After Struggle With Roche, Panel Casts Doubt on Tamiflu

The team, led by Chris Del Mar of Bond University in Australia, embarked on an update of the 2008 Cochrane review at the request of the U.K. National Institute for Health Research in August. Just a few weeks earlier, the team had received an email from a Japanese pediatrician, Keiji Hayashi, who questioned the validity of one of the papers on which the 2008 review was based. That paper—-published in 2003, with Laurent Kaiser of the Hôpital Cantonal de Genève as its first author—was a meta-analysis of 10 Roche-sponsored trials of oseltamivir. Only two of them had been published in peer-reviewed journals, Hayashi pointed out; the remaining eight were unpublished or published only in abstract form. Hayashi challenged the Cochrane team to "appraise the 8 trials rigidly."

According to Doshi's account, panel member Tom Jefferson—the first author of the 2008 review and the only member of that group also involved in the new analysis—asked the authors of the Kaiser paper to send him data on the eight trials. When they couldn't, Jefferson contacted Roche to obtain the information. The company said it would make the studies available, but only if Jefferson signed a confidentiality agreement that also contained a clause "not to disclose ... the existence and terms of this Agreement." Apparently, Doshi writes, Roche "intended not only to keep its data concealed, but also to conceal the fact that it was quieting people through a secrecy clause."

The team found that unacceptable, and after some more e-mails back and forth, it decided to exclude the Kaiser meta-analysis from its new review. They were left with a total of 20 trials that together allowed the conclusion that Tamiflu has "modest effectiveness against the symptoms of influenza in otherwise healthy adults," and that it "might be regarded as optional for reducing the symptoms of seasonal influenza"—but that "paucity of good data has undermined previous findings for oseltamivir’s prevention of complications from influenza. Independent randomized trials to resolve these uncertainties are needed."

The experience left the four members with serious doubts about the way the Cochrane Collaboration does its business. "The previous reviewers endorsed the conclusion that oseltamivir reduces complications such as pneumonia and bronchitis by implicitly trusting that the unpublished data were verifiable," Doshi writes. "This trust now seems naive."

In the point-by point response, Roche's James Smith, international medical leader for Tamiflu, says that Roche now publishes all of its clinical trials, but that this was not standard procedure in the industry 7 to 10 years ago. "At the time, it was considered that the studies that were published (2 abstracts and 2 full manuscripts) reflected accurately the benefits of the drug," Smith writes, "and that the additional studies provided little new information and would therefore be unlikely to be accepted for publication by most reputable journals."

Smith adds that offering the data under a confidentiality agreement is "commonplace within the scientific community to ensure the responsible use of data," and says that a group at the U.K.'s Medical Research Council (MRC) recently agreed to just such a deal. But the paper does not explain why the existence of the deal itself needed to remain a secret.

Roche also denies most of the allegations in Cohen's feature story—for instance, that the authors of Roche-sponsored studies were pressured by the company's marketing people to talk up the importance of influenza. Roche acknowledges that anonymous medical writers were involved in at least one of the papers, but the company argues this was standard practice at the time, before so-called Good Publication Practice Guidelines took effect in 2003.

Whether the blow to Tamiflu's image is permanent remains to be seen. If the MRC group that has access to Roche's full database is working on a similar review—and finds that the eight studies were properly conducted—it could well reach a different conclusion.