AbstractMethotrexate (MTX) is widely used in treating chronic inflammatory diseases and in chemotherapy for both adult and pediatric patients. Its pharmacokinetics are complex, as MTX is a substrate of multiple transporters. Particularly during high-dose chemotherapeutic use, it is associated with severe toxicities including reversible kidney impairment and hematological toxicity. Physiologically based pharmacokinetic (PBPK) modeling offers a mechanistic approach to predict drug pharmacokinetics and guide dosing, especially in pediatric applications where clinical data is limited. A whole-body parent-metabolite PBPK model of MTX and 7-hydroxymethotrexate was developed in PK-Sim® (version 12.1) for intravenous and oral administration in adults and children. Model development used 42 MTX and 16 metabolite plasma concentration-time profiles and showed strong predictive performance, with 53/58 area under the plasma concentration-time curve and 38/39 maximum plasma concentration values within a 2-fold range of observed data. Pediatric extrapolation incorporating transporter ontogeny yielded adequate predictions. Drug-drug interaction simulations for five perpetrator drugs met established acceptance criteria, and renal impairment simulations revealed markedly increased exposure in the high-dose setting. The model will be made publicly available to support dose optimization across clinical scenarios.IntroductionMethotrexate (MTX) is a cornerstone therapy for chronic inflammatory diseases, including rheumatoid arthritis (RA), Crohn’s disease, as well as psoriasis, and is also widely used in oncology. In chemotherapy, high-dose methotrexate (HDMTX) is an essential component of multi-agent regimens, for example in acute lymphoblastic leukemia (ALL)1,2. Depending on the indication, treatment setting, and patient age, MTX may be administered orally (p.o.), intravenously (i.v.), subcutaneously, intramuscularly, or, in specialized protocols requiring greater central nervous system exposure, intrathecally3. The main routes of application, however, are i.v. and p.o. administration using significantly different dosing regimens, ranging from low weekly doses of 5–30 mg in chronic inflammatory diseases to HDMTX doses up to 33,000 mg/m² within a single chemotherapy cycle1,2. This broad therapeutic use, combined with a narrow therapeutic window at high doses, makes precise control of systemic exposure crucial.MTX inhibits dihydrofolate reductase (DHFR), resulting in impaired folate metabolism and subsequent inhibition of cell proliferation and immunosuppression1. As more than 80% of MTX is eliminated unchanged via the kidneys, renal function is a key determinant of MTX exposure and toxicity risk. In HDMTX setting, kidney function is therefore monitored closely due to the risk of reversible acute kidney injury. Therapeutic drug monitoring (TDM) of MTX is commonly performed 24–48 h after dosing, with a typical target plasma concentration of 1 µmol/L at 48 h to avoid prolonged exposure and associated toxicities4. The administration of folinic acid (“leucovorin rescue therapy”) replenishes reduced folate pools and thereby “rescues” normal cells from the antifolate effects of MTX, primarily providing pharmacodynamic protection against toxicity rather than substantially altering MTX plasma pharmacokinetics. Rescue intervention is intensified when delayed MTX clearance and prolonged systemic exposure are detected by pharmacokinetic monitoring4,5. However, substantial interindividual variability in MTX plasma concentrations makes it difficult to estimate individual exposure, especially before the first dose. Clinically relevant variability arises from multiple sources and has been linked to renal function, numerous uptake and efflux transporters involved in MTX distribution and elimination, as well as genetic polymorphisms in transporter genes6. The simultaneous application of concomitant chemotherapeutic agents such as thiopurines and variability in aldehyde oxidase (AO) activity, which converts approximately 3–33% of MTX to 7-hydroxymethotrexate (7-OH-MTX) are further reasons for variability7,8. 7-OH-MTX itself contributes to the overall exposure profile and may impact efficacy and toxicity9. In pediatric patients, such as those treated with MTX for ALL or juvenile idiopathic arthritis, developmental changes of renal function and maturation of drug transporters introduce additional age-dependent variability in MTX and 7-OH-MTX pharmacokinetics10. In summary, these factors make it challenging to anticipate individual exposure and to evaluate the impact of co-medications on both adults and children.Mechanistic mathematical modeling provides an opportunity to address this complexity by creating a digital representation (“digital twin”) of patients that captures drug-specific and system-specific determinants of MTX exposure11. Physiologically based pharmacokinetic (PBPK) modeling, in particular, has become an established tool in drug development and regulatory decision-making to quantitatively predict drug pharmacokinetics, evaluate drug–drug interactions (DDIs), and support dose selection in special populations12. PBPK models combine drug properties (e.g., solubility, permeability, protein binding) with physiological parameters (e.g., organ size, blood flows, enzyme and transporter expression) to simulate absorption, distribution, metabolism, and excretion. Such models can be used to extrapolate between adult and pediatric populations, explore the impact of organ impairment or transporter modulation, and provide a mechanistic framework to inform and interpret TDM13,14.Given the clinical importance of MTX across age groups, the relevance of renal function, and the prominent role of transporters and DDIs, there is a need for a comprehensive parent–metabolite PBPK model that jointly describes MTX and 7-OH-MTX across routes of administration and clinically relevant scenarios. Therefore, this study has four main objectives: (i) to develop and validate a parent–metabolite PBPK model for MTX and 7-OH-MTX of i.v. and oral dosing across fasted and fed states in adults, (ii) to incorporate ontogeny functions, enable pediatric extrapolation, and assess model utility for pediatric MTX therapy, and (iii) to simulate clinically relevant transporter-mediated DDIs with cyclosporine A (CsA), ibuprofen (IBU), naproxen (NAP), probenecid (PB), and rifampicin (RFP), and lastly (iv), to quantify the impact of renal impairment on MTX exposure for both high-dose oncologic and low-dose rheumatologic regimens. The developed parent–metabolite PBPK model files will be made publicly available, and details not presented in the main body of the manuscript are provided in the accompanying comprehensive Supplementary Materials.MethodsSoftwareThe Open Systems Pharmacology Suite version 12.1 (www.open-systems-pharmacology.org) was used as an open-source PBPK modeling software, with its tool PK-Sim®. Parameters were optimized using integrated Monte-Carlo and Levenberg-Marquardt optimization algorithms. For digitization of concentration-time profiles the GetData Graph Digitizer was used (version 2.26). Further calculation as well as visualization were performed with the R programming language version 4.3.2 and RStudio version 2023.12.0 + 369 (RStudio Inc., Boston, MA, USA).Clinical dataPlasma concentration-time profiles of MTX and its main metabolite 7-OH-MTX obtained from adults and children were digitized from published literature according to Wojtyniak et al.15. Fractions excreted in urine (FE) were also collected and digitized accordingly. Clinical studies investigating i.v. and p.o. administration under fasted and fed conditions were extracted. The clinical dataset used for model development and evaluation comprised 43 clinical studies, consisting of 58 concentration–time profiles from 557 patients, including adults (n = 515) and children (n = 42). The profiles represent data as reported in the respective publications. The majority were study-level arithmetic mean profiles (n = 43), while a minority were median profiles (n = 2) and individual patient profiles (n = 13). The studies covered healthy volunteers (n = 7), inflammatory diseases (n = 22; e.g., RA, psoriasis) and oncologic indications (n = 14; e.g., ALL, osteosarcoma, diffuse large B-cell lymphoma). Dosing ranged from low dose (2.5 mg to 30 mg) to high dose setting (200 mg/m² to 33,600 mg/m²) and were administered as i.v. bolus (n = 5), i.v. infusion (n = 8), or p.o. tablets (n = 30).For the development of the base model solely data measured in adult populations was used, which was divided into training and test datasets based on the following criteria: training datasets included plasma concentration-time profiles with many measurement points over a long period of time, a wide dosing range covering i.v. and oral single as well as multiple dose regimens and, if possible, main metabolite 7-OH-MTX. Of the 58 included plasma concentration-time profiles, 12 were used as training profiles, while the remaining 46 were used as test profiles. The two median profiles were part of the test dataset and were therefore not used for parameter estimation. Training data was used for model development and test data was used for model evaluation. Datasets measured in children were used only for evaluating the model performance after pediatric scaling. A detailed description of the clinical dataset, including the type of data reported for each profile, is provided in Section S1.3, Table S4 of the Supplementary Materials.PBPK model building and evaluationMTX parent-metabolite model development was initiated with an extensive literature search for physicochemical properties of MTX and 7-OH-MTX16,17,18. A digital twin was created for each study population based on reported mean and mode data on age, sex, weight, height, body mass index as well as ethnicity. Missing demographic data was substituted by suggested values generated in PK-Sim® based on the included population databases. Relative expressions for enzymes and transporters of interest in the different organs were specified according to the PK-Sim® expression database and compared with expressions in Proteinatlas19. Simulations representing each study were established containing the corresponding digital twin, the plasma or renal exposure of MTX and, if available its metabolite 7-OH-MTX as well as application protocol. To account for physiological changes under fed conditions, gastric emptying time was set to 45 min (threefold higher than in fasted simulations), consistent with observations by Fisher et al.20. If drug-dependent parameter values could not be informed from the literature, they were estimated using mathematical optimization algorithms. First, Monte Carlo optimization was applied minimizing least squares differences between observed and simulated data using broad parameter boundaries. The Levenberg–Marquardt algorithm was used for final refinement. Observed data of the training dataset, consisting of plasma and urine concentration-time profiles, was used throughout these optimization procedures. In a stepwise procedure, parameters values of the parent compound and metabolite were estimated consecutively, adding the metabolite only after satisfactory results were achieved of MTX itself.The MTX parent-metabolite model was evaluated both graphically and quantitatively. Model performance was characterized by its potential to describe the training datasets as well as predict the test dataset. For this, predicted concentration-time profiles were compared visually to their corresponding observed data points. In addition, goodness-of-fit (GOF) plots were used to present the relation between predicted and observed concentration values. GOF plots were created to compare AUClast for all predicted versus observed profiles of MTX and 7-OH-MTX and Cmax values for profiles with p.o. application of MTX as well as p.o. 7-OH-MTX profiles. AUClast was calculated by noncompartmental analysis using the linear-up/log-down trapezoidal rule (linear interpolation for ascending concentrations and logarithmic interpolation for descending concentrations). Mean relative deviations (MRDs) for all plasma concentration predictions along with geometric mean fold errors (GMFEs) for all predicted AUClast and Cmax values were calculated according to Eqs. (1) and (2). MRD and GMFE values of ≤ 2 describe good model performance.$$ MRD = 10^{x} \:with\:x = \sqrt {\frac{{\mathop \sum \nolimits_{{i = 1}}^{k} \left( {log_{{10}} {\text{~c}}_{{obs,i}} - log_{{10}} \:c_{{pred,i}} } \right)^{2} }}{n}} $$(1)where:cobs, i = observed plasma concentration, cpred, i = predicted plasma concentration, MRD = mean relative deviation and n = number of observed values.$$GMFE={10^x}{\text{~with~}}x=\frac{{\mathop \sum \nolimits_{{i - 1}}^{m} \mid lo{g_{10}}(\frac{{{p_{obs}}}}{{{p_{pred}}}})\mid }}{m}$$(2)where:GMFE = geometric mean fold error, pobs = observed PK parameter such as AUClast or Cmax, ppred = corresponding predicted PK parameter such as AUClast or Cmax and m = number of studies.Global sensitivity analysis was performed with the two‑stage, moment‑independent δ method according to Borgonovo, De Carlo et al. and Cuquerella-Gilabert et al. implemented in the open‑source R package ospgsa21,22,23,24. For plasma MTX and 7‑OH‑MTX exposure, each parameter was ranked and classified by its direct (δ₁) and correlation‑aware (δ₂) importance. Full methods, settings and results are provided in Section S2.8. of the Supplementary Materials.Pediatric modeling and evaluationThe developed PBPK model for MTX was scaled and applied to pediatric populations. In the module for humans, mean values for children of all age groups are included in PK-Sim® to account for physiological and anatomical differences between children and adults, meaning age-related changes such as size, composition of tissue compartments, protein binding and maturation of metabolization and elimination processes were scaled to children. The included pediatric populations were based on each clinical study data of children receiving MTX as chemotherapy, respectively. The unbound fraction of MTX was fit within literature-reported ranges observed in children receiving HDMTX chemotherapy, to account for reduced albumin levels in the pediatric patient cohort reported by Dong et al. and Jamaladdin et al.25,26. Ontogeny of transporter breast cancer resistance-related protein (BCRP), multidrug resistance protein 3 (MRP3), and organic anion transporter 3 (OAT3) was considered to resemble the expression profiles of the transporters in different age groups. MRP3 ontogeny data was taken from Prasad et al., who investigated the ontogeny of protein expression of MRP3 in neonates, infants, children, adolescents and adults10. The study showed a significantly lower expression of MRP3 in infants and adolescents compared to adults independent of sex10. Ontogeny scaling factors for BCRP and OAT3 were readily available and taken from Hunt et al.27.The protein expression data taken from Prasad et al. was digitized and transferred to the protein expression of MRP3 in PK-Sim® aligned to the equation described in the PK-Sim® ontogeny database version 7.3, shown here in Eq. (3)10,28.$$A=\frac{{PM{A^n}}}{{\left( {PM{A^n}+~{A_{0.5}}^{n}} \right)}}$$(3)where:A = relative activity at PMA, A0.5 = PMA (in weeks) at 50% activity compared to adult, n = Hill coefficient, PMA = post-menstrual age in weeks (= gestational age at birth + age since birth).Modeling kidney impairmentThe developed model was tested in scenarios with varying degrees of kidney impairment. For this purpose, two studies were selected, and different stages of reduced renal function were modeled as decreased glomerular filtration rate (GFR) values in the study populations. The first is a low-dose oral MTX application in patients with RA by Namour et al.29. The second is a HDMTX application during chemotherapy in patients with malignant cancers by Stoller et al.30. GFR values were aligned with renal function categories as defined by the United States Food and Drug Administration (FDA) guideline for pharmacokinetics in patients with impaired renal function: I: normal renal function (≥ 90 mL/min), II: mild impairment (60 < 90 mL/min), III: moderate impairment (30 < 60 mL/min), IV: severe impairment and kidney failure not receiving dialysis (< 30 mL/min)31.DDI interaction modeling and evaluationDDI performance of the MTX model was assessed by simulating different transporter-based interactions with the perpetrator drugs PB, RFP and CsA as well as interactions involving displacement from plasma protein binding and a reduction in GFR during comedication with nonsteroidal anti-inflammatory drugs, such as IBU and NAP. For evaluation of the DDIs, predicted and observed plasma concentration-time profiles of the victim drug administered with and without the respective perpetrator drug were plotted and compared graphically. Moreover, DDI AUClast and Cmax ratios were calculated for each predicted and observed profile as specified by Eq. (4). Predictions of the DDI effect were validated using predictive measures proposed by Guest et al., allowing for 20% variability to account for uncertainties in observed ratios approaching value one32.