Julia Stolk 1, Liska Scheffers 1, Elizabeth de Lange 2, Alwin Huitema 1,3,4, Dannis van Vuurden 1, Hinke Huisman-Siebinga 3
1 Princess Máxima Center (Utrecht, The Netherlands), 2 Leiden Academic Center for Drug Research (Leiden, The Netherlands), 3 Netherlands Cancer Institute (Amsterdam, The Netherlands), 4 University Medical Center Utrecht (Utrecht, The Netherlands)
Objectives: Gemcitabine shows strong in vitro promise against H3K27M-altered diffuse midline gliomas (DMGs), a rare and aggressive pediatric brain cancer [1], [2]. However, translation to the clinic lacks proper consideration of relations between plasma pharmacokinetics (PK), brain tumor target site PK, and efficacy. LeiCNS-PK3.0, a comprehensive translational physiologically-based PK (PBPK) model, can predict central nervous system (CNS) disposition in mice, rats and humans, based on system-specific information of physiology and anatomy, and drug-specific physiochemical information [3], [4]. Therefore, LeiCNS-PK3.0 is useful to guide decision making regarding dosing regimens, even when clinical data is scarcely available, especially at target sites. This study aimed to predict gemcitabine PK profiles in brain and brain tumors in pediatric DMG patients using LeiCNS-PK3.0 and to compare brain and brain tumor exposure after different doses and infusion schedules.
Methods: The model structure of LeiCNS-PK3.0 was extended to describe gemcitabine-specific transport and metabolism processes (LeiCNS-PK-GEM). First, Michaelis Menten-equations were added to describe the bidirectional facilitated diffusion of gemcitabine by human equilibrative nucleoside transporter (hENT) 1 and 2 at the blood-brain barrier (BBB) and cell membranes. Second, first-order intracellular metabolism of gemcitabine into its inactive metabolite, dFdU was added. This rate was assumed to be proportional to the CDA mRNA expression in the brain compared to the liver, as previously performed [5]. Third, an extra compartment to describe intracellular conversion of gemcitabine into active metabolite dFdCTP was incorporated in the model, as well as first-order intracellular clearance of dFdCTP, as previously done [5]. This conversion into dFdCTP was described through Michaelis Menten kinetics. Finally, a DMG tumor consisting of five subcompartments (i.e. tumor extracellular fluid, tumor cell membrane, tumor intracellular fluid (ICF), tumor lysosomes and dFdCTP in tumor ICF) was added. This structure was similar to the compartments representing healthy tissue, but system parameters were specific for tumor physiology in DMG patients.
Gemcitabine plasma and total brain concentrations were obtained in a PK study with a patient-derived xenograft DMG mouse model. These data were used as input into LeiCNS-PK-GEM and used to estimate the maximum rates of hENT transport for gemcitabine over the BBB (vmax,hENT,BBB) and cell membranes (vmax,hENT,CM), the maximum rate of conversion to dFdCTP (vmax,DCK), and dFdCTP clearance (CL,dFdCTP). Parameter identification was carried out using the first-order conditional estimation with interaction (FOCEI) algorithm to identify a global optimum, using nlmixr2 (v5.0.0). Subsequently, the mouse LeiCNS-PK-GEM was translated into human adult and pediatric models, using associated plasma PK parameters for adults as model input [8], scaled allometrically to pediatric parameters. Observed plasma PK in adult patients from a previously composed dataset [8] and in pediatrics from a previously published phase I study of gemcitabine in children with relapsed refractory leukemia [9] were used to evaluate adult and pediatric model performance. A local sensitivity analysis was performed on the final human pediatric LeiCNS-PK-GEM model, where the effect on area under the curve (AUC) was assessed after adjusted parameter values by 10%. Lastly, gemcitabine disposition into the brain and brain tumor of children with DMG after intravenous (i.v.) doses of 1 to 4800 mg/m2 and infusion times of 1 to 480 minutes was simulated. Model development and simulations were performed using R (v4.4.1) and RStudio (v2026.01.0).
