Report From the Cardiovascular and Renal Drugs Advisory Committee
US Food and Drug Administration; June 15–16, 2005; Gaithersburg, Md
Day 1: Class Labeling for Antihypertensive Drugs
The Advisory Committee was asked to consider the desirability of class labeling for antihypertensive drugs. Antihypertensive drugs, with few exceptions, have no clinical outcomes claims in their labeling because approval of these agents is usually based on demonstration of efficacy at lowering blood pressure, not reducing morbidity or mortality. Initial placebo-controlled trials of antihypertensive agents were conducted more than 2 decades ago. Accordingly, it has been considered ethically inappropriate to perform placebo-controlled studies of modern antihypertensive agents. These factors have resulted in a clinical conundrum in which agents believed to be highly effective at reducing morbidity and/or mortality cannot claim such benefits. The Advisory Committee meeting was convened to consider how labeling should address the relationship between blood pressure and outcome.
The committee felt strongly that not having outcome data available in antihypertensive drugs labels prevents optimal education of practitioners regarding the benefits of important therapeutic agents. Several committee members pointed out that undertreatment of blood pressure remains a major health problem that might be addressed by more descriptive class labeling of antihypertensive agents. The committee agreed that scientific knowledge about antihypertensive drugs is extensive and that lowering blood pressure represents one of the best-studied surrogate outcome measures in clinical medicine. However, the committee expressed some concern regarding “unintended consequences” of class labeling and discussed the need for robust preapproval safety studies and postmarketing surveillance. The industry representative on the committee cautioned that class labeling might represent a disincentive to the pharmaceutical industry to engage in future clinical trials.
Specific Benefits Attributable to Blood Pressure Reduction
The committee reviewed the effects of blood pressure reduction on specific cardiovascular and renal outcome measures. After considerable discussion, the committee agreed that reduction in the incidence of stroke and myocardial infarction is the most clearly attributable benefits of blood pressure reduction. Evidence for reduction in cardiovascular mortality and renal disease is also clearly attributable to blood pressure lowering, but the committee felt that the evidence was somewhat less consistent. The committee agreed that new drugs seeking a hypertension indication should not automatically receive labeling for all the above benefits. Instead, it was suggested that language in the “class label” should describe the “generally expected” benefits for any drug that lowers blood pressure. The committee recommended that labeling statements should not conflict with national guidelines such as the Joint National Committee on Prevention, Detection, Evaluation and Treatment of High Blood Pressure (JNC7).
A strong consensus emerged that most of the benefits of antihypertensive agents are related to blood pressure reduction rather than class-specific or agent-specific effects. However, the committee pointed out that some important differences between classes exist and agreed that the benefits are not necessarily identical among members of any class. For example, better outcomes in preventing congestive heart failure are evident for ACE inhibitors, angiotensin receptor blockers (ARBs), and diuretics, whereas calcium channel blockers appear more useful in preventing stroke. Committee members felt strongly that advice regarding specific choices or preferred agents is best served by guidelines writers rather than drug labels. The committee discussed whether ACE inhibitors and ARBs should be considered as a single class. However, several members pointed out that ARBs lack bradykinin-mediated biological effects, and there was general agreement that it would be inappropriate to treat these 2 categories as the same class of drugs.
The committee also recommended that specific trial data, when adequately reviewed by the US Food and Drug Administration, are always relevant and should be included in the label when appropriate. This approach preserves the incentive of the industry to perform clinical outcomes trials. The committee suggested that labeling for each drug should distinguish whether the specific agent or the drug class contributed to the available outcome data. Most members of the committee felt that single trials are inadequately powered to describe benefits with a high degree of confidence, and many were comfortable with inclusion of some findings derived from carefully performed meta-analyses.
Relevance of Epidemiological Data
The committee felt strongly that epidemiological studies are less reliable and should not contribute significantly to labeling considerations. The committee also suggested that observational data on the relationship between blood pressure and risk should not be included in drug labels. The committee agreed that it is important, to a certain extent, to recognize the convergence of blood pressure with other risk factors (control of lipids, smoking cessation, weight loss, aerobic exercise, etc). However, members agreed that, while relevant, the role of co-existing risks such as hyperlipidemia in determining optimal blood pressure targets should not be included in drug labels. The committee recommended explicitly indicating when a specific agent or class has no available outcome data.
