Blood lead levels in children: epidemiology vs. simulations

The key problem in environmental health is to identify the potential health hazards at the lowest possible cost based upon available environmental data. Biokinetic models such as IEUBK Lead 0.99d are very promising in this respect. We attempted a comparison between epidemiological data and predictions of the model. As input, we used the existing exposure data for the Katowice Voivodship (administrative district, Poland). Epidemiological analysis was based on the results of the screening programme 'Prevention of the Environmental Lead Intoxication in Children Living in Katowice Voivodship'. The simulations consisted of predicted distributions of blood lead levels in children. They have been compared with observed distributions. Sensitivity analysis of simulations with respect to lead concentration in air, soil, water and diet has also been performed. The agreement between predicted and observed mean blood lead levels was quite good (relative difference of about 40%) as for the coarse exposure assessment employed. At the level of risk (fraction of population having blood lead levels exceeding 10 microg/dl) the difference is much higher (about a factor of 2). In order to explain this discrepancy we checked the goodness of fit for the log-normal distribution function (usually taken as a template distribution for lead in the population) in the right tail of the distribution. We noticed a systematic effect depleting the right tail of the actual distribution as compared with the log-normal one. Therefore one can expect that using (as a template) another skewed distribution better fitted in the right tail would improve the accuracy of risk assessment.