Re: WHO downgrades status of oseltamivir
There is another way we must look at this problem. Taking the example of the Nicholson et al (2000) trial quoted in the article - "Duration of illness was significantly shorter by 29 h" - What was the lower limit of 95% CI for this difference? These results were not provided by the authors, but by looking at their results, we can say that the lower limit would be close to one hour! Even if statistically significant, this difference is not of major consequence.
We have argued that in superiority trials, like the one above, we should have a "superiorlty margin" like the inferiority trials (Shafiq N, Malhotra S. Superiority trials: raising the bar of null hypothesis statistical testing. Evid Based Med. 2015 Oct;20(5):154-5) and shown how small effect sizes are published as statistically significant findings even in the best of journals (Shafiq N, Malhotra S. Superiority trials: statistical trickery or mass blindness? Postgrad Med J. 2016 Feb;92(1084):118-9).
The meta-analysis quoted in the article (Jefferson et al, 2009) clearly shows what we are trying to point out - "The hazard ratios for time to alleviation of influenza-like illness symptoms were in favour of treatment: 1.20 (95% confidence interval 1.06 to 1.35) for oseltamivir". The lower limit of 1.06 for the hazard ratio is so close to 1.0, the level of no difference, that it is meaningless to give importance to such findings.
Therefore, even for the endpoint on which oseltamivir had some efficacy (duration of symptoms), the small effect size made the results clinically not too important.
Competing interests: No competing interests