BUCKLE UP—huge HYDROXYCHLOROQUINE & #COVID19 meta-analysis of all useable trials & observational studies w/ 30,000—11,932 participants on HCQ, 8,081 on HCQ+azithromycin & 12,930 control. Result? Big fat null. Trials or not. HCQ+azithro even worse! https://www.clinicalmicrobiologyandinfection.com/article/S1198-743X(20)30505-X/fulltext

2) HCQ didn’t lower mortality— Relative Risk RR=0.83 (95% CI: 0.65-1.06, n=17 studies) for all studies and RR=1.09 (95% CI: 0.97-1.24, n=3 studies) for trials alone. The RR=0.83 but confidence intervals 0.65-1.06 overlapping 1 (no effect) means the results were not significant.

3) For HCQ+azithromycin, the results were even worse: Hydroxychloroquine with azithromycin was associated with an increased mortality: RR=1.27 (95% CI: 1.04-1.54, n=7 studies), meaning 27% increased risk (significant because CI didn’t overlap 1).

4) Summarized into a table, notice none of the HCQ alone vs control studies significant. While HCQ+azithromycin combo significantly increased overall.

Sidenote: alternative Risk Difference is also non-sig for HCQ. (RD is 1/Number Needed to Treat—useful if it were beneficial).

5) For the record, they looked at a crap ton of studies in their review. #metaHCQ

6) The authors also diligently looked at the risk of bias and stratified analyses by degree of potential bias. They only found a very borderline inverse association among the subgroup studies with the most serious degree of bias. Hence don’t cherry pick.

7) They even went to the lengths of doing a sensitivity omit-one-study-at-a-time repeated pooling to see if any one study had large effect or got diff results if excluding any—there was nothing. Pretty robust.

8) Similarly, HCQ+azithromycin also pretty robust overall excluding one study at a time in repeat meta pooling. Hence not one study was driving the overall results.

9) They also tried Bayesian meta, and got even more nothing burger in HCQ: pooled RR for mortality of 0.93 (95%CI: 0.72-1.14, n=17 studies). And for HCQ+azithromycin: pooled RR for mortality of 1.32 (95%CI: 0.97-1.68, n=7 studies). Still no deceased risk. Robust.

10) also the “meta-analysis reported a high heterogeneity for hydroxychloroquine alone, but this heterogeneity was lowered among RCT, studies with moderate (lower level) risk of bias”—which means results less ‘all-over-the-place’ among higher quality studies. This is good sign!

11) Bottomline: this 30,000+ personal meta analysis of all available quality trials and observational cohort studies found no decreased risk for HCQ or HCQ+azithromycin. (HT to authors @AnthonyGuihur @T_Fiolet @Damkyan_Omega @Nibor_Tolum @Mht_Saleh_yahya @nathanpsmad) #COVID19

12) As for the issue of HCQ dose, the lead author said there wasn’t enough data on dose reports consistently across all studies. Though I would be interested to see what data there is among those studies with dose—unless it is too limited to draw conclusions. https://twitter.com/T_Fiolet/status/1299108004747448320

13) the reason dose is maybe relevant is that a new study released August 24, reports that dose has a major effect on HCQ and Covid. Notably in Belgian study, low-dose HCQ was associated with lower mortality in hospitalized #COVID19 patients. https://www.sciencedirect.com/science/article/pii/S0924857920303423

14) That said the Belgian study is not a trial. And we know from experience that one study doesn’t prove a true phenomenon. I’ve taught meta analysis at Harvard SPH for over 15 years, and done many metas, including for COX-2 inhibitors. Metas more reliable than single studies.

15) And... of course there is ZINC. I don’t know of any HCQ+azithro+zinc trials, or even cohort studies of triple combo. Unless cohort/trial data, I can’t comment. We cannot assume effects without any large studies. Epidemiologists don’t do “believe true unless proven not”.

16) and believe me, I’ve been watching and reading all the latest HCQ studies and debates over the last many months. When the HCQ viral video came out, I tried to lay out the data available - to the tune of tons of trolling. But data is not equal. Trials, cohorts, and metas best. https://twitter.com/DrEricDing/status/1287912652006719488

17) p.s. obligatory “demon sperm” mention — the HCQ viral video was released the same day as the demon sperm thing as part of some coordinated right wing push for HCQ in which even Rudy Giuliani got involved, and set a record as one of the most viral videos in Twitter history.

18) I think that is all a distraction. There is however a coordinator movement for HCQ that clearly goes beyond scientific evidence. And fighting that misinformation is key right now when so many lives on the line.