A limited sampling method for estimation of the carboplatin area under the curve (AUC) from one or two plasma concentration determination is presented. The model was conceived and developed using 43 pharmacokinetic studies in 15 patients with ovarian cancer (model data set) who received carboplatin in combination with cyclophosphamide. Linear regression analyses comparing the AUC and the drug concentration at a single time point (0.25-10 h after the end of the infusion) as calculated from the fitted exponential equations gave correlation coefficients as high as 0.97, with maximal correlations falling within the interval of 2-3.25 h. The model was validated prospectively in 9 patients with ovarian cancer (validation data set) who received the same treatment as did the model data set (21 pharmacokinetic studies), testing the equation AUC = 0.52 x C2.75 h + 0.92. Observed and estimated AUCs were correlated in the validation data set (r = 0.91). The mean predictive error (MPE% +/- SE) was -4.4% +/- 3.1% and the root mean squared error (RMSE%) was 13.9%. Multiple regression analysis revealed that adding a second sample drawn at 0.25 h (AUC = 0.053 x C0.25h + 0.401 x C2.75h + 0.628) improved the MPE% to -2.2% +/- 2.1% and the RMSE% to 9.4% (r = 0.96). We conclude that the carboplatin AUC can be estimated from a single plasma sample at 2.75 h or, more precisely, from two plasma samples at 0.25 and 2.75 h. The methods described may prove to be a handy tool for the calculation of approximate AUCs in trials of a size that would discourage detailed pharmacokinetic studies.