Population pharmacokinetics of tacrolimus and CYP3A5-driven variability: implications for model-informed dose individualization in pediatric renal transplant recipients Authors: Lesly Yanira Xajil-Ramos, Jesús Alonso Gándara-Mireles, Mariela Guerra-García, Rodrigo José Vargas Rosales, Andrea Mariela Ruano Toledo, Dulce Sofia Ocaña-Garrido, Ismael Lares-Asseff, Oscar Cóbar, Sindy Méndez Soveranis, Leslie Patrón-Romero, Horacio Almanza-Reyes, Randall Lou-Meda Published: 2026 ABSTRACT AbstractBackgroundPopulation pharmacokinetics and pharmacogenomics are essential tools for understanding interindividual variability in tacrolimus exposure in pediatric transplant recipients. Identifying clinical and genetic factors associated with tacrolimus disposition may support individualized dosing strategies. This study aimed to develop a population pharmacokinetic (PopPK) model of tacrolimus in pediatric renal transplant patients and evaluate the impact of CYP3A5 genotype and relevant clinical covariates on pharmacokinetic parameters.MethodsA total of 165 tacrolimus whole-blood concentrations from 51 pediatric patients receiving immunosuppressive therapy were analyzed at ≥ 6 months following renal transplantation. Population pharmacokinetic analysis was performed using nonlinear mixed-effects modeling implemented in Monolix®, assessing associations between pharmacokinetic parameters and clinical, biochemical, and genetic covariates to explain interindividual variability.ResultsA two-compartment model with first-order absorption adequately described tacrolimus pharmacokinetics in the study population. Estimated population parameters were: Ka = 2.1282 h⁻¹, CL = 0.01633 L/h/1.44 m², V1 = 0.09417 L/1.44 m², V2 = 0.9558 L/1.44 m², and Q = 0.0514 L/h/1.44 m². CYP3A5 genotype was significantly associated with clearance, and estimated glomerular filtration rate (eGFR) influenced both clearance and central volume of distribution. Age was positively associated with the absorption rate constant.ConclusionsThe developed PopPK model provided and adequate characterization of tacrolimus pharmacokinetics in pediatric renal transplant recipients. CYP3A5 genotype, eGFR, and age were identified as relevant covariates associated with interindividual pharmacokinetic variability. These findings provide a basis for future evaluation of model-informed individualized dosing strategies in this population.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s40360-026-01157-5. Introduction Tacrolimus (TAC) is the cornerstone of immunosuppressive therapy worldwide and represents the mainstay of immunosuppressive therapy for preventing graft rejection in pediatric renal transplant recipients. However, its clinical use remains a significant challenge due to its narrow therapeutic window and marked interindividual variability in systemic exposure [1, 2]. TAC pharmacokinetics are influenced by multiple factors, including variable oral bioavailability and hepatic metabolism mediated by cytochrome P450 enzymes, particularly CYP3A5, whose expression is genetically determined [3, 4]. CYP3A5 polymorphisms significantly affect TAC clearance, resulting in substantial variability in dose requirements and risk of underexposure or toxicity [5]. Despite this evidence, current clinical practice continues to rely primarily on standardized dosing and trough concentration monitoring, which may not ensure optimal exposure, especially in pediatric patients. Population pharmacokinetic (PopPK) models have increasingly been used to characterize variability and support individualized dosing strategies in transplant populations. Although these models have improved TAC dose optimization in Asian, European, and North American cohorts [6, 7], their applicability to other populations and clinical contexts remains limited. Most TAC PopPK models have been developed in non-Latin American populations, raising concerns about their applicability to Latin American pediatric patients with different genetic backgrounds. Within Latin America, pharmacokinetic data are scarce and predominantly derived from Mexican populations, which may share some similarities but do not fully capture the genetic and clinical diversity of the region, thereby limiting direct extrapolation [8]. In addition, many existing models are based on data from the early post-transplant period or include a limited number of covariates, often focusing primarily on body size and CYP3A5 genotype, without fully integrating relevant clinical and biochemical factors in clinically stable patients. Furthermore, pediatric patients in the stable phase following transplantation may present different pharmacokinetic profiles compared with the early post-transplant period, yet this stage remains underrepresented in pharmacokinetic modeling studies [9]. Guatemala is characterized by marked ethnic diversity, with a substantial proportion of the population of Mayan ancestry, a group historically underrepresented in pharmacokinetic and pharmacogenomic research. Therefore, there is a need to develop population-specific pharmacokinetic models that integrate both genetic and non-genetic factors in clinically stable pediatric transplant recipients. The lack of PopPK models in pediatric populations with a high proportion of Amerindian ancestry represents a significant knowledge gap. Accordingly, this study aimed to develop a PopPK