A spokesman for Roche told in-PharmaTechnologist.com the review by the Cochrane Collaboration contained major mistakes and did not use all the patient data available, a claim the researchers blasted as “disingenuous.”
Cochrane published a review of Tamiflu and other neuraminidase inhibitors today in the British Medical Journal (BMJ), concluding the drug only reduced flu symptoms by half a day and was ineffective in reducing hospital admissions and serious complications from influenza.
“Having not yet received the full report, and only reviewing the BMJ summary, Roche has found significant errors in their analysis,” Anthony Vaughn, Roche spokesman, told In-PharmaTechnologist.com.
“Despite requesting the patient level data for our 77 trials, they have used only the summary tables from 20 of the clinical study reports, which is similar information used within their 2009 report.”
The Roche spokesman also criticized the Cochrane report’s statistical analysis.
“Without having seen their detailed analysis, there are a few key errors: for example, when analysing the time taken for patients to recover after flu, they have selected the mean value, which is incorrect for the statistical analysis, where the median should be used, and clearly demonstrates adults and children with flu get better a day and a half sooner with Tamiflu.”
The Cochrane researchers hit back at Roche, telling in-PharmaTechnologist.com it was “disingenuous” to say their analysis was flawed.
Roche was “factually incorrect” in claiming the researchers had requested individual patient data (IPD), said Chris Del Mar, Coordinating Editor of the Cochrane Acute Respiratory Infection Group and one of the review’s authors. The group had so far lacked the time and funding to do IPD analysis, although it planned to do so in future, he said.
Del Mar also defended the review’s statistical methods against Roche’s charge of “key errors.”
Roche’s response had tried “to trump randomised controlled trial (RCT) data with observational data,” a method which is “not appropriate,” said Del Mar.
“Observational data will always be vulnerable to biases, from both known and unknown factors. This is the very reason why RCTs have become the basic research method for answering intervention questions.
“Moreover we believe the methods they have used to attempt to redress observational study biases were inappropriate, while the crude data shows a higher proportion of patients died on treatment than died without treatment.”
The report’s author also told us Cochrane’s methods had been made available in advance for many months, but “we have not received any criticism from Roche until now.” By contrast, he said, GSK had made extensive suggestions which the researchers had addressed.
Roche’s spokesman further criticised Cochrane, saying it had admitted it was “inexperienced in dealing with large quantities of data and lack familiarity with clinical study reports.”
Del Mar told In-PharmaTechnologist.com this phrase was used out of context, and the authors were “describing the methods of meta-analysing CSRs [clinical study reports] as pioneering, and without precedent.” The Cochrane Collaboration usually analyses published research, rather than CSRs.
He added the Cochrane team would make available all the data it used in its analysis for others to conduct their own reviews.
Around 130m patients worldwide have taken Tamiflu, and governments have stockpiled it to guard against a future pandemic, with UK authorities spending £424m ($711m) on Tamiflu alone.