Predictive Capacity of Population Pharmacokinetic Models for the Tacrolimus Dose Requirements of Pediatric Solid Organ Transplant Recipients

Abstract

Background:

Transplant recipients require individualized tacrolimus doses to maximize graft survival. Multiple pediatric tacrolimus population pharmacokinetic (PopPK) models incorporating CYP3A5 genotype and other covariates have been developed. Identifying the optimal popPK model is necessary for clinical implementation in pediatric solid organ transplant. The primary objective was to compare the dose prediction capabilities of the developed models in pediatric kidney and heart transplant recipients

Methods:

Pediatric kidney or heart transplant recipients treated with tacrolimus and available CYP3A5 genotype data were identified. The initial weight-based tacrolimus dose and first therapeutic tacrolimus dose were collected retrospectively. Three published popPK models (Zhao et al., Chen et al., and Andrews, et al.) were used to predict the tacrolimus dose required to achieve a tacrolimus trough concentration of 10 ng/mL. Model dose predictions were compared with the initial and first therapeutic doses using Friedman’s test. The first therapeutic dose was plotted against the model-predicted dose.

Results:

The median initial dose approximately 2-fold lower than the first therapeutic dose for CYP3A5 expressers. The Chen et al. model provided the closest estimates to the first therapeutic dose for kidney transplant recipients; however, all three models tended to under-predict the observed therapeutic dose. For heart transplantation, Andrews et al. model predicted doses that were higher doses than the initial dose but the similar to the actual therapeutic dose.

Conclusions:

Weight-based tacrolimus dosing appears to underestimate the tacrolimus dose requirements. The development of a separate popPK model is necessary for heart transplantation recipients. A genotype-guided strategy based on the Chen et al. model provided the best estimates for doses in kidney transplantation and should be prospectively evaluated.

Introduction

Tacrolimus is the most frequently prescribed immunosuppressant (> 90 %) for pediatric kidney and heart transplant recipients.^1,2 Despite its extensive use, clinical dosing protocols for tacrolimus are suboptimal in rapidly achieving target tacrolimus exposure.³ High interindividual variability in tacrolimus exposure, most frequently assessed via trough concentrations, is associated with both treatment efficacy and treatment toxicity. Low exposure can increase the risk of graft rejection, while overexposure is associated with toxicity, such as acute kidney injury and failure.⁴ Most frequently, tacrolimus administration is initiated using weight-based dosing, which is then adjusted incrementally until the target trough exposure is achieved. Additional patient characteristics beyond weight can significantly impact tacrolimus exposure for all oral tacrolimus formulations, however, initial tacrolimus dosing recommendations are still not routinely personalized.^5–7

The patient characteristics that have been shown to contribute to tacrolimus exposure in pediatric kidney and heart transplant recipients include CYP3A5 genotype, hematocrit (Hct) concentration, patient age, and concomitant CYP3A inhibitor use. The CYP3A5 genotype determines the expression of the CYP3A5 enzyme; CYP3A5 enzyme expression is associated with increased tacrolimus clearance, resulting in lower tacrolimus exposure and ~50% higher oral tacrolimus dose requirements.^8–13 Hct concentrations are inversely associated with tacrolimus concentrations as tacrolimus is extensively distributed into erythrocytes.^14,15 Younger age has been associated with increased drug clearance, which might be due to the natural ontogeny of CYP3A enzymes or other processes, although age classifications differ significantly among studies in pediatrics.^8–11,13,16 Additionally, concomitant medication interactions, particularly with CYP3A inhibitors, such as azole antifungals, have been demonstrated to increase tacrolimus exposure, which would indicate lower dose requirements.^13,17,18 Despite multiple studies showing connections between several patient characteristics and tacrolimus dosing requirements, the best approach to incorporate these clinical characteristics into tacrolimus dosing decisions remains unknown, and weight-based dosing remains the current standard of care.^5–7

