Side Effects: A Prosecutor, a Whistleblower, and a Bestselling Antidepressant on Trial

by Alison Bass

  Firestein sighed. The case of the People ofNew York v. GlaxoSmithKline, Docket Number 04 CV 5304, would hinge in large part on the rulings made by Judge Miriam Cedarbaum. Firestein prayed that she had not alienated this all-powerful woman on her very first appearance in her courtroom.

  A FEW DAYS LATER, Firestein finally made contact with a source she'd been told might be helpful in the case against GlaxoSmithKline. The source was a woman by the name of Donna Howard, and all Firestein knew was that she had once worked for the psychiatry department at Brown University and might have some useful information about the Paxil study Keller and his coauthor published in 2001. By this time, Howard was no longer working for NAMI Rhode Island; she had been fired in May 2004. She was now on unemployment and looking for another job. In the meantime, she was volunteering at Taunton State Hospital. When Firestein finally reached Howard at her home, the two women talked for more than an hour.

  One particular fact stood out in Firestein's memory of their conversation: Howard told her that a few weeks after entering the Paxil study at Brown, one of the participants, a fourteen-year-old boy, had what she called a "psychotic break" on Thanksgiving weekend, 1994. He punched some pictures, broke the glass, and cut himself badly. He was taken to the emergency room of a Fall River hospital, where he received six stitches. Because he seemed acutely suicidal, the boy was then transferred to Bradley Hospital, a psychiatric hospital for children and adolescents in East Providence. The treating psychiatrists at Bradley had been told the boy was part of the Brown Paxil study, but they didn't know what, if any, drugs he was on. (He could have been taking Paxil, Tofranil [imipramine], or the placebo.) They needed more infor mation; how could they help their patient if they didn't have his basic medical history? Over the weekend, one of the treating psychiatrists finally reached a researcher who was involved with the Paxil study and asked him to break the blind so they could find out what medicine their patient was on. The researcher refused.

  "The doctors at Bradley were beside themselves," Howard recalled. "Here was a kid in an obviously bad state, and they had no way of finding out what, if anything, the kid was on."

  Firestein was appalled at what she was hearing. From the little she knew about medicine, it was an accepted principle that the needs of the patient came before the needs of a drug study or anything else. Otherwise, the doctor supervising the clinical trial could be seen as violating the basic tenet of medicine, the Hippocratic oath. According to the protocol itself for the Paxil study, "the blind was to be broken ... in the event of a serious adverse experience that the investigator felt could not be adequately treated without knowing the identity of the study medication."

  Ronald Seifer, director of research at Bradley Hospital and a member of Brown University's Institutional Review Board at the time, fired off an angry memo to Martin Keller, saying that the psychiatrists at Bradley had to have this kind of information in order to treat their patient. Shortly afterward, the blind was broken, according to the GlaxoSmithKline final report of this incident. It was discovered that the boy had been taking Paxil during the clinical trial. The Bradley doctors began to phase him off the drug, and on Tuesday, November 30, the fourteenyear-old was withdrawn from the Brown study.

  But despite the fact that this boy was clearly suicidal and required hospitalization, he was not included among the patients listed as having developed serious adverse effects in the published 2001 Paxil study. Other patients were similarly miscoded. One was a fifteen-year-old girl who had been withdrawn from the Brown study site in 1995 after becoming combative with her mother. According to internal university documents that Howard gave me, Brown researchers knew that this girl had become suicidal after taking Paxil. In a memo to the Institutional Review Board dated October 30, 1995, Martin Keller wrote that this teenager, who had been enrolled in the study in June 1995, "was hospitalized on 9/15/95 due to becoming very combative with her mother and threatening suicide." Yet instead of coding her behavior as an adverse effect related to Paxil, Keller in his memo says she was "terminated from the study for non-compliance." The Brown investigators may have coded her as noncompliant because she had stopped taking Paxil before having her meltdown. But they shouldn't have, according to several clinicians familiar with the study. The Brown researchers should have included all adverse effects experienced by their patients, regardless of what may have caused the problems. As a Harvard Medical School biostatistician later told me, "You shouldn't try to make these subjective attributions and exclude patients who don't fit into your thesis." As research has shown, the SSRI antidepressants can cause serious side effects, including suicidal behaviors and hostility, weeks after people stop taking them.

