Issues in Using Human Variability Distributions to Estimate Low-Dose Risk

BACKGROUND The National Research Council (NRC) Committee on Improving Risk Analysis Approaches Used by the U.S. EPA (Environmental Protection Agency) recommended that low-dose risks be estimated in some situations using human variability distributions (HVDs). HVD modeling estimates log-normal distributions from data on pharmacokinetic and pharmacodynamic variables that affect individual sensitivities to the toxic response. These distributions are combined into an overall log-normal distribution for the threshold dose (dose below which there is no contribution to a toxic response) by assuming the variables act independently and multiplicatively. This distribution is centered at a point-of-departure dose that is usually estimated from animal data. The resulting log-normal distribution is used to quantify low-dose risk. OBJECTIVE We examined the implications of various assumptions in HVD modeling for estimating low-dose risk. METHODS The assumptions and data used in HVD modeling were subjected to rigorous scrutiny. RESULTS We found that the assumption that the variables affecting human sensitivity vary log normally is not scientifically defensible. Other distributions that are equally consistent with the data provide very different estimates of low-dose risk. HVD modeling can also involve an assumption that a threshold dose defined by dichotomizing a continuous apical response has a log-normal distribution. This assumption is shown to be incompatible (except under highly specialized conditions) with assuming that the continuous apical response itself is log normal. However, the two assumptions can lead to very different estimates of low-dose risk. The assumption in HVD modeling that the threshold dose can be expressed as a function of a product of independent variables lacks phenomenological support. We provide an example that shows that this assumption is generally invalid. CONCLUSION In view of these problems, we recommend caution in the use of HVD modeling as a general approach to estimating low-dose risks from human exposures to toxic chemicals.