Many committee members felt that not enough information is known about the importance of dosing intervals to support specific claims, although this information is relevant to clinicians, and should be readily accessible in the pharmacokinetic and pharmacodynamic section of the label. There was additional discussion about differences in the impact of circadian rhythms among various medications, but the committee reached no conclusion about the relevance of such information in labels. Some members emphasized that a single blood pressure measurement does not provide an adequate assessment of efficacy. There was general agreement that ambulatory blood pressure studies provide more robust and consistent data and should be strongly encouraged in drug development. There was considerable discussion regarding optimal timing for increases in drug doses in initial hypertension management. Suggestions included increasing doses at 1-week intervals and including a statement regarding the likely need for multiple drugs to achieve good control. However, most committee members suggested that such details should be left to guideline writers rather than the drug label.
Optimal Choices for Initial Therapy
Additional discussion focused on the optimal choice of initial drug therapy. The committee agreed that most data suggest starting with diuretics; several members emphasized that this approach is also recommended in the guidelines. However, there was general agreement that the large Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) study did not support the superiority of any specific drug class for the primary end point. Thus, all 3 drug classes (diuretics, calcium channel blockers, and ACE inhibitors) are indistinguishable, and none should be given a “first-line” preference. Many committee members felt that this issue generally should be left to the guideline writers rather than labeling.
The committee discussed the issue of when to add a second drug, noting that labeling currently recommends starting a second drug only after a single drug has proven inadequate at its highest tolerated dose. The committee agreed that there is no substitute for clinical judgment—some patients reach dose-limiting toxicity at lower doses than others. Because clinicians need to make complex decisions on the urgency necessary to achieve blood pressure control, it would be difficult to convey this principle in labeling.
The committee considered which labels should include the results of ALLHAT. The committee agreed that the large size of ALLHAT contributed greatly to knowledge about blood pressure management, and a description of those results should be included in drug labels. Several members suggested that only the specific agents used in the ALLHAT study should have this information included in their labels. There was discussion regarding the uncertainty whether all diuretics are likely to show the same benefits as chlorthalidone, the agent used in ALLHAT. Similarly, uncertainty exists regarding the interchangeability of all calcium channel blockers. The committee discussed statistical methodology and its applicability based on ALLHAT to support noninferiority claims for studied agents such as amlodipine.
The committee discussed issues related to pediatric studies of blood pressure–lowering therapies. The committee did not support the concept that the agency should require studies of antihypertensive drugs in children before approval for use in adults. Members felt, however, that the agency should promote studies in children by continuing to grant additional exclusivity (a 6-month patent extension) for assessing the effects of antihypertensive drugs in children. In discussing the challenge of placebo-controlled trials in children, some members suggested that 1 to 3 months of treatment with a placebo would be acceptable in some cases.
The committee considered the issues surrounding drugs that increase blood pressure. The committee generally agreed that it is difficult to extrapolate data from blood pressure–reducing agents to predict the effect on clinical outcomes of drugs that increase blood pressure. The committee agreed that the issue is relevant, but concern was expressed about setting a regulatory standard for labeling, especially without direct clinical trial evidence. Members further identified the challenge that some drugs that increase blood pressure are commonly used in patients at lower risk, ie, those with lower baseline risk. The committee advised against “leaping to conclusions” about blood pressure–increasing drugs without solid evidence. However, members suggested that the greater the magnitude of blood pressure increases and the longer the duration of contemplated use, the more likely this effect should appear in a warning.
Day 2: Hydralazine-Nitrate Combination for Congestive Heart Failure in Blacks
The committee was asked to assess whether clinical trials adequately support a claim that BiDil (hydralazine 75 mg plus isorbide dinitrate 40 mg) improves outcome in patients with congestive heart failure (CHF). The committee reviewed 3 trials, the Veterans Administration Cooperative Vasodilator–Heart Failure Trials (V-HeFT I and II) and the African-American Heart Failure Trial (A-HEFT), to provide a recommendation to the agency regarding approval of BiDil for use in CHF patients. In considering this recommendation, the committee reviewed the results of a post hoc analysis of V-HEFT-I and II for the black subgroup. In committee discussions, a strong consensus emerged that post hoc analyses of these 2 studies were supportive but alone provided insufficient evidence for an approval recommendation. Accordingly, the committee considered the A-HEFT trial as the primary basis for evaluation of efficacy and safety.
