Burden of Proof for Presuming Sub-Linear Dose-Response in Cancer Risk Assessment. K. P. Brand and J. S. Evans, Harvard School of Public Health, Cambridge, MA; G. M. Gray, Department of Environmental Health and Harvard Center for Risk Analysis, Cambridge, MA; and L. Rhomberg, Harvard School of Public Health, Cambridge, MA
The EPA’s recent 1996 proposed guidelines for Cancer Risk Assessment are intended to be more receptive to departures from the default (linear no-threshold dose-response) assumption used in high-to-low dose extrapolation. The EPA proposes a separate high-to-low dose extrapolation approach for those agents having "sufficient" evidence of a sub-linear dose-response. Because this separate approach could allow for markedly less stringent regulatory standards, the question of what constitutes "sufficient" evidence is of keen interest. Evidence is unlikely to ever definitively identify the operative mode of low-dose carcinogenesis, thus it is acknowledged that some level of residual doubt is consistent with the judgment of "sufficient evidence" — but how much doubt is acceptable? The quantitative implications of different degrees of "burden-of- proof" are examined using a weighted-weibull model to encode the competition between two dose-response models: the default "linear" no threshold dose-response model (the incumbent); and a rival model which can express varying degrees of sub-linearity. The relative likelihood, denoted Osl (odds sub-linear), between these two rival models serves as a quantitative index of "burden- of-proof." The sensitivity of two decision-metrics, including low-dose risk, R, and EDq (dose below which risk is lower than, q, a specified level of acceptable risk), to varying Osl is examined. Decision-metrics are found to be critically dependent on the estimate of Osl, and relatively insensitive to the presumed extent of sub-linearity of the rival model. Less stringent regulatory standards, the question of what constitutes "sufficient" evidence is of keen interest.