$$DDIPK\:parameter\:ratio=\frac{{P{K_{parameterDDI}}}}{{P{K_{parameterControl}}}}$$(4)where:PKparameter = AUClast or Cmax, PKparameterDDI = AUClast or Cmax of victim drug during perpetrator co-administration and PKparameterControl = AUClast or Cmax of victim drug control.GMFE values for the predicted DDI AUClast or Cmax ratios were calculated according to Eq. (2).ResultsModel buildingWhole-body PBPK models for MTX and 7-OH-MTX were developed in PK-Sim®. The population characteristics and details of the clinical study data are listed in Table S4 of Section S1.3. as well as Table S16 of Section S4.1. in the Supplementary Materials.To mechanistically capture MTX disposition across routes and dose ranges, the final model incorporated transporter-mediated processes at key sites of absorption and disposition. Intestinal handling was represented by passive processes complemented by MRP3 (basolateral efflux to portal blood), cellular uptake by reduced folate-coupled transporter 1 (RFC1), and apical efflux via BCRP. Hepatic distribution included uptake via organic anion transporting polypeptide 1B1 (OATP1B1) and BCRP-mediated efflux processes. Metabolic formation of 7-OH-MTX from MTX in hepatocytes was implemented via aldehyde oxidase (AO). Renal elimination of MTX was described by glomerular filtration (implemented as a GFR fraction of 1.0) in combination with transporter-mediated tubular handling, including OAT3 and BCRP.MTX and 7-OH-MTX are converted intracellularly to active polyglutamates, which form an intracellular depot and bind to target enzymes1. In the model, this “trapping” mechanism was represented by reversible binding to the target enzyme DHFR, thereby capturing target-mediated distribution and retention behavior. Michaelis-Menten constants (KM) for metabolizing enzyme and transporter processes as well as binding affinity constants (Kd) were informed from literature where available. Turnover numbers (kcat) were estimated by fitting for transporters and metabolizing enzymes based on the clinical concentration–time data. For DHFR binding, the dissociation rate (koff) was fitted to reproduce the observed extent of intracellular retention, while the binding affinity of MTX to DHFR was fixed to literature values by Rajagopalan et al.33. An overview of drug-dependent parameters and the implemented model parameters are listed in Table 1.We were unable to identify published KM data for the metabolite 7-OH-MTX in the available literature. Given the reported similarity of MTX and 7-OH-MTX pharmacokinetics, key efflux processes were represented using MRP3 with an assumed KM equivalent to MTX, while the corresponding activity parameter kcat was fitted. In addition, elimination of the metabolite was simulated by an unspecific renal clearance process. Binding of 7-OH-MTX to DHFR with an assumed Kd similar to MTX and metabolite lipophilicity was optimized to support a consistent description of distribution. The structure of the model is depicted in Fig. 1.Fig. 1Full size imageSchematic overview of the PBPK parent-metabolite model. Left: schematic of the basic PBPK model structure. Right: digital patient twin with key model features. To illustrate transporter-mediated MTX distribution, the three pharmacokinetically most relevant organs are shown with the transporters implemented into the model. The model is built on compound-specific properties and human physiology, representing organs and tissues as compartments connected via the vascular system34. In the main structure, MTX is metabolized by AO to 7-OH-MTX. MTX interacts with BCRP, MRP3, OATP1B1, OAT3 and RFC1 and inhibits DHFR, while 7-OH-MTX also binds MRP3. Greyed out transporters were not included in the final model. An overview of modeled DDI-scenarios is provided through red lines indicating perpetrator drug interactions with relevant transporters and enzymes35,36,37. IBU and NAP act as perpetrator drugs by decreasing GFR and plasma protein binding of MTX38. Ontogeny of BCRP, MRP3 and OAT3 protein expression from Hunt et al. and Prasad et al. was implemented to scale the model to pediatric populations10,27. AO: aldehyde oxidase, BCRP: breast cancer resistance protein, DHFR: dihydrofolate reductase, MRP1-4: multidrug resistance-related protein 1–4, MTX: methotrexate, OATP1B1: organic anion transporting polypeptide 1B1, OAT3: organic anion transporter 3, RFC1: reduced folate-coupled transporter 1. Drawings by Servier Medical Art (licensed under CC BY 4.0)39.Table 1 Drug-dependent and general model parameters of the parent-metabolite PBPK model of MTX and 7-OH-MTX.Full size tableModel evaluationThe presented MTX parent-metabolite model allowed good data description of the training and test dataset across i.v. and p.o. administrations, confirmed by quantitative evaluation. Oral absorption of MTX is known to be dose-dependent, with decreasing absorption rates and a plateau at weekly doses above 15 mg, consistent with saturation of intestinal transporter-mediated processes (e.g., BCRP, MRP3) reported in the literature48. Exemplary simulations for MTX and 7-OH-MTX illustrating model performance under fasted and fed conditions are shown in Fig. 2.Quantitative evaluation confirmed adequate predictive performance. For fasted studies, mean GMFE values for AUClast and Cmax (across MTX and 7-OH-MTX) were 1.47 and 1.32, respectively, while fed studies yielded mean GMFE values of 1.21 (AUClast) and 1.21 (Cmax), indicating that the model captured the food effect without loss of performance. Across the entire dataset, 53/58 AUClast predictions and 38/39 Cmax predictions were within a two-fold range of the observed values (Fig. 3). Additionally, 29/58 AUClast predictions and 22/39 Cmax predictions fall within 1.25-fold of the observed values, corresponding to the conventional 80–125% bioequivalence range (Fig. 3). The overall MRD was 2.03 (range: 1.16–4.21), and the mean GMFE values (across MTX and 7-OH-MTX) were 1.43 for AUClast (range: 1.00–5.49) and 1.28 for Cmax (range: 1.00–2.28), supporting sufficient descriptive and predictive performance. The MRD for the training dataset in adult populations was 2.02 and the mean GMFE values were 1.50 for AUClast and 1.24 for Cmax. Within the training dataset, 10/12 AUClast predictions and 7/7 Cmax predictions were within a two-fold range of the observed values. For the test dataset in adult populations these numbers amount to 2.02 for the MRD (2.04 including pediatric populations) as well as 1.40 and 1.29 for AUClast and Cmax. Within the test dataset, 43/46 AUClast predictions and 31/32 Cmax predictions were within a two-fold range of the observed values. Additional graphical and quantitative model evaluation is shown in Section S2 and S4 of the Supplementary Materials: plasma concentration-time profiles are shown in Figs. S1-S14 and S26-S27, urinary excretion profiles are shown in Figure S13 and further GOF-plots are shown in Figure S16 and S28. Individual MRD and GMFE values for AUClast and Cmax for all 58 profiles are shown in Tables S5 – S8 as well as Tables S19 and S20.Global, correlation-aware sensitivity analysis confirmed the principal determinants of plasma AUC for both administration routes. MTX exposure was governed mainly by renal glomerular filtration and the plasma fraction unbound. After oral dosing, MTX exposure was additionally governed by OAT3- and MRP3-mediated transport. 7-OH-MTX exposure was governed by AO–mediated metabolite formation and BCRP-mediated efflux. Detailed description of the performed local and global sensitivity analysis is shown in Section S2.7. (Figures S17 and S18 and Table S9) and S2.8. (Figures S19 – S22 and Table S10 and S11) of the Supplementary Materials.Fig. 2Full size imageMTX and 7-OH-MTX model performance. Illustrated are exemplary semilogarithmic plots of predicted plasma concentration-time (a-e) and fraction excreted to urine profiles (f) of MTX and 7-OH-MTX in i.v. and p.o. single and multiple dose administrations. Solid and dashed lines represent model predictions, while full circles show corresponding observed data. 7-OH-MTX: 7-hydroxymethotrexate, fas: fasted, FE to urine: fraction excreted to urine, iv inf: intravenous infusion, md: multiple doses, MTX: methotrexate, n: number of patients in clinical study, po: per oral, sd: single dose, te: test data, tr: training data.Fig. 3Full size imageGoodness-of-fit plots of the MTX parent-metabolite model. Predicted versus observed AUClast of training (a) and test (b) data, Cmax of training p.o. studies (c) and test p.o. studies (d). Solid lines are lines of identity, whereas dotted lines indicate 1.25-fold and dashed lines 2-fold deviation from respective observed value. 