How well can the Caco-2/Madin-Darby canine kidney models predict effective human jejunal permeability?

The study aimed to predict effective human jejunal permeability (P(eff)) using a biophysical model based on parametrized paracellular, aqueous boundary layer, and transcellular permeabilities, and the villus-fold surface area expansion factor (k(VF)). Published human jejunal data (119 P(eff), 53 compounds) were analyzed by a regression procedure incorporating a dual-pore size paracellular model. Transcellular permeability, scaled by k(VF), was equated to that of Caco-2 at pH 6.5. The biophysical model predicted human jejunal permeability data within the experimental uncertainty. This investigation revealed several surprising predictions: (i) many molecules permeate predominantly (but not exclusively) by the paracellular route, (ii) the aqueous boundary layer thickness in the intestinal perfusion experiments is larger than expected, (iii) the mucosal surface area in awake humans is apparently nearly entirely accessible to drug absorption, and (iv) the relative "leakiness" of the human jejunum is not so different from that observed in a number of published Caco-2 studies.