model of TAC and to evaluate the influence of CYP3A5 genotype and relevant clinical covariates in pediatric renal transplant recipients. Materials and methods Study design A prospective population pharmacokinetic analysis was conducted, including 51 Guatemalan pediatric renal transplant recipients treated with tacrolimus (TAC) as part of their immunosuppressive regimen. Patients were followed at the Fundación para el Niño Enfermo Renal (FUNDANIER), Nephrology, Hypertension, Dialysis, and Transplantation Service of the Department of Pediatrics, Hospital Roosevelt, Guatemala City. Induction immunosuppressive therapy consisted of basiliximab or thymoglobulin according to donor type (living or deceased), in combination with corticosteroids. Post-transplant maintenance immunosuppression followed the institutional protocol and included tacrolimus, mycophenolate mofetil, and low-dose prednisone. All patients were in the stable post-transplant phase (≥ 6 months). TAC was administered orally at 0.2–0.6 mg/kg/day in two divided doses every 12 h and subsequently adjusted based on target plasma concentrations and clinical judgment of the treating team. All patients received the same immediate-release oral TAC formulation (generic capsules). Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice (GCP) guidelines. Given the inclusion of pediatric renal transplant recipients, additional safeguards were implemented to protect this vulnerable population. The study protocol was approved by the Independent Ethics Committee of Hospital Roosevelt, Guatemala City, Guatemala (Approval No. 179–2023). The study was classified as minimal risk, as procedures (blood sampling and genetic analysis) were part of routine clinical follow-up or involved predictable and controlled risks. Written informed consent was obtained from legal guardians, and assent was obtained from patients when appropriate. Participants were informed about study objectives, procedures, potential risks, and their right to withdraw at any time without affecting their medical care. All procedures were performed by trained personnel. CYP3A5 genotyping Genomic DNA was extracted from peripheral blood samples using the QIAamp DNA Blood Mini Kit cat. no. 51,104 (Qiagen), following the manufacturer’s instructions. Genotyping of the CYP3A5*1 and CYP3A5*3 alleles was performed using restriction fragment length polymorphism (RFLP) analysis, as previously described [10]. Polymerase chain reaction (PCR) amplification was carried out in a final reaction volume of 25 µL, containing 12.5 µL of GoTaq® Hot Start Colorless Master Mix (Promega), 0.6 µL of each primer (10 µM; Inqaba Biotechnical Industries, SA), 1.2 µL of genomic DNA (< 250 ng per reaction), and nuclease-free water. Amplified PCR products were digested with the SspI restriction enzyme (Thermo Fisher Scientific®) by incubation at 36 °C for 16 h. Genotypes were determined based on DNA fragment patterns observed after agarose gel electrophoresis: three fragments of 168, 148, and 125 bp for the *1/*3 genotype, two fragments of 168 and 125 bp for the *3/*3 genotype, and two fragments of 148 and 125 bp for the *1/*1 genotype [10]. Determination of tacrolimus whole-blood concentrations Blood samples (3 mL per sample) were collected by peripheral venipuncture and stored in tubes containing EDTA as anticoagulant. For each patient, up to four blood samples were obtained at randomly selected post-dose time points: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, and 12 h after oral administration of TAC. These sampling times were selected based on optimal sampling theory and previously described pharmacokinetic models [11]. A total of 165 blood samples from 51 patients were analyzed. TAC whole-blood concentrations were determined using chemiluminescent microparticle immunoassay (CMIA) on Abbott platforms (ARCHITECT/Alinity i®; Abbott Laboratories, USA), routinely employed for therapeutic drug monitoring in the Endocrinology and Specialties Laboratory of Roosevelt Hospital, where tacrolimus concentrations are measured as part of routine clinical follow-up of patients from the Roosevelt Pediatric Nephrology/FUNDANIER service. Model development Structural Model PopPK analysis was performed using nonlinear mixed-effects modeling in Monolix® software (version 2024R; Lixoft, France), with parameter estimation based on the stochastic approximation expectation-maximization (SAEM) algorithm. A two-compartment model with first-order absorption and elimination was used to describe TAC pharmacokinetics. Interindividual variability (IIV) was described using an exponential model assuming a log-normal distribution of individual parameters, according to the relationship: , where represents the individual parameter, , the typical population value, and the interindividual random effect (mean 0, variance ω2). Interindividual variability (IIV) was evaluated for clearance (CL), central volume of distribution (V1), peripheral volume of distribution (V2), absorption rate constant (Ka), and intercompartmental clearance (Q), and was retained only for parameters in which it significantly improved model fit and parameter stability. Alternative structural models were evaluated, and model selection was based on goodness-of-fit, parameter precision, and information criteria (AIC and BIC). Covariate selection Interindividual variability