The aim of this study was to determine the optimal sampling design for empirical Bayesian forecasting for nedaplatin, and also to develop a simple formula for estimating the area under the plasma concentration-time curve (AUC) of platinum which relates to hematological toxicity after nedaplatin dosing using limited sampling points. Plasma unbound platinum concentration data were retrospectively collected from 27 courses administered to 20 Japanese adult patients. To determine the optimal sampling points, 1 - 5 data point(s) were selected with all combinations and clearance in each patient was estimated by the empirical Bayesian method. As measures for the Bayesian predictive performance, mean prediction error and root mean squared error were estimated. These indices suggested that the sampling time(s) of 4 hours in case of the one-point sampling gives better estimates for individual clearance. As for the limited sampling strategy, a simple formula to calculate AUC, AUC = 0.039x dose + 11.6 x Cp4h - 0.88, was obtained, where Cp4h is the concentration at 4 hours after the end of infusin. These results should be helpful for adjusting dosage to achieve the target AUC.