Calculating preventable risk fractions for exposure-reducing interventions.

How can causal models be used to quantify the fractions of risk associated with environmental and occupational exposures that would be prevented by reducing exposures by different amounts? This paper provides a constructive answer. It introduces three key metrics - Interventional Probability of Causation (IPoC), Causal Assigned Shares (CAS), and Preventable Risk Fraction (PRF) curves - to help overcome the limitations of traditional association-based metrics, such as Population Attributable Fractions (PAFs), which are sometimes misused to answer interventional causal questions. The tools introduced here provide scenario-specific, individual-level predictions of risk reductions grounded in mechanistic causality rather than associations. Using case studies of benzene exposure and acute myeloid leukemia (AML), smoking and lung cancer, and blood lead levels and mortality, we demonstrate how PRF curves quantify the potential risk-reduction benefits caused by exposure reductions at both the individual and population levels, even under uncertainty or heterogeneity. Monte Carlo simulations capture inter-individual variability, and scenario analyses identify practical thresholds where additional exposure reductions yield minimal added benefit. These methods can provide evidence-based assessments of how specific exposure reductions affect risk. By shifting the focus from attribution to prevention of harm, this framework can potentially advance risk assessment, policy development, and legal decision-making. It offers a simple, easily visualized, transparent, and scientifically rigorous approach to identifying causally effective interventions and quantifying risk-reduction benefits.