'Fundamental data flaws' in study exonerating role of Covid-19 vaccine in cardiac arrests

By Bhaskaran Raman* 
A study from GB Pant hospital in Delhi was recently published in the PLOS ONE journal. Based on a cohort of 1,578 patients, the authors conclude that the Covid-19 vaccines used in India reduced mortality. While the publication itself is careful in its wording of the conditions under which this reduction in mortality was seen, the press coverage was quick to headline that Covid vaccination does not increase risk of heart attacks.
While experts who understand the clinical aspects of the study can comment on those, this writeup examines the statistics behind the cited study and points out three main flaws. With suitable analogies, we seek to make it understandable to the layperson.
We shall use the following scenario to construct the analogies. Suppose that a new energy drink product has been introduced, designed to say help the customers complete a 5-km running race. However, in some people, the energy drink could potentially cause sudden hiccups, making the person abandon the race.
The parallels are obvious: a new vaccine has been introduced to help people complete their life, but in some people, the vaccine could cause sudden adverse events, resulting in cardiac arrest and possibly death.

First flaw: extremely unhealthy/at-risk cohort

The first statistical flaw in the study is that the cohort chosen in the study is extremely unhealthy. 201 out of 1578 participants, i.e. more than 1-in-8 participants, died within 30 days! This is an absurdly high death rate, indicating an extremely unhealthy/at-risk cohort. Therefore the study results are most certainly not applicable for the general population.

In terms of our analogy, suppose the study on 5-km race completion was done only on overweight people above 100-kg in weight, it clearly does not apply to the general population. This is depicted in the diagram below.

Second flaw: survivor bias

The second major statistical flaw in the study is the survivor bias. The cohort only consists of those who survived long enough to enroll in the hospital registry (NORIN-STEMI). It specifically excludes those who died suddently, while dancing, singing, gymming, working, walking, etc. This excluded group never even had a chance to reach the hospital.
In our analogy, suppose 100 people consume the energy drink one hour before the start of the race. Suppose 50 of them develop hiccups within 5 minutes, and hence do not even start the race. And of the 50 who start the race, 10 did not complete. And say in a group of 100 people who do not consume the drink, 40 did not complete the race.

If one looks at the statistics with the survivor bias, 10/50 (20%) did not complete among those who consumed the drink, while 40/100 (40%) did not complete among those who did not consume. This appears as a two-fold benefit. However, on removing the survivor bias, we find that 60% of those took the drink did not complete the race, which is 1.5 times worse than the control group which did not consume the drink. This is depicted in the figure below.

Third flaw: healthy vaccinee bias

The third major flaw in the study is the possibility that the vaccinated cohort is healthier than the unvaccinated cohort. Indeed, the authors do acknowledge this, but phrase this as “healthy user effect” and attribute it to healthy behaviours. This behaviour-related claim has no evidence and perpetrates the mainstream propaganda that somehow vaccinated people are more responsible in life.
What is much more likely instead is that the vaccine was not given at all, to the extremely unhealthy with various medical contraindications. Indeed, official websites give some such possible contraindications. The authors should examine this concrete possibility from their data, rather than conjecture personal behaviour as a reason.
The healthy vaccine bias can be illustrated in our analogy as follows. Suppose the energy drink is contraindicated for people above 150-kg in weight. Suppose that in a set of 100 people, 50 are above 150-kg and 50 are below. So only 50 are given the energy drink, of which say 10 did not complete the race. In the control group of 100 people, say 5 out of 50 under-150-kg did not complete, and 35 out of 50 above-150-kg did not complete the race.
With the healthy-participant bias, the statistics would work out to: 10/50 (20%) did not complete among the energy drink users, while 40/100 (40%) did not complete among the control group. Therefore the drink appears to have a two-fold benefit. However, if you compare only the under-150-kg group, the energy drink users had a two-fold harm in terms of race completion rate (10/50=20% versus 5/50=10%). This is depicted in the diagram below.

Note that the study reports an almost two-fold reduction in all-cause mortality in the vaccinated group compared to the unvaccinated. This points to a strong healthy-vaccinee bias, as the Covid-19 vaccines were designed to reduce Covid-19 deaths, not all-cause deaths. Even the vaccine manufacturers have not claimed benefit of the vaccine in terms of all-cause death reduction!

Possible analysis with data

Given the above flaws, what is the way forward? A possible data analysis which can be done with data similar to that in the published study is the following. We could look at the time-since-last-Covid-vaccination for all heart attack cases who came to the hospital. 

One could look for a temporal indication, although this cannot possibly find long-term causal relationship between the Covid-19 vaccines and cardiac issues. For detecting or ruling out long-term correlation, the only high quality study is a randomised controlled trial.

---
*Professor at IIT Bombay; views are personal. He has authored the book “Math Murder in Media Manufactured Madness”, presenting simple math to illustrate various absurdities related to the mainstream Covid-19 narrative; available at: https://bhaskaranraman.in/