V-HEFT I compared survival in male veterans with CHF who were randomized to Hydralazine 75 mg plus isosorbide dinitrate (ISDN) 40 mg QID, prazocin 5 mg QID, or placebo. For all patients, there was a trend toward greater survival in the hydralazine-ISDN group compared with placebo (hazard ratio [HR], 0.78; log-rank P=0.09). However, in a post hoc analysis, the differences between active treatment and placebo in black patients (n =56) were larger and statistically significant, showing a mortality rate of 9.7% compared with 17.3% for placebo (P=0.04). In V-HeFT II, enalapril and hydralazine-ISDN were compared. For all patients, there was a trend toward greater survival in the enalapril arm by the log-rank test (P=0.08). For all patients, 2-year survival was greater in the enalapril-treated group (18.0% versus 25.0%; P=0.016). However, in black patients, survival was similar for enalapril and hydralazine-ISDN (HR=1.01; P=0.96). Conversely, for white patients, enalapril resulted in a 48% lower mortality rate compared with hydralazine-ISDN (P=0.02). These findings provided the basis for the A-HEFT trial, which was designed to compare hydralazine-ISDN (BiDil) with placebo in black patients receiving contemporary therapies for CHF.
The A-HeFT trial randomized 1050 patients to BiDil administered TID or placebo. The primary end point was an unusual composite score that included all-cause mortality, time to hospitalization for heart failure, and response to the Minnesota Living With Heart Failure questionnaire. By the sponsor’s and the statistical reviewer’s intent-to-treat analyses, BiDil was associated with an improved composite risk score (P=0.021 by the agency reviewer). All 3 components of the composite showed statistically significant benefits of therapy. All-cause mortality was reduced by 43% (P=0.012), time until hospitalization was prolonged (P<0.001), and quality of life score improved (P=0.003). The trial was terminated early at the recommendation of the Data Safety and Monitoring Committee (DSMB) on the basis of an analysis of mortality data from the first 1014 patients.
Although the trial met its primary prespecified end point, the committee spent considerable time discussing the weight of the statistical evidence and the conditions under which the trial was terminated. Dr Thomas Fleming, a biostatistician, discussed at considerable length the optimal methodology to adjust for interim analyses and early termination and suggested a conservative approach for adjustment of the efficacy probability value. This adjustment yielded a marginally significant probability value between 0.04 and 0.05. The committee discussed whether the strength of evidence provided by this probability value was sufficient to support approval. Typically, 2 trials meeting statistical efficacy are required. Several committee members pointed out that all 3 components of the composite efficacy score showed benefit, contributing to the robustness of the results. Others commented that the efforts to study a minority population with a disproportionate burden of disease also warranted consideration.
Questions From the Agency
Prior to considering questions from the Agency, the committee listened to 13 speakers at an open public hearing, some supportive and others critical of the concept of labeling a drug for use in a specific racial-defined group. Questions from the Agency included a review of the factors responsible for the non-significant results for the sponsor’s pre-specified per-protocol analysis (P=0.46). The committee discussed this discrepancy, specifically the per-protocol analysis exclusion of 60% of the intent-to-treat population. This occurred in large part because of the early termination of the trial, which introduced the potential for bias and reduced statistical power. The committee felt strongly that the intent-to-treat analysis is more valuable and should constitute the principal efficacy assessment.
Subjects enrolled before the second interim analysis, when sample size was reestimated, made up 30% of the total patients and 42% of the events and showed a nominal 7% lower risk of death on BiDil. Subjects enrolled after the second interim analysis showed a 62% lower risk of death on BiDil. The committee discussed these interim analyses and decisions made by the DSMB during meetings throughout the course of the trial. Several committee members warned of the risk of interpreting statistical strength of evidence when interim data are used to refine the original study hypothesis. Specifically, questions were raised about the proper weighting to give to different portions of the sample size, depending on the timing of the interim analyses. Concern was raised about the potential bias introduced during the various unblinded examinations of the data and the decision to stop the trial prematurely.
In interpretation of results, some committee members thought that recognition should be given for the fact that very few trials have been done in a solely black population. The committee discussed the challenges in enrolling an adequate sample of this restricted population. Some participants argued that compromises such as those in question, while affecting the strength of evidence, are necessary in light of the exigency of doing such a trial in a subpopulation by a small company with limited resources. Other committee participants disagreed, citing concern about lowering the bar on strength of evidence. The consultant patient representative provided an additional patient perspective of the advantages that such a trial provided to the black population while expressing appreciation for the complexity of the statistical analysis.
