Model-based meta-analysis of individual patient data for the characterization of intravenous 5-fluorouracil population pharmacokinetics.

Abstract

Purpose: The pharmacokinetics (PK) of 5-fluorouracil (5-FU) has been thoroughly studied over the past decades. Using an individual patient data model-based meta-analysis (IPD-MBMA) approach, this study aims to build a generalized model by combining IPD from multiple studies, investigating the role of lean body mass (LBM) in 5-FU PK, and exploring a model informed precision dosing (MIPD)-based approach for refining its therapeutic drug monitoring (TDM).

Methods: Raw PK data from three 5-FU PK studies (726 patients, 2,332 plasma concentrations) were harmonized and grouped into one homogeneous dataset, preserving traceability to the original data source. An overall population PK (popPK) model for intravenous 5-FU was developed using nonlinear mixed-effects modeling (NONMEM).

Results: A two-compartment model with both saturable (Michaelis-Menten elimination; V_max, K_m) and linear elimination (CL) best described 5-FU disposition, with a stratified error model to capture between-study differences. Body-surface area (BSA) significantly impacted CL, and V_max was 15% lower in women and decreased with age. In a parallel analysis, the sex effect on V_max was translated into an LBM effect, with metabolic capacity increasing with LBM. Based on a 12-cycle simulation, after 4 cycles of TDM-based dose adjustment, 95% of the patient should be optimally exposed in the context of colorectal cancer (area under the curve, AUC 20-29 mg∙h/L). A virtual MIPD-based approach predicted higher target attainment after one adjustment (70 - 68%) compared to classic TDM predictions (49-59%) in men and women, respectively. Finally, the administration of a bolus before the continuous infusion of 5-FU is not expected to materially affect its subsequent continuous infusion AUC.

Conclusion: Aggregating IPD enables building more generalizable popPK models while accommodating study-specific residual variability; broader raw-data sharing would strengthen current evidence and reveal remaining gaps. For 5-FU, BSA dosing leaves substantial interpatient variability, so LBM-based and model-informed strategies, including a priori dosing and MIPD-assisted TDM, warrant prospective evaluation.