Objectives: The aim of this study was to develop limited sampling models for estimating the area under the concentration-versus-time curve (AUC) of carboplatin.
Methods: Based on pharmacokinetic analyses of 14 patients who received 300 mg x m2 of carboplatin over a 90-min infusion following irinotecan, we developed limited sampling models with stepwise multiple linear regression analysis. We validated these models to be unbiased and precise using pharmacokinetic data of a second group of 14 patients. We also compared the observed and the predicted AUC in the patients using Calvert's formula with the patients' renal function.
Results: We developed the following models: AUC (mg x ml(-1) x min) = 0.784 x C4 + 1.30 (r2=0.930) and AUC=0.100 x C0.25 + 0.597 x C4 + 0.140 (r=0.992), where C0.25 and C4 denote unbound plasma concentrations (microg x ml(-1)) of carboplatin at 0.25 h and 4 h after the end of infusion, respectively. These models were validated to be unbiased and precise: a mean prediction error (MPE) with standard deviation (SD) = 2.41 (9.45)% and a root mean squared error (RMSE) = 9.42% for the one-sample model, and MPE with (SD) = 1.22 (5.56)% and RMSE = 5.49% for the two-sample model. We also calculated predicted AUC in the patients using Calvert's formula: MPE with (SD)= -5.87 (21.5)% and RMSE = 21.5%.
Conclusions: These estimations were, as expected, more accurate than the prediction using Calvert's formula based on patients' renal function. The result of this study confirmed the idea that the pharmacokinetic parameters derived from limited sampling models would be more suitable for pharmacokinetic analysis of carboplatin than those obtained using Calvert's formula.