in TAC pharmacokinetic parameters was explored by assessing relationships between individual parameter estimates from the base model and clinically relevant covariates. Covariates were first screened graphically, and those showing potential associations were evaluated using the Wald test, applying a threshold of p < 0.01 for inclusion and p < 0.001 for elimination. Their impact on model performance was assessed using changes in AIC and BIC. Covariates were incorporated sequentially and retained when they significantly improved model fit and reduced interindividual variability. A complementary sensitivity analysis was performed in which CYP3A5 genotype was recoded according to expressor status (*1 carriers vs. *3/*3) to evaluate the robustness of the pharmacogenetic effect. Continuous covariates evaluated included age, body weight, height, body mass index (BMI), nutritional status, time since transplantation (months), serum creatinine (mg/dL), blood urea nitrogen (mg/dL), serum albumin (g/dL), estimated glomerular filtration rate (eGFR), aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT), and alkaline phosphatase. Estimated glomerular filtration rate (eGFR) was calculated using the Schwartz equation. Categorical covariates included sex and CYP3A5 genotype (CYP3A5*1/*1, CYP3A5*1/*3, and CYP3A5*3/*3). Concomitant immunosuppressive therapy included mycophenolate mofetil and low-dose prednisone according to the institutional post-transplant protocol. All patients were in the stable post-transplant phase (≥ 6 months), during which corticosteroids are maintained at low doses. Given the uniform use of prednisone across patients, its influence was not evaluated as a covariate for drug-drug interaction analysis. In routine clinical practice, transplant patients receive concomitant medications, particularly anti-infective agents, as well as antihypertensive and gastrointestinal therapies. However, their use was heterogeneous and time-varying, and detailed time-aligned information relative to pharmacokinetic sampling was not consistently available. Therefore, these variables were not included as time-dependent covariates in the model. A summary of the most frequently used concomitant medications in the study cohort is provided in Supplementary Table S2. Pharmacokinetic parameter estimation Parameter precision was assessed using Standard Errors (SEs), along with evaluation of parameter stability and physiological plausibility. Model evaluation Model development followed three sequential steps: structural model selection, covariate selection, and elimination of non-significant covariates. The final model was selected based on goodness-of-fit diagnostics, parameter precision, and information criteria (AIC and BIC). Model robustness was further evaluated using bootstrap analysis to assess parameter stability and predictive performance. Visual predictive check (VPC) Predictive performance was evaluated using visual predictive checks (VPCs). Simulated concentration-time profiles were generated from the final model, and the 5th, 50th, and 95th percentiles were compared with observed data. Observed percentiles were overlaid onto the 90% prediction intervals derived from simulations. Agreement between observed and simulated data supported the accuracy, robustness, and validity of the developed PopPK model. Simulation Simulations were performed using the final PopPK model in Monolix® with a Monte Carlo simulation approach to predict individualized TAC dosing required to achieve target trough concentrations (C0) within the therapeutic range (5–7 ng/mL), according to the institutional post-transplant protocol [2, 12]. These simulations accounted for patient-specific clinical and genetic characteristics. Two patients with supratherapeutic TAC concentrations were selected. Individual concentration-time profiles were simulated across a dose range of 1–12 mg at a 12-h dosing interval. Additional simulations evaluated alternative dosing intervals (6, 8, 12 and 24 h). The dose-interval combination achieving target concentrations at steady state was identified for each patient. Results Study population A total of 51 pediatric renal transplant recipients (29 males and 22 females) were included, contributing 165 whole-blood TAC concentrations, with up to four samples obtained per patient. All participants completed the study and were included in the final analysis. Baseline demographic, genetic, and clinical characteristics are summarized in Table 1. The mean BMI was 20.7 ± 4.2 kg/m², and CYP3A5 genotype distribution was consistent with Hardy–Weinberg equilibrium. Table 1Patient baseline characteristicsCharacteristicPatients(n = 51)DemographicsSex (F/M)22/29Age (years)16.47 ± 3.2Weight (kg)45.4 ± 12.9Height (cm)147 ± 11.5BMI (kg/m2)20.7 ± 4.2 Genetic variables CYP3A5 genotype frequency *1/*17 (13.8%) *1/*317 (33.3%) *3/*327 (52.9%)CYP3A5 Alleles frequency *131 (30.4%) *371 (69.6%)CYP3A5 phenotype frequencyExpresser24 (47.1%)Non-expresser27 (52.9%) Clinical and biochemical variables Post-transplant time (months)42.9 (6.14–164.38)Serum creatinine (mg/dL)1.67 ± 0.8Blood urea nitrogen (mg/dL)21.4 ± 10.1Serum albumin (g/dL)4.72 ± 0.4eGFR (mL/min/1.73 m2)64.5 ± 25.2AST (U/L)20.4 ± 8.4ALT (U/L)15.8 ± 15.9GGT (U/L)21.0 ± 9.6Alkaline phosphatase (U/L)217.7 ± 106.9Tacrolimus dose (mg/kg/day)0.11 ± 0.07Tacrolimus trough concentration (C0, ng/mL)5.53 ± 2.08Data are presented as mean standard deviation or median (range) Patient baseline characteristics Data are presented as mean standard deviation or median (range) Structural model A structural model assuming oral administration, without absorption delay, first-order absorption, two-compartment distribution, and linear elimination, with no absorption lag time required in the final model, provided the best description of TAC whole-blood concentrations and showed the lowest AIC and BIC values among the evaluated models. Inclusion of a lag time did not improve model fit or parameter stability. The final model showed an OFV of 888.10, an AIC of 927.03, and a BIC of 962.13. Figure 1 shows observed TAC concentrations (n = 165) as a function of time. Baseline (time 0) concentrations ranged from 1 to 9 ng/mL, whereas maximum observed concentrations (Cmax) reached up to 58 ng/mL in some samples. One- and three-compartment models showed inferior performance, with higher OFV and suboptimal data fitting. Estimated population pharmacokinetic parameters for the selected model are presented in Table 2. In the final pharmacokinetic model, all pharmacokinetic parameters were linearly scaled to BSA, using a mean value of 1.44 m2 for the study population, which improved the characterization of interindividual variability. Spaghetti plot (Concentration [ng/mL] vs. Time [hours]) of 165 observed tacrolimus whole-blood concentrations Table 2Population pharmacokinetic (PopPK) parameters of tacrolimusParameterPopulation estimate (%RSE)Interindividual variability (%CV)Bootstrap median (95% CI) kapop (h− 1) 2.1282 (27.8)21.22.05 (0.82–3.01) Clpop (L/h/1.44m2) 0.01633 (16.3)20.10.0162 (0.011–0.022) V1pop (L/1.44m2) 0.09417 (19.1)33.30.097 (0.052–0.118) Qpop (L/h/1.44m2) 0.0514 (18.8)39.20.054 (0.028–0.091) V2pop (L/1.44m2) 0.9558 (28.4)30.30.88 (0.72–1.85) Residual error model Value Proportional residual error (%) 0.782 (8.82) Additive residual error (ng/mL) 0.912 (12.83) Population pharmacokinetic (PopPK) parameters of tacrolimus Covariate selection The final model retained CYP3A5 genotype, eGFR, and age as significant covariates. As illustrated in Fig. 2, a strong positive association between age and Ka was observed (r = 0.92). The relatively large coefficient for age likely reflects the scale of the covariate, as age was included in its original units without normalization. Therefore, the magnitude of the effect should be interpreted qualitatively rather than quantitatively. This is consistent with exponential covariate modeling, where unscaled continuous covariates may lead to large coefficient estimates without implying disproportionate biological effects. eGFR was retained as a covariate on both CL and V1, reflecting its contribution to model performance. Specifically, inclusion of eGFR resulted in improved model fit, as reflected by lower OFV (888.10 vs. 889.08), AIC (927.03 vs. 933.01), and BIC (962.13 vs. 964.18) compared with the model without this covariate. Although the improvement was moderate, it was consistent across criteria and therefore supported its retention. Individual pharmacokinetic parameters vs. covariates (Age vs. Ka) Although exploratory analysis suggested a weak negative correlation between eGFR and V1, the final model identified a positive association after covariate adjustment, likely reflecting confounding effects between covariates. The relationships between pharmacokinetic parameters and covariates were described using explicit model equations to facilitate interpretation of covariate effects on each parameter. The final model equations were defined as follows: These equations allow direct interpretation of both the magnitude and direction of covariate effects on pharmacokinetic parameters. Table 3 summarizes the results of the covariate analysis on population pharmacokinetic parameters. The exploratory correlation coefficient (r) represents the simple correlation between each continuous covariate and the individually estimated pharmacokinetic parameter, allowing identification of potential initial relationships. The model-based effects (β, 95% CI) correspond to estimates obtained from the final SAEM-based model fitting and are expressed on a logarithmic scale for CL, V1, and Ka. Table 3Covariates associated with tacrolimus pharmacokinetic parameters in the final modelAssociated variablePK parameterR (exploratory)Model effect (β, IC 95%)InterpretationCYP3A5 (genotype)CL- (categorical)Reference *1/*1 → CL = 0.021 L/h/1.44 m²; *1/*3: β = +0.80 → ×2.23 vs. *1/*1; *3/*3: β = −0.91 → ×0.40 vs. *1/*1Expressers (*1 carriers) ↑CL; Non-expressers (*3/*3) ↓CLeGFRCL+ 0.058β = +0.0445 (IC 95% +0.0089 to + 0.0818) → S. Higher eGFR → modestly higher CLeGFRV1−0.19β = +0.0666 (IC 95% +0.006 to+ 0.290) → S. Higher eGFR → higher V1 after covariate adjustmentAgeKa+ 0.92β = +3.12 (IC 95% +0.43, + 5.81) → S.Older age → higher Ka (faster absorption) Covariates associated with tacrolimus pharmacokinetic parameters in the final model β = +0.0666 (IC 95% +0.006 to + 0.290) → S. The reference value of CL for the CYP3A5*1/*1 genotype reflects the covariate-adjusted parameter estimate derived from the final model, and therefore differs from the population typical value (CLpop). Among the main findings, CYP3A5 genotype exerted a relevant influence on CL. Compared with the CYP3A5*1/*1 genotype (0.021 