3 ms·
Probability and Risk: Is vaccine efficacy a statistical illusion?
- gus_massa 5y agoThe article explain why studies with a placebo branch are important. You have a lot of known unknown and unknown unknown. In particular, as discussed in the article, counting the number of cases/death is hard. The main problem with the article is that all the vaccines are approved after a RCT were the vaccine is compared to a placebo. For example, look at the slide #6 in https://investors.biontech.de/static-files/53f0968a-279b-4f82-a2fc-d67dcb6e4e91 https://investors.biontech.de/static-files/53f0968a-279b-4f8... and more details in https://en.wikipedia.org/wiki/Pfizer%E2%80%93BioNTech_COVID-19_vaccine https://en.wikipedia.org/wiki/Pfizer%E2%80%93BioNTech_COVID-...
- charbonneau2 5y ago> The main problem with the article is that all the vaccines are approved after a RCT were the vaccine is compared to a placebo The author talks about effectiveness wrt mortality rates. No statistically significant difference was found during RCTs: > During the blinded, controlled period, 15 BNT162b2 and 14 placebo recipients died https://www.medrxiv.org/content/10.1101/2021.07.28.21261159v1.full.pdf+html https://www.medrxiv.org/content/10.1101/2021.07.28.21261159v...
- charbonneau2 5y agoAs was mentioned in the comment section of the article, it would be interesting to see if this effect really exists (by trying to shift the numbers backwards). Furthermore, I think it is a striking reminder that even (seemingly) simple data like mortality can be subject to strong bias.