Predicting the solubilities of drugs in ethanol + water mixtures has been improved using a trained version of the Jouyban-Acree model. To check the accuracy of the improved model, solubility data for various drugs in ethanol + water mixtures at different temperatures were collected from the literature and the predicted solubilities were compared to the corresponding experimental values by computing the mean percentage deviations (MPDs). The overall MPDs for the predicted solubilities (from a total of 113 data sets including 1318 data points) using the previous trained model and the improved model were 52.9 (±44.2) % and 40.7 (±21.6) %, respectively; this mean difference was statistically significant (P < 0.0005).

Key words: solubility prediction, cosolvency, water-ethanol, Jouyban-Acree model