Tuesday, February 23, 2010

The Secret Avandia Tapes: Comedy or Smoking Gun?

Gardiner Harris broke this story yesterday in the New York Times:

My own opinion is that this is more of a bit of Pharmagossip than anything really substantive that moves forward our understanding of the ethical issues at the medicine/Pharma interface; but who can withstand the attraction of a really juicy bit of gossip?

According to the timeline provided by Cleveland Clinic cardiologist Dr. Steve Nissen in the last post, it's now May 10, 2007. On May 1, Nissen and his colleagues submitted a manuscript of a meta-analysis showing excess cardiac risks with the diabetes drug Avandia (rosiglitazone) to the New England Journal of Medicine. On May 17, highlighting the public health importance of the findings, NEJM will publish the paper on line. I gather from the story that Nissen knew by May 10 that NEJM was about to accept the paper if they had not already formally accepted. What Nissen did not know on May 10 was that an NEJM reviewer, Dr. Steven M. Haffner of the University of Texas-San Antonio medical center, had faxed a copy of the manuscript to GlaxoSmithKline, with whom he had a financial relationship (in violation of all the rules of journal publishing and academic peer review--see previous post).

On May 10, four executives from GlaxoSmithKline, the manufacturers of Avandia, met with Nissen in his office. Presumably Nissen knew that they knew that a paper had been submitted to NEJM. Nissen did not know that they knew the paper's exact contents. Nissen knew that they would not like the conclusions of the paper. Nissen did not know that GSK had internally crunched the numbers and that their own statisticians agreed that Nissen et al. had got the numbers right.

Now, add to the who-didn't-know-what that Nissen decided to tape the meeting without telling the GSK execs--which is legal in Ohio, we are told. Harris says that Nissen "fear[ed] he would face pressure and criticism from executives" and therefore decided to record the meeting. I don't know if Nissen had read John Le Carre's novel, The Constant Gardener, or had seen the movie, but if so, he might have feared further that if GSK had its way, he'd soon be sleeping with the fishes. For whatever reason he decided on the recording gambit.

GSK told the Times they were "disturbed" to learn that the meeting had been recorded without their knowledge. GSK did not indicate that they were disturbed that their lackey had violated all the rules of journal peer review by sending them a fax of Nissen's paper.

So, with all this high drama, what do we now know from the tape? Well, not all that much, but a few nuggets.

First, GSK tried to buy off Nissen from publishing his paper by promising that important new data would be available very soon, and hinting that Nissen might look stupid if he published his own, incomplete data before awaiting the new, definitive study. (Again, no hint that they had already internally confirmed the correctness of Nissen's numbers.) Specifically, on June 6, 2007, GSK published an interim analysis of another trial called RECORD. GSK's chief medical officer, Dr. Ronald L. Krall, is captured on the tape playing a "what if" game with Nissen. Nissen's group had calculated that the hazard ratio for excess heart disease due to Avandia was 1.64. "Let's suppose," said Krall, "RECORD was done tomorrow and the hazard ratio was 1.12."

The lesson here is that when RECORD was later published, the hazard ratio was 1.11. Krall would not have known this fact on May 10 if the study protocol had been followed correctly; as of that date the study would still have been blinded. So the tape suggests that GSK was playing fast and loose with the study methodology, looking for ways to cover its rear end with the disclosure of Avandia's risks imminent.

Krall's next "what if" move was--"Suppose we did [a] patient-level analysis and it looked very different from what you have?" Nissen shot back that he agreed that a patient-level analysis, returning to the actual raw data, would be superior in reliability to his meta-analysis that relied only on grouped study data. But he also had enough confidence in his meta-analysis numbers to claim that it was not mathematically possible that the patient-level data could deviate from his calculations by a significant degree--certainly not enough to remove evidence of a serious risk. In any event, if GSK ever actually ran the patient-level data and were not simply bluffing, those data have never been made public.

No comments: