Formulation design space of indomethacin tablets was investigated using a nonlinear response surface method incorporating multivariate spline interpolation (RSM-S). In this study, a resampling method with replacement was applied to evaluate the reliability of border on the design space estimated by RSM-S. The quantities of lactose, cornstarch, and microcrystalline cellulose were chosen as the formulation factors. Response surfaces were estimated using RSM-S, and the nonlinear design space was defined under the restriction of more than 3 kgf hardness and more than 70% dissolution 30 min before and after an accelerated test. The accuracy of the resampling method was elucidated and high correlation coefficients were produced. However, the distribution of the border on the design space generated by the resampling method was far from normal, and the confidence interval of the border was estimated using a nonparametric percentile technique. Consequently, the reliability of the design space was decreased by approaching the edge of the experimental design. RSM-S and this resampling method might be useful for estimating the reliability of nonlinear design space.
Copyright © 2011 Wiley-Liss, Inc.