Thread by @GidMK

This letter has been published in the AJT, claiming that

a) The Elgazzar study is not withdrawn
b) Even if it is, this makes no difference to meta-analyses of ivermectin

Some thoughts about why this is very misleading and/or simply wrong 1/n

https://twitter.com/profnfenton/status/1435642777757523969?s=20

2/n For the first claim, it is rather fascinating to see the defense that the authors have chosen. As a reminder, most of this study was plagiarized, and the dataset the authors UPLOADED THEMSELVES was fake gidmk.medium.com/is-ivermectin-…

3/n Even if you dislike me personally for whatever reason, several independent experts on fraud confirmed that this data cannot possibly have come from a real RCT i.e. steamtraen.blogspot.com/2021/07/Some-p…

4/n It's not really a matter of "independent adjudication" being an issue - the study has been independently adjudicated, and the only response from the authors has been to claim that the data they said was their data isn't really, and then go silent

5/n As to whether this impacts the meta-analyses of ivermectin: obviously it has a large impact

This was THE LARGEST STUDY included in the analysis in question. It is clear that it made a BIG difference

6/n The authors have re-run their model, and come up with this substantially modified forest plot

They say "Hence, the leading outcome conclusions (ie, for mortality and prophylaxis) are robust to the removal of the study by Elgazzar,3 contrary to the press claims"

7/n But this isn't scientific. They haven't rerun the actual study, they've just run one model excluding Elgazzar

I noted this issue in a thread AGES ago

https://twitter.com/GidMK/status/1416290862724644867?s=20

8/n It's a HUGE issue for a very simple reason - once you've gotten rid of the likely fraudulent Elgazzar trial, the ENTIRE MORTALITY BENEFIT comes from one very bad trial

This means that even the most basic sensitivity analysis removes the effect entirely

9/n So when the authors say that the "leading outcome conclusions" remain unchanged, they are factually mistaken. Indeed, it appears that their entire result now relies on a single study with very substantial issues

10/n Whether or not this means that you can still get a marginally statistically significant result in your model, it means that the certainty of that result is ENORMOUSLY downgraded, and thus the conclusions (below) are clearly reversed

11/n The Niaee trial is worrisome in particular because, despite numerous efforts, the authors have refused to share any data with either myself or
@K_Sheldrick

At this point, that's a huge red flag

12/n Finally, I will note that the letter is also rather poorly timed, as just this morning we confirmed that another study the authors have included in their analysis is almost certainly fake 🤷‍♂️

13/n I think the most rigorous scientific step, at this point, is for the authors to revise their actual study

I can't imagine leaving a meta-analysis containing likely fraudulent research up as-is myself, but perhaps that's just me

14/n A minor addition - another letter published alongside this makes a similar argument, but in a moment of what I can only call scientific absurdity includes another trial with serious concerns in the REVISED analysis

15/n It's rather, uh, strange to see people arguing that potential fraud has not impacted their results while still including potential fraud

You can read about this study here: buzzfeednews.com/article/stepha…