Several articles in the April 7 issue of BMJ were devoted to the epic battle fought by BMJ and Cochrane to gain unfettered access to the Tamiflu® and Relenza® data. Not just some data, but all the data. Considering that only 2 out of 10 Tamiflu studies were published in 2010, this request for complete data sets made good sense. Was industry afraid that their results would not hold up to scrutiny by independent statistical re-analysis? Were there secrets locked up in company vaults?
Eventually Roche and GSK provided complete data sets and permitted Cochrane statisticians to check the clinical trial data. Overall, this effort confirmed that the efficacy of the neuraminidase inhibitors is moderate and unimpressive: for Tamiflu, treatment shortens the time to first alleviation of symptoms by 16.8 hours (p<0.001); while prophylaxis reduces symptomatic influenza in household contacts and protects individuals post-exposure. Importantly, there was no reduction in hospital admissions and no effect on influenza complications (although the companies would dispute this). Regarding safety, most adverse events were trivial GI complaints but a very small minority of patients had significant neuropsychiatric AEs. Obviously, the Cochrane researchers felt that the benefits did not outweigh the risks, and certainly not the costs with stockpiling.
For those of you who believed that neuraminidase inhibitors were potent influenza drugs, these conclusions may come as a surprise. However, most of us already knew that these drugs had at best a small benefit. The potential for CNS side effects was not a new finding, either. After all, FDA warned prescibers of neuropsychiatric problems in the label. So, overall, no big surprises, really.
We could stop this blog here but to this ‘war for data access’ brought some rarely discussed practice patterns to the fore:
• There is huge mistrust of industry practices related to data generation and interpretation. There is an implicit assumption that trial data are presented in a particularly industry-friendly fashion. However, the Cochrane efforts detected no bias or poor statistical work on part of the sponsors.
• The tendency in industry to prioritize the publication of ‘favorable’ data while delaying the release of other studies is a disgrace and has to stop. Content and timing of publications should not be decided by marketing strategists.
• EMA had less data for review than FDA which had full data sets. Seems that FDA did its own analysis of the raw data while EMA worked from company documents. This seems to be the underlying reason for differences in labeling details which prompted the Cochrane analysis in the first place.
• As with all anti-infectives, the benefit : risk ratio worsens over time as ‘resistance creep’ reduces the efficacy numerator. Rare but serious side events are hard to accept when benefits are small and keep getting smaller. Without lowering mortality from influenza, or reducing complications or hospitalizations, these drugs are clearly not life-savers when administered after the onset of symptoms.
