Bradford Hill criteria for causation

Occupational physicians could not prevent disease at work without a sound way to decide which hazards actually caused sickness and injury. Austin Bradford Hill put that question to them in 1965, in the first President's Address to a newly formed Section of Occupational Medicine. He asked in what circumstances one could pass from an observed association to a verdict of causation. The talk was printed that May in the Proceedings of the Royal Society of Medicine.

His answer was a list of nine aspects to weigh: strength, consistency, specificity, temporality, biological gradient, plausibility, coherence, experiment and analogy. For strength he used Percival Pott's observation of scrotal cancer in chimney sweeps, a risk nearly 200 times that of other occupations; small associations, he said, could more easily come from bias or confounding. A dose response made a causal link more likely. For coherence he pointed to changes in the bronchial lining and to animal tests of cigarette smoke, which together told a coherent story about cigarette smoke and cancer. Evidence that disease fell after an exposure was removed could give the strongest support of all.

Later writers renamed the list the Bradford Hill criteria. Temporality, the requirement that exposure come before disease, is perhaps the only one that epidemiologists universally treat as essential; Rothman and Greenland, who doubted the value of the others, called it inarguable. Hill closed by reminding his audience that all scientific work is incomplete and open to revision as knowledge advances.

He spoke twelve years after Watson and Crick proposed the double helix, when the biology between most exposures and diseases was still a black box. A 2015 review in Emerging Themes in Epidemiology described his nine points as the most frequently cited framework for causal inference in epidemiologic studies and examined how molecular and genomic data now bear on each of them.