L/h/1.44 m²), carriers of the CYP3A5*1/*3 genotype exhibited an approximately 2.23-fold increase in clearance, whereas individuals with the CYP3A5*3/*3 genotype showed an approximately 60% reduction in CL (factor × 0.40) (Table 3). As a complementary sensitivity analysis, CYP3A5 genotype was recoded according to expressor status, grouping carriers of at least one functional allele (*1/*1 and *1/*3) as expressers (n = 24) and CYP3A5*3/*3 individuals as non-expressers (n = 27), given the relatively small size of the *1/*1 subgroup (Table 4). Under this alternative classification, expressers exhibited higher TAC clearance than non-expressers. The estimated effect of expressor status on clearance was β = +0.62 (95% CI 0.18 to 1.05; p = 0.006), corresponding to an approximately 1.86-fold increase in clearance. Model performance remained stable in the complementary analysis, with OFV, AIC, and BIC values of 888.56, 926.49, and 960.88, respectively, compared with the original model (OFV = 888.10; AIC = 927.03; BIC = 962.13). These findings support the biological consistency of the pharmacogenetic effect and suggest that the apparent discrepancy between CYP3A5*1/*1 and *1/*3 in the primary model may reflect subgroup variability rather than a true biological difference. Table 4Complementary sensitivity analysis according to CYP3A5 expressor statusCovariateParameterEstimate (β)95% CIexp(β)InterpretationCYP3A5 expressor statusCL+ 0.620.18 to 1.051.86Expressers show higher clearanceeGFRCL+ 0.040.01 to 0.081.04Modest positive effecteGFRV1+ 0.070.02 to 0.111.07Positive effect on V1AgeKa+ 3.120.43 to 5.81-Faster absorption with age Complementary sensitivity analysis according to CYP3A5 expressor status Pharmacokinetic parameter estimation Population pharmacokinetic parameters estimates are presented in Table 2. Parameter precision was acceptable for all parameters, as indicated by low relative standard error (%RSE) values. Model evaluation Observed concentrations showed good agreement with both individual and population predictions, with no evidence of significant structural bias (Fig. 3). These results support the ability of the model to adequately describe both the central tendency and the variability of TAC whole-blood concentrations. Goodness-of-fit plots for tacrolimus pharmacokinetics. (A) Observed concentrations versus individual predicted concentration. (B) Observed concentrations versus population predicted concentration Visual predictive check The VPC (Fig. 4) showed acceptable agreement between simulated and observed concentration percentiles, supporting the predictive performance of the final model. Visual predictive check (VPC) Bootstrap analysis Relative standard error (RSE) values were acceptable for all parameters in the final PopPK model. To further evaluate model stability and robustness, a bootstrap analysis with 200 replicates was performed. Good agreement was observed between the final model parameter estimates and the distributions obtained from the bootstrap analysis, supporting the internal stability and consistency of the developed population pharmacokinetic model (Table 2). Simulation Simulation-based dose adjustment was performed for two patients (P1 and P2) who exhibited supratherapeutic TAC concentrations under the regimen of 12 mg every 12 h (Fig. 5A). P1 had a body weight of 45.1 kg and P2 weighed 41.3 kg. In both cases, the dose corresponded to approximately 0.5 mg/kg/day. For P1, using the individual pharmacokinetic parameters obtained from the population pharmacokinetic analysis (Ka = 2.65 h-1, CL = 0.037 L/h/1.44 m2, V1 = 0.096 L/1.44 m2, Q = 0.069 L/h/1.44 m2, and V2 = 1.62 L/1.44 m2), simulations were conducted at different dosing intervals. The simulated profiles showed that a dose of 9 mg every 12 h achieved steady-state whole-blood concentrations within the therapeutic range, with a clinically relevant reduction in peak concentrations and overall exposure compared with the original regimen (Fig. 5B). For P2, whose individual pharmacokinetic parameters were Ka = 1.69 h-1, CL = 0.029 L/h/1.44 m2, V1 = 0.054 L/1.44 m2, Q = 0.032 L/h/1.44 m2, and V2 = 0.68 L/1.44 m2, dose simulations indicated that a reduced dose of 4 mg every 12 h produced concentration-time profiles within the target therapeutic range while markedly decreasing systemic exposure relative to the initial dose (Fig. 5C). In both patients, dose individualization based on pharmacokinetic simulations was predicted to normalize trough concentrations and reduce overall exposure. Simulations enabled evaluation of concentration-time profiles under different dosing scenarios, integrating patient-specific covariates and allowing assessment of peak exposure and steady-state attainment. Tacrolimus trough concentrations and simulation-based dose adjustment in two representative patients. (A) Observed tacrolimus trough concentrations (C0) in the study population, highlighting patients P1 and P2 with supratherapeutic levels under the standard regimen. (B) Simulated concentration-time profiles of tacrolimus at a dose of 9 mg under different dosing intervals in patient 1 (P1) by Simulx (Monolix Suite®). (C) Simulated concentration-time profiles of tacrolimus at a dose of 4 mg under different dosing intervals in patient 2 (P2) by Simulx (Monolix Suite®) Discussion TAC PopPK has been widely described across diverse