7-OH-MTX: 7-hydroxymethotrexate, AUClast: area under the plasma concentration-time curve between first and last concentration measurement, Cmax: maximum plasma concentration, MTX: methotrexate.Pediatric model evaluationFor pediatric applications, the adult parent-metabolite PBPK model was extrapolated to children using the PK-Sim® pediatric physiology framework, accounting for age-dependent anatomical and physiological differences based on the underlying population database49,50,51. The pediatric evaluation dataset comprised 40 children (1–17 years) treated for malignant diseases. The range of the applied MTX doses was between 16.45 mg and 30,240 mg. To capture age-dependent changes in MTX disposition, ontogeny of intestinal MRP3 was implemented to account for intestinal maturation, and ontogeny of renal BCRP and OAT3 was implemented to account for renal maturation during pediatric scaling10,27. The fraction unbound of MTX was scaled to 61.06% compared to 50% in adult populations to account for age specific changes in albumin expression25,26. For the individual presented in the study by Slørdal et al., a very high concentration of the metabolite was observed. To account for this unexpected increase in 7-OH-MTX formation, the individual activity of AO was fitted for this individual. With these adjustments, the model provided satisfactory predictive performance in the pediatric dataset, with an overall mean GMFE for AUClast of 1.52 as well as 1.47 for Cmax.Quantitative results are shown in Table 2 and exemplary simulations are illustrated in Fig. 4. Additional graphical and quantitative model evaluation is shown in Section S4 of the Supplementary Materials: full clinical study data of all pediatric studies in Table S18, plasma concentration-time profiles are shown in Figures S26 and S27, and further GOF-plots are shown in Figure S28. Individual MRD and GMFE values for AUClast and Cmax for all pediatric profiles are shown in Tables S19 – S20.Table 2 Quantitative evaluation of pediatric model performance.Full size tableFig. 4Full size imagePediatric model performance. The profiles show exemplary semilogarithmic plots of pediatric plasma concentration-time data (a-c) as well as the AUClast GOF-Plot of the pediatric model (d). Solid lines are lines of identity, whereas dotted lines indicate 1.25-fold and dashed lines 2-fold deviation from respective observed value. In (d) points indicate MTX profiles and triangles indicate 7-OH-MTX profiles. 7-OH-MTX: 7-hydroxymethotrexate, AUClast: area under the plasma concentration-time curve between first and last concentration measurement, fas: fasted, iv inf: intravenous infusion, MTX: methotrexate, n: number of patients in clinical study, po: per oral, sd: single dose.Kidney impairmentThe presented PBPK model predicts a fraction of MTX excreted to urine of 45% for low-dose p.o. studies (p.o. tablet of 15 mg), 70% for low-dose i.v. studies (i.v. bolus of 15 mg) and more than 80% for studies administering HDMTX therapy (i.v. infusion of 350 mg and i.v. infusion of 8067 mg) at 24 h post-dose. To assess the model’s sensitivity to renal function, we simulated the impact of reduced GFR using kidney function categories according to the FDA guideline for pharmacokinetics in patients with impaired renal function31. Two representative clinical regimens were selected and simulated for 72 h following single dose administration: a low-dose oral administration (7.5 mg, single dose, fed) and a high-dose intravenous infusion (50 mg/kg, infusion time of six hours, single dose). Results are shown in Fig. 5.Reducing GFR had a comparatively modest impact on MTX exposure after low-dose oral administration, but markedly increased and prolonged MTX exposure in the high-dose intravenous setting: for low-dose oral application, normal kidney function (I) reaches about 9.81 × 10− 3 µmol/L at 24 h post-dose, whereas severe renal impairment (IV) reaches around 1.52 × 10− 2 µmol/L, corresponding to roughly a 1.5x increase. However, in the high-dose intravenous setting, (I) is at approximately 1.03 µmol/L at 24 h post-dose, while (IV) reaches 84.2 µmol/L, which corresponds to an increase of about 80-fold. Consistent with these concentration–time profiles, AUClast increased progressively with declining GFR, with a substantially steeper exposure increase observed for the high-dose regimen (Fig. 5d) compared with the low-dose regimen (Fig. 5c).Fig. 5Full size imageImpact of reduced GFR on MTX exposure in representative high- and low-dose regimens. Simulations were performed by varying GFR according to kidney function categories (I-IV) defined in the FDA guideline regarding pharmacokinetics in patients with renal impairment31. (a, b) Simulated MTX plasma concentration–time profiles. (c, d) Corresponding AUClast values by GFR stage. Panels (a, c) represent a low-dose p.o. regimen (7.5 mg, single dose, fed; Namour et al.)29. Panels (b, d) represent a high-dose i.v. infusion regimen (50 mg/kg, single dose; Stoller et al.)30. AUC: area under the plasma concentration-time curve, I: normal renal function (> 90 mL/min), II: mild impairment (60–89 mL/min), III: moderate impairment (30–59 mL/min), IV: severe impairment and kidney failure not receiving dialysis (< 30 mL/min), iv inf: intravenous infusion, MTX: methotrexate, po: per oral, sd: single dose.DDI model evaluationThe final MTX parent–metabolite PBPK model was evaluated in five clinically relevant DDI scenarios representing two mechanistic classes: (i) direct transporter-/enzyme-mediated interactions implemented as competitive inhibition of processes relevant for MTX disposition (AO, BCRP, MRP3, OATP1B1, OAT3) for the perpetrator drugs CsA, PB, and RFP, and (ii) indirect interactions primarily affecting renal elimination and protein binding for the non-steroidal anti-inflammatory drugs (NSAID) such as IBU and NAP. The implemented DDI mechanisms are summarized in the model schematic (Fig. 1).For the transporter-mediated DDIs, the model captured the observed exposure changes across PB, RFP, and CsA scenarios with a mean GMFEAUClast of 1.82, and all three interactions fulfilled the acceptance criteria defined by Guest et al.32. PB co-administration resulted in a pronounced increase in MTX exposure consistent with inhibition of renal tubular handling, while the RFP scenario produced a more moderate increase in systemic exposure consistent with OATP1B1-related effects. For CsA, which affects multiple pathways, the model accurately reproduced the overall interaction magnitude for the parent-compound MTX when the relevant transporter and enzyme processes were inhibited. Here, the model captured a marked reduction in metabolite formation consistent with inhibition of AO through CsA, however, the model did not meet the criteria defined by Guest et al. for describing the interaction at the metabolite level.For NSAID co-medication, the interaction was implemented by (i) reducing the GFR fraction to match the reported decrease in renal MTX clearance by around 40% for concomitant NAP and 18% for IBU application and (ii) adjusting MTX plasma protein binding (fraction unbound) from 50% to 44% for simultaneous NAP and 42% for IBU dosing according to literature values38. With this combined representation, both NSAID scenarios (IBU and NAP) met the Guest et al. criteria, with a mean GMFEAUClast of 1.4632.Across all five DDI scenarios, overall performance was adequate with a mean GMFEAUClast of 1.69. To account for interindividual variability in transporter activity across DDI studies, BCRP activity (kcat) was individualized in a two-step procedure by first fitting the control arm and subsequently applying the individualized value to the corresponding DDI arm. GOF-ratio-plots of AUClast and Cmax are shown in Fig. 6.For the evaluated DDI scenarios across both mechanistic classes, the clinical studies showed only limited changes in MTX exposure. For example, IBU increased MTX AUC by 35.8% (AUC ratio 