The predictions of two source-to-dose models are evaluated with observed data collected in a village polluted by an operating secondary lead smelter. Both models were built up from several sub-models linked together and run using Monte-Carlo simulation. The first model system provides the distribution of the media-specific lead concentrations (air, soil, fruit, vegetables, and blood) in the whole area investigated. The second model provides an estimate of the concentration of exposure of specific individuals living in the study area. The predictions of the first model system were improved by performing a sensitivity analysis and using field data to correct the default value provided for the leaf wet density. However, in this case study, the model system tends to overestimate the exposure due to exposed vegetables. The second model was tested for nine children with contrasting exposure conditions. It managed to capture the blood levels for eight of them. In the last case, the exposure of the child by pathways not considered in the model may explain the failure of the model. The interest of this integrated model is to provide outputs with lower variance than the first model system, but further tests are necessary to conclude about its accuracy.