Causal Inference in Cancer Epidemiology

Judgments about causality are central to the development of interventions intended to reduce exposure to risk factors that cause cancer. Because causation is not directly observable in medicine, scientists and philosophers have had to develop sets of constructs and heuristics that define “cause” operationally. The criteria in this framework, often attributed to the British medical statistician Sir Austin Bradford Hill or to the 1964 Report of the US Surgeon General on tobacco, include consistency, strength of association, specificity, temporality, coherence/plausibility/analogy, biological gradient, and experiment. This chapter reviews these criteria in depth and considers the challenges of applying them to population research on cancer. It discusses the concepts of causation in the context of the multistage nature of cancer, the “counterfactual” notion of causation, the component cause model for understanding diseases with multiple causes, and the “weight of the evidence” approach for integrating information from multiple lines of research.