Bradford Hill criteria for assessing causation and the definitions used in this review
| Criterion | Bradford Hill criteria (1965) | Application in this review |
|---|---|---|
| 1. Strength of the association | The strength of a supposed association between an intervention and an outcome is determined by the appropriate statistic used to measure the protective effect of an intervention (eg, relative risk or OR). This is the most important factor determining causation | A statistically significant change (p<0.05) in alcohol consumption or alcohol-related harms, in the expected direction. The exact magnitude of the association was assessed on a study by study basis |
| 2. Consistency | Has it been repeatedly observed by different persons, in different places, circumstances and times? | Whether different studies conducted in different locations, in different populations, by different investigators and at different times have reported similar findings |
| 3. Specificity | Specificity is present when the intervention is exclusive to the outcome and when the outcome has no other known cause or associated risk factors; cautions that this criterion should not be overemphasised and that if specificity is not apparent, this does not preclude causation | If pricing was the only reason that alcohol consumption or alcohol-related harm could have fallen, this adds to the argument for causality. However, if a price intervention was one of a number of alcohol policy interventions, then this criterion is not satisfied |
| 4. Temporality | Refers to temporal relationship of association between exposure and disease outcome; to infer causality, exposure must precede outcome | The pricing intervention studied must have taken place before a change in alcohol consumption or harm was observed |
| 5. Dose–response | If the association is one in which a dose–response curve or biological gradient can be observed, this adds to the case for causality | If interventions leading to a larger increase in prices had a greater effect on alcohol consumption and alcohol-related harm than interventions where the price change was small, or if studies demonstrate that different minimum prices have differing effects, in the expected direction |
| 6. Plausibility | A likely biological mechanism linking the intervention to the observed findings helps to explain causality; plausibility depends on biological knowledge of the day | Studies that found an association between price and population-level alcohol consumption and that heavier drinkers tend to purchase the cheapest alcohol could demonstrate plausibility |
| 7. Coherence | When the evidence from different disciplines sources ‘hangs well together’ and does not conflict with other generally known facts, this criterion is met | Describes whether studies conducted in different settings or disciplines had complementary findings. Will not be demonstrated by a single study in isolation but rather the evidence base as a whole |
| 8. Experiment | Experimental evidence from laboratory studies or RCTs could potentially provide strongest support for causation This criterion often provides the strongest support for causation and describes whether there is empirical evidence for the association | In addition to laboratory studies and RCTs, natural experiments with before-and-after measures could also show the effectiveness of minimum unit pricing in a ‘real-world’ setting |
| 9. Analogy | Causality is supported by analogy if there are similar associations or causal relationships in other areas of relevance, weakest form of evidence of causality | Other areas of relevance include whether higher taxation on alcohol is associated with reduced alcohol consumption and alcohol-related harm, and may require drawing on additional literature outside of the main systematic review |
RCTs, randomised controlled trials.