Cancer Dose-Response Assessment of Ethylene Oxide Based on Animal and Human Data. R. L. Sielken Jr., Sielken, Inc., 3833 Texas Avenue, Suite 230, Bryan, Texas 77802
The best available human epidemiological data for ethylene oxide (EO) cancer dose-response modeling are the Union Carbide Corporation study of 1,896 workers exposed during EO manufacturing and the NIOSH study of 18,239 workers exposed to EO in the sterilization industry. Poisson regression analyses are presented for lymphatic and hematopoietic cancer, leukemia, and the combination of lymphocytic and non-Hodgkin’s lymphoma. Different alternatives are evaluated for the background response rate, sex, dose-metric, latency period, lag period, and functional form of the impact of the EO dose. Best fits and sensitivity analyses are presented. Added cancer risks are estimated for both occupational and environmental exposures. Animal bioassay data include the Bushy Run Research Center EO inhalation study of male and female rats, the NIOSH study of male rats, and the NTP study of male and female mice. There are several factors and alternatives for each factor in the tree of potentially relevant dose-response analyses of the animal bioassay data (e.g., data set, response, severity of response (onset vs death), number of animals at risk, dose-response model, dose metric, and method of interspecies extrapolation). Interesting findings from the dose-response evaluation of the several thousand combinations of alternatives are briefly noted. Additional impacts on interspecies extrapolation are also noted. For example, the use of physiologically-based pharmacokinetic modeling instead of default assumptions of species equivalence on the (body weight)2/3 or (body weight)3/4 scales reduces estimates of human cancer risk by 4 to 13 fold. Interspecies extrapolation incorporating interspecies differences in background transition rates from stage to stage in multistage carcinogenic processes instead of default assumptions can reduce estimates of EO human cancer risks by orders of magnitude. Interspecies extrapolation can also be significantly impacted by nonlinearity in the formation of DNA adducts at high doses, induction of DNA repair at low doses in rats, human adaptation to higher endogenous levels of DNA adducts than in rats and mice, and not much interindividual variability in humans in adduct levels — either DNA adducts or hemoglobin adducts. High-to-low-dose extrapolation incorporating the required close proximity of two genetic events and the resultant quadratic dose-response relationship instead of default assumptions of linearity reduces estimates of human cancer risks by orders of magnitude at low doses. We are on the road to incorporating more science and more of the available data into the cancer dose-response assessment of ethylene oxide. We have already identified large orders of magnitude differences between current regulatory risk assessment of ethylene oxide and the new perspectives obtained on that road.