Not until 2006: Lawsuit on interpretation of Paxil's suicide risk in drug trials

An inappropriate analysis of clinical trial data by researchers at GlaxoSmithKline obscured suicide risks associated with paroxetine, a profitable antidepressant, for 15 years, according to court documents released last month. Not until 2006 did GSK alert people to raised suicide risks associated with the drug, marketed as Paxil and Seroxat.

An analysis of internal GSK memos and reports, which were released to US lawyers seeking damages, suggests that the company had trial data demonstrating an eight-fold increase in suicide risk as early as 1989. Harvard University psychiatrist Joseph Glenmullen, who studied the papers for the lawyers, says it's "virtually impossible" that GSK simply misunderstood the data - a claim the company describes as "absolutely false."

Glenmullen's report rests on documents obtained by lawyers in Los Angeles, who are bringing around 30 cases against GSK linking suicides and suicide attempts to the use of Paxil. The report was under seal at a district court in Sacramento, California, until 18 January, when the judge agreed to make parts of it public.

The analysis focuses on the "washout" phase preceding a trial, when subjects stop taking most or all medications to avoid confusion with results from the trial itself. Because the washout occurs before patients randomly receive either the drug or the placebo control, adverse events during this time can't be attributable to the trial and so are seldom if ever included in final results.

However, GSK researchers submitting data on Paxil to the US Food and Drug Administration in the late 1980s and early 1990s included suicides and suicide attempts from the washout period in the results for the placebo arms of trials, but not from the Paxil arms. Glenmullen alleges that these extra "placebo" suicides negated suicides attributed to Paxil in the trials, making the drug appear safer than it really was. He says that if the washout results had been excluded, the data would have showed that Paxil increased eight-fold the risk of suicidal behaviour in adults.

GSK spokeswoman Mary Anne Rhyne says inclusion of the washout data "was intended to present the full picture of events that occurred in all phases of the clinical trials - starting from the time patients were enrolled, before they were randomised." She says that even without the washout data, Paxil still came out as safe as the placebo in this trial. She accused Glenmullen of incorrectly analysing the data to reach the opposite conclusion, but didn't respond to a request for numerical proof that Glenmullen's verdict was wrong.

Glenmullen suggests that the FDA would have acted differently had the use of the washout data been made more explicit. Rhyne says that material still under seal shows the FDA to be fully aware of how the washout data was being used. But Glenmullen quotes Martin Brecher, the FDA official who reviewed Paxil's safety, as agreeing during a pre-trial hearing that the use of the washout data was "scientifically illegitimate."

Independent researchers say it was wrong to use washout data as GSK did. "I can't imagine circumstances in which it would be appropriate," says Bruce Psaty of the University of Washington in Seattle.

Advice: Weigh the tradeoffs carefully in deciding whether to take antidepressants.

Browse for similar stories in our index at the very bottom of this page, or read an antidepressant adverse drug reaction story.

Thanks to Helen Haskell for sharing Jim Giles' source story in the Feb. 6 issue of the New Scientist.