Methods to better predict personalized tacrolimus dose requirements can decrease the time to achieve therapeutic concentrations and decrease variability in tacrolimus exposure.^19,20 Although prior investigators have developed population pharmacokinetic (popPK) models to predict personalized tacrolimus dosing for pediatric kidney and heart transplant recipients, the models do not include the same patient characteristics, and the clinical applicability of these models remains unclear.^{15,16,18,19,21,22} A popPK model including the CYP3A5 genotype has not been developed for pediatric heart transplant recipients, despite the development of three popPK models including the CYP3A5 genotype for kidney transplant recipients (Table 1). Andrews et al. developed a popPK model for pediatric kidney transplant recipients, which was evaluated prospectively in a clinical trial. However, the trial was discontinued early as the model did not accurately predict initial therapeutic doses. The investigators subsequently modified the PopPK model with additional prospective study data.^21,22 The revised model included weight and CYP3A5 genotype as covariates affecting tacrolimus clearance. Zhao et al. also published a pharmacogenetic popPK model for pediatric kidney transplant that includes weight, CYP3A5 genotype, and Hct.¹⁵ Finally, Chen et al. published a popPK model in pediatric kidney transplant that includes weight, CYP3A5 genotype, postoperative day (POD), and concomitant administration of Wuzhi capsule, a CYP3A inhibitor.¹⁸

Table 1.

Population pharmacokinetic model equations for tacrolimus clearance

Model	Parameters	CL/F equation	Model parameters
Andrews et al.22	Weight CYP3A5 genotype	CL/F = 34.5× (weight/70)0.56 × 1.46 (if CYP3A5 expresser)	Two compartment model
Zhao et al.15	Weight CYP3A5 genotype Hematocrit	CL/F = 13.9 × (weight/70)0.75 × 2.26 (if CYP3A5 expresser) + 7.11 × 1.74 (if Hematocrit <33%)	Two compartment model with first order absorption and lag time
Chen et al.18	Weight CYP3A5 genotype Post-operative day (POD) Wuzhi (WZ) capsule	CL/F = 27.2 × (weight/70)0.75 × (POD/46)−0.201 × (1+(θ×−0.28)) × 1.6 (if CYP3A5 expresser) θ= 1 if WZ co-prescribed; θ=0 if not co-prescribed	One-compartment model with first order elimination

None of these popPK dose prediction models have been validated in an external population. Further, prior comparisons to elucidate how the addition of new covariates improves model performance have not been performed. The purpose of this study was to assess the predictive capability of the aforementioned popPK models for determining tacrolimus doses and corresponding tacrolimus trough concentrations in an external cohort. If these models are determined to be predictive of tacrolimus dose requirements, the clinical application of these dosing strategies could improve the initial tacrolimus dose selection, decreasing the risk of non-therapeutic tacrolimus concentrations in this patient cohort.

Materials and Methods

Data Collection

Patients were eligible for inclusion in this retrospective study if they: 1) received a kidney or heart transplant at Michigan Medicine between June 1, 2014, and December 31, 2019, 2) received tacrolimus as part of their initial immunosuppression regimen, and 3) were <18 years old at the time of transplant. Clinical data, including tacrolimus dosing and trough concentration data, as well as patient demographics relevant to the respective popPK models (Hct, POD, and weight at transplant), were extracted from medical records or corresponding research systems (DataDirect, EMERSE).²³ Information regarding the concomitant administration of a CYP3A inhibitor (e.g., azole antifungals) at the time of tacrolimus administration was also collected. Additional information on post-transplant immunosuppression is published elsewhere.³ Briefly, both kidney and heart transplant recipients were administered immediate-release tacrolimus according to a weight-based dosing protocol (kidney: 0.1 mg/kg/dose BID or TID, heart: 0.05 mg/kg/dose BID) combined with mycophenolate or azathioprine. Kidney transplant recipients received either standard or steroid-avoidant immunosuppressive treatment, while all heart transplant recipients received standard steroid-based immunosuppressive treatment. The initial target trough concentrations were 10–12 ng/mL for kidney transplant recipients and 10–15 ng/mL for heart transplant recipients. This study was approved by the local institutional review board.

