In US EPA's evaluation of ethylene oxide (EO) in 2006, the calculation of the excess risk of lymphohematopoietic (LH) cancer incidence was flawed. The calculation was inappropriately based on an exposure-response model for LH mortality instead of LH incidence. This is especially inappropriate for EO because EO exposure may not increase LH incidence except at high doses. The observed increases in LH mortality with EO exposure in males in the NIOSH epidemiology study, although not statistically significant, can be explained at all but the highest doses by exposure-dependent changes in the survival time between LH onset and LH mortality without any changes in LH incidence. Furthermore, EPA's life-table calculations of excess risk of incidence used formulas that are only appropriate for mortality. All of these concerns strongly suggest that EPA should limit their excess risk calculations to mortality unless they have data from an epidemiology study of incidence from which to derive an exposure-response model. What excess risks are calculated and how they are calculated is important for a scientifically-defensible regulatory assessment of EO and other substances.