Human Interindividual Variability in Cancer Risks—Technical and Management Challenges

Existing risk assessment procedures for carcinogens are intended to be “conservative” in the uncertainty dimension—giving estimates that are expected to be higher than true risks for typical people. However, these procedures do not consider the likely variability in susceptibility among individual people. This paper updates previous estimates of the likely extent of this variability for metabolically activated, genetically-acting carcinogens based on recent information on human interindividual variability in metabolic activation, detoxification, and DNA repair. The resulting expected skewness of cancer risk distributions is estimated using Monte Carlo simulations of both variability and uncertainty.

Some risk management implications are:

1. When evaluating the fairness of a particular risk distribution, managers need to gain familiarity with a three-dimensional characterization—X level of risk, for the Yth percentile individual (addressing variability) with Z degree of confidence (addressing uncertainty).

2. To the extent that variability distributions are skewed (e.g., with a long tail extending to high values) population mean risks will tend to exceed risks for median individuals. Together with the skewness in uncertainty distributions, this implies that “expected value” estimates of aggregate population risks—the estimates of interest for cost benefit analyses—are likely to be closer to traditional upper confidence limit risk estimates than has often been assumed in the past.