GlaxoSmithKline is breaking out the data to respond to the Nissen and Wolski NEJM paper on the possible cardiovascular risks of Avandia (rosiglitazone). In a letter published by The Lancet (PDF), the company's chief medical officer, Ronald Krall, defends the drug (and the company):
"GlaxoSmithKline did similar meta-analyses in 2005 and 2006 and found hazard ratios in the same direction as Nissen and Wolski. However, all these results are highly dependent on the methods used and the studies included, given the small number of events reported. For example, the actual number of myocardial infarctions in the Nissen and Wolski meta-analysis yields a very low frequency of events (0·6%), and the absolute difference in rates of myocardial infarctions between rosiglitazone and controls is less than 0·1%.
These observations support a view expressed by Nissen and Wolski them-selves: “a meta-analysis is always considered less convincing than a large prospective trial designed to assess the outcome of interest.”
He then goes back over the data in the three large trials that bear on the question. Reanalyzed data from the ADOPT study still do not show a statistically meaningful cardiovascular risk for rosiglitazone versus the other two diabetes drugs in the trial (metformin and glibenclamide). (There's no placebo group - this is one of those head-to-head comparisons of a drug versus its strongest competitors, a type of study that some people believe never takes place). The second completed study, DREAM, looked at co-administration of rosiglitazone and the ACE inhibitor ramapril. There were four groups - placebo only, rosi and placebo, ramapril and placebo, and rosi plus ramapril. The first three showed no difference in cardiovascular events, but the last one did, for unknown reasons.
These two studies are in the Nissen/Wolski meta-analysis, of course, but as I noted originally, it was the sum of the smaller studies that gave them their cardiovascular warning. But when the statistically less powerful trials show one thing that isn't borne out by the larger ones, the issue is (at the very least) still in doubt. The letter also points out that the company's database mining of managed-care patients taking rosi has shown no increase in cardiovascular risks.
Other controlled studies are ongoing, the (now highly awaited) RECORD and another one called ACCORD. Both are designed from the start to address cardiovascular outcomes (which are a major complication in diabetic patients). Krall's letter lifts the veil a tiny bit on RECORD, saying that the independent review board has now completed an interim analysis of its cardiovascular data and concluded that the trial should continue. This would not be the case, you'd have to presume, were the numbers to clearly show increased CV deaths in the treatment group.
My take on this is that the company has a pretty strong case so far, certainly strong enough to wait for the ongoing trials to settle the issue. What never fails to disappoint me, though, is the way that stories like this are jammed into ready-made templates. Depending on the editorial writer, the appearance of the NEJM paper became "FDA Corrupt, Broken: Snores While Dangerous Drugs Kill Thousands", or "Giant Drug Company Sells Heart Attack Poison, Doesn't Give Hoot". Or maybe just "Drug Approval System Completely Broken - Again".
Now, Steve Nissen does sound the alarm a lot, but I have no doubt that his intentions are honorable. His paper, to me, was the equivalent of saying "Hey, you people may have a problem here. Did you know that?" GSK's response, then is "Yeah, we've looked at that, too, but we don't see it. Are you sure your numbers are good?" Meanwhile, the studies which should answer the question for good are already years into their runs. If this is our standard for a broken drug approval system, we've certainly become mighty fastidious over the years. For what it's worth, The Lancet agrees.