Genotyping

Secondary use samples of extracted DNA were obtained from the University of Michigan Histocompatibility Laboratory for patients who met the inclusion criteria. Genotyping was performed for the following single nucleotide polymorphisms (SNP): CYP3A5*3 (rs776746), CYP3A5*6 (rs10264272), and CYP3A5*7 (rs41303343) (Thermo Fisher Scientific, Waltham, MA), as previously described.²⁴ Based on CYP3A5 genotypes, patients were grouped into CYP3A5 expressers (those with 1 or zero variant alleles) and CYP3A5 non-expressers (those with two variant alleles).²⁵

Pharmacokinetic estimates

Tacrolimus dose and trough concentrations, timing of dose administration, and timing of trough measurements were obtained for each patient within the first year post-transplant. The patient’s initial tacrolimus trough, tacrolimus daily dose at the time of the initial trough, POD of the initial trough, and concomitant interacting medications were extracted for initial dose comparisons. As all patients in the Zhao et al. model were ≥ 2 years old and this model excluded medications known to interact with tacrolimus, patients who were < 2 years old were excluded from subsequent analysis. Furthermore, patients receiving concomitant CYP3A inhibitors at the time of tacrolimus administration corresponding to the initial tacrolimus trough concentration were also excluded from subsequent analysis. Tacrolimus CL/F estimates were calculated for each patient using each CL/F equation from each popPK model (Table 1). For the Chen et al. and Zhao et al. models, POD and Hct of the initial tacrolimus trough concentration were respectively used to determine the CL/F estimate for each patient. The initial tacrolimus total daily dose (mg) estimated to achieve a trough concentration of 10 ng/mL was predicted for each patient using the respective popPK-predicted CL/F estimates via the equation: Daily dose (mg/kg) = ((185 ng *h/mL*CL/F)/1000),²² based on the Andrews et al. model, resulting in a tacrolimus AUC_0–12 of 185 ng*h/mL that corresponds to a tacrolimus trough of 10 ng/mL. This exposure to dose translation is based on the assumption that the calculated daily dose is divided and administered twice daily.²²

The first steady-state therapeutic tacrolimus trough concentration, defined as the second consecutive tacrolimus trough within ± 1 ng/mL of the target range, with stable tacrolimus dosing (no change to the prior four doses) was identified for each patient, and the total daily tacrolimus dose, trough concentration, POD, and concomitant interacting medications were extracted for this time point for first therapeutic dose comparisons. The estimated CL/F was recalculated for the Chen et al. model using the POD of the first observed steady-state therapeutic trough concentration.

Analysis

The initial daily tacrolimus dose predictions for the three popPK models were compared with the actual initial dose from the weight-based protocol for kidney and heart transplant recipients. The popPK model-predicted doses were also compared with the observed first therapeutic dose for kidney and heart transplant recipients. The Friedman test was used to compare the total daily initial dose, total daily first therapeutic dose, and total daily model-predicted doses. For the significant findings from the Friedman test, post-hoc comparisons among each group were planned using the Wilcoxon test with Bonferroni correction. The daily tacrolimus dose was compared between CYP3A5 expressers and non-expressers within each dose group using the Mann–Whitney U test. The median number of days required to achieve the first steady-state therapeutic trough was calculated for kidney and heart transplant recipients. The time to achieve the first therapeutic trough was compared between CYP3A5 expressers and non-expressers using the Mann-Whitney U test. All data analyses were performed using SPSS version 27 (IBM, Armonk, NY, USA). Figures were generated using Stata Version 17 (StataCorp, College Station, TX, USA) and included linear fits of the observed therapeutic dose to the individual model predicted dose by transplant type.

