The Rosiglitazone Story — The FDA Meeting Summary
Clifford J. Rosen has written a perspective in the New England Journal of Medicine summarizing his views on the FDA Advisory meeting on rosiglitazone (a.k.a. Avandia), which he was chair of.
I agree with his main conclusion, that rosiglitazone is probably not as good as some alternatives, but I have to say, his article’s stance on observational studies and meta-analyses was rather poorly presented. The main data he cites for showing that pioglitazone is better than rosiglitazone for cardiovascular risk is Gerrits et al., basically an observational meta-analysis. On the other hand, he essentially seems to dismiss the observaitional studies by Wellpoint, a health care insurer, which was much larger. I’m giving him the benefit of the doubt that he’s being self-consistent, but the way the article is written wrongly portrays his references. Consider that he says,
Moreover, we are facing a troubling paradox: preliminary data that were presented at the meeting and published by Gerrits et al. suggest that among the thiazolidinediones — a class of drugs that has been shown to improve metabolic control — rosiglitazone may increase cardiovascular risk whereas pioglitazone may reduce it.
Later on, when he mentions the Wellpoint study, he concludes,
The contrasts among the levels of evidence and the results regarding the safety of rosiglitazone raised new questions about relative and absolute risks but also highlighted the weaknesses of observational studies examining events that are common and whose rates are likely to be increased only slightly by a given drug, even in a large cohort (such as that used by WellPoint, which comprised 160,000 patient records).
Note that he never mentions that the Gerrits study is essentially an observational study, too, comprised of data pulled from a large health care insurer. So, the way he wrote the article casts a falsely bad impression on the studies showing little increase in cardiovascular risk for rosiglitazone compared to the alternatives, while not casting the same sort of critical eye at the studies that seem to support his conclusions.
A better (and I find, more convincing) argument would have been that pioglitazone has been demonstrated, in a clinical trial, to have fewer cardiovascular events than placebo, whereas rosiglitazone has not been shown conclusively to have less or more. So, if faced with a choice, it’s easy to recommend pioglitazone over rosiglitazone.
That argument would be much clearer than what he did, which was to simply cite a whole slough of meta-analyses and observational studies (the only non-meta-analysis data he cites at all are two clinical trials, both basically mum on cardiovascular risk), cast doubt on some of them, and then leave the reader to weigh one against the other while making the (unwarrented) conclusion that “a new ‘wonder drug,’ approved prematurely and for the wrong reasons by a weakened and underfunded government agency subjected to pressure from industry, had caused undue harm to patients.” Weighing meta-analyses that conflict with each other is for Ph.D. level statisticians, and I (nor, do I think, most of the readers of the NEJM) am not up to it. I’m still not terribly convinced that rosiglitazone has absolutely hurt patients; I think it’s premature to say so (otherwise, why not pull it from the market?), but certainly there are much more effective alternatives. Clifford Rosen just didn’t make a good case for it.
