Mixture experimental design applied to solubility predictions

The solubility of theophylline (as a model drug) was studied in four-component systems, using an a priori experimental strategy. Ethanol, polyethylenglycol, propylenglycol, and water were chosen as cosolvents. A reduced cubic model was postulated to describe the solubility as a function of mixture composition. A priori criteria in combination with an exchange algorithm were used to select, from a set of 31 candidate points, the optimal design with the least number of experiments. A weighting was assigned to each of the 31 experiments, on the basis of the cost, in order to obtain a design that would be optimal also from an economic point of view. Such a methodology made it possible to obtain, with the minimum number of experiments and with a low cost, a model which was validated and found suitable for accurate prediction of solubility.