Andrews, Colm D., Edward P. K. Parker, Elsie Horne, et al. 2026. “OpenSAFELY: Effectiveness of COVID-19 Vaccination in Children and Adolescents.” Epidemiology 37 (1): 141. https://doi.org/10.1097/EDE.0000000000001908.
A single sentence from this study, “Myocarditis and pericarditis were documented only in the vaccinated groups,” is being widely circulated without context. Most people sharing it have not read the paper, nor have they considered the methodological limitations that the authors themselves emphasize.
A major issue is ascertainment bias. Although the raw number of vaccinated and unvaccinated participants appears similar, this is misleading. Vaccinated children had substantially longer follow‑up time, while many unvaccinated children were censored early because they later became vaccinated. This creates large differences in person‑time, which is the correct denominator for rare adverse events.
The study notes that “both members of the matched pair were censored if the matched control was vaccinated,” but this does not equalize follow‑up. Vaccinated children still accumulate follow‑up time before their matched control gets vaccinated, whereas the unvaccinated child is removed from the analysis immediately once they receive a dose. A vaccinated child may contribute months of observation; their matched unvaccinated counterpart may contribute only weeks. This asymmetry inflates the opportunity to detect myocarditis in vaccinated children and suppresses it in unvaccinated ones.
This sits within a broader pattern of differences between groups. Unvaccinated children in this dataset had less medical monitoring, fewer clinical encounters, less testing, and were less likely to be evaluated for myocarditis or pericarditis. The authors explicitly warn that these conditions are more likely to be diagnosed in vaccinated individuals and that unvaccinated children had less follow‑up and different health‑seeking behavior, all of which bias detection toward the vaccinated group.
Another major limitation is that the study excluded clinically vulnerable children. This means the analysis compares two groups of unusually healthy children, which reduces the apparent benefit of vaccination and makes any adverse event appear proportionally larger. Including clinically vulnerable children, ie, those actually at risk of severe COVID‑19, would substantially change the risk–benefit balance.
This study is now being used out of context to support an antivaccine narrative. But the same flawed reasoning could be applied in the opposite direction: in this study, the only children who required ICU care were unvaccinated. That would be an equally invalid conclusion, because the number of ICU events is far too small to support meaningful comparisons, just as the myocarditis and pericarditis counts are too small, and too affected by follow‑up and ascertainment differences, to be interpreted as true differences in risk.
