A general approach for the prediction of the intestinal absorption of drugs: regression analysis using the physicochemical properties and drug-membrane electrostatic interaction.

A general method for predicting the intestinal absorption of a wide range of drugs using multiple regression analysis of their physicochemical properties and the drug-membrane electrostatic interaction was developed. The absorption rates of tested drugs from rat jejunum were measured by the in situ single-pass perfusion technique. The drugs used in this study were divided into three groups for regression analysis, and a smaller "test" set of compounds was used to assess the predictive capacity of the regression equation. When the analysis was applied to each respective group of drugs (i.e., anionic, cationic, and nonionized compounds), obtained regression coefficients were 0.569, 0.821, 0.728 by using the organic solvent (n-octanol)/buffer partition coefficient, 0.730, 0.734, 0.914 using the permeation rate across a silicon membrane, and 0.790, 0.915, 0.941 using an EVA membrane, respectively. However, smaller regression coefficients of 0.377, 0. 468, and 0.718 were obtained when these three groups of drugs were put together for prediction. Meanwhile, correlation was improved remarkably when drug-membrane electrostatic interactions, namely, hydrogen-bonding donor (Halpha) and acceptor (Hbeta) activity or index of electricity (Ec), were added to the other parameters of lipophilicity and permeation rate across the EVA membrane (r = 0.880 and 0.883, respectively). Moreover, the equation obtained from these regression analyses was applicable even to the prediction of the absorption of the zwitterionic drugs. These results suggest that including the electrostatic interaction parameters in addition to lipophilicity and permeability across artificial membranes would afford a better prediction for the intestinal absorption of the vast majority of drugs.