Take Your Tamiflu!

To the Editors:

In “Flu Warning: Beware the Drug Companies!” [NYR, May 12], Helen Epstein faults Laurent Kaiser’s analysis of Tamiflu. In addition, by citing examples where apparently independent researchers were paid by drug companies, she implies that he favored Roche’s drug because of conflicts of interests.

For these serious allegations, however, she presents little evidence.

Kaiser’s analysis was not paid by Roche.

Epstein’s other arguments are equally fallacious. She posits that the postulated flaws in Kaiser’s analysis (analyzing several studies together) resulted in an overestimation of Tamiflu’s effectiveness. Why? She herself notes that overestimation and underestimation are equally possible.

Even more importantly: Marc Lipsitch, who has reanalyzed essentially the same data, including those that Roche withheld from Epstein, specifically states that lumping several studies together did not change the findings, and confirms Kaiser’s results.

Epstein skips over this by saying that Lipsitch “didn’t address the other problems the Cochrane team had identified.” What problems?

• The Cochrane team claimed that the studies summarized by Kaiser had an unusually high rate of confirmation of suspected influenza by virus cultures. First of all, this is untrue; within an epidemic, what looks like influenza is mostly influenza; confirmation rates can exceed 80 percent as compared to Kaiser’s 70 percent. Second, how would the detection rate skew the results in Tamiflu’s favor? Tamiflu has no effect in cases which are not due to the influenza virus. The high confirmation rate shows that the studies were properly done in circumstances where Tamiflu is indicated, i.e., during influenza epidemics.

• Three serious adverse events, “possibly related to Tamiflu,” were claimed to be present in documents unearthed by the Cochrane collaborators but not reported in the Kaiser analysis. Even if true, this is not exactly earth-shaking news. Almost anything that happens during a drug trial (including a traffic accident, lung cancer in a heavy smoker, or serious sunburn, to quote a few examples from recent drug trials in which we participated) is “possibly related.” Much more needs to be known before concluding that the three events were caused by Tamiflu. To imply otherwise is fooling the reader.

Kaiser’s findings have been vindicated by Lipsitch’s analysis. There is no evidence in support of Epstein’s suppositions. Apologies are due.

Bernard Hirschel, MD
Daniel Lew, MD
Stephan Harbarth, MD
Christian van Delden, MD
Karl-Heinz Krause, MD
Didier Pittet, MD
Division of Infectious Disease
Hôpitaux universitaires de Genève
Geneva, Switzerland

Helen Epstein replies:

The six doctors who sign the above letter in support of Dr. Kaiser say his analysis of the Tamiflu trials was “not paid by Roche.” Why then does the acknowledgements section of Dr. Kaiser’s article state: “This study was supported by F. Hoffmann-La Roche Ltd, Basel, Switzerland”? Moreover, Kaiser et al. also acknowledge that the statistical analysis was conducted by a Roche employee.¹

Combining the data from several studies as Kaiser did is poor epidemiological technique by any standard. Do Dr. Hirschel and his colleagues really believe this doesn’t matter?

Marc Lipsitch’s reanalysis of the Tamiflu trials has yet to be published in a peer-reviewed journal. But according to a poster presentation of the results, he found a 40 percent reduction in severe complications, whereas Dr. Kaiser claimed a 55 percent reduction. Both results suggest the drug may have been effective, but given that the two groups were using the same data, the difference in effect is puzzling.

A clue to this discrepancy can be found in this sentence from Kaiser’s paper: “One limitation of these studies was the lack of standardized clinical guidelines across participating centers for diagnosing complications.” This lack of a standard definition of “complications” makes interpretation of these trials virtually impossible.² Here’s why: one serious influenza complication is “pneumonia,” but this means different things to different doctors. Diagnosis could be based on attempts to grow the bacteria responsible from bronchial washouts or sputum specimens, or it could be done solely by clinical symptoms and/or by radiography. What are Kaiser and Lipsitch claiming Tamiflu is protecting people from?

How much of the “influenza-like illness” that doctors see in their clinics is really due to the influenza virus? In the Kaiser trial clinics, the authors claim that 70–80 percent were and Dr. Hirschel, citing no references, maintains that this is not unusual. In my piece, I cite data from a ten-year Centers for Disease Control clinical survey that found that roughly 15 percent of influenza-like illnesses were ultimately confirmed by lab testing to be caused by influenza.³ Unless Professor Hirschel can explain the fivefold higher confirmation rates in the Kaiser trials, readers may reasonably assume that the patients in those trials were selected for reasons that Roche has declined to make clear.

Regarding “adverse events” in the trials, the primary issues have to do with ethical behavior and scientific integrity. If adverse events occurred that could have been due to Tamiflu side effects, they should be reported in the articles. The three cases of adverse events I refer to were not incidental. According to Roche’s own internal documents, there were ten adverse events in those two trials. While seven were of the kind Dr. Hirschel describes—coincidental and unlikely to be Tamiflu-related—three were “possibly” caused by the drug, a strong statement for Roche itself to make. This rate of serious adverse events is consistent with the pattern seen in Japan⁴ and is also consistent with four toxicology studies of Tamiflu, which show the drug kills laboratory animals that are administered at progressively higher doses.