This study evaluates the correlation of some widely used skin permeability predictive models with a recently proposed empirical model based on human in vivo dermatopharmacokinetic data. Drug fluxes through the skin have been calculated using in vitro- and in vivo-based models, and observed in vivo data, and the values compared.
Most in vitro-based models underestimate the in vivo data by 1–100-fold. The discrepancy between observed data and prediction reaches the maximum (1000–10,000-fold underestimation) for nicotine (with the smallest molecular weight and log Koct), nitroglycerin (with the largest number of hydrogen bond acceptor groups), and for oxybutynin (with the largest molecular weight and log Koct) where there was a 1000-fold flux overestimation. However, most models correlated well with the in vivo data and the in vivo-based model (p < 0.05).
The vehicle effect and using non-steady state in vivo data in the flux calculations partly account for the observed discrepancies between predicted and observed values. Nevertheless, these results reveal the need for further refinement of skin permeability predictive equations, using the steady state in vivo data, and consideration of formulation effect.