In general toxicity testing, maximal acceptable concentrations are derived from no-observed adverse effect levels (NOAEL) in rodents. Risk assessment then considers safety factors for the interspecies difference, and intraspecies variability. This approach can be used for assessing maximal acceptable concentrations for chemicals inducing direct immunotoxicity, resulting in e.g. reduced resistance to infections. As for predictive testing of chemicals in terms of sensitization, laboratory animal data are mostly used for risk assessment as well. Generally, the assessment of risk for chemicals that induce contact sensitivity is limited to hazard identification, and risk management is restricted to labeling. An alternative type of evaluation of the risk of adverse effects due to exposure to immunotoxic chemicals may be the so called parallellogram approach. In this parallellogram there are four cornerstones, one of which is the health effect of exposure to a chemical, assessed as an endpoint (e.g. infection model) in experimental animals, and another the quantitative prediction of this endpoint in humans. The other cornerstones are assays of parameters that are relevant to the mechanism of the adverse effect in experimental animals and humans, and are used for species comparison. Species comparisons between the animal species used for hazard identification and humans are crucial for extrapolation of animal data to the human situation. This approach can be used to provide relevant information on the dose-response relationship in humans. In concert with information on actual exposure, such data can then be used for the characterization of risk for adverse health effects in humans. Such approaches have been used for chemicals that exert direct immunotoxic activity (bis(tri-n-butyltin)oxide (TBTO)), and may hold promise for the risk evaluation of chemicals that exert skin sensitizing properties.