Bradford Hill Considerations
Bradford Hill's nine considerations are viewpoints for judging whether an observed association is causal, offered as aids to judgement rather than as a checklist or a statistical test.
Austin Bradford Hill set them out in a 1965 presidential address to the Royal Society of Medicine. The nine are strength of association, consistency across settings and investigators, specificity, temporality, biological gradient or dose-response, plausibility, coherence with what else is known, experiment, and analogy. Hill was explicit that none could deliver indisputable evidence for or against causation and none could be demanded as a prerequisite; only temporality is strictly necessary, since a cause must precede its effect.
The setting was tobacco and lung cancer, where randomising people to smoke was impossible. What carried the argument was consistency across many populations and designs, a steep gradient with cigarettes per day, and falling risk after cessation. The same reasoning runs through pharmacovigilance, where causality assessment leans on temporality and on whether an event resolved when the drug stopped and returned when it was restarted.
Used properly the considerations decide what evidence to seek next rather than producing a score. Temporality and biological gradient discriminate hardest. Strength earns its place because a very large association is hard for residual confounding to manufacture, whereas a hazard ratio of 1.1 in a cohort sits well inside what unmeasured differences between groups can produce.
The failure mode is treating the list as a scorecard where six ticks out of nine equals causation. Plausibility is the cheapest to satisfy and the most abused: a receptor exists, a pathway is drawn, therefore the effect is real. Almost every compound sold on the grey market has a plausible mechanism story, and plausibility is bounded in any case by whatever biology was known at the time.