Severely flawed meta analysis, incorrect at first glance.
Authors cherry-pick to include only 4 studies reporting non-zero mortality and they claim a mortality RR of 1.11 [0.16-7.65]. However, they report incorrect values for Niaee et al., claiming an RR of 6.51 [2.18-19.45]. The correct RR for Niaee et al. is 0.18 [0.06-0.55] (as below). After correction, their cherry-picked studies show >60% mortality reduction.
Similarly, for viral clearance and NCT04392713, they report 20/41 treatment, 18/45 control, whereas the correct day 7 clearance numbers are 37/41 and 20/45 (sum of clearance @72hrs and @7 days), or 17/41 and 2/45 @72 hrs.
The duration of hospital stay for Niaee et al. is also incorrectly reported, showing a lower duration for the control group.
All of the errors are in one direction - incorrectly reporting lower than actual efficacy for ivermectin.
Authors claim to include all RCTs excluding prophylaxis, however they only include 10 of the 24 non-prophylaxis RCTs (28 including prophylaxis).
Authors actually reference meta analyses that do include the missing RCTs, so they should be aware of the missing RCTs.
For additional errors, see .
Only one of these errors has been partially fixed 4 days later - the Niaee RR was corrected, but the associated conclusion was not. Other errors have not been corrected.
Roman et al., 5/25/2021, preprint, 6 authors.