1.36) and RFP increased MTX AUC by 26.7% (AUC ratio 1.27), suggesting that, overall, the effect of the perpetrator drugs on MTX exposure is modest. Further details are presented in Section S3 of the Supplementary Materials: full description of the simulated clinical studies in Tables S12 – S13, drug-dependent parameters of the perpetrator drugs in Tables S14 – S16, as well as quantitative DDI evaluation in Table S17. Plasma concentration-time profiles are shown in Figs. S23 – S25.Fig. 6Full size imageGoodness-of-fit plots of the DDI models. (a) predicted vs. observed AUClast ratios, (b) predicted vs. observed Cmax ratios. Solid lines are lines of identity, whereas dotted lines indicate 1.25-fold and dashed lines 2-fold deviation from respective observed value. The sweeping line indicates measures proposed by Guest et al., allowing for 20% variability to account for uncertainties in observed ratios approaching value one32. GOF plots: goodness-of-fit plots, AUClast: area under the plasma concentration-time curve between first and last concentration measurement, Cmax: maximum plasma concentration, DDI: drug-drug interaction.DiscussionIn this work, a whole-body parent–metabolite PBPK model was developed and evaluated to describe MTX and 7-OH-MTX across clinically relevant routes of administration and dose levels (2.5 mg – 30,240 mg). Its applicability for pediatric extrapolation, renal impairment scenarios, and transporter-mediated DDIs was demonstrated. The model showed good descriptive and predictive performance across the compiled dataset, with overall GMFE values of 1.43 (AUClast) and 1.28 (Cmax), and 53/58 AUClast and 38/39 Cmax predictions within a 2-fold range of observed values. Collectively, these results support the model as a consistent mechanistic framework across both low-dose and high-dose MTX use cases. Intrathecal, subcutaneous, and intramuscular administration were not implemented in the present model. This was based on several considerations: (i) the model is intended for further development toward TDM support in HDMTX treatment, where i.v. infusion is the predominant route of administration, (ii) model development was deliberately focused on achieving robust predictive performance for both i.v. and p.o. MTX, and (iii) systematic pharmacokinetic sampling data for the other routes are limited.A major challenge in mechanistic MTX modeling is the large number of transporters involved in intestinal absorption, hepatic distribution, and renal handling, while kinetic parameters (KM, kcat, Ki) are incomplete or heterogeneous across sources45. For example, MTX is not only a substrate of MRP3, but also of other members of the ABC subfamily C transporters, such as multidrug resistance-related protein 2 and 4 (MRP2 and MRP4). The transport processes these mediate are, however, already represented in the model. MRP2-mediated apical efflux (enterocytes, hepatocytes and renal tubular cells) is covered by BCRP, MRP4 is expressed on basolateral membranes in hepatocytes (overlapping the implemented MRP3) and apically in the renal proximal tube (covered by BCRP). Implementing several transporters with shared localization, direction and KM of similar order of magnitude would yield to substantial parameter non-identifiability and was therefore not considered further56. For this reason, model development prioritized the transporters and enzymes AO, BCRP, DHFR, MRP3, OAT3, OATP1B1, and RFC1, which show dominant expression patterns compared with other transporters in the same families balancing physiological resemblance and parameter identifiability issues56. Transporter polymorphisms may contribute to interindividual variability in MTX disposition, but the available evidence is inconsistent across transporters, study populations, and clinical endpoints57. While some associations have been reported, findings are often not reproducible, which currently limits their use in mechanistic PBPK modeling. Datasets required for PBPK model development, especially in pediatric populations, are largely lacking, so transporter polymorphisms could not be parameterized in the present model. Population pharmacometric studies do suggest that genetics can influence MTX disposition (for example, for SLCO1B1), but these results do not yet provide the transporter-specific scaling information needed for direct PBPK implementation57.For 7-OH-MTX, limited quantitative information on transporter affinities and kinetics necessitated simplifying assumptions. Metabolite formation was implemented via AO, and 7-OH-MTX kinetics were captured using a reduced mechanistic representation, including assumed similarities to MTX for selected interactions such as affinity to MRP3 and a fitted renal clearance component. The model’s inability to describe the MTX–CsA interaction at the metabolite level indicates that the simplified representation of the metabolite’s pharmacokinetic pathway has limitations, particularly for complex DDIs that require accurate characterization of multiple transporter interaction sites. Reported conversion of MTX to 7-OH-MTX ranges from approximately 3% to 33%, suggesting substantial inter-individual variability in AO activity7. In line with this, Hutzler et al. reported at least a 17-fold variability in AO activity across human hepatocyte donors58. Foti et al. further demonstrated that AO single-nucleotide polymorphisms can alter AO enzyme function, supporting a potential genetic contribution to AO-related variability59. However, there is currently no data directly linking AOX1 genetic variants to altered MTX-to-7-OH-MTX conversion. Therefore, the pronounced variability observed in clinical routine remains mechanistically unresolved and, at present, can only be addressed indirectly in the model. Additional experimental data, particularly on AO-related MTX binding parameters and on AOX1 genotype–phenotype relationships, would improve mechanistic fidelity, especially for the transporter-mediated distribution and elimination of 7-OH-MTX. Despite these constraints, the model yielded satisfactory predictive performance for the metabolite with GMFE values of 1.62 (AUClast) and 1.41 (Cmax), supporting its use as a parent–metabolite framework.DDI simulations reproduced the observed direction and magnitude of clinically relevant interactions across multiple perpetrator classes. The strongest interaction in the evaluated set was captured for PB, consistent with OAT3 inhibition reducing renal MTX elimination as well as reduction of metabolite formation in concomitant CsA therapy, consistent with strong inhibition of AO as metabolizing enzyme. More modest MTX exposure changes observed for RFP and CsA were represented through inhibition of OATP1B1 and predominantly BCRP-related processes, respectively. It is important to note that these perpetrator drugs affect additional transporters beyond those explicitly represented, such as p-glycoprotein (P-gp), organic anion transporter 1 and 4 (OAT1 and OAT4) and MRP4, and that such unmodeled pathways contribute to residual discrepancies. Inevitably, the simplification of the model limits its ability to reflect the full extent of the observed interaction between perpetrator and victim drug. For NSAIDs, the literature indicates that reduced MTX renal clearance is multifactorial38. Accordingly, the interaction was implemented pragmatically via a reduced GFR fraction and a minor protein binding adjustment, which resulted in satisfactory performance within accepted DDI prediction criteria. A mean GMFEAUClast value of 1.69 suggests that the model captures the clinically relevant directionality of transporter and renal interactions.Building on adult model performance and the DDI evaluation, pediatric extrapolation was enabled through the PK-Sim® pediatric physiology framework, combined with explicit implementation of BCRP, MRP3, and OAT3 ontogeny and age-dependent scaling of the MTX fraction unbound to account for developmental changes in albumin expression. This is consistent with reports of increased MTX unbound fractions in children receiving HDMTX (approximately 10–80%)10,25,26. PK-Sim®’s pediatric physiology framework accounts for age-dependent organ function, including renal function, but does not fully capture maturation of individual renal transport processes. This is important in pediatric PBPK modeling because transporter activity may mature independently of overall kidney function and can influence active secretion and reabsorption, especially