Prediction of percutaneous absorption from physicochemical data: a model based on data of in vitro experiments

Correlations between in vitro percutaneous absorption data and physicochemical properties of industrial chemicals are evaluated in order to develop predictive mathematical models based on said properties. Percutaneous diffusion of 16 pounds of occupational interest, eight of which were polycyclic aromatic hydrocarbons (acenaphthene, anthracene, benzo(a)anthracene, chrysene, phenanthrene, fluorene, naphthalene, pyrene), six organophosphorus insecticides (acephate, chlorpyrifos, dimethoate, fenitrothion, methamidophos, omethoate) and two phenoxycarboxylic herbicides (2,4-D, MCPA), were tested in vitro using monkey (Cercopithecus aetiops) skin. The test apparatus consisted of nine static diffusion cells with normal saline, gentamycin sulphate and 4% bovine serum albumin as receiving solution. Test compounds were applied at various concentrations in 30 microliters of acetone solution and determined, in the receiving phase, by chemical analysis. Values for ln Kow (octanol/water partition coefficient) were correlated with experimentally determined values of the permeability constant Kp (r = 0.90, P < 0.001) and lag time (r = 0.81, P < 0.01). Analysis of variance in a model of multiple linear regression between Kp, ln Kow and water solubility [water] of the compounds, showed that the data had a highly significant fit (P < 0.0001). A more general model which also included molecular weight (MW) and vapour pressure was evaluated as well, but the two variables made no substantial difference. Multiple regression analysis between lag time, ln Kow and [water] was significant (P < 0.0001), whereas introduction of vapour pressure and MW as independent variables did not significantly improve the predictive effect on lag time. Our experimental system, therefore, enables the values of Kp and lag time to be predicted with reasonable precision on the basis of ln Kow and [water] values, using the algorithm derived from the multiple linear regression equation.