Uncertainty in risk characterization of weak carcinogens

Epidemiologic inference is subject to uncertainty that is inherent to observational approaches. It was shown by the present analysis that an SMR of 1.4, as observed for 2,3,7,8-tetrachlorodibenzo-para-dioxin in the pooled analysis by IARC, would be observed in the absence of carcinogenic risk, if the smoking prevalence of cohort members was as low as 80%. It was also shown that a 1.4-fold increase in lung cancer risk among nonsmokers, which roughly corresponds to a daily consumption of one cigarette, might be extremely difficult to identify if the subjects are exposed to strong carcinogens such as those that result from cigarette smoking. On the other hand, a genotoxic agent that increases the lung cancer risk of smokers 1.4-fold, could increase the risk to nonsmokers by as much as 6.3-fold. In light of these uncertainties in epidemiologic inference, attempts to estimate the absolute risk in human populations by epidemiologic studies should be made with caution.