Inside Paxil Study 329, Courtesy the Justice Department

I've previously discussed the now-infamous Study 329, which took discouraging data on the efficacy and safety of paroxetine (Paxil) in kids and spun it into an article claiming excellent results:
http://brodyhooked.blogspot.com/2011/11/will-brown-university-investigate.html

Thanks to the U.S. Justice Dept. complaint in the suit recently settled by GlaxoSmithKline for a record $3B: http://www.justice.gov/opa/documents/gsk/us-complaint.pdf
--we can follow the history of this study in more detail, based on the internal GSK documents discovered during the proceedings, and see just how the data were manipulated for marketing purposes.

Study 329, directed by Dr. Martin Keller of Brown University, was one of 3 clinical studies in children and adolescents that were all interpreted by GSK scientists between August and October, 1998 to be discouraging. Study 329's protocol specified two primary endpoints, and on neither measure did Paxil do better than placebo. The study also logged in 11 serious adverse reactions to Paxil, much more than in the placebo group, including 5 with agitated or suicidal behavior, the major risk for which eventually the FDA issued a black-box warning for the SSRI class of antidepressants.

Undaunted, GSK contracted in April 1998 with Scientific Therapeutics Information, Inc. to prepare an article for journal publication based on 329. As we know from other sources, STI writers basically wrote the paper and later did revisions, with minimal input from the supposed scientific authors, meaning that the paper was largely if not completely ghostwritten. As part of the laundering process, STI decided to downplay the primary endpoints and instead inserted 8 "efficacy measures," none of which had been specified in the original study protocol. By what's called data-dredging, STI was able to show that Paxil was statistically better than placebo on 4 of the 8 newly invented measures.

Initially the purported authors sent the paper to JAMA, claiming in the abstract that Paxil was "a safe and effective treatment for major depression in adolescents." In December, 1999 JAMA rejected the paper, and reviewers' comments indicated that the scientific reviewers had seen through the fog and realized that there were no solid data to show the superiority of Paxil. One internal memo then indicated that GSK and Dr. Keller agreed to send the manuscript to "a less demanding journal."

Even the less-demanding Journal of the American Academy of Child and Adolescent Psychiatry originally couldn't swallow the revised manuscript that Keller and company sent them in June 2000. Their reviewers detected some of the same problems as the earlier JAMA review.  GSK had STI go back to work responding to the reviewers' comments and eventually JAACAP accepted the manuscript in February 2001.

According again to the Federal complaint, even with the recommended changes, GSK and STI (with the willing acquiescence of Dr. Keller and his co-authors, we presume) managed to slip in a number of incorrect and misleading statements. The abstract stated that the drug "is generally well tolerated and effective for major depression in adolescents." The article mentioned the primary endpoints but failed to make clear that by neither was Paxil statistically superior to placebo. The article falsely implied that Paxil was superior on one of the primary endpoints by deliberately conflating one of the later "efficacy measures" with that endpoint. The article consistently downplayed all the measures where Paxil failed to show superiority to placebo and focused on the few, invented-later measures where a statistically significant result had been found.

More important, the 11 patients with serious adverse reactions due to Paxil, and the 5 of them with specifically suicidal or agitated symptoms, magically disappeared. In the revised manuscript the investigators suddenly decided that only one of the reactions (headache) was actually caused by Paxil, and the other bad outcomes were unrelated to the drug. When the FDA got its hands on the raw data from 329, it eventually determined that 10 of 93 patients taking Paxil had experienced a potentially suicidal reaction--a far different and scarier picture than that portrayed in any of the drafts or in the final manuscript.

To be sure that child psychiatrists heard the correct marketing message, GSK sponsored 8 "Forum" meetings at lavish resorts such as Rio Mar Beach Resort in Hawaii and Renaissance Esmerelda Resort in Palm Springs, CA. The psychiatrists who attended had their airfare and hotel paid plus a $750 honorarium, and in many cases their spouses were also paid for (a practice supposedly prohibited in the 2002 PhRMA code of conduct). The hired speakers who told them how wonderful Paxil was for treating kids received $2500 honoraria in addition.

Talking of speakers' bureaus, another media summary of the complaint:
http://finance.yahoo.com/news/glaxosmithklines-gsk-3-billion-whistleblower-154800879.html
--reported that GSK had enrolled 49,000 health professionals to be on its speakers' bureau for Paxil and other drugs. Two conclusions are possible: 1) there are 49,000 really excellent and scientifically informed speakers out there whose talents are needed to inform their fellow practitioners; or 2) a "speaker's bureau" is really nothing but a disguised bribe to get all those docs to prescribe a lot of the drugs they are paid to speak about.

Hat tip to Dr. Roy Poses of Health Care Renewal blog for sending me the link to the full complaint (all 76 pages, for anyone who wants some fun reading).