Comparison of cancer risks projected from animal bioassays to epidemiologic studies of acrylonitrile-exposed workers.

Bioassay findings have demonstrated that acrylonitrile (ACN) is a rodent carcinogen, but the available epidemiologic evidence provides little support for the human carcinogenicity of ACN. This discordance between laboratory animal and human study findings is explored by determining post hoc the statistical power of 11 epidemiologic studies of ACN-exposed workers to detect the all-site and brain cancer excesses that are projected from rodent drinking water bioassay data. At reasonable estimates of the level and duration of exposures among the occupational cohorts, a majority of the human studies had sufficient power (> 80%) to detect the projected excesses, yet such responses were consistently absent. We conclude, subject to certain caveats, that the upper bound estimate of ACN's inhalation cancer potency of 1.5 x 10(-4) per ppm is too high to be consistent with the human ACN experience.