242 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS , .:, --• % : x • - . ... . -. _ • . . :• • Figure lb. Scanning electron micrograph of palmitic acid membrane surface. was 5.76 mg/cm2h, whereas for the model membrane, this value was about two orders of magnitude lower, i.e., 5.20 x 10 -2 mg/cm2h. DIFFUSION EXPERIMENTS Table I lists the compositions of the lipid mixtures used in the preparation of the model membranes. Figure 4 shows typical permeation profiles across the model membranes for a few of the markers tested. It is observed that plateauing of the profiles occurs earlier for the more hydrophobic markers such as progesterone or estradiol as compared to the more polar cortisol or sucrose. Indeed, the permeation profiles for cortisol, and sucrose were linear over a 24-hr time period. Table II lists the permeability coefficients of the drug markers across the model membrane. The permeability coefficients were calculated by using the linear portion of the permeation profiles after appropriate corrections for changes in donor phase concentration with time were carried out according to the treatment of Flynn et al. (14). Table II also lists the corresponding permeability values across human skin (15), along with the octanol-water partition coefficients of the markers (16). Figure 5 shows the correlation between the permeability coefficients across model membranes and octanol-water partition coefficients. Figure 6 shows a plot of the permeability coefficients across the model membranes against those reported across human skin for the various markers. DISCUSSION Scanning electron microscopy revealed complete coverage of the filter surface by the

ARTIFICIAL MEMBRANES 243 Figure 2a. TEM of cross-section of lipid matrix of palmitic acid membrane. liposomal lipids. It was also evident from transmission micrographs that these lipids were organized as bilayer structures. The water vapor transmission studies showed that the lipid matrix on the surface of the filter provided significant resistance to water transport. The high reproducibility of the water flux across the model membranes ruled out the possibility of major random defects in the model membranes. The water flux across the 60 •m membrane was 0.052 mg/cm2h and agrees well, after accounting for thickness effects, with a value of 0.4 mg/cm2h for in vivo water flux across human skin with a stratum corneum thickness of roughly 10 •m (10). The electron micrographs as well as the water vapor transmission rates were obtained using model membranes that were not treated with BSA. Preliminary diffusion studies with non-BSA-treated model membranes indicated that although the permeability values of hydrophobic markers across these membranes could be ranked according to their hydrophobicities, hydrophilic markers such as sucrose exhibited permeation rates that were excessive and highly reproducible (Table II). Thus, it appeared that although the lipid matrix in the model membrane exhibited barrier properties to markers that were hydrophobic, hydrophilic compounds such as sucrose permeated easily, perhaps across another pathway in the membrane. It is possible that a polar pathway exists in the bilayer matrix, as evidenced by the extremely high reproducibility of both water vapor transmission rates and permeation of sucrose, and not the result of random defects in the membrane. The extensive investigations into the structure of stratum corneum by Elias et al. (17-20), as well as those by Downing et al. (21-23), have revealed that the cohesive property of the stratum corneum is attributable