Results: Plasma PK models accurately described the observed concentrations in mice, adults, and children. Initially, vhENT,BBB and vmax,hENT,CM were assumed to be equal to observed values in hepatocytes [6], and vmax,DCK and CL,dFdCTP to values from a previously published PBPK model for gemcitabine and dFdCTP [5]. Yet, model predictions showed an overestimation of whole brain PK in mice compared to the observations, and underestimation of dFdCTP PK in brain ICF, as compared to the ratio between observed concentration of dFdCTP and gemcitabine in blood cells [7]. Parameter optimization resulted in values of 14.1 ng/min for vmax,hENT1,BBB, 1.41·10² ng/min for vmax,hENT2,BBB, 7.09·10² ng/min for vmax,hENT1,CM, 7.09·10³ ng/min for vmax,hENT2,CM, 24.0 ng/min for vmax,DCK and 7.90·10-⁴ ml/min for CL,dFdCTP, respectively. After optimization, visual predictive check (VPC) indicated that LeiCNS-PK-GEM adequately predicted the observed mouse brain PK. Sensitivity analysis showed that the model is robust and not dependent on certain input parameters (<3% AUC changes with varying parameter inputs). Simulations revealed that dFdCTP accumulation in tumor ICF depends on both gemcitabine dose and, even more so, infusion time, and that much higher accumulation could be reached when infusion time is extended. Adjusting the infusion duration of an i.v. infusion of 1800 mg/m² from 30 minutes to 360 minutes revealed a 3.9-fold increase in gemcitabine AUC in tumor ICF (i.e. 5.1 to 19.7 µM·h), and a 3.2-fold increase of active metabolite dFdCTP AUC in tumor ICF (i.e. 54.5 to 175.1 µM·h). To illustrate, doubling the dose from 1800 mg/m² to 3600 mg/m² (with an infusion time of 30 minutes) yielded an ICF dFdCTP AUC of 82.6 µM·h, an increase of only 1.5-fold. An i.v. dose of 2100 mg/m² over 30 minutes was used in a prior trial in pediatric DMG patients that showed lack of efficacy [10]. Simulations of this regimen showed a tumor dFdCTP exposure of 59.8 µM·h. Yet, the maximum tolerated dose (MTD) of 3600 mg/m² over 6 hours, as investigated in pediatric leukemia patients [9] would have resulted in a simulated tumor dFdCTP AUC of 169.8 µM·h in DMG patients, 2.8-fold higher. Conclusions: LeiCNS-PK-GEM related gemcitabine plasma PK to CNS and brain tumor PK, allowing for the prediction of tumor exposure of gemcitabine and dFdCTP in mice, human adults and human pediatrics. Predictions showed the importance of infusion duration for the dFdCTP exposure in brain and brain tumor. Predictions for DMG patients indicated that previous gemcitabine trials in DMG may have shown lack of efficacy due to low target site exposure due to inadequate dosing over time. Prolonged infusion schedules can enhance dFdCTP tumor exposure, potentially improving therapeutic effect. LeiCNS-PK-GEM based simulations could guide innovative dosing regimen in children with DMGs and help translate promising preclinical in vitro and in vivo results to the clinic, contributing to the development of effective treatments for this devastating disease. References: [1] D. Wang et al., “Fimepinostat Impairs NF-kB and PI3K/AKT Signaling and Enhances Gemcitabine Efficacy in H3.3K27M-Diffuse Intrinsic Pontine Glioma,” Cancer Res., vol. 84, no. 4, pp. 598–615, 2024 [2] C. Bastiancich et al., “Gemcitabine and glioblastoma: challenges and current perspectives,” Nov. 2018, Elsevier Ltd. [3] M. Hirasawa et al., “The Extension of the LeiCNS-PK3.0 Model in Combination with the ‘Handshake’ Approach to Understand Brain Tumor Pathophysiology,” Pharm. Res., vol. 39, no. 7, pp. 1343–1361, Nov. 2022 [4] M. A. A. Saleh et al., “Using the LeiCNS-PK3.0 Physiologically-Based Pharmacokinetic Model to Predict Brain Extracellular Fluid Pharmacokinetics in Mice,” Pharm. Res., vol. 40, no. 11, pp. 2555–2566, Nov. 2023 [5] T. Matsumoto et al., “A physiologically based pharmacokinetic and pharmacodynamic model for disposition of FF-10832,” Int. J. Pharm., vol. 627, Nov. 2022 [6] T. Shimada et al., “Saturable Hepatic Extraction of Gemcitabine Involves Biphasic Uptake Mediated by Nucleoside Transporters Equilibrative Nucleoside Transporter 1 and 2,” J. Pharm. Sci., vol. 104, no. 9, pp. 3162–3169, Sep. 2015 [7] E. J. B. Derissen et al., “Intracellular pharmacokinetics of gemcitabine, its deaminated metabolite 2′,2′-difluorodeoxyuridine and their nucleotides,” Br. J. Clin. Pharmacol., vol. 84, no. 6, pp. 1279–1289, Jun. 2018 [8] R. J. Boosman et al., “Is age just a number? A population pharmacokinetic study of gemcitabine,” Cancer Chemother. Pharmacol., vol. 89, no. 5, pp. 697–705, Nov. 2022 [9] P. G. Steinherz et al., “Phase I study of gemcitabine (difluorodeoxycytidine) in children with relapsed on refractory leukemia (CCG-0955): A report from the children’s cancer group,” Leuk. Lymphoma, vol. 43, no. 10, pp. 1945–1950, Nov. 2002 [10] M. Van Den Bent et al., “H3 K27M-Altered glioma and diffuse intrinsic pontine glioma: Semi-systematic review of treatment landscape and future directions,” Nov. 2024, Oxford University Press.
Reference: PAGE 34 (2026) Abstr 12257 [www.page-meeting.org/?abstract=12257]
Poster: Oral: Preclinical and Translational modelling to support drug discovery and development