Trusting the wrong data
According to the Guardian (and various other sources), the Pfizer Covid vaccine “may be” less effective in people with obesity, based on an Italian study of antibody levels in 248 vaccinated healthcare workers, 26 of whom were classified as obese and 56 more as overweight.
In the actual randomised clinical trial of the Pfizer vaccine, there were 13,000* people classified as obese and a further 13,000 classified as overweight. The trial considered obesity as a factor that might affect the vaccine success. In people with obesity, the estimated efficacy based on actual Covid cases was 95.4%. In those without, it was 94.8%. This is not mentioned at all in the Guardian story.
It’s quite common for clinical trials to end up with a very non-representative subset of the population where you want to use the treatment. That’s much less true of the US Covid vaccine trials (apart from being US-biased). They made serious efforts to recruit a wide range of people, and did pretty well. Any risk factor that’s common in the United States is likely to have been common in the trial population, so it’s already represented in the results — even if the results weren’t specifically reported based on that factor, as they are for obesity.
There’s some good research with careful estimation efforts being made to check that efficacy in practical use matches up to the trial, but just measuring antibodies in a small convenience sample isn’t worth a headline.
* There are a bunch of different analysis populations in the trial, so the number will be different in different places. They’re all big.
Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »