A multivariate statistical technique was applied to designing a tablet for the sustained release over 24 h of diltiazem hydrochloride, a model highly water-soluble drug. Tablets of a hydrophilic matrix composed of dextran derivatives and hypromellose were prepared. The formulations were optimized using a nonlinear-response surface method incorporating thin-plate spline interpolation (RSM-S). A bootstrap (BS) resampling method was used to estimate the confidence intervals of the optimal formulations. The response surfaces estimated by RSM-S visualized the effects of the formulation factors, and the optimal release profile for diltiazem was predicted quantitatively as a function of the quantities of the formulation factors, using RSM-S. The simultaneous optimal solutions and their confidence intervals were estimated using RSM-S and BS resampling. The results clearly indicate nonlinear relationships between the formulation factors and the response variables. The observed responses of the optimal preparation coincided well with the predicted responses. The optimal hydrophilic matrix tablet allowed almost zero-order release of diltiazem hydrochloride for 24 h. In conclusion, an oral sustained-release tablet formulation, active over a long period, was successfully optimized using RSM-S, and the reliability of the optimal solution was evaluated using BS resampling.