Epidemiology: Association and Causation

Key points

• • Finding that an exposure and a disease occur together (an association) does not by itself prove that the exposure causes the disease.
• • An association may be false because of chance, bias in the study, or confounding by a third factor linked to both the exposure and the disease.
• • An association is more likely to be real when chance, bias and confounding are made unlikely, and further judgement is then needed to weigh whether it is causal.
• • A set of considerations (the Bradford Hill viewpoints) helps weigh causation, including the strength of the link, its consistency across studies, and a dose-response gradient, but they aid judgement rather than prove cause.
• • Other pointers to causation are that the exposure comes before the disease (temporality), that the link is biologically plausible, and that removing the exposure reduces the disease.
• • Temporality (cause before effect) is the one essential requirement; the others add weight but are not each essential.