in younger children. Consistent with this, Hunt et al. showed that adding renal transporter ontogeny profiles improved predictive performance for transporter substrates, particularly in neonates and infants, even after accounting for other maturational physiology27. The pediatric simulations showed adequate agreement with observed data (overall mean GMFEAUClast 1.52, overall mean GMFECmax 1.47), supporting the mechanistic plausibility of the age-dependent predictions. Notably, the pediatric datasets used here predominantly included older children (with mean ages above five years). This likely reduced the observable impact of ontogeny in the current evaluation, as renal OAT3 and BCRP expression approaches mature levels during early childhood27. However, as outlined above, the model is intended for further development toward TDM support, where children under five years are frequently encountered and ontogeny effects are expected to be more pronounced. Future work should therefore prioritize expanded pediatric parameterization using larger individual-level datasets and assess whether inclusion of additional transporter and secretory ontogeny pathways improves predictive accuracy across pediatric age groups.Renal function emerged as the dominant determinant of MTX exposure in the high-dose setting. Consistent with the increased MTX exposure predicted under reduced renal function (Fig. 5), leucovorin rescue is clinically intensified and prolonged in renally impaired patients. This follows from delayed MTX elimination prolonging the period over which folate rescue must be sustained, rather than from any effect of leucovorin on MTX disposition5. Simulations across chronic kidney disease categories illustrated that reductions in GFR can markedly prolong high MTX concentrations after HDMTX, whereas the relative impact is much smaller at low dosing. Model description of urinary excretion of MTX varies in a way that is consistent with MTX pharmacokinetics and reflects several contributing factors: First, in the low-dose setting, intracellular “trapping” of MTX through polyglutamation plays a more prominent role than in high-dose therapy. Polyglutamated MTX must be converted back to the free (non-polyglutamated) form before it can be transported out of the cell, which contributes to slower elimination48. Second, at high doses, MTX “flooding” creates a strong concentration gradient, leading to rapid urinary excretion, influenced further by transporter and possibly AO saturation. This finding aligns with the clinical rationale for intensive monitoring during HDMTX and supports the use of mechanistic simulations to contextualize measured concentrations and explore “what-if” scenarios (e.g., varying degrees of renal impairment) when interpreting TDM results.Compared with previous MTX PBPK implementations in PK-Sim® and other published models, the present model was developed in humans and evaluated against a broad set of concentration–time profiles, with particularly strong coverage of i.v. administrations that are central to HDMTX regimens and TDM-guided clinical decision making35,36,60,61,62. The presented approach enables a mechanistic interpretation of MTX disposition, which allows scaling to children. This work also complements alternative approaches in the literature, including a minimal PBPK model (reduced compartments) focused on HDMTX predictions and population PK models (two- to three-compartment) designed to identify covariates affecting MTX plasma levels and support dosing optimization63,64,65,66,67.Limitations of the presented model include: (i) reliance on aggregated literature profiles rather than large individual-level datasets with population data showing high interpatient variability, as many patients have late-stage cancer or chronic inflammatory disease with liver or kidney dysfunction that may alter transporter and enzyme activity, (ii) reduced model complexity and focus on a few representative transporters, potentially overlooking other relevant processes, (iii) transporter-related variability is represented through physiology and ontogeny, but not through explicit pharmacogenetic stratification that may reduce the model’s ability to capture extreme phenotypes and may contribute to residual unexplained variability and (iv) limited reflection of further application routes in clinical context such as the intrathecal, subcutaneous or intramuscular administration of MTX.Although the present model provides a mechanistic framework for describing typical MTX pharmacokinetics across clinical populations, it does not by itself quantify unexplained interindividual variability. This would require individual-level concentration-time data analysis with population estimation methods that quantify between-subject variance components68. This is distinct from the least-squares fitting of aggregated profiles used here. Such a population model could then be applied a posteriori, via Maximum a Posteriori estimation or related model-informed precision dosing approaches to individualize parameter estimates from measured MTX concentrations for therapeutic drug monitoring69.In conclusion, the presented parent–metabolite PBPK model provides a mechanistic description of MTX and 7-OH-MTX in adults and children and supports the evaluation of renal impairment and clinically relevant DDIs. A full description is provided in the Supplementary Materials, and the model is publicly available.Data availabilityThe PBPK model is comprehensively described in the Supplementary Materials and will be made publicly available for open access (GitHub repository on http://models.clinicalpharmacy.me).ReferencesHider, S. L., Bruce, I. N. & Thomson, W. The pharmacogenetics of methotrexate. Rheumatology 46, 859 (2007).Slørdal, L., Kolmannskog, S., Prytz, P. S., Moe, P. J. & Aarbakke, J. Pharmacokinetics of methotrexate and 7-hydroxy-methotrexate after high-dose (33.6 g/m2) methotrexate therapy. Pediatr. Hematol. Oncol. 3, 127–134 (1986).Article PubMed Google Scholar Maryam, H., Meenal, M. Methotrexate. National Library of Medicine (2024). https://www.ncbi.nlm.nih.gov/books/NBK556114/.Nagamine, A. et al. Target concentration achievement for efficacy and safety of patients with osteosarcoma treated with high-dose methotrexate based on individual pharmacokinetics: a retrospective study. Oncol. Lett. 25, 70 (2022).Article PubMed PubMed Central Google Scholar Ramsey, L. B. et al. Consensus guideline for use of glucarpidase in patients with high-dose methotrexate induced acute kidney injury and delayed methotrexate clearance. Oncologist 23, 52–61 (2018).Article CAS PubMed Google Scholar Roth, M., Obaidat, A. & Hagenbuch, B. OATPs OATs and OCTs: the organic anion and cation transporters of the SLCO and SLC22A gene superfamilies. Br. J. Pharmacol. 165, 1260–1287 (2012).Choughule, K. V., Joswig-Jones, C. A. & Jones, J. P. Interspecies differences in the metabolism of methotrexate: an insight into the active site differences between human and rabbit aldehyde oxidase. Biochem. Pharmacol. 96, 288–295 (2015).Article CAS PubMed PubMed Central Google Scholar Dervieux, T. et al. HPLC determination of thiopurine nucleosides and nucleotides in vivo in lymphoblasts following mercaptopurine therapy. Clin. Chem. 48, 61–68 (2002).Article CAS PubMed Google Scholar Grim, J., Chladek, J. & Martinkovi, J. Pharmacokinetics and pharmacodynamics of methotrexate in non-neoplastic diseases. Clin. Pharmacokinet. 42, 139–151 (2003).Article CAS PubMed Google Scholar Prasad, B. et al. Ontogeny of hepatic drug transporters as quantified by LC-MS/MS Proteomics. Clin. Pharmacol. Ther. 100, 362–370 (2016).Article CAS PubMed PubMed Central Google Scholar Dette, C. M. U. et al. Advancing rare disease therapeutics through digital twins: opportunities in drug development and precision dosing. Comput. Struct. Biotechnol. J. 28, 592–608. https://doi.org/10.1016/j.csbj.2025.11.047 (2025).Article CAS PubMed PubMed Central Google Scholar Britz, H. et al. Physiologically based pharmacokinetic models of probenecid and furosemide to predict transporter mediated drug-drug interactions. Pharm. Res. 37, 250 (2020).Article CAS PubMed PubMed Central Google Scholar European Medicines Agency: EMA/CHMP/458101/2016 - Reporting of physiologically based pharmacokinetic (PBPK) modelling and simulation (2019). https://www.ema.europa.eu/en/reporting-physiologically-based-pharmacokinetic-pbpk-modelling-simulation-scientific-guideline.U.S. Food and Drug Administration: FDA-2016-D-3969 - Physiologically Based Pharmacokinetic Analyses — Format and Content Guidance for Industry (2018). