et al., medRxiv, doi:10.1101/2020.09.29.20203869 (Preprint) (meta analysis - not included in study count)
Combination of the four underpowered prophylaxis RCTs to date showing statistically significant results, RR 0.78 [0.61-0.99].
The actual effect of HCQ is likely to be higher for several reasons: the trials did not adjust for losses to follow-up or other deviations from protocol. There was very long treatment delays in the postexposure prophylaxis trials - in one trial, about a third of participants were enrolled 4 days after exposure with an additional shipping delay of ~46 hours on average, in the other, participants were enrolled up to 7 days after exposure, with an unknown additional delay before treatment, and results suggesting that exposure detection was delayed.
For other reasons see: [1, 2, 3, 4].
Garcia-Albeniz et al., 10/1/2020, preprint, 5 authors.
risk of COVID-19 case, 22.0% lower, RR 0.78, p = 0.04.