PURPOSE: Cell-based permeability screens are widely used to identify drug-P-glycoprotein (PGP) interaction in vitro. However, their reliability in predicting the impact of PGP on human drug pharmacokinetics is poorly defined. The aim was to determine whether a quantitative relationship exists between PGP-mediated alterations in Caco-2 permeability and oral pharmacokinetics in mice. METHODS: Two indicators of drug efflux were measured in Caco-2 for a group of 10 compounds, the ratio of A-B and B-A transport (R9B-A/A-B)) and the ratio of A-B transport in the presence and absence of a PGP inhibitor, GF120918 (R(GF)). These data were correlated with ratios of oral plasma levels in either mdr1a(-/-) or mdr1a/1b(-/-) and wild-type mice (R(KO/WT in vivo)) calculated from literature data on these compounds. RESULTS: A significant, positive correlation (r2 = 0.8, p < 0.01) was observed between RGF and R(KO/WT in vivo). In contrast, R(B-A/A-B), a more commonly used in vitro measure, showed a much weaker correlation with in vivo data (r2 = 0.33, p = 0.11). A strong correlation with R(GF) was also observed after correction of in vivo data for PGP effects on IV clearance. CONCLUSION: The increase in A-B drug permeability following inhibition of PGP in Caco-2 allows a reasonable prediction of the likely in vivo impact that PGP will have on plasma drug levels after oral administration.