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/physiologically-based-pharmacokinetic-analyses-format-and-content-guidance-industry.Wojtyniak, J., Britz, H., Selzer, D., Schwab, M. & Lehr, T. Data digitizing: accurate and precise data extraction for quantitative systems pharmacology and physiologically‐based pharmacokinetic modeling. CPT Pharmacometr. Syst. Pharmacol. 9, 322–331. https://doi.org/10.1002/psp4.12511 (2020).Article CAS Google Scholar Chemicalize Methotrexate Entry. https://chemicalize.com/app/calculation (2026).Chen et al. Pubchem 2025 update. Nucleic Acids Res. (2025). https://pubchem.ncbi.nlm.nih.gov/compound/5484402#section=Chemical-and-Physical-Properties.Chemicalize 7-Hydroxymethotrexate Entry. https://chemicalize.com/app/calculation (2026).Uhlén, M. et al. Tissue-based map of the human proteome. Science (1979) 347, 8596 (2015).Fisher, R. S., Rock, E. & Malmud, L. S. Effects of meal composition on gallbladder and gastric emptying in man. Dig. Dis. Sci. 32, 1337–1344 (1987).Article CAS PubMed Google Scholar Selzer, D., Rüdesheim, S., Dette, C. M. U., Marok, F. Z. & Lehr, T. Ospgsa: global sensitivity analysis for open systems pharmacology models (2026). https://github.com/Clinical-Pharmacy-Saarland-University/ospgsa.Borgonovo, E. A new uncertainty importance measure. Reliab. Eng. Syst. Saf. 92, 771–784 (2007).Article Google Scholar De Carlo, A., Tosca, E. M., Melillo, N. & Magni, P. A two-stages global sensitivity analysis by using the δ sensitivity index in presence of correlated inputs: application on a tumor growth inhibition model based on the dynamic energy budget theory. J. Pharmacokinet. Pharmacodyn. 50, 395–409 (2023).Article PubMed PubMed Central Google Scholar Cuquerella-Gilabert, M. et al. Leveraging two‐stage δ global sensitivity analysis method to inform parameter estimation in PBPK models. Pharm Stat 25, 859 (2026).Jamaladdin, N. et al. Key pharmacokinetic parameters of 74 pediatric anticancer drugs providing assistance in preclinical studies. Clin. Pharmacol. Ther. 114, 904–913 (2023).Article CAS PubMed Google Scholar Dong, W. C. et al. Simultaneous determination of unbound plasma concentration of methotrexate and 7-hydroxymethotrexate in children patients receiving high-dose methotrexate therapy. Drug Des. Devel Ther. 19, 4383–4396. https://doi.org/10.2147/DDDT.S516431 (2025).Article PubMed PubMed Central Google Scholar Hunt, J. P. et al. Maximum likelihood estimation of renal transporter ontogeny profiles for pediatric PBPK modeling. CPT Pharmacometr. Syst. Pharmacol. 13, 576–588 (2024).Article CAS Google Scholar Open Systems Pharmacology Suite Community. PK-Sim® Ontogeny Database Documentation, Version 7.3 (2024). https://github.com/Open-Systems-Pharmacology/OSPSuite.Documentation/blob/master/PK-Sim%20Ontogeny%20Database%20Version%207.3.pdf.Namour, F. et al. Pharmacokinetics, safety, and tolerability of GLPG0259, a Mitogen-Activated Protein Kinase-Activated Protein Kinase 5 (MAPKAPK5) inhibitor, given as single and multiple doses to healthy male subjects. Drugs R D. 12, 141–163. https://doi.org/10.2165/11633120-000000000-00000 (2012).Article CAS PubMed PubMed Central Google Scholar Stoller, G. R., Samuel, A. J., James, C. D., Robert, J. L. & Bruce, A. C. Pharmacokinetics of high-dose methotrexate (NSC-740). Rep. Div. Cancer Treat. NCI. 2, 285–290 (1976).Google Scholar U.S. Food and Drug Administration: FDA-2010-D-0133 - Pharmacokinetics in Patients with Impaired Renal Function — Study Design, Data Analysis, and Impact on Dosing 8 (2024). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/pharmacokinetics-patients-impaired-renal-function-study-design-data-analysis-and-impact-dosing.Guest, E. J., Aarons, L., Houston, J. B., Rostami-Hodjegan, A. & Galetin, A. Critique of the two-fold measure of prediction success for ratios: application for the assessment of drug-drug interactions. Drug Metab. Dispos. 39, 170–173 (2011).Article CAS PubMed Google Scholar Rajagopalan, P. T. R. et al. Interaction of dihydrofolate reductase with methotrexate: ensemble and single-molecule kinetics. Proc. Natl. Acad. Sci. U S A. 99, 13481 (2002).Article CAS PubMed PubMed Central Google Scholar Jones, H. & Rowland-Yeo, K. Basic concepts in physiologically based pharmacokinetic modeling in drug discovery and development. CPT Pharmacometrics Syst. Pharmacol. 2, 1–12 (2013).Article Google Scholar Schaller, S. et al. Evaluation of BCRP-related DDIs between methotrexate and cyclosporin A using physiologically based pharmacokinetic modelling. Drugs R D. 25, 1–17. https://doi.org/10.1007/s40268-024-00495-1 (2024).Article PubMed PubMed Central Google Scholar Hwang, S. et al. Comprehensive evaluation of OATP- and BCRP‐mediated drug–drug interactions of methotrexate using physiologically‐based pharmacokinetic modeling. Clin. Pharmacol. Ther. 116, 1013–1022 (2024).Article CAS PubMed Google Scholar Aherne, G. W., Piall, E., Marks, V., Mould, G. & White, W. F. Prolongation and enhancement of serum methotrexate concentrations by probenecid. BMJ 1, 1097–1099 (1978).Article CAS PubMed PubMed Central Google Scholar Tracy, T. S. et al. The effects of a salicylate, ibuprofen, and naproxen on the disposition of methotrexate in patients with rheumatoid arthritis. Eur. J. Clin. Pharmacol. 42, 121–125 (1992).Article CAS PubMed Google Scholar Servier Medical Art by Servier is licensed under a Creative Commons Attribution 4.0 Unported License. https://creativecommons.org/licenses/by/4.0/ (2026).Skibińska, L., Ramlau, C., Załuski, J. & Olejniczak, B. Methotrexate binding to human plasma proteins. Pol. J. Pharmacol. Pharm. 42, 151–157 (1990).PubMed Google Scholar Slørdal, L., Sager, G., Jæger, R. & Aarbakke, J. Interactions with the protein binding of 7-hydroxy-methotrexate in human serum in vitro. Biochem. Pharmacol. 37, 607–611 (1988).Article PubMed Google Scholar Yousefi, G., Foroutan, S. M., Zarghi, A. & Shafaati, A. Synthesis and characterization of methotrexate polyethylene glycol esters as a drug delivery system. Chem. Pharm. Bull. (Tokyo). 58, 147–153 (2010).Article CAS PubMed Google Scholar Mir, O. et al. Hyper-alkalinization without hyper-hydration for the prevention of high-dose methotrexate acute nephrotoxicity in patients with osteosarcoma. Cancer Chemother. Pharmacol. 66, 1059–1063 (2010).Article CAS PubMed Google Scholar Chen, Z. S. et al. Transport of methotrexate, methotrexate polyglutamates, and 17beta-estradiol 17-(beta-D-glucuronide) by ABCG2: effects of acquired mutations at R482 on methotrexate transport. Cancer Res. 63, 4048–4054 (2003).CAS PubMed Google Scholar Inoue, K. & Yuasa, H. Molecular basis for pharmacokinetics and pharmacodynamics of methotrexate in rheumatoid arthritis therapy. Drug Metab. Pharmacokinet. 29, 12–19 (2014).Article CAS PubMed Google Scholar Patik, I. et al. Functional expression of the 11 human Organic Anion Transporting Polypeptides in insect cells reveals that sodium fluorescein is a general OATP substrate. Biochem. Pharmacol. 98, 649–658 (2015).Article CAS PubMed PubMed Central Google Scholar Ifergan, I., Jansen, G. & Assaraf, Y. G. The Reduced Folate Carrier (RFC) is cytotoxic to cells under conditions of severe folate deprivation. J. Biol. Chem. 283, 20687–20695. https://doi.org/10.1074/jbc.M802812200 (2008).Article CAS PubMed PubMed Central Google Scholar Bedoui, Y. et al. Methotrexate an old drug with new tricks. Int. J. Mol. Sci. 20, 5023. https://doi.org/10.3390/ijms20205023 (2019).Article CAS PubMed PubMed Central Google Scholar Tanaka, G. & Kawamura, H. Anatomical and physiological characteristics for Asian reference man: Male and female of different ages: Tanaka model. Report Number: NIRS-M-115. In Division of Radioecology, National Institute of Radiological Sciences. Hitachinaka 311 – 12 Japan. (1996).Valentin, J. Basic anatomical and physiological data for use in radiological protection: reference values. Ann. ICRP. 32, 1–277 (2002).Article Google Scholar National Center for Health Statistics. Third National Health and Nutrition Examination Survey, (NHANES III) (1997). https://wwwn.cdc.gov/nchs/nhanes/nhanes3/default.aspx.Ishii, E. et al. Oral mucositis and salivary methotrexate concentration in intermediate-dose methotrexate therapy for children with acute lymphoblastic leukemia. Med. Pediatr. Oncol. 