Determinants of sudden deaths among adults in India: The problem of overmatching in case-control studies

Dear Sir,

A recent study1 in the Indian Journal of Medical Research has ruled out any association between the COVID-19 vaccines and sudden cardiac deaths. This is quite reassuring at first glance as the majority of us have taken the vaccine. However, on studying the article critically, there are some concerns which we hope the authors may be able to explain.

There is a data discrepancy in Table 1. The ‘N’ value is different between ‘Vaccination Status’ and ‘Received COVID-19 vaccine before the death of cases’. The number of people who did not receive the vaccine is the same and appears to be accurate, but the number of jabbed is quite different. To explain, ‘Any vaccination’ is 577 among ‘cases’, but in the numbers in the rows below, 266 + 22 add up to only 288. Where did the remaining 289 disappear? The authors may be requested to explain this discrepancy.

Another concern, which is a major one, is the study design and the selection of controls. The limitation of comparative analytical study is that if the exposure under question is universal, it is impossible to identify it as a cause, however, strong the association. For example, if smoking was universal and all inhabitants on this earth had been smokers, it would have been impossible to identify smoking as a cause of lung cancer.

Against this background, the selection of controls is the biggest flaw in this article which renders the whole study null and void. The authors of the study say that the controls were selected from the neighbourhood of the cases. Given the mass vaccination drive implemented with gusto, the neighbours in all likelihood would have taken the vaccines in equal measures as the cases (sudden death victims). In case–control study, this oversight is termed, ‘overmatching’,2 i.e. matching for the variable under investigation, in this case, the vaccine. This will efface any cause-effect association, however, strong.

The confirmation of cause effect should depend on the totality of evidence such as robust randomized controlled trials, cohort designs backed by with post-mortem studies including histopathology and biochemical markers.