Defining causal relationships between viral infections and human diabetes

Abstract

Type 1 diabetes is a multifactorial, chronic disease with a long induction period, and little is known about its etiology. This poses specific challenges to the study of viral infections as a potential causes. Statistical associations may be due to either chance, bias, or causality. Study designs (in combination with methods of viral detection) have important impact on ability to make causal inference under various hypothetical mechanisms of disease induction. Causes of multifactorial disease may be categorized as necessary, sufficient, both, or neither. Most single risk factors for type 1 diabetes is likely to be neither. Causality can never be inferred with certainty, but practical criteria can be employed to assess various aspects of the available evidence. Koch's postulates and Hill's criteria are well known, but often misinterpreted. Many established causal relations do not fulfil most of these criteria. Subjective judgment will always be part of the causal inference process, but modern analytical methods may help validate the process. While summarizing existing evidence with respect to potential causality may help to justify specific intervention trials, such as vaccination, the risk of side effects should be seriously considered.