Causal interpretation requires appropriate study design

Excerpt

To the Editor: I read with interest the paper by Richardson et al. published in Diabetologia [1]. The paper reports a sort of case–control study, but gives very little information about the age or other demographic or clinical characteristics of the cases and controls, so it is not possible to properly assess the appropriateness of the control group. At the very least, it is unlikely to represent an unbiased sample of the population from which the cases arose (which is not described). For example, would one expect neonates to have had the opportunity to be infected with enterovirus? Presumably not. Therefore, except for neonatal cases (of which there were none), neonates should not be used as controls. …