populations, most commonly using one- or two-compartment models with first-order absorption in pediatric transplant recipients [6, 13]. The two-compartment model identified in this study is consistent with previous reports from Latin American, Asian, and European cohorts [8, 14–17]. Compared with studies that include a limited number of covariates [18], the present model integrates both genetic and clinical factors, supporting its applicability to this specific population. Regarding model performance, the model demonstrated good stability, supported by agreement between final estimates and bootstrap results. The relatively high interindividual variability observed for these parameters is expected in transplant populations, given the substantial physiological and clinical heterogeneity characteristic of these patients [13, 19, 20]. Covariate analysis identified CYP3A5 genotype, eGFR, and age as significant predictors (Table 3). CYP3A5 expressers (CYP3A5*1/*1 and CYP3A5*1/*3) showed increased CL, consistent with previous studies, supporting its role as a key determinant of TAC disposition and dose requirements. The magnitude of this effect was consistent across pediatric populations and geographic regions, including Latin American, European, and Asian cohorts, where CYP3A5 expression has been systematically associated with increased TAC clearance and higher dose requirements [14, 16, 21]. Consistently, in our study, individuals carrying the functional CYP3A5*1 allele (expressers) exhibited a significant increase in clearance (β = +0.80), corresponding to approximately 2.2-fold lower plasma exposure compared with non-expressers (CYP3A5*3/*3). This finding is consistent with previous evidence demonstrating enhanced metabolic capacity in CYP3A5 expressers and the consequent need for individualized dose adjustments [8, 14, 22]. However, the observation that the CYP3A5*1/*3 subgroup exhibited higher clearance than CYP3A5*1/*1 was not the biologically expected pattern. To address this, a complementary sensitivity analysis was performed grouping CYP3A5*1/*1 and *1/*3 individuals as expressers. Under this classification, the expected biological pattern was observed, with higher clearance in expressers compared with non-expressers. This supports the interpretation that the apparent higher clearance in the CYP3A5*1/*3 subgroup compared with *1/*1 in the primary model may be due to subgroup variability rather than a true biological effect. In Mexican pediatric cohorts, a stronger effect on CL has been reported, with clearance values approximately 2- and 1.5-fold higher in carriers of the CYP3A5*1/*1 and CYP3A5*1/*3 genotypes, respectively, compared with non-expressers, highlighting interpopulation variability [8]. Nevertheless, CYP3A5 genotype remains a consistent determinant of TAC disposition and clearance across studies, including Latin American and global cohorts [23]. This supports its central role in dose individualization across populations [24–26]. This pattern has also been observed in stable North American renal transplant recipients, where a two-compartment model with first-order absorption and a lag time best described TAC pharmacokinetics. In that study, higher clearance values were observed in rapid and intermediate metabolizers compared with slow metabolizers, and model-based simulations showed that TAC dose requirements are largely genotype-dependent, with dose increases of approximately 1.5- and 2-fold in intermediate and rapid metabolizers, respectively [27]. Furthermore, estimated glomerular filtration rate (eGFR) showed a positive association with the central volume of distribution (V1) (β = +0.0666; 95% CI: 0.006 to 0.290). Although the exploratory correlation suggested a weak negative relationship, the adjusted model identified a positive independent effect, likely reflecting confounding by other covariates included in the model, suggesting that higher eGFR may be associated with a wider distribution of the drug within the central compartment. Although the effect of eGFR on clearance (CL) was of smaller magnitude (β = +0.0445), it remained statistically significant in the final model. Although the estimated effect of eGFR was modest, its inclusion consistently improved model fit, as reflected by reductions in OFV, AIC, and BIC. Therefore, according to the predefined covariate selection strategy, eGFR was retained in the final model. This covariate should be interpreted as a contributor to overall model performance rather than as a major clinical determinant of TAC pharmacokinetics. This effect likely reflects indirect physiological mechanisms rather than a direct mechanistic role in TAC elimination, which is primarily hepatic [28]. In this context, eGFR may act as a surrogate marker of overall clinical status, encompassing factors such as fluid balance, tissue perfusion, body composition, and systemic inflammation, which may influence TAC distribution. Consequently, this covariate may contribute to the model´s descriptive performance by capturing interindividual variability related to complex physiological processes not directly measured in this study. However, the effect of eGFR on V1 should be interpreted with caution, as it may also reflect residual variability not explained by other covariates included in the model [28, 29]. In addition, age was positively associated with the absorption rate