Advancing Precision Dosing of 5-FU: Population PK Model Development, Limited Sampling Strategies, and Fit-for-Use Application

BACKGROUND: A significant proportion of patients receiving 5-fluorouracil-based chemotherapy have been reported to experience grade ≥ 3 toxicities, which range from 21 to 76% depending on the treatment regimen. This may in part be due to interpatient variability in pharmacokinetics. More recently, only about 20% of the patients were reported to achieve the target area under the concentration-time curve from time zero to infinity of 20-30 mg*h/L. OBJECTIVE: We aimed to evaluate existing population pharmacokinetic models and develop a final pharmacokinetic model. In addition, we aimed to identify optimal limited sampling strategies to accurately estimate individual area under the concentration-time curve from time zero to infinity, facilitating model-informed precision dosing using a fit-for-use application. METHODS: Patient data from four prospective clinical studies were obtained. Published 5-fluorouracil population pharmacokinetic models were evaluated. A final model was developed, and limited sampling strategies were identified. A user-friendly model-informed precision dosing application was created for initial and subsequent dose adjustments. RESULTS: Using the published models, population predictions resulted in underestimation of observed concentrations and misspecification of a 24-h infusion regimen were observed, with the two-compartment Michaelis-Menten model outperforming the other model structures. Of this model, the parameters were re-estimated and covariates analysis was conducted. Body surface area significantly influenced the maximum rate of reaction and was implemented using allometric scaling with an exponent of 1.14 (relative standard error 24%). For a bolus in combination with a 46-h continuous infusion, the best limited sampling strategy was T = 0.1 (end of bolus), 1 and 3 h after treatment initiation; while for the 46-h continuous infusion, the optimal limited sampling strategy was T = 0.8, 2, and 5 h. Model-informed precision dosing algorithms were implemented in an model-informed precision dosing application. CONCLUSIONS: The external evaluation of 5-fluorouracil population pharmacokinetic models against a pooled prospective population resulted in a generic two-compartment Michaelis-Menten pharmacokinetic model with body surface area as a covariate on the maximum rate of reaction. The limited sampling strategy and the shiny model-informed precision dosing application demonstrated the potential for accurate dose individualization of 5-fluorouracil.