last value below
an assay limit in a patient who initially suppressed HIV RNA
below an assay limit but subsequently demonstrated virologic
rebound (two consecutive HIV RNA levels above the assay limit).
Subjects who do not achieve suppression below the assay limit
during the study (or within a predefined shorter time period
allowing for earlier treatment switching) can then be defined as
having a time to loss of response of zero. See appendix A for an
algorithm that the Division typically uses to define loss of
virologic response. Other algorithms may be appropriate
depending on the study designs. Analysis of the total duration
below the assay limit may also be presented, usually as a
secondary analysis.
For the primary
analysis, the Division usually considers patients who die, are
lost to follow-up, or introduce new study treatment due to
toxicity/intolerance (or any other reason) as treatment failures
at the time of those events. Patients who have introduced a new
antiretroviral drug regimen are also usually considered to be
nonresponders unless such changes only involve alterations in
background therapy that are permitted in the protocol. Other
analyses based on virologic failure alone should also be
performed. Patients who experience a CDC-C clinical event, but
who otherwise are maintaining complete viral suppression on
randomized therapy, can be considered to be responding.
Since statistical methods are insufficient for time
to event analyses for noninferiority comparisons, assessing
proportions of patients below the assay limit at a prespecified
time point is recommended for noninferiority trials. However,
the Division prefers the use of Kaplan Meier plots (using the
algorithm in appendix B) to estimate the proportions of
treatment responders at 48 weeks. See Statistical Considerations
below.
·
Clinical endpoints
Adequate and
well-controlled trials showing clinical benefit, as measured by
the occurrence of new AIDS defining events (CDC Class C events)
or death, will continue to be considered acceptable support for
traditional approval. Results of such studies can be described
in the package insert, and could influence the approved
indication(s).
It is imperative
that all new Class C CDC-defined events be thoroughly documented
and analyzed in registrational trials, even those in which the
primary endpoint is virologic. Since clinical events may
sometimes be difficult to assess, the Division recommends that
all study protocols have independent assessment and adjudication
of all Class C events, using case report forms and additional
medical records to fully document their occurrence when
appropriate.
The same
considerations for trials supporting accelerated approval also
apply to traditional approval. It should be emphasized that
studies for traditional approval should be analyzed after the
last patient randomized has completed 48 weeks of treatment (if
still on therapy). Therefore, for many participants, there will
be data points past 48 weeks. As much extended data as possible
should be included and evaluated in the NDA.
When assessing
superiority for time to loss of virologic response, the log rank
test for differences in the Kaplan-Meier (KM) curves, using all
available follow-up data, should be performed. For
noninferiority comparisons, differences in the KM proportions
below the assay limit at a prespecified time point with
associated confidence intervals should be assessed.
Chuang-Stein, C., and R. Demasi, 1998,
"Surrogate Endpoints in AIDS Drug Development: Current Status,"
Drug Information Journal, 32:439-448.
Hill, A.M., R. DeMasi, and D. Dawson, 1999,
"Meta-Analysis of Atiretroviral Effects on HIV-1 RNA, CD4
Cell Count and Progression to AIDS or Death," Antiviral
Therapy, 3:139-145.
Hogg, R.S., B. Yip, and C. Kully, 1999,
"Improved Survival Among HIV-Infected Patients After Initiation
of Triple-Drug Antiretroviral Regimens," CMAJ,160:659-65.
Marschner, I.C., A.C. Collier, R.W. Coombs,
et. al., 1998, "Use of Changes in Plasma Levels of Human
Immunodeficiency Virus Type 1 RNA to Assess the Clinical Benefit
of Antiretroviral Therapy," J Infect Dis, 177:40-7.
Murray, J.S., M.R. Elashoff, L.C. Iacono-Connors,
et al., 1999, "The Use of Plasma HIV RNA as a Study Endpoint in
Efficacy Trials of Antiretroviral Drugs," AIDS,13:797-804.
Palella, F.J., K.M. Delaney, A.C. Moorman,
et al., 1998, "Declining Morbidity and Mortality Among Patients
with Advanced Human Immunodeficiency Virus Infection, N.
Engl. J. Med., 338:853-60.
A. General Considerations
In this guidance a new assay
refers to any assay that has not been approved by the FDA or to
an approved assay that is being used in a manner different than
described in its labeling. The assay that is used in the
clinical trial should be identical to the assay that is used to
assess the performance characteristics. Clinical studies can
use HIV RNA measurements either to quantify the amount of HIV
RNA in patient samples (e.g., copies/mL) or to classify a
patient sample as above or below a specific value. Therefore,
considerations for both quantitative and qualitative uses will
be addressed in the following subsections.
It is recommended, but not required, that
FDA-approved HIV RNA assays be used to support clinical trials.
Currently approved assays are the Roche Amplicor HIV-1 Monitor
standard and ultrasensitive tests and the
NucliSens HIV-1 QT.
However, when experimental and/or investigational HIV RNA
measurement assays are used to support clinical trials,
sufficient data characterizing assay performance should be
provided. This permits an independent evaluation of an assay’s
limitations. Review of assay performance by the Division will
focus on the interpretability of data generated by the assay
with respect to the particular clinical trials in the NDA.