  The underreporting didn't stop there. In their 2001 paper, Keller and his coauthors reported that "serious adverse effects occurred in 11 patients in the paroxetine group, 5 in the imipramine group and 2 in the placebo group." After comparing internal Brown documents with the final report of study 329 that GlaxoSmithKline posted on its Web site, it becomes apparent that the researchers had not included among the patients with serious adverse effects yet another teenage girl who had left the study after trying to kill herself. According to a memo that Keller himself wrote to the Brown Institutional Review Board on January 30, 1995, this patient (number 70) ingested eighty-two Tylenol pills in an apparent overdose attempt on January 19. Patient 70 was admitted to a hospital and terminated from the study shortly afterward, according to Keller's memo to Brown IRB. Yet this teenage girl was not included in the group reported to have experienced serious adverse events while in the study. Instead, in another memo that Keller had written to the Brown IRB in 1995, she was described as having been terminated from the study for being "noncompliant."

  In an even more mysterious turn, patient 70 was described in GlaxoSmithKline's final report of the study as being a twelve-year-old boy. This boy was enrolled in the clinical trial a month after the teenage girl identified in Keller's memos as patient 70 overdosed on Tylenol and withdrew from the study. The boy with the same patient number was removed from the study on March 22, 1995, after developing tachycardia (rapid heartbeat) while taking the tricyclic antidepressant known as imipramine, according to the company's final report. There was no mention in the company's posted final report or the 2001 journal paper that the original patient 70 was a young girl who had ingested eightytwo Tylenol pills in a clear bid to kill herself.

  During her phone conversation with Firestein, Howard alleged that the data in the Paxil study had been changed to satisfy the study's sponsor, SmithKline Beecham. "Everybody knew we had to keep SmithKline happy and give them the results they wanted," she said.

  Howard told Firestein that she still had in her possession from her time at Brown a number of internal Brown memos showing how researchers may have miscoded the data in the Paxil study. Would Firestein like copies of them?

  Sure, Firestein replied.

  Before hanging up, the litigator had one more question: would Howard would be willing to testify in court if it came to that? Her heart in her throat, Howard said yes. She then asked, "Do I have to fly? Can I take a train?" Firestein laughed and said how Howard got down to New York was up to her.

  n August 3, 2004, Charles "Chuck" Grassley, the Republican senator from Iowa and chairman of the Senate Finance Committee, sent a sternly worded letter to Christopher Viehbacher, president of the U.S. Pharmaceuticals division of GlaxoSmithKline. In his four-page dispatch, the senator wrote that he was troubled by the New York attorney's allegations that Glaxo had concealed important information about Paxil. He said he had heard similar accusations from Dr. Andrew Mosholder.

  Mosholder was the FDA official who had been prevented from testifying about his findings at the agency's February 2 hearing. Grassley's Finance Committee was investigating that gagging incident, and in the course of its probe, Mosholder had told the senators that "GlaxoSmithKline, in his opinion, was attempting to `sugar-coat' the adverse effects of Paxil on children by `miscoding' suicidal ide
ations and/or suicidal behavior."

  In his letter to Viehbacher, Grassley said that he was concerned that some drug companies" may not have provided the FDA with all the information at their disposal. He concluded by demanding that GlaxoSmithKline hand over information about all of the clinical trials on antidepressants and other drugs it had initiated between 1990 and the present. Glaxo had until August 27 to comply.

