Accurate Risk-Based Chemical Screening Relies on Robust Exposure Estimates

The massive undertaking reported in Wetmore et al. (2012) represents an important step forward as we integrate innovative in vitro chemical screening efforts such as ToxCast into risk assessment approaches. However, the authors overstate the degree to which their exposure estimates represent the “highest estimated U.S. population exposures” and consequently underestimate the number of chemicals for which current exposures exceed levels associated with biological activity. In this study, the researchers estimated human oral doses expected to produce steady-state blood concentrations at which ToxCast chemicals were active in vitro. They then compared these “active” doses with estimates of human exposure that they characterized as the “highest estimated U.S. population exposures” and the “estimated upper limit of human exposure.” These estimates came primarily not only from pesticide registration documents but also from exposure surveillance data in National Health and Nutrition Examination Survey (NHANES). The authors attempted to capture higher exposures of susceptible subpopulations by selecting the 95th percentile estimates from subgroups presented in the Centers for Disease Control (CDC) exposure report (2009) and characterized these exposures as representing “most highly exposed subpopulations” in Figure 4B and the study abstract. However, 95th percentile estimates of NHANES subpopulations defined by age, gender, and ethnicity do not capture extreme variability in the high end of exposure distributions. In publicly available NHANES data, it is not unusual for maximum values to exceed 95th percentiles by a factor of 100 or more. For example, maximum values of several pesticides, chlorinated solvents, mercury, and several phthalates are more than 100 times higher than 95th percentile values in NHANES data cycles from 1999 to 2010, and one 8-year-old girl in the 2007–2008 test had urinary mono-n-butyl phthalate levels almost 900 times higher than the 95th percentile (NCHS/CDC, 2012). NHANES is designed to be a representative sample of the 300 million-person U.S. population, and so the top 5% of NHANES exposures represent over 15 million people. NHANES data clearly demonstrate that “upper limits of human exposure” are well above 95th percentile levels and may reflect exposure to sizable subgroups. Thus, the highest exposures cannot necessarily be considered rare events and have broad public health significance. We appreciate Wetmore’s acknowledging the need for better population exposure data. Most of the exposure estimates used in this screening came from EPA pesticide Reregistration Eligibility Decision (RED) documents which have not been validated against NHANES and other biomonitoring data. RED exposure estimates, which are intended to represent “high end” exposure, should be systematically evaluated as in some cases exposure estimates from REDs may underestimate these values. For example, the diazinon RED exposure estimate does not consider cancelled uses, such as residential exposure, even though diazinon is still detected in homes, so exposure from former residential uses continues. Although Wetmore et al. (2012) validated RED estimates with NHANES 95th percentiles for triclosan and cacodylic acid, these may not be representative. For these two chemicals, there is better exposure data than for most; for example, the triclosan RED appears to have relied on NHANES data in its derivation, whereas most chemicals do not. For chemicals with limited exposure data, it is important to use available data and to clarify that without exposure data there is no basis for reassurance that exposure levels are below activity levels. No estimate of human exposure was included for 24% of the chemicals in the study, and so no comparison toxicological sciences 128(1), 295–296 (2012) doi:10.1093/toxsci/kfs143 Advance Access publication April 20, 2012