Thus, the Division’s review of assay performance data does not
imply that the given assay is validated, or FDA-approved, for
patient prognosis and/or monitoring. Furthermore, this review
does not imply that the given assay is automatically acceptable
for future clinical trials.
B. Investigational HIV RNA Assays
Assay
design rationale, essential methodology, and performance
characteristics are important components of information
submitted to support new assays. Assay performance
characteristics studies should be conducted on specimens that
are representative of the HIV target subtype (Clade/s), and from
the same tissue reservoirs (serum, plasma, other) assessed in
the clinical trial. Specimen stability (handling, processing,
and storage protocols) data should show no significant change of
HIV RNA material as measured by the assay. Generally, these
data should be derived from protocol-based experiments.
Protocols and quality assurance/quality control information for
the assay should be submitted with the data.
1. Quantitative Assays:
Performance Studies
To support
clinical virology data that rely on quantitative assessments,
the methodology and/or technology used to generate those
data should be adequately described in the application. An HIV
RNA quantitative assay should be able to accurately and
precisely report HIV RNA copy numbers over a defined range.
Assay performance characteristics should include, but not be
limited to, information and/or data that define the assay
accuracy, precision, sample stability, and effects of certain
interfering substances.
Accuracy can be
assessed by calculating the mean of repeated observations of a
given known sample and comparing the mean to the known input
value. Assay accuracy should be determined across the proposed
range of the assay. Precision can be assessed by calculating
the mean square error (MSE) and converting to a percent
coefficient of variation (CV). Precision should be determined
across the proposed range of the assay. The quantitative limit
of the assay will be determined by the lowest input value where
the assay maintains its accuracy and precision. A quantitative
upper limit may be similarly defined.
Ultimately, the quantitative limit should be supported by data
that characterized the assay performance characteristics.
Laboratory strains and unique clinical HIV test specimens should
be used to derive the performance characteristics of the assay.
Each of these test specimens should first be independently and
adequately quantitated (i.e., by comparability to an acceptable
standard) prior to being used to define the performance
characteristics of the new assay.
2. Qualitative Assays:
Performance Studies
Assays
that are used in a qualitative manner should have the ability to
distinguish between known HIV seropositive clinical specimens
and known HIV seronegative specimens with 95 percent
confidence. A threshold or screening cut-off value (qualitative
limit), expressed in HIV RNA copy numbers per mL should be
determined. An assay result would be expressed as either a
³
or < the HIV RNA copy number qualitative limit. However, a
result that is below the screening cut-off value does not imply
that the specimen is HIV negative, it implies only that the
specimen has less virus material than that needed to distinguish
the specimen from a known negative with 95 percent confidence.
Assay performance characteristics should include, but may not be
not limited to, information and/or data that define the assay
range of specificity, range of sensitivity, sample stability,
and effects of certain interfering substances.
The Division
recommends that the range of specificity of the assay be defined
as the 95 percent confidence interval of reported observations
from 500 random seronegative blood or plasma donors. The range
of sensitivity of the assay may be defined as the 95 percent
confidence interval of reported observations from 200 unique
seropositive samples. Each of these seropositive samples should
be quantified by an independent method and then diluted to the
proposed qualitative limit prior to assessing the assay range of
sensitivity. It is expected that the two ranges will not
overlap.
Ultimately, the
qualitative limit can be supported by and derived from the assay
performance data. Laboratory strains and unique clinical HIV
test specimens should be used to derive the performance
characteristics of the assay. Each of these test specimens
should first be independently and adequately quantitated (i.e.,
by comparability to an acceptable standard) prior to being used
to define the performance characteristics of the new assay.
1. For 2 and 3 below, discard all visits with no data. In what
follows, visit means visit with an observed viral load. All
available visits, including off-schedule visits and post-week 48
visits, should be used for the calculation. Data should not be
interpolated for visits or time points with missing data.
2. Subjects who never achieved confirmed HIV RNA levels below
the assay limit (on two consecutive visits) before any of the
following events will be considered to have failed at time 0.
b. Introduction of a new antiretroviral drug to the
regimen. Exceptions may be made for certain prespecified changes
in background therapy where the reason for change is due to
toxicity or intolerance of background therapy and not the study
drug or control.
3. For all subjects who have confirmed HIV RNA levels below an
assay limit (two consecutive visits below an assay limit), the
time to failure is the earliest of the choices below, with
modification specified in 4:
c. Time of confirmed levels above an assay limit.
Confirmed is defined as two consecutive visits greater than an
assay limit or one visit greater than an assay limit followed by
loss to follow‑up.
4. If the time to virologic failure defined above is immediately
preceded by a single missing scheduled visit or multiple
consecutive missing scheduled visits, the time of virologic
failure is replaced by the time of the first such missing visit.
*Subjects who were
never suppressed but were loss to follow-up or introduced new drug
therapy before week 48 should be classified into drug change or
discontinuation categories. Prespecified background drug changes
where the reason for change is due to the background drug can be
ignored in the analysis with consent from the Division.