Abstract
This investigation concerned the use of quantitative risk assessment for estimating cancer mortality at low-level exposures. We empirically tested whether extrapolating by linear no-threshold models predicted implausible risks at low-level exposures. Cadmium in cigarette smoke was the low-level exposure, and extrapolation was based on potencies estimated from an occupational study and a rodent cancer bioassay. Inhaled cadmium in mainstream and sidestream smoke was estimated from published laboratory experiments. Smoking-specific lung cancer and all-cause mortality rates were estimated from large population-based studies. The mortality rates, amount of inhaled cadmium, and potency values were used to construct life tables for calculating lifetime lung cancer risk with and without a contribution from cadmium in cigarette smoke. The epidemiologic data predicted that 1 to 18 lung cancer deaths per 10 000 smokers may be attributable to inhaled cadmium in cigarette smoke, or approximately 0.2% to 1.6% of smoking-induced lung cancer deaths. Upper 95% bounds on these figures are 7 to 95 lung cancer deaths or 1.6% to 8.8% of smoking-related deaths. The rodent data predicted that 80 to 416 lung cancer deaths per 10 000 smokers (95% upper bounds: 136–707) or 13% to 47% (23–81%) of smoking-induced lung cancer mortality may be attributable to cadmium in cigarette smoke. Linear extrapolation from human data appears to provide plausible estimates of risk at low doses. Considering the large number of carcinogens present in cigarette smoke, the extrapolation from rodents appears to overestimate human risks. Whether this discrepancy results from differences in potency for cadmium chloride aerosol as opposed to cadmium in particulate form, or from humans having greater sensitivity to cadmium's carcinogenic effect, or both, remains unclear.