Meta analysis of late stage studies (and one early treatment study with only 2 deaths), showing HCQ RR 0.83 [0.65-1.06], before exclusions RR 0.80 [0.65-1.0].Authors claim "HCQ alone is not effective", but the result directly contradicts this, RR 0.83 [0.65-1.06], i.e., inconclusive but much more likely to be effective than not.There are many errors in this meta analysis which introduce critical bias, for example:- Very biased sample of studies, including <4% of early treatment studies (only 1), and <30% of late treatment studies, focused on negative studies.- Arshad et al. (propensity matched HR 0.49, p=0.009) was excluded because the authors claim a "critical" risk of confounding bias due to steroid use, however steroids were controlled for in the multivariate and propensity analyses [1].- For Skipper et al., authors use an RR of 1.01, however the study had one hospitalized control death and one non-hospitalized HCQ death. Since the HCQ death was non-hospitalized, it may not be caused by COVID-19, or the patient did not receive standard care, therefore this should not be treated as equal to the control death. Further, medication adherence was only 77%, the HCQ patient may not have taken the medication (Skipper et al. neglects to answer this question). In any case, including a trial with only 1-2 deaths is likely to increase bias.- Cavalcanti et al. received the lowest bias rating, despite having treatment delayed up to 14 days after symptoms, randomizing 14% of patients in the ICU, having significant protocol deviations, unusually low medication adherence, randomization that resulted in 64.3% male patients (HCQ) vs. 54.2% (control), and excluding patients already receiving longer and potentially therapeutic doses of the study treatments.- Sbidian el al. received the lowest bias rating, however many more control patients are still in hospital at 28 days suggesting there will be a significant improvement when extending past 28 days.- The RECOVERY trial received the lowest bias rating, despite using a very high dose likely responsible for the increased mortality. Results of this trial are not relevant to use at normal dosages.- Inclusion criteria required RT-PCR confirmed cases, but this was disregarded when including Horby et al. (very negative, excessive dose) and Skipper et al.- Authors do not consider different treatment delays, risk level of patients, differences in dosage, or usage of Zinc.Also see many other reports of problems and fatal flaws: [2, 3, 4, 5, 6]This analysis is also missing several recent studies, for a more up-to-date analysis see [7].