Inhibition of P-glycoprotein (PGP) resulting from the co-administration of substrate drugs represents a potential source of drug-drug interactions. Although in vitro screens can readily identify such interactions, the accuracy with which they mimic interactions in tissues or their value in predicting interactions in vivo is unresolved. This was addressed for the model PGP substrate digoxin by comparing the modulation of its permeability across Caco-2 cells and ex vivo human and rodent intestine by drugs for which pharmacokinetic data on interactions with digoxin in man is available. All five compounds (talinolol, omeprazole, verapamil, quinidine, cyclosporin) dose-dependently increased absorptive (A-B) digoxin permeability with maximal increases of 2.2-4.5-fold across Caco-2. Quantitatively similar increases were observed in ex vivo human and mouse intestine and studies in mdr1a(-/-) intestine confirmed that these interactions are mediated solely by PGP. In vitro changes in digoxin permeability were qualitative indicators of the increase in digoxin C max for these compounds in man, although accounting for the luminal drug concentrations expected for a given oral dose was a critical consideration. Based on a limited dataset these data suggest that Caco-2 accurately mimics intestinal digoxin interactions and may be useful in predicting the threshold dose at which interactions become clinically significant. Further studies across a wider range of drugs are needed to determine the broader applicability of in vitro data for quantitative prediction of clinical drug interactions.