236 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS ß ' ' ' I ' ' ' ' ! ' ' 0 10 20 30 Time (h) Figure 4. Flux versus time profile for theophylline from saturated methanol donors. there is an increase in slope and variability, signalling membrane deterioration. The change in skin properties with time is shown even more graphically in the flux plot (Figure 3). Similar profiles were observed for nude mouse skin (8). Human skin is more stable than fuzzy rat skin to water-induced changes in permeability (9). Solvents affected the ability of the skin to maintain its barrier properties. Methanol was found to be the most damaging of the solvents studied. While methylparaben flux from methanol was essentially constant from about 10 to 23 hours, rapid deterioration of the membrane's barrier function then occurred (Figure 4). Several guidelines for the remaining experiments were established from these initial findings. The skin was dermatomed to 450 I•m for the remainder of the experiments to maximize sample yield (approximately 12-15 dorsal skin samples per animal). A min- imum of six replicates (three replicates from two separate animals run on different days) was measured for each treatment (solvent/solute system). In most cases four treatments (12 cells) were run simultaneously. The data for each treatment were pooled only after testing for significant differences. The experiments were run for about 24 hours, with continuous sample collection every two or three hours.

SKIN PENETRATION 237 0.10 0.08 0.06 0.04 0.02 0.00 0 25 50 75 100 % Saturation Figure 5. Methylparaben flux as a function of the percent saturation (activity) for propylene glycol treat- ment. In previous work on permeation through polydimethylsiloxane membranes from inter- active solvents (alcohols), it was shown that at high paraben concentrations flux re- mained constant or declined as concentration was increased further (5). This was a consequence of the decrease in alcohol activity, which resulted in an attenuation of solvent/membrane interaction. The same pattern was not observed with permeation through fuzzy rat skin. The flux profiles for methylparaben delivered from 1-propanol and propylene glycol donors were linearly related to solute activity (Figures 5 6). This indicates that the interaction was not reduced by the decrease in solvent activity, as was found with the synthetic membrane. Since flux is a linear function of concentration, it is necessary to utilize but a single concentration at which to make comparisons between solvents. One approach is to select a fixed concentration of permeant in all solvents. However, this method neglects the fact that solute activity will vary between solvents, thus biasing the results. A better approach is to compare data for saturated systems, in which the solute activity is constant and maximal. With this design, the flux from all of the systems should be