Thermodynamic modeling of activity coefficient and prediction of solubility: part 2. Semipredictive or semiempirical models

The solubility of stearic acid, ranitidine hydrochloride, and stavudine were predicted in selected organic solvents. The experimental solubility data of stearic acid and ranitidine hydrochloride were reported in previous work of the authors and stavudine's solubility was measured in this work. Equilibrium aqueous solubility of crystalline stauvudine was determined at controlled temperatures by stirring and filtration, with spectrophotometric quantification. The new model developed in Part 11 of this communication was modified as a semipredictive model with two adjustable parameters. Predicting the solubility data with the NRTL model using just one experimental point resulted in a big error while the modified new model and the UNIQUAC model showed much smaller errors. A new method was proposed in this work for predicting the solubility data of all polymorphs of a given compound using the experimental solubility data of one of the polymorphs of the same chemical compound. Although in general, the UNIQUAC model predictions were marginally superior, the new model is simpler and does not require the molecular parameters such as Van der Waals area and volume. The solubility prediction in a mixture of solvents using the NRTL and UNIQUAC models was also discussed. (C) 2006 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 95:798-809, 2006.