Abstract:
Use of stochastic rather than deterministic estimates of exposure parameters offers the promise of more realistic dose predictions. Separation population variability from uncertainty due to lack of information allows generation of confidence limits on dose distributions and analysis of predictive capability. Few students, however, have compared model-generated estimates to biomonitoring data, due to the scarcity of appropriate data sets.
In this study predicted doses were compared to measured body burdens for two cases. The first involved measurements of organophosphate pesticides in soil and dust in and around houses of farm workers in Eastern Washington, and measurements of the workers’ children’s urinary pesticide metabolites. The second case was a copper smelter in Tacoma, Washington, near which arsenic levels in soil and dust were monitored. Urinary arsenic was measured in children living within 0.5 miles of the smelter in two studies two years apart.
A nested Monte Carlo simulation incorporating variability and uncertainty was run for each case, modeling soil and dust exposure pathways. Dose distributions were calculated with confidence intervals, and predicted values were compared to biomonitoring data. The arsenic results indicated that plausible parameters, utilizing best available knowledge, can result in agreement between model prediction and observations. The current arsenic model utilized several nontraditional input distributions, however, which may not be seen in a typical regulatory exposure assessment. Therefore, these positive results should be used with caution. In the pesticide case, actual doses were underpredicted by two orders of magnitude, suggesting an additional, unknown exposure pathway; as a result, little could be concluded about the validity of the pesticide soil and dust pathway parameters. The arsenic case lends support to the promise of Monte Carlo analysis as an exposure assessment tool, but additional cases in which soil and dust are the dominant pathways are needed.