Is the clotting issue with the Oxford vaccine being blown out of proportion?

[Laroxe]

My criticism of the estimate in the paper is that the denominator for the estimate of the incidence of CVT after COVID-19 vaccination is likely wrong.

The paper looked at anonymized electronic health records from 59 healthcare organizations primarily in the USA. These organizations cover 81 million patients according to the authors. Their data looks at the population of people who received at least one dose of a COVID-19 vaccination before March 25, 2021. According to the CDC, 95 million people had received at least one dose of a COVID-19 vaccination by March 25. This amounts to about 29% of the US population. However, their dataset only has N = 489,871 patients who received a COVID-19 vaccination (0.6% of the 81 million patients). They observed 2 cases of CVT in this cohort, which is the source of the 4.1 per million statistic. However, only 490 thousand vaccinated individuals in this cohort is an implausibly low number; the expected number of vaccinated individuals should be ~23 million, which would lower the incidence of CVT to ~0.09 per million in the two weeks after vaccination.

This discrepancy in the observed number vaccinated versus the expected number vaccinated likely comes about because vaccinations in the US are being distributed through a number of means, not necessarily through one's primary healthcare provider (e.g. thorough pharmacies or mass vaccination sites run by the government). Therefore, patients who got vaccinated through these means would not have a record of vaccination in these electronic health records. This information would only be entered if the patient had to go to their healthcare provider for another reason. Thus, the method has a selection bias for vaccinated people who experienced adverse events. This error could have come about because the authors are from Oxford University in the UK, so they may not be familiar with how the US healthcare and vaccine distribution system has been operating.

Assuming that ~29% of the patient population was vaccinated (as opposed to the 0.6% they observe), the risk of CVT in the sample would drop significantly (by a factor of ~50, which would bring the incidence of CVT in the vaccinated cohort below the expected incidence of CVT).

I will write to the authors of the study to notify them of this major flaw in their data.

I think your first sentence summarizes the real problem, people are trying to make sense of very low frequency events and currently each new data set seems to muddy the water rather than clarify things. Hopefully given time the picture will become more accurate. Remember we still have the very basic argument about whether this is even a real effect, the data we have is so contaminated by other effects. There seems to be some basic differences in many of the groups studied these might include age, gender, risk of exposure, dose of exposure, ethnicity etc, so we end up with some big differences in findings.

 

[bhobba]

I will write to the authors of the study to notify them of this major flaw in their data.

I think you are correct. In fact, taking into account your analysis, I am not sure it would pass peer review. That is the problem with this pandemic - things move so fast papers before peer-review are discussed. They then get discussed in the media. A talk show discussed this paper last night where the host observed just who do you believe with so much conflicting information around. I well remember the egg I had on my face when I argued with Chemisttre about the value of face masks. We had experts, a number with Nobel's in Epidemomology like Peter Doherty, who said face masks are useless. I was wrong - Chemisttre was right:
https://www.nature.com/articles/d41586-020-02801-8

Despite the evidence of their value, there are still people that say they are useless. It's a problem - maybe even one of the biggest issues with this pandemic.

As an aside, I did not know that about the way things are done in the US either. You live and learn. Here in Aus we have a centralised database of everyone's Covid vaccine - what little has been done .

Thanks
Bill

 

[Ygggdrasil]

Useful slide from the April 23 ACIP meeting that puts the relative risk of thrombotic thrombocytopenia from the Johnson & Johnson vaccine into context with other vaccine-related adverse events:

(source)