Limited-sampling models for irinotecan pharmacokinetics-pharmacodynamics: prediction of biliary index and intestinal toxicity.

PURPOSE

To construct limited-sampling models (LSMs) for irinotecan (CPT-11) pharmacokinetic (PK) measures.

MATERIALS AND METHODS

The recommended phase II dose of weekly CPT-11 administered as a 90-minute infusion is 145 mg/m2 with granulocyte colony-stimulating factor (G-CSF) and maximal antidiarrheal support. Diarrhea is the dose-limiting toxicity. Previously, we demonstrated a highly significant correlation between severity of diarrhea and biliary index (BI), the estimated biliary exposure of SN-38. BI was the product of total CPT-11 area under the concentration-time curve (AUC) and the relative area ratio of SN-38 to SN-38 glucuronide (SN-38G). At 145 mg/m2, PK data were available for 40 patients; 36 were also assessable for intestinal toxicity. Total plasma concentrations of CPT-11, SN-38, and SN-38G from 1.0 to 11.5 hours from the start of the infusion were evaluated. CPT-11 concentration at each time point, t, is denoted by CPT-11t. LSMs were constructed by stepwise linear regression, on a training set (n = 25), and were validated on a test set (n = 15).

RESULTS

LSMs for CPT-11, SN-38, and SN-38G AUCs displayed excellent fit to the training set (R2 = .87, 0.75, and .98, respectively). AUCCPT-11 = 5.25 x CPT-11(3.0) + 20.60 x CPT-11(11.5) + 1,520.53; AUCSN-38 = 5.38 x SN-38(3.5) + 33.61 x SN-38(11.5) - 7.73; and AUCSN-38G = 10.73 x SN-38G9.5 + 20.66 x SN-38G11.5 + 142.69. BI at each time point (denoted BIt) was calculated as CPT-11t x (SN-38t/SN-38Gt). Several promising LSMs were defined by BI3.5 paired with BI7.5 (R2 = .63) or BI9.5 (R2 = .76), or BI5.5 paired with BI9.5 (R2 = .76). Predicted BI from each model (categorized into low, intermediate, or high) accurately predicted observed intestinal toxicity grade (P < or = .005).

CONCLUSION

These LSMs appear to accurately predict PK parameters of CPT-11. Notably, BI, predicted from several LSMs, accurately predicted intestinal toxicity and thus may be useful for future trials that use adaptive control with feedback.