New Lancet report associates hydroxychloroquine and death, in total contradiction with previous research (Rosenberg, JAMA).
Critical issue is indication bias: is HCQ slated for the most severe patients as regulators in America and Europe recommend compassionate emergency use?

You will only be able to disentangle such major confounding factors if you very carefully track severity, as Rosenberg did, concluding HCQ is not an issue. He has included respiratory rate, CT scan, 3-strata O2 saturation and 4 more clinical features, that fully exonerate HCQ.

Mehra has decided to go for a most rudimentary and inadequate binary severity scoring, in particular obfuscating the critical capture of one of the largest COVID19 risk factors: severity of oxygen impairment (the under/above 94% O2 binary, at too high a threshold).

As 42% of non-survivors went under mechanical ventilation vs 6% survivors, if the study does not control properly for oxygen deprivation, any mortality causal attribution to HCQ becomes null and void. Which is exactly what happened. Null and void. Not Lancet’s finest hour.

The author had the foresight to caveat his work with the following sentence that is the only thing we’ll remember from that study:
“a cause-and-effect relationship between drug therapy and survival should not be inferred”