Purpose: We developed limited sampling models (LSMs) for predicting the area under the curve (AUC) of irinotecan (CPT-11) and its metabolites SN-38 and SN-38 glucuronide (SN-38G).
Patients and methods: Regression models were developed based on data from a phase I clinical trial involving 34 patients with advanced solid tumor malignancies who received CPT-11 as a 90-min infusion on an every 3-week dosing schedule. Multiple stepwise regression procedures were supplemented by all possible subsets regression analysis. Alternative clinically based and empirically derived LSMs were determined via model validation assessment including bootstrap simulation testing.
Results: The best LSMs for CPT-11 AUC included concentrations recorded at the end of infusion and 4 h later with an option to include a blood draw at 7.5 h from infusion start. For SN-38 and SN-38G AUC, optimal LSMs included the additional metabolite concentration at 48 h after infusion. The LSMs were able to predict most patient AUC values to within 10% of the true value.
Conclusion: CPT-11 AUC can be modeled with acceptable accuracy using only two or three plasma concentration time-points. A variety of LSM alternatives provided comparable accuracy in predicting AUC. Given the wide variety of LSM alternatives, clinical considerations and patient burden become more important performance parameters than statistical considerations for the choice of time-points in constructing LSMs.