Estimation of Exposure and Human Health Risks Using Monte Carlo Simulations. Peter T. Katsumata and William E. Kastenberg, Mechanical, Aerospace and Nuclear Engineering Department, School of Engineering and Applied Science, 48-121 Engineering IV, University of California, Los Angeles, CA 90024-1597
By assuming and combining a series of average, conservative, and worst-case values in the exposure models, the majority of health risk assessments provide a conservative point estimate of risk. Major limitations are introduced from this procedure due to the inherent uncertainties in the parameters in the exposure models, including the environmental contaminant concentration. The inherent uncertainties in the cancer potency factors is also a contributor to uncertainty in the risk. Recent developments in uncertainty propagation methods, which utilize Monte Carlo simulations, have aided in the assessment of these uncertainties. Programs such as Crystal Ball ® and @Risk ® allow the user to assign uncertainty distributions to the parameters in the models and propagate these uncertainties to the risk. The resulting risk distribution shows the uncertainty in the risk. The risk distribution can then be used to cite a conservative risk, which may be the 95th percentile, or a best estimate risk, which may be the median. To illustrate the value of this method, it is applied to a number of Superfund Sites which have already been assigned a Record of Decision (ROD). Uncertainty distributions were developed for the key parameters such as exposure duration, exposure frequency, body weight, and cancer potency factor. The risk values cited in the ROD's utilize the current US EPA point estimate (deterministic) approach using so called "default" values. These risk values are compared to the results of the uncertainty propagation (probabilistic) approach, which uses a Monte Carlo analysis. The health risk assessments utilizing the deterministic approach were found to predict risks which were much larger than the 95th percentile of the probabilistic approach. The risks calculated were then compared with proposed remedial actions, and cost/benefit measures were determined.