CDER associate director for medical policy Robert Temple
sharply rebutted 8/2 a charge in Scientific American
that the agency’s 6/05 approval of NitroMed’s BiDil (hydralazine/isosorbide) represented poor science and a regulatory cave-in to race-based commercial opportunism. The charge was made by Hamline University School of Law professor Jonathan Kahn
in an article in the magazine’s August issue entitled “Race in a Bottle.”
In it, Kahn calls BiDil’s development “a tangled tale of inconclusive studies, regulatory hurdles and commercial motives” and contends that “no firm evidence exists that BiDil actually works better or differently in African-Americans than in anyone else. The FDA’s approval of BiDil was based primarily on a clinical trial that enrolled only self-identified African-Americans and did not compare their health outcomes with those of other ethnic or racial groups.”
In an extensive comment to FDA Webview
, Temple said most of Kahn’s complaints had already been addressed by him and CDER cardiorenal drugs director Norman Stockbridge
in the January issue of the Annals of Internal Medicine
. Temple said Kahn is incorrect in his article to dismiss evidence that BiDil does not work very well in whites; there was evidence of this in subset analyses of two
well-controlled VA studies, but “this does not mean that BiDil could not work in some whites, only that it hasn’t been shown to do so and is pretty clearly less effective than it is in blacks.” At the time of approval, Temple was quoted in an FDA news release as hoping that characteristics in whites that could identify responders could be found. “It conceivably could be shown to work some in whites or to work in particular whites,” Temple told us, “but it would take a very large study to show a small effect or to identify the characteristics that lead to a response.”
Temple took issue with Kahn’s assessment of the “data-dredged” positive clinical data NitroMed used from the enalapril active-controlled Vasodilator Heart Failure Trial (V-HeFT 1) study of the hydralazine/isosorbide combination. Breaking it down by race, only 49 African-Americans on hydralazine/isosorbide supported BiDil’s effectiveness — “a tiny number,” Kahn said. He said such data dredging “is fraught with statistical peril: if the number of research subjects in each category is too small, the results for the subgroups may be meaningless.” Temple acknowledged the well-known risks of post-hoc analyses (data dredging) but noted (below) the strengths of the BiDil analysis of the two well-controlled studies.
Kahn thought the weak data were used to support an “ethnic” patent that won’t expire until 2020. “Thus was BiDil reinvented as an ethnic drug,” Kahn wrote, and FDA then approved NitroMed’s African-American Heart Failure Trial (A-HeFT). When this study, conducted entirely in self-identified blacks, produced “strikingly positive” results, reporting a 43% lower mortality rate for BiDil subjects than for those on placebo, it prompted the trial’s suspension and NitroMed’s stock to soar.
Kahn’s assessment of the BiDil trials is wrong, Temple told FDA Webview, “for the same reason in his idea that the drug is ethnic in name only. The issue has been addressed to a substantial extent in the two V-HeFT studies. V-HeFT 1, a negative study overall, showed a nominally significant effect on survival in just 128 black patients, and very little hint of an effect in 324 white patients. The small number of black patients is not a ‘demerit,’ as Kahn seems to think, but evidence of how impressive the effect was. Drugs that work very well don’t need huge sample sizes to show that they work.
“I’m sure Kahn is not aware of it,” Temple added, “but the CONSENSUS study, the first to show that an ACE inhibitor could improve survival, had just 253 patients overall. That did not stop it from being impressive. We did not consider the subset analysis in V-HeFT 1 persuasive, as it was a post hoc analysis. V-HeFT 2, however, strongly supported it. In that study, BiDil was compared with enalapril. The overall result favored enalapril nearly significantly (i.e., the BiDil looked a lot like a placebo). But the enalapril and BiDil groups were almost identical in the black population, while enalapril was nominally significantly superior to BiDil in whites, again, making it look like a placebo. Thus there were two analyses of well-controlled studies that strongly supported a black-white difference. We did not find these data sufficient to approve the drug for blacks (although we were asked to) and we asked for A-HeFT, which of course showed a dramatic effect in self-identified blacks. With the data from the V-HeFT trials as guidance, we estimated that a study to discover a modest effect in whites (say, a third that in blacks) would take something like 16,000 patients.
