August 17, 2006

Vioxx

Merck's previous win in the Vioxx suit gets thrown out because the judge was concerned about the new criticism of the NEJM study.

What happened is an idiot's guide to forensic computing.  Greg Curfman, executive editor of NEJM, was going to give a deposition in the trial of Frederick Humeston, an Idaho postal worker (or he was just curious about the data after Vioxx was pulled-- depends on which story you read) and so pulled the manuscript.  Back in 2000 you'd submit a paper copy and a disk; NEJM says they worked off paper, so the first time they looked at the disk was Oct 5, 2004 (days after Vioxx was withdrawn.)

 

Here's the fishy part: on the disk was a table called "CV events," which was blank. 

Time stamps in the software indicated that the table was deleted two days before the manuscript was submitted to The New England Journal on May 18, 2000. "When you hover the cursor over the editing changes, the identity of the editor pops up, and it just says 'Merck,'" Curfman says.

What's so terribly misleading about this and NEJM's "Expression of Concern" is this statement:

We determined from a computer diskette that some of these data were deleted from the VIGOR manuscript two days before it was initially submitted to the Journal on May 18, 2000.

This isn't true.  First, the missing MIs were never in the table to begin with.  Second, the table was deleted, but the data itself was still in the paper. 

 

Now it is obvious the study attempts to minmize the thromboembolic risks.  What do you expect from an academic study? Let me assure you-- if you think drug reps are biased, go find yourself a professor.   So I acknowledge the criticism that the study is misleading.  But.
 

But it's the social policy angle that gets me, the moralistic high ground of journal editors who are far worse than study authors.  The gateway to hell is peer reviewed. 

The article says Curfman was deposed by plaintiff's lawyers.  Was Curfman paid by them?  It doesn't mean he's biased, but if you have to disclose Pharma sponsorship, don't you think you should disclose lawyer sponsorship?  (and I am looking to find out if he was indeed paid.)

As I have absolutely no interest whatsoever in the actual outcome of these trials-- my interest is really about how doctors butcher science and promote themselves to senators-- but, we should take a look at what this revelatory missing data says.

What they found was that with the inclusion of the missing data, the rate of heart attacks would have been 5 times greater than naproxen, not 4 times.  0.5% vs. 0.1%.

Just to put this in perspective, of course, you should know that the missing data was three more heart attacks, raising the number of patients with MI from 17 to 20 (out of 4000+ patients), vs. 4 in tha naproxen group.

BTW, "five times" and "four times" may sound like big differences, but they do not even approach statistical significance in this study. 

BTW, strokes were the same in both groups.  Not that anyone cares, of course.