A modified optimization technique, based on the response surface methodology, was developed for selecting pharmaceutical formulations. In general optimization methods, it is difficult to insure that the optimum formulation is strictly obtainable. Thus, the combined use of random number techniques and Andrews' plots with general optimization methods was investigated for seeking the optimum formulation. The method developed in this study was applied to the optimization of a sustained-release formulation based on the interpolymer complex of polyvinylpyrrolidone with carboxyvinyl polymer. Indomethacin was selected as a model drug for which sustained-release formulations are desirable. Experimental results obtained for the optimum formulation agreed well with the predictions, indicating the usefulness of this approach.