Fourteen years ago, a leading drug maker published a study showing that the antidepressant Paxil was safe and effective for teenagers. On Wednesday, a major medical journal posted a new analysis of the same data concluding that the opposite is true.
That study — featured prominently by the journal BMJ — is a clear break from scientific custom and reflects a new era in scientific publishing, some experts said, opening the way for journals to post multiple interpretations of the same experiment. It comes at a time of self-examination across science — retractions are at an all-time high; recent cases of fraud have shaken fields as diverse as anesthesia and political science; and earlier this month researchers reported that less than half of a sample of psychology papers held up.
“This paper is alarming, but its existence is a good thing,” said Brian Nosek, a professor of psychology at the University of Virginia, who was not involved in either the original study or the reanalysis. “It signals that the community is waking up, checking its work and doing what science is supposed to do — self-correct.”
The authors of the reanalysis said that many clinical studies had some of the same issues as the original Paxil study, and that data should be made freely available across clinical medicine, so that multiple parties could analyze them.
The dispute itself is a long-running one: Questions surrounding the 2001 study played a central role in the so-called antidepressant wars of the early 2000s, which led to strong warnings on the labels of Paxil and similar drugs citing the potential suicide risk for children, adolescents and young adults. The drugs are considered beneficial and less risky for many adults over 25 with depression.
Over the years, thousands of people taking or withdrawing from Paxil or other psychiatric drugs have committed violent acts, including suicide, experts said, though no firm statistics are available. Because many factors could have contributed to that behavior, it is still far from clear who is at risk — and for whom the drugs are protective.
The maker of Paxil, GlaxoSmithKline, said it stood by the original conclusions, given what was known at the time. The company also noted that it had provided all the data for the new analysis, “an unprecedented level of data sharing that speaks to our absolute commitment to transparency.”
The team that reanalyzed the data included several longtime critics of the original study, including a psychiatrist who has been a paid expert witness in lawsuits against Glaxo. But with the company’s permission they spent about a year poring over Glaxo’s files on the study, combing through summaries, internal trial reports and a sample of what is known as patient-level data, the detailed descriptions of what happened for each person in the original trial.
The original study began in the late 1990s, when antidepressant makers started testing the drugs in young people. Antidepressant trials are an extremely tricky enterprise, in part because anywhere from a third to more than half of subjects typically improve on placebo. Choices about how to measure improvement — and how to label side effects — can make all the difference in how good a drug looks.
And so it was in the Paxil study. The original research, led by Dr. Martin Keller of Brown University, tracked depression scores over eight weeks in three groups of about 90 adolescents each, one taking Paxil, one on placebo pills and one taking imipramine, an older generic drug for depression. The Paxil group did no better than the other two groups on the study’s main measure — a standard depression questionnaire — but did rate higher on other, “secondary” measures, like another scale of mood problems, the authors reported.
Researchers consider secondary measures like these as akin to circumstantial evidence, potentially meaningful but not as strong as the primary ones.
The drug’s manufacturer — SmithKline Beecham, now a part of GlaxoSmithKline — submitted the trial and others to federal regulators, who told the company that the drug was on track for approval for use in adolescents.
But critics began picking apart the study soon after it was published in the Journal of the American Academy of Child and Adolescent Psychiatry, charging that it was not at all convincing, and that serious side effects had been played down.
Dr. Keller and his co-authors responded at the time that the testing of antidepressants in young people was a new area, that the paper was upfront about its use of secondary measures and that charges of bias were baseless. Glaxo stood by the team’s conclusions.
Prescriptions of antidepressants to young people surged in the wake of the study, increasing by 36 percent between 2002 and 2003, according to one analysis. The growth slowed after regulators ordered the black-box warnings on labels.
The reanalysis delivers the same critique as before — no clear effectiveness, and mislabeling of serious side effects — only from the inside, using voluminous data from the study itself. Its authors include Jon Jureidini, of the University of Adelaide in Australia, an early critic, and Dr. David Healy, a professor of psychiatry at Bangor University in Wales, who, with the help of a BBC reporter, Shelley Jofre, first noticed and made public the serious side effects in the early 2000s and who has acted as an expert witness against Glaxo.
In an interview, Dr. Healy said that five of six adverse events labeled “emotional lability” in the original study involved suicidal thinking or behavior but were not presented as such. The patient-level files provided detail on what, exactly, happened in those cases: One teenager was hospitalized after taking 80 Tylenol tablets. Another overdosed on Paxil and other medications after a “disagreement with her mother.” Others suffered “severe suicidal ideation,” and one was “admitted due to severe suicidal and homicidal ideation, towards his parents.” No completed suicides occurred.
“When I first heard about this new analysis, I suspected it might be biased,” said Dr. Erick Turner, an associate professor of psychiatry at Oregon Health & Science University, who was not involved in the report. “But I did my own analysis and found, as they did, no significant effect.”
Dr. Turner added, “The only way to really know about adverse events is to dig into the patient-level data.”
Dr. Keller and his co-authors strongly disputed the reassessment of their work. In a joint statement, he and his team said they incorporated secondary measures before knowing which patients were taking Paxil and which were not — not afterward, as the new analysis claims, for some of the measures. “In summary, to describe our trial as ‘misreported’ is pejorative and wrong,” they conclude.