Without a randomized controlled trial we may never really know the answer to that question though maybe the complex systems / molecular biology people will get it worked out one of these days.
We will however get to find out if the data on which the RECORD trial (Rosiglitazone Evaluated for Cardiovascular Outcomes in Oral Agent Combination Therapy for Type 2 Diabetes - don't ask me how that works out to "RECORD") was based had been properly characterized before being analyzed. In "Regulatory Action on Rosiglitazone by the U.S. Food and Drug Administration", the announcement of new tighter restrictions on Avandia and the end of the TIDE trial, it was also disclosed that the FDA "is requiring an independent readjudication of end points at the patient level" of the RECORD trial's data. While the RECORD trial apparently established the noninferiority (NI) of Avandia versus the standard treatment, claims of design flaws that biased the results in favor of Avandia, data cherry-picking, and worse were subsequently raised (see "The Doctor Doth Protest Too Much, Methinks") .
What all this dredging and now reclassification-to-be-followed-by-redredging will reveal is anyone's guess and at best it will only lead to a posthumous vindication. The more important question is how will the battle between those who hold epidemiology to be a source of actionable knowledge and those who hold it to be a source of testable hypotheses play out. On one side of the debate, at least in the Avandia context, are idealistic academics. Their presentations to FDA were filled with lots of exclamation points, suspicions and worries about the rich and powerful exploiting the poor, sick and vulnerable. On the other side are the practitioners, pragmatists and industry scientists. They (to ruin my metaphor) played Spock to to the anti-Avandia group's Captain Kirk.
Will Avandia wind up with a pardon? If so, this episode may in hindsight turn out to have been the last stand for a methodology that for 40 years reliably generated the "knowledge" needed to support a particular world view; a methodology which has now gone decades without a big hit (e.g. asbestos). If not, and should Dr. Nissen's view prove to be the accepted one, the conclusion that the approval of Avandia was "one of the worst drug safety tragedies in our lifetime" resulting in up to 47,000 needless heart attacks could reignite a "science" that for three decades ushered in wave after wave of mass tort litigation. Whichever way it turns out I'm guessing it'll be Avandia's lasting legacy.