An orthogonal collocation on finite elements-based regression method was applied to help design optimal drug-dosage regimens. The approach, suitable for two-compartment models, would allow clinicians to design multiple boluses followed by a constant-rate infusion of a medicament to patients in order to assure a desired plasma concentration. The algorithm was tested on theophylline, a drug that has been described by both linear and Michaelis-Menten elimination pharmacokinetics. In the linear case, increasing the number of boluses, from 1 to 4, decreased the normalized square root of the integral square error from 1.80 to 0.58 when a target concentration of 10 mu g/mL was selected in the central compartment. When applied to the nonlinear metabolism, the procedure, implemented in Mathematica (R) (Wolfram Research, Inc.). effectively computed optimal dose sizes, injection times and infusion rates. Estimations, based on linear interpolation, provided a good time-saving alternative to the full optimization methodology. (C) 2008. Elsevier Ltd. All rights reserved.