ABSORPTION IN HUMAN SKIN 223 Log Poet is the only experimentally determined component in the Potts-Guy algorithm predictive of skin penetration. As a measure of lipophilicity, it is thereby confirmed as a critical factor determining the diffusion of solutes through biological membranes in general. A large body of octanol-water partitioning values has been compiled by Hansch and Leo in the Pomona College Medicinal Chemistry Project (PCMCP) log P database over a period of 25 years (3). In the process, more than 30,000 published experiments have been reviewed and also tested for validity. While partitioning values between water and a variety of lipophilic solvents, and also with soil, are included in that database to describe polarity, the core of that information system consists of octanol-water partition behavior. Among the latter, the numbers found to be the best after review of the experimental methods are highlighted as "star-values," and at this time over 9000 such "star-values" make up the so-called starlist of measured log Pact, which is constantly under review and expansion. The approximately 30,000 measured partition values, however, represent only a small portion of the total inventory of hundreds of thousands of chemicals presently available for industrial or commercial use. Those log Pact values on record have been experimen- tally determined on a priority basis, focusing primarily upon the most prominent pharmaceuticals, pesticides, and frequently used solvents. Log Pact for fragrance chem- icals, on the other hand, are only by chance included in that list. For compounds for which no experimental data exist, an accurate method had been elaborated at the PCMCP for calculating log P by summation of fragment values, combined with factors that take into account any fragment interaction. Including an estimate of error, they are accessible alongside the measured values, and thus data required for the description of polarity of virtually all chemicals have become readily retrievable. In order to ascertain the validity of log Poet values so calculated, they have been correlated here with the corresponding measured values published in the literature for 33 fragrance compounds, in order to justify their use in the Potts-Guy algorithm. THE POTTS-GUY SKIN PENETRATION MODEL In recent time a number of mathematical models have been developed that are mech- anistically and biologically founded, in the endeavor to render the penetrability of the skin by chemicals predictable, without requiring laborious and expensive experimenta- tion. In a review and evaluation of such models, the U.S.E.P.A. selected an equation by Potts and Guy (1) as the most appropriate for predicting skin penetration of priority pollutants (4). It is based upon physicochemical interpretation of molecular transport across biological membranes, and adequately accounts for the percutaneous absorption of compounds of the most diverse structures. Using measured permeability coefficients, Kp, from the literature for a wide selection of nonelectrolytes, from aqueous solution through human epidermis, as a basis, Potts and Guy have identified, through linear regression analysis, the two basic physicochemical parameters that determine transport through the skin: molecular size, expressed as the molecular weight MW, and the octanol-water partition coefficient log Pac•, a measure of lipophilicity. The derived predictive equation log Kp (cm/hr) = -2.72 + 0.71 log Pac• - 0.0061 MW (1) was shown to be applicable to chemicals of diverse structures and functionality in

224 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS general, ranging in molecular weight from 18 to 750, and in lipophilicity with a log Poc• between - 3 to + 6. To test the validity of the Potts-Guy algorithm when it is applied to fragrance chem- icals, the values calculated according to Eq. (1) for 20 structures commonly encountered in fragrance compositions have been correlated with experimental Kps from the litera- ture, determined through human skin in vitro from aqueous solution under steady-state conditions. The multiple regression analysis by Potts and Guy, correlating the predicted with experimental log Kps, was also performed on the 20 fragrance chemicals, for direct comparison with the fit obtained for Flynn's much wider assortment of 93 chemicals (1). SAMPLE SELECTION The 20 compounds chosen for analysis of the predictive power of the Potts-Guy algo- rithm were selected based on the criteria that ß The measured log Poc• value does not exceed 4 ß An experimental permeability coefficient determined with human skin under steady- state conditions from aqueous solution is available from the literature ß The octanol-water partition coefficient has been determined experimentally ß They are listed in the standard reference on perfumery chemicals by S. Arctander (5). The list of chemicals selected for the validation of calculated versus measured log P values, on the other hand, was expanded to include 33 entries, as an experimental Kp is no longer a requirement there. All chemicals are listed in Table I. STATISTICS Statistical analysis was conducted using the Minitab TM statistical package (Minitab Release 6.1.1., Minitab Inc., State College, PA, 1987). RESULTS RELATIONSHIP BETWEEN EXPERIMENTAL AND PREDICTED Log P Linear regression analysis (Figure 1) shows near-perfect concordance between measured and calculated partition coefficients obtained from the PCMCP database for 33 common fragrance compounds (r 2 = 0.97, p 0.001), thereby legitimizing the use of calcu- lated values from that database in the Potts-Guy algorithm where measured values are not available. PREDICTIVE VALUE OF THE POTTS-GUY ALGORITHM Validity of the Potts-Guy algorithm predictive of skin permeability described in Eq. (1) was demonstrated when applied to fragrance chemicals in particular. Using measured log P values, the correlation of estimated permeability constants Kp with experimental values for 20 compounds, through human skin from an aqueous medium for steady-state