17, 429–432 (1989).Article CAS PubMed Google Scholar Ylinen, E., Jahnukainen, K., Saarinen-Pihkala, U. M. & Jahnukainen, T. Assessment of renal function during high-dose methotrexate treatment in children with acute lymphoblastic leukemia. Pediatr. Blood Cancer. 61, 2199–2202. https://doi.org/10.1002/pbc.25137 (2014).Article CAS PubMed Google Scholar Winograd, B. et al. Renal excretion and pharmacokinetics of methotrexate and 7-hydroxy-methotrexate following a 24-h high dose infusion of methotrexate in children. Eur. J. Clin. Pharmacol. 30, 231–238 (1986).Article CAS PubMed Google Scholar Kearney, P. J., Light, P. A., Preece, A. & Mott, M. G. Unpredictable serum levels after oral methotrexate in children with acute lymphoblastic leukaemia. Cancer Chemother. Pharmacol. 3, 859 (1979).Peters, S. A. & Dolgos, H. Requirements to Establishing Confidence in Physiologically Based Pharmacokinetic (PBPK) models and overcoming some of the challenges to meeting them. Clin. Pharmacokinet. 58, 1355–1371 (2019).Article CAS PubMed PubMed Central Google Scholar Taylor, Z. L. et al. Systematic review of pharmacogenetic factors that influence high-dose methotrexate pharmacokinetics in pediatric malignancies. Cancers (Basel). 13, 2837. https://doi.org/10.3390/cancers13112837 (2021).Article CAS PubMed PubMed Central Google Scholar Hutzler, J. M., Yang, Y. S., Brown, C., Heyward, S. & Moeller, T. Aldehyde oxidase activity in donor-matched fresh and cryopreserved human hepatocytes and assessment of variability in 75 donors. Drug Metab. Dispos. 42, 1090–1097 (2014).Article PubMed Google Scholar Foti, A., Dorendorf, F. & Leimkühler, S. A single nucleotide polymorphism causes enhanced radical oxygen species production by human aldehyde oxidase. PLoS One. 12, e0182061. https://doi.org/10.1371/journal.pone.0182061 (2017).Article CAS PubMed PubMed Central Google Scholar Ogungbenro, K. & Aarons, L. Physiologically based pharmacokinetic modelling of methotrexate and 6-mercaptopurine in adults and children. Part 1: methotrexate The CRESim & Epi-CRESim Project Groups. J. Pharmacokinet. Pharmacodyn. 41, 159–171. https://doi.org/10.1007/s10928-014-9354-4 (2026).Le Merdy, M., Mullin, J. & Lukacova, V. Development of PBPK model for intra-articular injection in human: methotrexate solution and rheumatoid arthritis case study. J. Pharmacokinet. Pharmacodyn. 48, 909–922 (2021).Article PubMed PubMed Central Google Scholar Brandhonneur, N., Noury, F., Bruyère, A., Saint-Jalmes, H. & Le Corre, P. PBPK model of methotrexate in cerebrospinal fluid ventricles using a combined microdialysis and MRI acquisition. Eur. J. Pharm. Biopharm. 104, 117–130 (2016).Article CAS PubMed Google Scholar Batey, M. A. et al. Population pharmacokinetics of adjuvant cyclophosphamide, methotrexate and 5-fluorouracil (CMF). Eur. J. Cancer. 38, 1081–1089 (2002).Article CAS PubMed Google Scholar Campagne, O. et al. Population pharmacokinetics of methotrexate and 7-hydroxymethotrexate and delayed excretion in infants and young children with brain tumors. Eur. J. Pharm. Sci. 193, 106669 (2024).Article CAS PubMed Google Scholar Fukuhara, K., Ikawa, K., Morikawa, N. & Kumagai, K. Population pharmacokinetics of high-dose methotrexate in Japanese adult patients with malignancies: a concurrent analysis of the serum and urine concentration data. J. Clin. Pharm. Ther. 33, 677–684 (2008).Article CAS PubMed Google Scholar Korell, J. et al. A Population pharmacokinetic model for low-dose methotrexate and its polyglutamated metabolites in red blood cells. Clin. Pharmacokinet. 52, 475–485 (2013).Article CAS PubMed Google Scholar Pesenti, G., Foppoli, M. & Manca, D. A minimal physiologically based pharmacokinetic model for high-dose methotrexate. Cancer Chemother. Pharmacol. 88, 595–606. https://doi.org/10.1007/s00280-021-04305-2 (2021).Article CAS PubMed PubMed Central Google Scholar Mould, D. & Upton, R. Basic concepts in population modeling, simulation, and model-based drug development—part 2: introduction to pharmacokinetic modeling methods. CPT Pharmacometrics Syst. Pharmacol. 2, 1–14 (2013).Article Google Scholar Darwich, A. S. Model-informed precision dosing: background, requirements, validation, implementation, and forward trajectory of individualizing drug therapy. Annu. Rev. Pharmacol. Toxicol. 61, 225–245 (2021).Article CAS PubMed Google Scholar Download referencesAcknowledgementsThis study was funded by the European Union as part of the project “Improve Safety in Polymedication by Managing Drug-Drug-Gene Interactions – SafepolyMed”. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the Health and Digital Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.FundingOpen Access funding enabled and organized by Projekt DEAL. Thorsten Lehr was supported by the SafePolyMed project which receives funding from the European Union’s Horizon Europe Research and Innovation Programme under Grant Agreement No. 101057639. Matthias Schwab was in parts supported by the Robert Bosch Stiftung Stuttgart, Germany and funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy – EXC 2180–390900677. Vita Dolžan was in part funded by the Slovenian Research and Innovation Agency (ARIS) (grant P1-0170). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the Health and Digital Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.Author informationAuthors and AffiliationsClinical Pharmacy, Saarland University, Campus C4 3, 66123, Saarbrücken, GermanyCharlotte Maria Ursula Dette, Fatima Zahra Marok, Veronika Alberg, Christiane Dings, Dominik Selzer & Thorsten LehrDr. Margarete Fischer-Bosch Institute of Clinical Pharmacology, 70376, Stuttgart, GermanyMatthias SchwabDepartments of Clinical Pharmacology, Pharmacy and Biochemistry, University of Tübingen, 72076, Tübingen, GermanyMatthias SchwabCluster of Excellence iFIT (EXC2180) “Image-Guided and Functionally Instructed Tumor Therapies”, University of Tübingen, 72076, Tübingen, GermanyMatthias SchwabDepartment of Hematology and Oncology, University Children’s Hospital, University Medical Centre Ljubljana, Ljubljana, SloveniaBarbara Faganel KotnikFaculty of Medicine, University of Ljubljana, Ljubljana, SlowenienBarbara Faganel KotnikPharmacogenetics Laboratory, Institute of Biochemistry and Molecular Genetics, Faculty of Medicine, University of Ljubljana, Ljubljana, SloveniaVita DolžanExperimental and Clinical Pharmacology and Toxicology, Center for Molecular Signaling (PZMS), Saarland University, 66421, Homburg, GermanyDaniela YildizJosé Carreras Center for Immuno- and Gene Therapy and Internal Medicine I, Saarland University Medical School, 66421, Homburg/Saar, GermanyJörg BittenbringDepartment of Pediatric Oncology & Hematology, Saarland University Medical School, 66421, Homburg, GermanyMarc RemkeAuthorsCharlotte Maria Ursula DetteView author publicationsSearch author on:PubMed Google ScholarFatima Zahra MarokView author publicationsSearch author on:PubMed Google ScholarVeronika AlbergView author publicationsSearch author on:PubMed Google ScholarChristiane DingsView author publicationsSearch author on:PubMed Google ScholarDominik SelzerView author publicationsSearch author on:PubMed Google ScholarMatthias SchwabView author publicationsSearch author on:PubMed Google ScholarBarbara Faganel KotnikView author publicationsSearch author on:PubMed Google ScholarVita DolžanView author publicationsSearch author on:PubMed Google ScholarDaniela YildizView author publicationsSearch author on:PubMed Google ScholarJörg BittenbringView author publicationsSearch author on:PubMed Google ScholarMarc RemkeView author publicationsSearch author on:PubMed Google ScholarThorsten LehrView author publicationsSearch author on:PubMed Google ScholarContributionsConceptualization, C.M.U.D, F.Z.M. and T.L.; funding acquisition, T.L.; investigation, C.M.U.D and F.Z.M.; writing-original draft, C.M.U.D; preparing figures and tables, C.M.U.D and F.Z.M.; writing-review and editing, F.Z.M., V.A., C.D., D.S., M.S., B.K., V.D., D.Y., J.B., M.R., T.L. All authors have read and agreed to the published version of the manuscript.Corresponding authorCorrespondence to Thorsten Lehr.Ethics declarationsCompeting interestsThe authors declare no competing interests.Additional informationPublisher’s noteSpringer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary InformationBelow is the link to the electronic supplementary material.Supplementary Material 1 (download PDF )Rights and permissionsOpen Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.Reprints and permissionsAbout this article