constant (β = +3.12; 95% CI: +0.43 to + 5.81), suggesting that older patients exhibit faster drug absorption. This finding is physiologically plausible, as maturation of gastrointestinal function with age enhances drug dissolution, intestinal permeability, and splanchnic blood flow, leading to increased absorption rates. These observations are consistent with established pharmacokinetic principles describing age-related physiological maturation [3, 30]. Additionally, previous studies have shown that corticosteroids can induce CYP3A-mediated metabolism of TAC; however, this effect is generally modest and highly dependent on dose and CYP3A genotype. In particular, steroid-related increases in TAC clearance have been reported to range between approximately 3% and 12%, suggesting a limited clinical impact when considered in isolation. Moreover, in stable post-transplant patients, where steroid doses are reduced and exposure is more homogeneous, their influence on TAC pharmacokinetics appears to be attenuated, with other factors such as CYP3A5 genotype playing a more dominant role [27, 31]. When compared with other populations, the Guatemalan cohort showed both similarities and relevant quantitative differences. In Latin American studies, particularly in Mexican pediatric patients, a two-compartment structure was also observed; however, higher clearance values were reported, although the relationship between CYP3A5 genotype and CL remained consistent across populations [8]. In contrast, in Asian and European cohorts, higher clearance and distribution volumes have generally been described, along with differences in model structure and absorption parameters [15, 17, 32]. Additionally, North American models report variability in Ka and structural assumptions [33]. Overall, although the qualitative behavior of TAC is consistent across populations, the Guatemalan cohort exhibited lower clearance and distribution parameters, supporting the presence of a population-specific pharmacokinetic profile. These differences are particularly relevant given TAC´s narrow therapeutic window. Reduced CL or Q may increase systemic exposure, reflected by higher area under the concentration–time curve (AUC), raising the risk of toxicity [34, 35], whereas elevated CL, particularly in CYP3A5 expressers, may result in systemic underexposure and increased risk of rejection [36, 37]. In Guatemala, where a high proportion of individuals of Mayan ancestry remains underrepresented in pharmacokinetic and pharmacogenomic research, these findings underscore the clinical importance of accounting for population-specific pharmacokinetic variability to optimize efficacy and safety. In this context, although TAC exhibits approximately linear pharmacokinetics within the therapeutic range, simulation-based approaches provide additional clinical value beyond simple proportional dose adjustments [29]. In pediatric transplant recipients, drug exposure is influenced by multiple interacting factors, including interindividual variability, CYP3A5 genotype, renal function, and age, that are not easily captured by linear calculations alone. Simulations allow the integration of these covariates within a unified framework and enable prediction of the full concentration-time profile under different dosing scenarios. This facilitates assessment not only of trough concentrations (C0), but also of peak exposure, overall exposure, and the time to reach steady state, which are clinically relevant for balancing efficacy and toxicity. In the present study, simulation–based dose adjustment illustrated the clinical applicability of the developed PopPK model by illustrating its potential to support prediction of individualized dosing regimens aimed at reducing systemic overexposure and achieving target trough concentrations. Strengths of the study This study represents the first PopPK analysis of TAC in pediatric renal transplant recipients in Guatemala and one of the few conducted in Latin America. A key strength is the integration of genetic, clinical, and biochemical covariates, allowing a comprehensive characterization of interindividual variability beyond models limited to body size or genotype alone. The prospective design and optimized sampling strategy provide informative pharmacokinetic data despite the constraints of pediatric studies. Pharmacometric methods, including SAEM estimation, model selection based on AIC/BIC, and bootstrap evaluation, supported the internal stability and consistency of the developed model. The identification of CYP3A5 genotype, eGFR, and age as relevant predictors highlights the clinical applicability of the model for individualized dosing. Additionally, the study provided population-specific evidence in a historically underrepresented group, supporting the development of precision dosing strategies adapted to the regional context. Limitations of the study The relatively small sample size may limit generalizability and the evaluation of less frequent covariates or interactions. In addition, the cohort included only clinically stable patients (≥ 6 months post-transplant), which may limit extrapolation to the early post-transplant period or unstable clinical conditions. The evaluation of concomitant medications was limited by the heterogeneous and time-varying nature of these data in routine clinical practice, precluding