Causation: The elusive grail of epidemiology

Abstract

The paper discusses the evolving concept of causationin epidemiology and its potential interaction with logic and scientific philosophy. Causes arecontingent but the necessity which binds them totheir effects relies on contrary-to-fact conditionals,i.e. conditional statements whose antecedent is false.Chance instead of determinism plays a growing role inscience and, although rarely acknowledged yet, inepidemiology: causes are multiple and chancy; a priorevent causes a subsequent event if the probabilitydistribution of the subsequent event changesconditionally upon the probability of the prior event.

There are no known sufficient causes in epidemiology.We merely observe tendencies toward sufficiency ortendencies toward necessity: cohort studies evaluatethe first tendencies, and case-control studies thelatter.

In applied sciences, such as medicine andepidemiology, causes are intrinsically connected withgoals and effective strategies: they are recipewhich have a potential harmful or sucessful use; theyare contrastive since they make a differencebetween circumstances in which they are present andthose in which they are absent: causes do notexplain Event E but event E rather thaneven F. Causation is intrinsically linked with thenotion of ``what is pathological''.

Any definition of causation will inevitably collapseinto the use made of epidemiologic methods. Theprogressive methodological sophistication of the lastforty years is in perfect alignment with a gradualimplicit overhaul of our concept of causation.