  Glaxo officials would decline to comment on whether there was any connection between Senator Grassley's letter and their decision to pursue a settlement with the New York State attorney general's lawsuit. But a few days after Grassley faxed his letter to Viehbacher, Rose Firestein received a call from Wick. The drug company was ready to resume talks.

  Looking back, Firestein thought the constant drumbeat of publicity might also have played a role in Glaxo's willingness to come to the table. Reporters from the New York Times and the Wall Street Journal, among other major papers, repeatedly cited the AG's lawsuit as a catalyst for the "expanding debate over the incomplete disclosure" of drug test results. And the AG's office continued to widen its net. The day after Senator Grassley dropped his letter bomb, the pharmaceutical giant Johnson and Johnson publicly acknowledged that Spitzer's office had asked for detailed information about six of its drugs, including Risperdal and Procrit, an antianemia medication. And on August 5, the Wall Street Journal, having somehow obtained a draft FDA document, reported that the federal agency had finally found evidence of a link between antidepressant drugs and suicidal tendencies in young people. The FDA's latest reanalysis appeared to be consistent with earlier conclusions reached by Mosholder.

  In an e-mail to me, Mary Anne Rhyne, a spokeswoman for GlaxoSmithKline, said she couldn't comment on whether any or all of these high-profile events contributed to the company's decision to resume talks. Whatever the reason, on August 11, Wick and Dwight Davis, along with Frank Rockhold and an in-house attorney for Glaxo, took a cab downtown to meet with Firestein and her colleagues at 120 Broadway. This time the Glaxo lawyers were escorted to a different conference room, a little bigger than their previous gathering place, but just as drab and airless. It too had half-empty metal bookcases lined up on one side of the room, boxes stacked in the corners, and no windows. No one seemed to mind.

  "I'm glad we're not in the bad karma room," Wick joked. Everyone around the table laughed.

  GlaxoSmithKline had already reconciled itself to the idea of establishing an online registry of its clinical trial data as part of a settlement with the New York AG. Indeed, it was the company's announcement of plans to post some clinical data that had prompted Tom Conway to end their first meeting on June 30 so abruptly. Now, six weeks later, it looked as if the talks might once again founder over the same sticking point: the level of detail the AG's office wanted included in the company's public registry.

  Firestein went through the list of demands one by one. The registry, she said, should include detailed summaries not only of all the company's Phase 3 and Phase 4 clinical trials but also of some of the safety studies that it performed on healthy volunteers. Drug companies routinely carried out these Phase 1 and Phase 2 trials to see how the target drug metabolized inside the human body (these were known as pharmacokinetic studies), what its cause of action might be, and what adverse side effects it generated. While these Phase 1 and 2 studies were sometimes submitted to the FDA, they were usually not published in peer-reviewed journals or made available to the public.

  Not surprisingly, the representatives from Glaxo balked at this demand.

  "Nobody had ever suggested that they publish the adverse events from pharmacokinetic studies," Firestein recalled. "So that was a huge conversation."

  At their first meeting in June, Firestein had also made it clear that she wanted to set a specific standard or threshold as to which adverse side effects would be included in the posted summary for each clinical study.

  But should that threshold be 5 percent, 10 percent? In other words, what percentage of patients would have to suffer that particular side effect-whether it was nausea, a headache, or suicidal behavior-for it to be included in the study summary?

  This, Firestein knew, was a key issue. If Glaxo only posted adverse side effects that occurred in, say, 10 percent or more of the study participants, that might mask some serious problems with a particular drug. In one clinical trial of a popular antidepressant, 9.7 percent of the study participants had developed troubling suicidal thoughts and behaviors. So if the standard settled upon was 10 percent, some serious side effects could be hidden from public view.

  "We felt there had to be a clear litmus-paper standard for what constitutes an adverse event, what constitutes a serious adverse event, and which ones you have to report," Firestein recalled. "We spent a lot of time talking about at what point does something become a serious adverse event that's reportable."

  The AG team also wanted each study summary to contain detailed information on efficacy results-in other words, how the drug in question did on the study's primary and secondary outcome measures. The posted summary should also include information on whether those outcome measures had been changed once the study began.