their reliable inclusion as time-dependent covariates. Similarly, genetic analysis was restricted to CYP3A5 variants, and other potentially relevant polymorphisms were not assessed. Finally, model validation was limited to internal methods (bootstrap and VPC) without external validation in an independent cohort. Finally, model validation was limited to internal methods without external validation in an independent cohort. Therefore, the findings should be interpreted as a preliminary and require confirmation in larger independent populations before broader clinical implementation. Future research Future research should focus on the prospective and multicenter validation of the developed PopPK model in larger and more heterogeneous pediatric cohorts to confirm its reproducibility, as well as to evaluate its performance across different post-transplant phases. Further research should also explore the implementation of model-informed precision dosing (MIPD) strategies that integrate pharmacokinetic and pharmacogenomic information into clinical practice. These approaches should be evaluated not only in terms of pharmacokinetic performance but also regarding clinically relevant outcomes, including rejection, toxicity, and graft survival [29]. Expanding genetic analyses to include additional polymorphisms involved in TAC metabolism and transport (e.g., CYP3A4, ABCB1) may help explain residual variability. In parallel, improved characterization of drug-drug interactions and the use of optimized sampling strategies, particularly during the absorption phase, could further refine model performance. Finally, external validation in independent cohorts and assessment of the feasibility of implementing individualized dosing strategies in routine clinical settings, particularly in resource-limited environments, will be essential to support clinical translation. Conclusions In this study, a PopPK model of TAC was developed in Guatemalan pediatric renal transplant recipients, showing adequate stability and predictive performance. A two-compartment model with first-order absorption and elimination provided an appropriate description of TAC pharmacokinetics. CYP3A5 genotype, eGFR, and age were identified as relevant covariates associated with interindividual pharmacokinetic variability, supporting the relevance of integrating genetic and clinical factors into dosing strategies. The results suggest the presence of population-specific pharmacokinetic characteristics, highlighting the limitations of directly extrapolating dosing approaches from other populations. These findings support the potential application of model-informed dosing approaches in this population and provide a basis for future validation and clinical implementation studies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors thank the National Secretariat of Science and Technology (SENACYT) of Guatemala for the financial support provided through the project Findecyt/Gestiona I+D 01-2024. The authors thank the EducaCTi funding program of the National Secretariat of Science and Technology (SENACYT) of Guatemala for covering the publication costs of this article (contract number: FINDECYT/EducaCTi 05-2024). The authors also express their sincere gratitude to the Laboratory of Endocrinology and Specialties of Roosevelt Hospital, particularly Gretel Andrea Lemus Arias and Pablo Alfonso Tzorín Velásquez, for their valuable contribution to the processing and analysis of blood samples for tacrolimus concentration determination, which was essential for the successful development of this research. Author contributions Conception: Lesly Yanira Xajil Ramos. Interpretation or analysis of data: Lesly Yanira Xajil Ramos, Jesús Alonso Gándara Mireles, Andrea Mariela Ruano Toledo, Dulce Sofía Ocaña Garrido, Mariela Guerra-García, Sindy Méndez Soberanis, Rodrigo José Vargas Rosales. Preparation of the manuscript: Lesly Yanira Xajil Ramos, Jesús Alonso Gándara Mireles, Mariela Guerra-García, Rodrigo José Vargas Rosales, Leslie Patrón-Romero, Horacio Almanza-Reyes. Revision for important intellectual content: Randall Lou Meda, Jesús Alonso Gándara Mireles, Oscar Cóbar Pinto, Ismael Lares-Asseff. Supervision: Randall Lou-Meda, Jesús Alonso Gándara Mireles, Ismael Lares-Asseff. Funding This research was carried out with co-funding support from the National Secretariat of Science and Technology (SENACYT) of Guatemala, project Findecyt/Gestiona I + D 01-2024. The publication of this article was funded by the EducaCTi funding program of the National Secretariat of Science and Technology (SENACYT) of Guatemala, contract number: FINDECYT/EducaCTi 05-2024. Data availability The data supporting this study´s findings are available on request from the corresponding author. Code availability The pharmacokinetic model code (Monolix) and analysis scripts are available from the corresponding author upon reasonable request. Declarations Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Lesly Yanira Xajil-Ramos, Email: leslyanxr@profesor.usac.edu.gt. Jesús Alonso Gándara-Mireles, Email: alonso_930@hotmail.com. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data supporting this study´s findings are available on request from the corresponding author. The pharmacokinetic model code (Monolix) and analysis scripts are available from the corresponding author upon reasonable request.