We examined the use of optimal sampling theory in the determination of the pharmacokinetics of piperacillin in febrile, neutropenic cancer patients. Patients were studied prospectively as part of a randomized, double-blind clinical trial of piperacillin and amikacin versus imipenem and placebo. The results from the analysis of 5 optimal samples were compared with those derived from 15 concentration determinations (10 samples, with the 5 optimal samples assayed in duplicate). The use of a standard least-squares estimator as opposed to a bayesian estimator, with normal prior distributions placed on beta and serum clearance, was also examined. Finally, the use of duplicate determinations in improving the precision of parameter estimation was studied. Plasma concentrations obtained at time points determined by optimal sampling theory, when analyzed with a bayesian estimator, produced estimates of pharmacokinetic parameter values that were in good agreement with those derived from the 15-determination set. Duplicate assay did not improve the precision of parameter estimation. Estimation of plasma clearance was quite robust, irrespective of the estimator used, probably because this evaluation was performed at steady state. Optimal sampling theory is a promising technique that can be employed to determine patient-specific estimates of pharmacokinetic parameter values in target populations.