Critics of the FDA's approval of BiDil for treating heart failure in self-identified black persons have taken 2 positions: 1) Approval for blacks was premature because further testing should have been performed before approval to discover what characteristics in black and white patients might predict responses (1, 2), and 2) the FDA should have approved BiDil, but for the broader population (3). We note that Dr. Kahn, in his letter, states the latter is what he meant—not that there should have been no approval.

Our paper focused on why delaying approval would have been intolerable, given the dramatic effect shown in the black population and the extreme difficulty of assessing effects further in white patients or discovering clinical predictors of response. But approval for the black population alone was not based, as Bibbins-Domingo and Fernandez assert (2), on the argument that BiDil's approval was urgently needed to address racial disparities in health. Approval for all patients would have accomplished that end equally well. We approved the drug for the black population because the evidence did not support broader approval, a conclusion also reached by 7 of 9 Cardiovascular and Renal Drugs Advisory Committee members.

We are entirely aware of the risks of subset analysis and have regularly endorsed Yusuf and colleagues' warnings about such analyses (4). But a replicated subset finding is not the same as a single subset finding, and the findings in V-HeFT I and II powerfully indicate a differential effect in black and white patients. We clearly did not assert that the absence of an effect in white patients has been proved, only that there is good evidence of a differential effect and that the effect in white patients, if any, is far smaller. Bibbins-Domingo and Fernandez argue that the findings in V-HeFT II, essentially identical outcomes with enalapril and hydralazine hydrochloride–isosorbide dinitrate in black patients, could have reflected a poor response to enalapril (another racial difference in response?), but that hypothesis doesn't support their view. What was striking in V-HeFT II was the evidence of no effect of hydralazine hydrochloride–isosorbide dinitrate in white patients, in whom enalapril was known to work. Moreover, the possibility that the equal effect of the treatments in black patients represented no effect of either drug is thoroughly rebutted by A-HeFT. The placebo-controlled V-HeFT I also showed a strong difference in effect, with a nominally significant effect versus placebo in the black population, an effect similar in size to that seen in black patients in A-HeFT.

The postulated adverse consequences of our approval alleged by Bibbins-Domingo and Fernandez are hard to follow. Whether physicians should try the combination in white patients is not for us to say. Certainly there is no legal bar to such off-label use, and labeling does not contraindicate that use. Although some subset of the white population may prove to respond, without any evidence of this in studies that detected drug effects in a responsive population, we don't believe labeling should recommend it.

Our threshold for making subset distinctions is high because of the potential for error, despite the current great interest in meaningful individualization of therapy. But errors can have 2 directions. When evidence of a difference is strong, it seems irresponsible to dodge its implications. The LIFE study (Losartan Intervention For Endpoint reduction in hypertension) (5) illustrates this. On the basis of that study, losartan (which was superior overall to atenolol in reducing stroke) was given a specific labeling claim for reduction of stroke. Because the finding was reversed in black patients, with atenolol nominally significantly superior to losartan, labeling notes that the LIFE study provides no evidence that the benefits of losartan apply to black patients.

Dr. Kahn focuses mainly on commercial matters that are not pertinent to regulatory approval, but he touches on 2 scientific issues. The bioequivalence concerns he notes are pertinent to approval of generic drugs but do not weaken the inferences that can be drawn from differential effects in white and black patients seen in V-HeFT I and II. Dr. Kahn also proposes a "reasonably powered" study in patients with "relevant" biological characteristics, such as nonischemic versus ischemic heart failure, diabetes, or alcohol abuse. It is hard to know what he has in mind, but any study that sought to examine effects in many subgroups of this heart failure population, in whom we have no idea about which biological characteristics are relevant (and Kahn left out hypertension and cardiomyopathy as causes), would be vast and, without some plausible hypothesis, utterly unappealing to a sponsor. Indeed, what he is proposing poses the very problem that Bibbins-Domingo and Fernandez are concerned about.