A flawed paper claims COVID19 vaccines prevent cardiovascular complications for a YEAR after the shot

Tips on how to read the medical literature

This post gives a few tips for how to read medical papers. In a nutshell, you have to think about what researchers are doing, and search the supplement and beyond to find the bit of data that would be most useful. I will show you that for this paper (which was hidden online only), and just how damaging it is.

Here is the paper:

Basically, this paper has been used to claim that getting the COVID19 vaccine lowers the risk of a host of bad CV outcomes. Here is the key figure from the authors paper. Less than one means vaccination is better than being unvaccinated. Wow COVID vaccines work so well, and the effect keeps lasting!! (or so it seems…)

If you believe this then COVID vaccines PREVENT myocarditis. Even though they clearly CAUSE some myocarditis, the myocarditis they prevent is more than enough to offset it. And, they just keep working even a year later!

But how did they create the figure?

In order to make the figure they draw from 3 observational databases. They know the data some people got vaccinated, and for unvaccinated people they randomly assign a start date (PS this is an issue too). Then they propensity score adjust (adjust for confounder variables), and ask if vaccination is associated with these outcomes. And voila, the figure.

But the authors did something else in the paper. They asked if vaccination was associated with negative control outcomes.

What are negative control outcomes. Bapu Jena and I advocated for them 10 years ago in JAMA. They are basically things that CANNOT be linked to the exposure. Endpoints that CAN’T POSSIBLE change by covid vaccination— like wax in your ear canal.

COVID vaccines cannot cause more or less wax in the ear canal. That makes zero sense. So if your observational study finds that getting the jab leads to more office visits for wax in ear canal, it tells you the study is FLAWED. The people getting jab are fundamentally different. In this case, far more health care seeking.

Where is that figure? The Negative control outcomes

I looked in the paper and in the 177 page supplement, and it says only this.

Then I logged onto the web interface and found it. WOW!!! Look

All the falsification tests are linked. COVID shots mean you are way less likely to get pressure ulcers and way more likely to have wax in your ear.

This is wild. These should all be green. The authors have proven that despite propensity score matching the paper is hopelessly confounded.

Then they claim they calibrated for this, but that doesn’t solve the issue. Its just a bandaid on top of a hemorrhage.

Honestly, the paper is dreadfully bad. It is obvious that people who get vaccinated are different than those who don’t, and these patterns of covariates cannot be easily adjust for because they pertain to domains of life that are not readily captured in health records.

The paper is so bad, I would not have published it, and probably why it appears in a mediocre journal.

In the future, when we run vaccine RCTs we should take low risk populations, and randomize and have long follow up without crossover to measure these things. 20 year olds didn’t need COVID vaccines urgently. So we could have randomized 200k to vaccine or no vaccine and followed for a year. That was an evidence generation failure that could be a lesson for future drug and vaccine development.

If you enjoyed this paper dissection, read others (sample one below) and subscribe and support this Substack