The Human Cost of a Misleading Drug-Safety Study

A reexamination of old data for Paxil found that the antidepressant is more dangerous than the authors let on. How much harm has been done in the 14 years since it was published?

Sara Bostock once sent me a picture of her and her daughter Cecily in happier times, and it’s a happy shot indeed: mother and daughter, 40-something and 20-something, outdoors in the sun, looking radiant. With their beaming smiles—the same smile, really, for they look so alike—they appear thrilled to be mother and daughter.

One night in 2002, a couple years after that photo was taken, Sara woke in the night thinking she’d heard a bump in her kitchen. When she went to investigate, she found her 25-year-old daughter on the kitchen floor in a pool of blood. Next to Cecily on the floor was a large and bloody chef’s knife. In her chest were two knife wounds. One was shallow; the other was fatally deep.

Sara has always thought that her daughter was killed that night by an antidepressant called Paxil. Cecily, a bright, generally cheerful Stanford graduate, had been taking Paxil for two weeks. Five months before, she had become moderately depressed, and, as Sara would recall it, entered a psychiatric system newly enamored of chemical models of depression and chemical solutions. In search of the drug that would work well, doctors had put her on one after another that worked badly. Of these, Sara says, Paxil was the worst. It made Cecily more and more agitated, increasingly unlike herself. Finally she ended it in what Bostock has called a death “completely unexpected, out of character and violent.”

Cecily’s doctor was hardly alone in prescribing Paxil that year. A year before, in 2001, a much-publicized paper described a clinical trial that showed Paxil to be safe and effective in teenagers as well as adults. Study 329, as it became known, helped spur a huge increase in Paxil prescriptions. In 2002 alone, over 2 million prescriptions were written for children and teens, and many more for adults. (As a 25-year-old, Cecily Bostock was, from a neurophysiological perspective, on the cusp between adolescence and adulthood.) Did Cecily’s psychiatrist choose Paxil because of Study 329? It’s hard to believe the answer is not at least partly yes.

The study is now again in the news, as a new reanalysis of the its original data—including about 77,000 pages of formerly inaccessible patient records—shows that Paxil was neither effective nor safe. The reanalysis, published in the scientific journal BMJ, found that the study, underwritten by the drug’s maker, GlaxoSmithKline (or GSK), created a false picture of safety partly by misclassifying suicidal acts (such as taking 80 Tylenol) as less-alarming behavior or side effects. Other researchers who have looked at Study 329’s data have concluded likewise.

Count this as shocking but unsurprising, for GSK has been admonished and fined many times since 2001, including once for $3 billion, for exaggerating Paxil’s safety and marketing it improperly for use in adolescents. Yet this BMJ study deals an especially sharp blow, for it’s only rarely that researchers are able to crack open the tightly sealed file cabinets of drugmakers and look at raw trial data. This illustrates why they want to do so: It appears to be a direct demonstration of how a company and researchers can misinterpret the data to make a bad drug look good.

Benedict Carey covers this story well in yesterday’s New York Times, though I disagree with his assertion that this reexamination of original data is something novel. Reexamining important data is part of science. It was delayed here, as it has been for most drug-trial studies, because drug companies are often allowed to keep their raw data secret. Even when outside researchers can get to the data —and it’s a wonder they managed to do so in this case—it’s both hard to find money for independent examination and hard to conduct that examination well. The BMJ study happened because of a special program called RIAT (for restoring invisible and abandoned trials) to spearhead such efforts.

Perhaps the most essential read, aside from the study itself, is the BMJ’s feature examining the study’s history. The article is deeply reported, sad, riveting, and damning. It condemns not just the company and the original study but an entire ecosystem of researchers, journals, and universities that have resisted the sort of correction the BMJ paper seems to offer. The most chilling part is that even though GlaxoSmithKline has been found to have acted questionably again and again in its promotion of Paxil, many of the institutions and researchers involved in the study and subsequent promotion of the product have remained unapologetic and are silent or unresponsive to efforts to have the original paper retracted or the record otherwise corrected.

At one point the article describes the increasing pressure put on Andrés Martin, a Yale professor and the current editor of the journal that published the 2001 study, to retract the paper. Martin has repeatedly stood his ground, stating in 2012, after GSK was fined $3 billion, that his inquiries had found “no basis for editorial action against the article.” Ivan Oransky, the cofounder of the Retraction Watch blog, told the BMJ that Martin’s silence is straight from the “typical scientific playbook.” The universities whose researchers authored the original 2001 study, including Yale and Brown, have similarly refused to censure the paper or publish any open inquiries into its making. Virtually everyone involved in producing Study 329 has either declined to comment or insisted the original study is sound.

As Carey notes in his Times story, thousands of children, teens, and young adults attempted or committed suicide while on Paxil—but it’s impossible to know how many, if any, did so because of Paxil. This new BMJ paper makes it seem more likely than ever than some did. And of her daughter, Sara Bostock has no doubt.

When I first wrote about the controversies over use of SSRI’s in teens, a decade ago, a psychiatrist and journal editor critical of the practice told me, “If a drug company’s sales representative’s lips are moving, he’s lying.” It was meant as hyperbole—and it was damned good hyperbole. The current silence of the journals and institutions involved with Paxil’s long promotion reminds us that sealed lips can be just as deadly to the truth.