Thread by @kakape

This (German) article by @hfeldwisch raises important questions about the Gangelt publication by @hendrikstreeck et al, that I've been wondering about. It argues that the study underestimates how deadly #sarscov2 is because not all the deaths were counted. medwatch.de/2020/11/26/die…

@hfeldwisch @hendrikstreeck I had wondered about this before because

1. We know that deaths lag cases and studies like this Princess Diamond one (eurosurveillance.org/content/10.280…) take that into account when calculating IFR
2. @hfeldwisch and others had warned before that the Gangelt study was not doing that

@hfeldwisch @hendrikstreeck The study was published in @NatureComms last week. Headline finding was an estimate for the infection fatality rate of #SARSCoV2 of 0,36%.
That was based on 7 deaths in the community of Gangelt compared to 1956 (estimated) infections.
nature.com/articles/s4146…

@hfeldwisch @hendrikstreeck @NatureComms The reporting from @hfeldwisch suggests that several more people of those 1956 counted in the number of infections in the study have since died of #covid19 and so should be counted when calculating the infection fatality rate.

@hfeldwisch @hendrikstreeck @NatureComms On top of this, an IFR calculated from such a small sample comes with a lot of caveats and uncertainty. Not that helpful when there are good studies around done on much larger samples, as @C_Althaus has pointed out in this thread:

https://twitter.com/c_althaus/status/1332076519095754752?s=21

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus Main point I wanted to make here is that the infection fatality rate of #SARSCoV2 is one of the topics that has very quickly become polarized in this pandemic. That is because lower IFR means less reason for restrictions.
Two things on this:

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus 1. In such a situation research needs to be particularly careful and well-done. That is not what we have seen. Take the questions about this Gangelt study. Or look at some of the meta-analyses from John Ioannidis, and at the excellent threads on these by @GidMK.

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus @GidMK 2. It is kind of ironic because the estimates of IFR have been remarkably stable compared to some other things in this pandemic:

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus @GidMK Here is @WHO on 19.2., for instance, estimating it between 0,3 and 1% (lower estimate was actually based on a mistake and should have been higher) who.int/docs/default-s…

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus @GidMK @WHO Here is a preprint of a meta-analysis done by @GidMK from 7.7. estimating IFR at 0,68% (0,53-0,82)
medrxiv.org/content/10.110…

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus @GidMK @WHO Here is an estimate in the Lancet based on data from Geneva: 0,64% (0,38-0,98)

thelancet.com/pdfs/journals/…

@hfeldwisch @hendrikstreeck @NatureComms @C_Althaus @GidMK @WHO There is a lot of heterogeneity, of course. So this report by @imperialcollege from 29.10. estimates two IFRs:
0.23% (0.14-0.42) in low-income countries
1.15% (0.78-1.79) in high.income countries

imperial.ac.uk/mrc-global-inf…