BRADFORD HILL CRITERIA | EXPLANATION | EXAMPLE |
---|---|---|
Strength of association | Strong associations are most likely to be causal | High PA confirmed by accelerometry associated with 40% to 63% reduction in all-cause mortality31 |
Consistency of association | Similar outcomes in different populations and in studies that might be prospective or retrospective | Higher fitness associated with reduced all-cause mortality in large prospective23 and retrospective75 cohort studies |
Specificity of association | If an association is greatly increased in a specific group exposed to an intervention, the case for causation is strengthened | In a large cross-sectional study, cardiorespiratory fitness was strongly associated with reduced waist circumference and metabolic syndrome15 |
Temporality | A necessary criterion. Exposure must precede outcome | All prospective cohort studies show benefit of PA in event and mortality reduction |
Biologic gradient | Causal association is more likely if a dose response is demonstrable | Walking at higher volume or intensity associated with progressive reduction in CVD events and all-cause mortality19 |
Plausibility | Compatible with current knowledge. Not always a necessary criterion, as this knowledge might be new | Occupational energy expenditure has dropped by 140 calories per day since the 1960s. This might be a contributing factor to increased obesity76 |
Coherence | Observed outcome should be congruent with other known characteristics of disease biology or natural history | PA prevents oxidative stress, reduces inflammation, and improves endothelial function77 |
Experiment | Occasionally causation can be demonstrated by a controlled intervention in a clinical trial | Exercise intervention superior to PCI for event-free survival in randomized single-blind trial over 12 mo78 |
Analogy | Probably the weakest argument for association, but worth consideration if the outcome is adverse | Resistance training has similar benefits for glycemic control to aerobic training79 |
CVD—cardiovascular disease, PA—physical activity, PCI—percutaneous coronary intervention.
Data from Lucas and McMichael73 and Hill.74