Use of a Composite End Point
There was discussion about the use of a composite end point. Several committee members suggested that this approach might not represent the wisest choice. Most panel members agreed that the robustness of the data was clearly undermined by early termination, pointing out that this always reduces access to valuable data. The committee also discussed the impact of early termination on the ability to assess the contribution of each of the components of the composite end point. The committee discussed the potential toxicity of hydralazine (drug-induced lupus development). Some members suggested that the threshold of concern is decreased when considering a nonlethal side effect for a therapy shown to prolong life. It was pointed out that historical data regarding safety and efficacy for these components do exist and can be considered.
Policy Considerations for Combination Drug
There was considerable discussion regarding the regulatory precedent for development of combination drug products. Typically, when considering such combinations, the agency requires demonstration of the effects attributable to each component. However, the committee expressed comfort in relaxing this requirement for the BiDil combination in view of the well-known effects of both components. Several participants pointed out the ethical challenges associated with new studies of components of a drug combination shown to reduce mortality. The committee discussed the importance of dose information while recognizing that it may be difficult to study multiple doses in trials with clinical end points. The committee agreed that this challenge is particularly daunting as the number of components increases.
Population Likely to Benefit
A-HeFT enrolled only the patients in whom BiDil appeared to provide optimal benefits in V-HeFT I and II, namely self-identified blacks. The committee was asked to consider whether the evidence supported an absence of benefit in white patients. Furthermore, the committee was asked to opine about whether labeling should restrict use to black. Several open public hearing speakers and some committee members pointed out that the US black population is heterogeneous and suggested that it remains unclear whether these differences are genetic, social, economic, or health delivery related. Others commended the agency for requesting this study in the black population. Many committee members agreed that labeling should include information about the A-HEFT study population including only blacks.
There was some committee disagreement about the conclusions supported by A-HeFT for other patient populations. Several committee members commented that the agency has a responsibility to precisely describe the population for which an approved drug has and has not shown benefit. Members expressed that, when the evidence of effectiveness comes from a population that we can define, such as self-identified race, the finding is significant. There was discussion about the future of genomic-based medicine and the hope that pharmacogenomic information will be useful in identifying drug responders in the future. Several members suggested that self-identified race represents a crude but useful surrogate for genomic-based medicine.
Discussions cited additional information about the white population available through V-HeFT I and II, both of which illustrated a different response for white and black populations. From a statistical standpoint, there was discussion about the unfavorable results in the white population in V-HeFT, coupled with proactive exclusion of whites in A-HeFT, leading to the reasonable conclusion that there was less benefit to risk in this population. There was also considerable discussion among committee members regarding the evidence for efficacy in subgroups classified by the severity of CHF. The committee concluded that A-HeFT demonstrated benefits for BiDil only in patients with New York Heart Association class III heart failure.
The committee voted unanimously in favor of approval of BiDil for the treatment of heart failure. One of the 9 voting participants thought that the approval should extend to the general population. A second member expressed concern about labeling based on self-designation of race and the heterogeneity among this population. Other members were not as conflicted, citing the disproportionate burdens of heart failure in blacks and the challenges of performing clinical trials in this subgroup. Other comments included the importance of developing effective treatment in the black population in light of disparities in health care. There was additional discussion about the statistical strength of evidence of A-HeFT, with general agreement about limitations in interpreting the individual components (heart failure hospitalizations, quality of life, and mortality). Despite these concerns, the committee felt that the consistency of favorable effects on components of the end point helped to overcome the limitations of using a single trial for approval. Further recommendations included encouraging the agency to emphasize, in labeling, that the trial was performed in the context of optimal use of ACE inhibitors and β-blockers, providing clear direction that this treatment should be considered an adjunct to standard therapy.
Dr Nissen was a consultant to AstraZeneca, Atherogenics, Lipid Sciences, Wyeth, Novartis, Pfizer, Sankyo, Takeda, Kowa, Isis Pharmaceuticals, Sanofi-Aventis, Novo-Nordisk, Eli Lilly, Kos Pharmaceuticals, Glaxo Smith Kline, Forbes Medi-tech, Roche, and Merck–Schering Plough; has given lectures at the invitation of AstraZeneca and Pfizer; and has received research support for clinical trials from AstraZeneca, Eli Lilly, Takeda, Sankyo, Sanofi-Aventis, Pfizer, Atherogenics, and Lipid Sciences. All honoraria related to these relationships are paid directly to charity by the sponsors.