Citation

Dekkers OM. Eur. J. Epidemiol. 2019; 34(6): 533-535.

Affiliation

Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark. o.m.dekkers@lumc.nl.

Copyright

(Copyright © 2019, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s10654-019-00507-4

PMID

30887378

Abstract

Epidemiologists face two fundamental and interrelated problems when judging causality: knowledge is fallible, and studies are imperfect. In medicine, this will always leave a degree of uncertainty in scientific judgements. From an epistemological point of view, even randomized trials cannot be regarded as the ultimate proof to establish a causal relation. Given this inherent uncertainty it is no surprise that much attention has been drawn to the question how we can move from an association to a valid judgement of causation. It was exactly this question that urged Austin Bradford Hill more than 50 years ago to his well-known and still worth-reading paper, in which his nine viewpoints (often referred to as Hill’s criteria) to judge causality were described.

In their paper, published in this issue of the European Journal of Epidemiology, Olsen and Jensen call it a time for revision of Hill’s criteria. They point towards the broad notion of new methodological developments in epidemiology that bring about the need for such a revision. No empirical argument is provided, in a sense that is assessed when and why these criteria were unable to display their role. Olsen and Jensen specifically argue for adding a consequence criterion to Hill’s list, ‘for epidemiological practice to be of use in real life’.

Very likely, Hill would have welcomed such discussion about his viewpoints; nowhere in his paper he claims that the presented list to judge causality is final, complete or sufficient. The only condition from Hill’s list that can be regarded as necessary, is temporality, as in medicine causes precede their consequences; however, as a feature to distinguish a mere association from causality, temporality is hardly helpful. Though not included as criterion ...

Language: en