“Kahn never addresses the issue and apparently believes we should simply have granted a broad approval, even in the face of the quite strong negative evidence in whites. His reason apparently is an objection to what he believes is the implication that race is a ‘genetic’ characteristic. Why that would be so objectionable is not clear to me (there are diseases associated with Sephardic Jews, with Mediterranean origin), but the BiDil labeling and FDA public statements make no such connection and nowhere suggest that BiDil is a ‘pharmacogenomic’ drug.
“BiDil is an example of individualized therapy but individualization can be based on many factors, not just genetic ones. It is clear from ICH E-5 that we contemplate racial/ethnic differences that could have many explanations, genetic being only one. Thus, we know from dozens of studies that blacks respond less well than whites, on average, to ACEIs, ARBs, and BBs, presumably related to their lower rate of high-renin hypertension. Labeling for most of those drugs describes this difference. The drugs are labeled for all populations because there clearly is some effect in blacks and because the drugs work well when combined with diuretics, with the combination about the same in both groups. No one, to my best knowledge, knows why there is a difference in the kind of hypertension in whites and blacks.
“Kahn’s discussion of what he thinks are the commercial reasons for the black population is unevaluable by me because, unlike Kahn, I do not imagine that I know people’s motives well enough to assert what they are. We were, however, asked to approve the drug for all populations before the black population discussion, so someone, at least, wanted a broader claim.
“The fact that BiDil is a combination of two old drugs is wholly irrelevant to the combination’s value, as is the observation that people are substituting generics.
“The discovery of the dramatic survival effect of BiDil is self-evidently important and its value in a black population generally thought to be underserved by available treatments should not be so casually dismissed,” Temple said. “Kahn is incorrect in thinking that V-HeFT 1 showed ‘great promise.’ The overall p-value was 0.09, a nice trend but leaving a lot of work (the p-value for the white population, by the way, was 0.47). Also, V-HeFT was not the study that put enalapril on the map, as Kahn implies. In fact, the design of that study (comparing BiDil with enalapril) was determined by the fact that enalapril was standard therapy (CONSENSUS and other studies).”
Kahn’s brief discussion of data-dredging showed that he doesn’t understand both the risk of this practice and subgroup analysis, Temple said. “The risk is not that the subgroups are too small; the risk is that there are many possible subgroups, posing the problems of multiplicity, which must be taken into account by adjusting the p value, and bias, as the analyses are done with data in hand. As noted above, however, a similar finding in a second study greatly strengthens a subgroup finding.”
In his Scientific American article, Kahn asserts that “researchers have good reason to believe that BiDil would also be effective in non-black patients,” but Temple said there is “no hint” in the article as to what those reasons might be. “In any case,” Temple said, “our law calls for evidence, not belief. In this case, there is quite strong evidence of a differential response and available data in whites shows little, if any, effect vs placebo (V-HeFT 1) and significant inferiority vs enalapril (V-HeFT 2). It is hard to think of a legal or medical reason for extending approval to non-blacks in the face of this evidence. The fact that we approve drugs broadly in the absence of evidence of difference is quite another matter.
“Saying FDA’s approval ‘was based on accepting NitroMed’s argument that the drug should be indicated only for African-Americans because the trial population was African-American’ is a funny way of putting it. We told NitroMed that a study in blacks (A-HeFT) would be an acceptable basis for supporting the effectiveness of the drug in blacks, given the A-HeFT experience. This determination was made well before there was any marketing application.”
As for Kahn’s criticism that the approval of BiDil reflects insufficient appreciation of race as a “powerful and volatile category” that can lead to “substandard medical treatment,” Temple countered: “Not approving BiDil for blacks would represent denial of a life-saving treatment, clearly bad medicine. Whether approval for others would be good or bad is not knowable at present, but no data support broader approval.
“It is surely true that the serious health disparities between races will not be eliminated by drugs such as BiDil, but then no one has ever suggested that they would be. BiDil, or the generic equivalent, simply provides a very beneficial, life-prolonging treatment for blacks with heart failure. This cannot possibly be undermined by the decision not to approve it for others. It is therefore an obviously good thing for the black community. Kahn might perhaps pay some attention to that.”