  The Glaxo negotiators weren't happy with any of these demands. "They didn't want to be told what the content of their registry should be," Firestein said. "They wanted to decide what the content would be" (In an e-mail to me, the spokesperson for the drug company declined to comment on these negotiations, saying it would be "inappropriate.")

  After several hours of intense but cordial discussion-"there was no shouting or yelling," Firestein recalled the meeting broke up. Wick said he would take the AG's demands back to corporate. He'd be in touch, he said.

  Walking back to their offices, Conway remarked to Firestein, "I have a feeling this case is going to settle. I think we're going to get what we need"

  Firestein managed a smile. But she felt completely drained. The intensity of the negotiations, the strain of being on the hot seat with everyone in the room judging her performance-it was all a bit much. At that particular moment, she felt as if the gray matter in her brain had turned to jelly.

  The next few weeks passed in a blur. Wick called her back a few days later and agreed to some of the demands, but not others. The two attorneys negotiated by phone, talking almost every day. Firestein would relay what Wick had said to Stark, Conway, and Baker, see what they thought, and then call him back. Firestein talked with Wick, then with Dwight, then with Wick again. Each time one of the Glaxo attorneys said no, they couldn't do that, Firestein would remind him that if the AG's office took this case to trial and won, it would get that particular clause. So why not agree to it now and spare both sides the expense and hassle of a prolonged court battle?

  In the end, the drug company agreed to post detailed summaries of some of their Phase 2 safety studies and all of their Phase 3 and Phase 4 drug studies that were completed after December 27, 2000, plus any earlier clinical studies that were material to a physician's medical judgment in prescribing a Glaxo drug. The drug company remained adamantly opposed, however, to the idea of publicly registering the design or protocol of any ongoing clinical studies. That had been another condition Firestein and her colleagues had proposed in their August 11 meeting, and the drugmaker seemed dead-set against it. Pretrial protocols were a closely guarded proprietary secret, the Glaxo attorneys argued. Posting them would only give competitors an unfair window into the company's pipeline of upcoming drug products.

  After talking it over among themselves, the assistant attorneys general agreed to concede on the pretrial registry. Since it was not central to their complaint-which had focused on GlaxoSmithKline's incomplete disclosure of already completed drug trials -"we felt we wouldn't have been able to get that through litigation," Firestein explained. Knowing that the editors of the world's top medical journals were about to require pretrial registration as a condition for publication made their decision that much easier.

  By the third week of August, Firestein and
Wick had worked out the details of what GlaxoSmithKline's clinical registry would look like. It would conform to the general principles of how a drug study should be reported, laid out years ago by the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. Known in the trade as the ICH E3 guidelines, this was a detailed guidebook to how clinical studies should be conducted and presented to regulatory authorities. But the ICH E3 guidelines said nothing about having to present such data to the public.

  What Glaxo agreed to went far beyond the ICH E3 guidelines. Each study summary on its registry would contain more than twenty categories of information, including specific data on the primary and secondary outcome measures of the study, the type and severity of any adverse side effects, and whether the study had been terminated early and why.

  In short, Glaxo's registry would include a great deal more substance than the brief and often spotty summaries of drug study results that the FDA and other pharmaceutical companies had thus far been willing to make public.

  All that was left to resolve was the amount of money G1axoSmithKline would have to put on the table to settle the lawsuit. Firestein and her colleagues had already decided not to insist on a huge fine. They were far more interested in getting the drugmaker to launch a comprehensive registry. But there had to be some money on the table. To get a sense of what the damages should be, Firestein calculated how much money might have been spent on Paxil prescriptions for children and adolescents in New York State in 2002. Firestein would not disclose to me the ballpark figure she came up with. However, data from the American Medical Association indicate that Paxil prescription sales in New York State for patients under eighteen years old came to about $5.5 million in 2002.