Limited sampling models for simultaneous estimation of the pharmacokinetics of irinotecan and its active metabolite SN-38

Irinotecan (CPT-11) is a novel topoisomerase I inhibitor with clinical activity in human malignancies. The objective of this study was to develop efficient limited sampling models (LSMs) to estimate simultaneously the area under the plasma concentration versus time curves (AUC) for both CPT-11 and its active metabolite SN-38. A total of 64 pharmacokinetic sets (≥24-h sampling) were obtained in phase I studies at doses ranging from 50 to 750 mg/m2 (0.5-h i.v. infusion). The patients were randomly assigned to a training data set (n=32) and a test set (n=32). Multiple linear regression analyses were used to determine the optimal LSMs based on the correlation coefficient (r), bias (MPE%, percentage of mean prediction error), and precision (RMSE%, percentage of root mean squared prediction error). Of these LSMs, the ones including maximal concentrations of CPT-11 (0.5 h, the end of the i.v. infusion) and metabolite SN-38 (≈ 1 h) were favored along with predictive precision and clinical constraints. Several bivariate models including a 6-h time point as the last sampling time (or 7 h) were found to be highly predictive of either the CPT-11 AUC or the SN-38 AUC. The chosen sampling time points were the ones that allowed the best compromise between the accurate determination of either compound alone with the same sampling times. The simultaneously best prediction of both CPT-11 and SN-38 AUCs was obtained with sampling time points harvested at 0.5, 1, and 6 h (or 7 h). With these sampling time points a trivariate model was selected for the determination of CPT-11 AUC namely, CPT-11 AUC (ng h ml−1)=0.820×C0.5h+0.402×C1h+15.47 ×C6h+928, and a corresponding model was selected for the determination of metabolite AUC, i.e., SN-38 AUC (ng h ml−1)=4.05×C0.5h−0.81×C1h+23.01×C6h−69.78, whereC(t) is the concentration in nanograms per milliliter of either compound at a given timet. These models performed well with the test data sets for CPT-11 AUC (r=0.98, MPE%=−1.4, RMSE%=13.9) and for SN-38 AUC (r=0.95, MPE%=−6.5, RMSE%=37.7). In addition to the determination of AUCs (and hence clearance), these models also allow the determination of the maximal concentrations of both compounds, which might be needed for pharmacodynamics studies. Other bi- and trivariate models including other time points are also presented. These LSMs not only will facilitate ongoing and future clinical trials by significantly reducing the number of blood samples needed for pharmacokinetics studies but will hopefully contribute to a better knowledge of pharmacokinetic-pharmacodynamic relationships for both CPT-11 and its active metabolite SN-38.