Another Covid-vaccine themed paper caught my eye (and again, h/t Deutsches Ärzteblatt). It claims to deal with “accelerated waning of the humoral response to COVID-19 vaccines in obesity”. Peer-reviewed and published in nature medicine, awesome! Please, you 47 authors (or many more, because the list of authors includes the PITCH consortium), tell us what you did:
We studied the relationship among body mass index (BMI), hospitalization and mortality due to COVID-19 among 3.6 million people in Scotland using the Early Pandemic Evaluation and Enhanced Surveillance of COVID-19 (EAVE II) surveillance platform. We found that vaccinated individuals with severe obesity (BMI > 40 kg/m2) were 76% more likely to experience hospitalization or death from COVID-19 (adjusted rate ratio of 1.76 (95% confidence interval (CI), 1.60–1.94).
Amazing! It is not about how many angels can dance on the head of a pin, but about, given an observed group of pin-head-dancing angels, the relation of the angels’ body weights to their dancing capabilities. We are dealing with a new kind of Covid-vaccine paper: nobody is interested in vaccine efficacy anymore. It is now all about the vaccines being excellent, of course (the authors are quick to note, lest anyone take offense, that “COVID-19 vaccines reduce the risk of symptomatic infection, hospitalization and mortality due to COVID-19”), but with varying degree of excellence (from “ridiculously excellent” to “amazingly excellent“) for different groups of people.
I am not in the mood for vivisection, so I will just emphasize three points.
First, the authors claim to have studied both hospitalizations and deaths, but there actually is almost no reporting on deaths in the paper. They explicate (“Results” section, p. 2 in the pdf version):
Between 14 September 2020 and 19 March 2022, there were 10,983 individuals (0.3%, 6.0 events per 1,000 person-years) who had a severe COVID-19 outcome: 9,733 individuals were hospitalized and 2,207 individuals died due to COVID-19.
This is proof that peer review has been sloppy. Compare to the “Methods” section (p. 10 in the pdf):
All the COVID-19-related hospital admissions or deaths were selected between 14 September 2021 and 19 March 2022.
Now, what is the correct year, 2020 or 2021? The former does not make sense because it precedes the start of the vaccination campaign (8 December 2020), so we go for the latter. Summing up all Covid deaths in Scotland from calendar week 37 of 2021 (starting 13 September) until calendar week 11 of 2022 (ending 20 March), we get 2,994. The 2,207 deaths mentioned above refer to a (subset of the) vaccinated population, altogether 3.6 million people. The remaining 1.9 million scots account for the remaining 787 deaths. Assuming that the latter are all unvaccinated (which is surely not the case), we compute a raw vaccine effectiveness of around 32% (or maybe 37% if we restrict to the 18+ age groups). Basically, the vaccines are useless, and the 47+ authors discover with amazement what we all knew before, that obese people are at greater risk of all kinds of stuff.
Second, Bayesian datacrime may be present (everything happening during the 14 days post vaccination is being disregarded) but IMHO is playing a minor role here. Most vaccinations happened well before the observation period (if my above assumption of 14 September 2021 being the start date is correct): the maximum possible number of person-years to be observed would be something like 1.8 million (half a year for 3.6 million people), and the number of 10,983 individuals with severe outcome would produce a rate of 10,983 / 1,800 = 6.1 events per 1,000 person-years, very close to the 6.0 stated in the paper.
Third, and most importantly, what an amazing dataset the EAVE II seems to be! It includes 99% of Scots, not only regarding Covid stuff, but also regarding all kinds of “confounders”, i.e., variables of interest. Consider this most amazing section on EAVE II statistical analysis (“Methods” section, p. 10 in the pdf):
We calculated the frequency and rate per 1,000 person-years of severe COVID-19 outcomes for all demographic and clinical factors. Generalized linear models (GLMs) assuming a Poisson distribution with person-time as an offset representing the time at risk were used to derive rate ratios (RRs) with 95% CIs for the association between demographic and clinical factors and COVID-19-related hospitalization or death. aRRs were estimated adjusting for all confounders, including age, sex, SIMD, time since receiving the second dose of vaccine, pre-existing comorbidities, the gap between vaccine doses, previous history of SARS-CoV-2 infection and calendar time. SIMD looks at the extent to which an area is deprived across seven domains:
income, employment, education, health, access to services, crime and housing. SIMD was allocated based on an individual’s home postcode, with quintiles of population ranging from 1 for the most deprived 20% to 5 for the least deprived 20% of the population.
I am generally not happy with government-related entities having so much control over citizens and their data, but think about what could be done with a dataset like this (Fabian, this might even be better than Minnesota)! Basically, the whole Covid vaccine business could be exposed as the scam that it is. And that’s why only the good people, like the chosen 47, are going to gain access.
"And that’s why only the good people, like the chosen 47, are going to gain access."
I do not share your optimism that all 47 authors are going to gain access.