Comparison of experimentally determined and mathematically predicted percutaneous penetration rates of chemicals

The aim of the study was to evaluate the predictive potential of three different mathematical models for the percutaneous penetration of industrial solvents with respect to our experimental data. Percutaneous penetration rates (fluxes) from diffusion cell experiments of 11 chemicals were compared with fluxes predicted by mathematical models. The chemicals considered were three glycol ethers (2-butoxyethanol, diethylene glycol monobutyl ether and 1-ethoxy-2-propanol), three alcohols (ethanol, isopropanol and methanol), two glycols (ethylene glycol and 1,2-propanediol), one aromatic hydrocarbon (toluene) and two aromatic amines (aniline and o-toluidine). For the mathematical prediction of fluxes, models described by Fiserova-Bergerova et al. (Am J Ind Med 17:617–635 1990), Guy and Potts (Am J Ind Med 23:711–719 1993) and Wilschut et al. (Chemosphere 30:1275–1296 1995) were used. The molecular weights, octanol–water partition coefficients (LogP) and water solubilities of the compounds were obtained from a database for modelling. The fit between the mathematically predicted and experimentally determined fluxes was poor (R ² = 0.04–0.29; linear regression). The flux differences ranged up to a factor of 412. For 4 compounds, the Guy and Potts model showed a closer fit with the experimental flux than the other models. The Wilschut et al. model showed a lower flux difference for 4 compounds as compared to experimental data than the models of Fiserova-Bergerova et al. and Guy and Potts. The Fiserova-Bergerova et al. model showed for 3 compounds a lower flux difference to experimental data than the other models. This study demonstrates large differences between mathematically predicted and experimentally determined fluxes. The percutaneous penetration as determined in diffusion cell experiments may be considerably overestimated as well as underestimated by mathematical models. Although the number of compounds in our comparison study is small, the results point out that none of the mathematical model has significant advantages.