Juan Antonio Garzón Lamarque 1, Ana Rosa Rincón Sánchez 1, Monzerrat Pardo Zepeda 2, Fernando Antonio Sanchez Zubieta 2, Martín Umpierrez 3, Manuel Ibarra 3, Elba Romero Tejeda 4
1 IBMMTG, Department of Molecular Biology and Genomics, University Center of Health Sciences, University of Guadalajara (Guadalajara, México), 2 Pharmacological Research Laboratory, Civil Hospital of Guadalajara “Dr. Juan I. Menchaca” (Guadalajara, México), 3 Department of Pharmaceutical Sciences, Faculty of Chemistry, Universidad de la República (Montevideo, Uruguay), 4 Department of Pharmacobiology, University Center of Exact Sciences and Engineering, University of Guadalajara (Guadalajara, México)
Introduction/objectives:
High-dose methotrexate (HDMTX) is integral to pediatric acute lymphoblastic leukemia (ALL) consolidation therapy but exhibits high variability in exposure and hematological toxicity. To develop and validate a semi-mechanistic nonlinear mixed-effects population PK/PD model of HDMTX in pediatric ALL, to characterize the exposure-response relationship on neutrophil dynamics, and to explicitly model the pharmacodynamic contribution of concomitant 6-mercaptopurine metabolites.
Methods:
A prospective-retrospective observational study was performed in 54 pediatric ALL patients treated per the MAS-ALL18 consolidation protocol. HDMTX regimens were stratified by risk: 2 g/m² over 4 h, 2.5 g/m² over 24 h, and 5 g/m² over 24 h. The MTX PK dataset comprised 531 plasma concentrations from 54 patients over 186 cycles. The PK/PD dataset included 450 longitudinal absolute neutrophil count (ANC) measurements from 39 patients across four cycles. External validation used an independent cohort (10 patients with 108 MTX concentrations and 92 ANC observations). Modeling used MonolixSuite® 2024R1 (Lixoft, Simulations Plus). One- to three-compartment models with linear elimination and BSA-based allometric scaling were evaluated. BLQ observations (HDMTX: 0.04 µM; ANC: 0.1×10⁹ cells/L) were handled with the Beal M4 method [1]. Covariate selection used stepwise forward-backward elimination (p: 0.05/0.01).
Using EBEs, subject-specific HDMTX exposure profiles were integrated into a mechanistic PK/PD framework combining the Hawwa et al. 6-MP model [2] to simulate intracellular 6-thioguanine nucleotide (6-TGN) concentrations (fixed parameters) and the Friberg myelosuppression model [3] to describe ANC time-course and drug effect. Drug effects were incorporated into the proliferative compartment through an Imax function for HDMTX (fixed to 0.99) and a linear slope for 6-TGN, with parameters estimated via maximum a posteriori using priors from Jost et al. [4].
Filgrastim and infections were coded as binary time-dependent covariates linked to mean transit time (MTT).
Model evaluation included goodness-of-fit diagnostics, bootstrap analysis, and pcVPC. Predictive performance was assessed using MPElog, RMSElog, and GMFE, which were computed in R with 95% bootstrap CIs.
Results:
A two-compartment model with first-order elimination best described MTX pharmacokinetics. Clearance was associated with BSA and pre-infusion serum albumin according to:
CLi = 8.83 × (BSAi/1.73)^0.75 × (Albumini/4)^0.272
Typical estimates (4-h reference): CL 8.83 L/h (RSE 2.9%), V1 25.05 L (4.5%), Q 0.327 L/h (14.7%), V2 3.89 L (9.2%). IIV on CL was 10.5% CV, IOV 22.1% CV, and proportional residual error 0.417.
24-hour infusions showed higher V1 (β = 0.244, RSE 22.3%), lower Q (β = −0.758, RSE 20.4%), and higher V2 (β = 0.698, RSE 22.8%), versus the 4-hour group; clearance was unaffected.
The integrated PK/PD model adequately described longitudinal ANC profiles across four cycles. Baseline circulating neutrophils (Circ₀) were 2.19 × 10⁹ cells/L (RSE 7.5%) and MTT was 19.93 days (RSE 7.0%). The feedback parameter γ was 0.647 (RSE 10.9%). G-CSF (filgrastim or infections) reduced MTT by 63.5% (θ = 0.635; RSE 11.9%). EC50 differed between infusion regimens, with lower values in the 24-hour infusion group (22.8 µM; RSE 56.3%) compared with the 4-hour group (44.1 µM; RSE 36.3%).
pcVPC showed adequate concordance at the 50th and 95th percentiles but persistent underprediction at the 5th, indicating limited characterization of severe neutropenia. The γ estimate (0.647) exceeded the Hopf bifurcation threshold (γ* = 0.568), warranting caution for extrapolation beyond observed scenarios.
PK validation based on PRED showed no significant systematic bias: MPElog 0.04 (95% CI −0.13 to 0.21), RMSElog 0.94 (0.65-1.30), and GMFE 1.85 (1.64-2.12). Predictive performance improved across cycles: cycle 1 GMFE 2.67 (95% CI 1.84-4.51) and cycle 4 GMFE 1.50 (95% CI 1.30-1.76).
PD validation (n=9) showed consistent overprediction (GMFE 2.80, 95% CI 2.35-3.44).
Conclusions:
This study presents the first validated integrated PK/PD framework combining a novel HDMTX population model, the Hawwa 6-MP model, and the Friberg myelosuppression structure in pediatric ALL consolidation. BSA and albumin significantly influenced methotrexate clearance. Infusion duration primarily affected distribution parameters (V, Q), consistent with prehydration-driven extracellular expansion, without modifying clearance or AUC; exposure-driven toxicity did not depend on infusion duration. Interoccasion variability exceeded interindividual variability, underscoring that a single dose-finding assessment is insufficient and cycle-specific TDM is warranted. Prolonged infusions showed lower EC50 despite similar clearance, suggesting enhanced time-dependent pharmacodynamic efficiency. Filgrastim and infections modulated MTT, supporting the biological plausibility of neutrophil recovery. Individual ANC prediction remained limited in external validation (n=9), underscoring the need for larger cohorts. Nevertheless, the framework provides a mechanistic platform for model-informed consolidation therapy, enabling cycle-specific therapeutic drug monitoring and simulation-based dosing optimization. Prospective PD validation in larger patient populations remains the critical next step toward clinical implementation.
References:
[1] Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001; 28:481–504.
[2] Hawwa AF et al. Population pharmacokinetics of 6-mercaptopurine in children with acute lymphoblastic leukemia. Clin Pharmacol Ther. 2008; 84:86–94.
[3] Friberg LE et al. Model of chemotherapy-induced myelosuppression with parameter consistency across drugs. J Clin Oncol. 2002; 20:4713–4721.
[4] Jost F et al. Model-based simulation of maintenance therapy of childhood acute lymphoblastic leukemia. Front Physiol. 2020; 11:217.
Reference: PAGE 34 (2026) Abstr 11929 [www.page-meeting.org/?abstract=11929]
Poster: Drug/Disease Modelling - Oncology