5.2 Bradford Hill criteria for causation

Establishing causation in epidemiology is crucial for understanding disease origins and developing interventions. The Bradford Hill criteria provide a framework for evaluating evidence of causal relationships between exposures and health outcomes.

These nine criteria, including strength, consistency, and temporality, guide researchers in assessing the likelihood of causation. While not a definitive checklist, they help distinguish between causal associations and those due to chance or bias in observational studies.

Bradford Hill Criteria for Causation

Overview of the Criteria

• The Bradford Hill criteria, proposed by Sir Austin Bradford Hill in 1965, are a set of nine principles used to establish epidemiologic evidence of a causal relationship between a presumed cause and an observed effect
• The nine criteria are: strength, consistency, specificity, temporality, biological gradient, plausibility, coherence, experiment, and analogy
• Serve as a framework to guide the evaluation of evidence and the determination of the likelihood of a causal relationship
• Not meant to be a definitive checklist for establishing causality
• The presence or absence of any one criterion does not necessarily prove or disprove causality
• The weight given to each criterion may vary depending on the specific situation and the available evidence

Applying the Criteria in Practice

• Researchers use the Bradford Hill criteria to assess the strength of evidence for a causal relationship in observational studies (cohort studies, case-control studies)
• The criteria help to distinguish between associations that are likely to be causal and those that may be due to chance, bias, or confounding
• Applying the criteria requires a comprehensive evaluation of the available evidence from multiple studies and sources
• The criteria are often used in conjunction with other epidemiological concepts (confounding, bias, effect modification) to establish a causal inference

Strength, Consistency, and Specificity Criteria

Strength of Association

• Refers to the magnitude of the association between the exposure and the outcome, with stronger associations providing more support for causality
• Often measured by the relative risk (RR) or odds ratio (OR), with higher values indicating a stronger association
  • Example: A study finds that smokers have a 10-fold increased risk of lung cancer compared to non-smokers (RR = 10)
• However, a strong association alone does not necessarily imply causality, as confounding factors or bias may also result in strong associations
• The strength of association must be considered in the context of other criteria and potential sources of bias

Consistency of Findings

• Refers to the reproducibility of the association across different studies, populations, and settings
• Consistent findings from multiple, well-designed studies using different methods and populations provide stronger evidence for causality
  • Example: Numerous studies in different countries and populations consistently show an association between smoking and lung cancer
• Inconsistent results, on the other hand, may suggest that the association is not causal or that there are unidentified confounding factors or effect modifiers
• Consistency across studies helps to rule out chance findings and strengthens the case for causality

Specificity of Association

• Refers to the extent to which the exposure is associated with a particular outcome, rather than multiple outcomes
• A specific association between an exposure and a single outcome provides stronger evidence for causality, as it is less likely to be due to chance or confounding
  • Example: Asbestos exposure is specifically associated with mesothelioma, a rare cancer of the lung lining
• However, many exposures can lead to multiple outcomes, and the absence of specificity does not necessarily rule out causality
• Specificity is a supportive but not necessary criterion for causality

Temporality, Biological Gradient, and Plausibility Criteria

Temporality

• Refers to the requirement that the exposure must precede the outcome in time for a causal relationship to be established
• This is a necessary but not sufficient criterion for causality, as the exposure must occur before the outcome to be considered a potential cause
  • Example: Smoking must occur before the development of lung cancer for it to be considered a potential cause
• Reverse causality, where the outcome precedes the exposure, can be ruled out if temporality is established
• Temporality helps to establish the direction of the association and is a key criterion for causal inference

Biological Gradient (Dose-Response Relationship)

• Refers to the presence of a monotonic relationship between the level of exposure and the risk of the outcome
• A dose-response relationship, where increasing levels of exposure are associated with increasing risk of the outcome, provides support for causality
  • Example: Studies show that the risk of lung cancer increases with the number of cigarettes smoked per day and the duration of smoking
• The shape of the dose-response curve (linear, threshold, or U-shaped) can provide insights into the nature of the causal relationship
• A clear dose-response relationship strengthens the case for causality, but its absence does not necessarily rule out a causal association

Biological Plausibility

• Refers to the biological and epidemiological reasonableness of the proposed causal relationship
• The proposed causal mechanism should be consistent with existing knowledge of the biology and pathophysiology of the exposure and outcome
  • Example: The biological mechanisms by which smoking causes lung cancer, such as DNA damage and inflammation, are well-established
• However, plausibility is limited by current scientific knowledge, and a lack of a known biological mechanism does not necessarily rule out causality
• Plausibility is a supportive criterion that can strengthen the case for causality, but it is not a necessary condition

Coherence, Experiment, and Analogy Criteria

Coherence with Existing Knowledge

• Refers to the consistency of the proposed causal relationship with existing knowledge and theories in the field
• The causal interpretation should not seriously conflict with established facts or principles in biology, pathology, or epidemiology
  • Example: The association between smoking and lung cancer is coherent with the known carcinogenic effects of tobacco smoke
• Coherence does not require complete understanding of the causal mechanism, but the proposed relationship should fit within the current scientific framework
• Coherence helps to integrate the proposed causal relationship with existing scientific knowledge

Experimental Evidence

• Refers to the evidence from experimental studies, such as randomized controlled trials (RCTs), that support the causal relationship
• Experimental evidence, where the exposure is manipulated and the outcome is observed, provides the strongest support for causality
  • Example: RCTs demonstrating that smoking cessation reduces the risk of lung cancer provide strong evidence for causality
• However, experimental studies are not always feasible or ethical, particularly for harmful exposures or rare outcomes
• In the absence of experimental evidence, well-designed observational studies can provide support for causality

Analogy to Other Causal Relationships

• Refers to the use of similar causal relationships in other contexts to support the plausibility of the proposed causal relationship
• If a similar exposure-outcome relationship has been established in another context, it can provide support for the plausibility of the proposed causal relationship
  • Example: The causal relationship between asbestos exposure and lung cancer is analogous to the relationship between smoking and lung cancer
• However, the absence of analogous relationships does not necessarily rule out causality, as each exposure-outcome relationship is unique
• Analogy can provide additional support for causality, but it is not a necessary criterion