Monday, June 25, 2007

Burying Data, Chapter 317

In the sad yet unsurprising category, Ed at Pharmalot has two reports (1, 2) on Pfizer burying negative data about Lipitor. How would they do such a thing? Well, Pfizer conducted a study comparing Lipitor to Zocor and found positive results. Naturally, a favorable press release was issued, which opened by stating...

A retrospective analysis of a large U.S. managed care database showed that patients who took Pfizer's cholesterol-lowering medicine Lipitor® (atorvastatin calcium) Tablets had a significant 14 percent reduction in the risk of cardiovascular events, including heart attacks and strokes, compared with patients who took simvastatin.

An "independent" academic was also featured in the press release, stating...

"This analysis is important for physicians, employers and formulary directors at managed care companies who are making real-world treatment decisions for patients," said Dr. Robert Vogel, an author of the study and professor of medicine at the University of Maryland. "This further supports the cardiovascular benefits previously seen with Lipitor."

Well, turns out the analysis was wrong. The difference between Lipitor and Zocor was not statistically significant -- someone made an error during data analysis. No big deal that someone screwed up the analysis in my book -- it happens from time to time. Ed noted that, of course, there was no press release accompanying the new, less favorable data analysis. Though Lie-Zer, er, Pfizer has issued platitudes regarding its commitment to transparency, this behavior does not exactly engender faith in their commitment to honesty, eh? Here's Ed Silverman's rant...
So Jeff [Kindler, Pfizer CEO], what happened? If the marketing team is no longer in charge of everything, why hasn’t Pfizer gotten around to issuing that bad news press release? Your company has had 11 days to correct its mistake. Will that happen, or are accountability and transparency just convenient buzzwords?
Well said. I'd like to ask similar questions regarding the convenient release of study results that found Zoloft was ineffective for posttraumatic stress disorder ten years after the study was conducted, yet positive trials for Zoloft in PTSD took much less time to get published. How Pfizer handled other data regarding Zoloft is also dubious.


Unknown said...

I agree with you that they should have issued a press release. I do think you should mention that they filed a Form 8-K noting the error.

Also, their secondary endpoint was still statistically significant.

The overriding issue for me, however, is that some drug company trolling through some database and supposedly statistically adjusting for the differences between the atorvastatin and simvastatin groups just isn't that useful.

M. Mann

CL Psych said...

Yep, I should have noted the SEC filing. Trolling through databases is where one can find anything one wants at some point, and ignore anything one does not want to find. So much for a priori analysis and publishing all relevant data.