This is from James Watson (the researcher behind that open letter about the fraudulent hydroxychloroquine paper in The Lancet):
This recommendation is basically saying that HCQ has either no effect or a negligible effect -> should not be used
This would suggest that the uncertainty around the primary outcomes of interest are very well characterised. Have they been for HCQ (or any drug) in prevention of COVID-19?
The top outcome for the guidelines panel is mortality (separate debate whether this is right choice)
Only 1/6 trials had deaths (Mitja et al): 5 deaths in HCQ arm (n=1116) and 8 deaths in control (n=1198). This gives the highly noisy estimated OR of 0.67 with a CI of 0.2-2
At the lower end this would be a fantastically good drug – at the upper end this would be a killer drug!
So how the heck does this endpoint get decided as “High quality evidence” that HCQ confers “No important difference in mortality”? That’s bonkers – this is orders of magnitude underpowered to say anything of interest.
Compare this to the same WHO guideline on dexamethasone:
The RECOVERY trial had 930 deaths out of 3883. The odds ratio for death was 0.82 (0.72-0.94). This was graded as “Moderate evidence with serious risk of bias”.
???
[…]
I’m definitely not a HCQ true believer – but we really need these guidelines to make some sense
The guideline was published in The BMJ.