The use of human skin models for performing cutaneous bioavailability studies has been little investigated. For instance, only few studies have been reported on human skin models dealing with vehicle effects on percutaneous penetration. The present study aimed at evaluating the influence on caffeine's and < alpha >-tocopherol's cutaneous bioavailability of cosmetic vehicles such as a water-in-oil emulsion, an oil-in-water emulsion, a liposome dispersion and a hydrogel applied at finite dose using the reconstructed human skin models EpiDermR and EpiskinR. The results were compared with those obtained in human skin ex vivo using similar experimental conditions. It was demonstrated that the rank order of solute permeability could be correctly predicted when the preparation was applied at a finite dose in human skin models, at least when solutes with far different physicochemical properties such as caffeine and < alpha >-tocopherol were used. If only slight effects of cosmetic vehicle on skin bioavailability were observed in human skin ex vivo, they were less predictable using skin models. Especially, alcohol-containing vehicles seemed to behave differently in EpiDerm as well as in Episkin than on human skin ex vivo. Stratum corneum intercellular lipid composition and organization of human skin models differ to some extent from that of human stratum corneum ex vivo, which contributes to less pronounced barrier properties, together with the increased hydration of the outermost stratum corneum layers of the models. These features, as well as still unknown factors, may explain the differences observed in vehicle effects in human skin ex vivo versus human skin models. Copyright < copyright > 2002 S. Karger AG, Basel.