Our new publication using data from the adult population in Austria is out in @OFIDJournal. We found 17 total COVID-19 deaths with/from Covid 19 among 500k previously infected adults in Austria between 10/1/2021-12/31/2021; 6 were vaccinated x2 w/mRNA; adjusted hazard ratio (aHR) of 2mRNA vaccines= 0.24 (0.09 - 0.65)
We found substantial healthy vaccinee bias with (adjusted) 36% lower (aHR 0.64 [0.56-0.74]) non covid deaths among 2x vaccinated (it's so important that all observational vaccination effectiveness studies report a negative control like this; see our @NEJM correspondence www.nejm.org/doi/full/10.1056/NEJMc2306683)
Yes, confidence intervals are overlapping, so how much of the "benefit" vs covid mortality is actually a result of better underlying health among the vaccinated is unclear.
Results were similar for extended observation period Jan-Jun 2022 (46 total covid deaths; 28 were vaccinated; aHR of 2mRNA vaccines= 0.47 [0.25 - 0.87]), and lower associated effectiveness vs infections and less healthy vaccinee bias during this period (see supplement Table 8)
Our study was such an interesting demonstration of the limits of observational data for estimating vaccine effectiveness or overall societal utility of a vaccination, particularly in the case where the absolute risks are so low.
Highlights from the discussion re our findings among Austrians with previously documented covid infections:
"In view of the very low absolute COVID-19 mortality, one may question the overall clinical significance and the cost-effectiveness of these vaccinations for prevention of COVID-19 deaths [21]. Due to very sparse outcome events, we refrained from calculating numbers needed to vaccinate and cost-effectiveness analyses of these vaccinations, but strongly recommend further studies to address these issues. It should be noted that our analyses do not consider potential vaccination harms or other potential negative health consequences following SARS-CoV-2 infection, except for COVID-19 deaths [22, 23]. Our results are also observational in nature and subject to biases such as healthy vaccinee bias, which our results suggest is present. This would likely overestimate any true vaccine effectiveness.
In individuals with one and two vaccine doses, we observed a 27% and 36% lower non-COVID19 mortality risk, respectively, when compared to unvaccinated individuals... This suggests that the vaccinated in our study may have been fundamentally different from the unvaccinated and likely had better underlying health. This would have tended to bias our results in the direction of overestimating vaccine effectiveness and highlights the potential challenges of using retrospective observational data to draw causal conclusions about interventions.
Potential behaviour changes in response to vaccination and/or SARS-CoV-2 infection, as well as test results, may introduce further biases. Furthermore, we did not have access to data on co-morbidities, socioeconomic status, and medications, preventing adjustment for these factors.
The fact that we found one dose of vaccination to be associated with more protection than two alone suggests residual bias.
Even during the Delta wave, which saw the largest number of deaths attributed to COVID-19 in Austria after the introduction of vaccines, the risk of COVID-19 deaths in the previously infected population was extremely low, and this was sustained even during the first half-year of 2022, when Omicron was predominant.
Our findings are subject to several limitations inherent to observational studies of COVID-19 vaccine effectiveness... These include limitations in data reporting, misclassification between COVID-19 and nonCOVID-19 deaths, and variation on access to and indication for SARS-CoV-2 tests
Further studies on various SARS-CoV-2 infection outcomes are warranted to provide estimates on the public health significance and cost effectiveness of SARSCoV2 vaccines in previously infected individuals."
In other words, because most people have already been infected, we need randomized data (or data not prone to underlying bias) of mRNA vaccines looking at both benefits and harms to make evidence-based covid vaccination recommendations.
I am so grateful for this research collaboration with talented Austrian researchers, in particular Alena Chalupka and Stefan Pilz. Also working with Dr. Ioannidis was such a joy. He provided so much helpful feedback. This is of course not always the case for someone with such a prestigious position, but was absolutely the case for him.
academic.oup.com/ofid/advance-article/doi/10.1093/ofid/of…
P.S. I guess OFID only allows one degree after names. I did not lose my MD and John Ioannidis, to my knowledge, did not lose his DSc. Just sayin