Results

One hundred and twelve patients met the initial study inclusion criteria: 61 kidney transplant recipients and 50 heart transplant recipients. Eight kidney transplant recipients and 12 heart transplant recipients were < 2 years at the time of transplant and were excluded from additional analysis. Accordingly, the final cohort included 53 kidney transplant recipients and 38 heart transplant recipients for initial dose prediction comparisons. Twelve kidney transplant recipients and one heart transplant recipient did not have a therapeutic steady-state trough concentration or were receiving a concomitant CYP3A inhibitor at the time of the first steady-state therapeutic trough. As a result, these patients were excluded from the therapeutic dose prediction analysis. Four patients received tacrolimus TID dosing at the time of the first therapeutic dose, and the total daily dose received was used for comparison. The baseline characteristics of the included patients are shown in Table 2.

Table 2.

Patient Demographics

	Kidney (N=53)	Heart (N=38)
Age (mean ± S.D)	11.1 ± 4.7	12.2 ± 4.1
Age ≥ 6 years	43 (81.1%)	23 (60.5%)
Male sex	31 (58.4%)	22 (57.9%)
Hct (mean ± S.D)	27.6 ± 5.0	31.1 ± 4.5
Hct ≥ 33%	9 (16.9%)	10 (26.3%)
Weight (kg) (mean +/− S.D)	38.6 +/− 20.6	46.2 +/− 21.3 kg
Initial tacrolimus daily dose (mg/kg) (mean +/− SD)	0.20 +/− 0.04	0.08 +/− 0.02
Days to initial therapeutic tacrolimus trough (median, 5th–95th percentile)	10 (3.1–60)	12 (2.9–65)
Race
White	41 (77.4%)	29 (76.3%)
African American	5 (9.4%)	5 (13.2%)
Other/Unknown	7 (13.2%)	4 (10.5%)
CYP3A5 Phenotype
Expresser (CYP3A5*1 carrier)	12 (22.6%)	6 (15.8%)
Non-expresser	41 (77.4%)	32 (84.2%)

Kidney Transplant

The median (5^th–95^th percentile) initial tacrolimus dose for kidney transplant recipients the institutional weight-based dosing protocol was 7 mg daily (2.3 mg-14 mg), which did not differ among CYP3A5 expressers and non-expressers (Table 3). The median therapeutic daily dose for kidney transplant recipients was 9.4 mg (2.4–15.9 mg) and differed between CYP3A5 expressers and non-expressers. A significant difference was found between the initial dose from the institutional weight-based dosing protocol and popPK-predicted doses among all patients (p<0.001). Post-hoc comparisons revealed that the first therapeutic dose (p=0.003), the Andrews et al. model-predicted initial dose (p<0.001), and the Zhao et al. model-predicted initial dose (p<0.001) differed significantly from the actual initial dose, while the Chen et al. model-predicted initial dose did not differ significantly from the actual initial dose (p=0.4). As shown in Table 3, the first therapeutic dose was higher than the prescribed initial dose, with a 2-fold dose increase in CYP3A5 expressers and a 14% increase in CYP3A5 non-expressers. The Andrews et al. and Zhao et al. models predicted lower doses than the prescribed initial dose.

Table 3.

Comparison of the alternate population pharmacokinetic model-predicted and institutional weight-based dosing protocol- initial doses to the first observed therapeutic dose for patient status post kidney transplantation by the CYP3A5 phenotype

Model – Dose Type	Factors	Median (5th–95th Percentile) Daily Dose (mg)			P-value *
		CYP3A5 Expresser N=12	CYP3A5 non-Expresser N=41	Overall N=53
Protocol – Initial	Weight	6 (3.2-NC)	7 (2–14)	7 (2.3–14)	--
Protocol – Therapeutic	TDM	13 (3-NC)	8 (2.0–16)	9.4 (2.4–15.9)	p < 0.001
Andrews	Weight, CYP3A5 genotype	6.5 (3.8-NC)	4.5(2.4 – 7.1)#	4.8 (2.5–8.3)
Zhao	Weight, CYP3A5 genotype, Hct	5.3 (4.0-NC)	3.6 (2.0–5.1)#	3.9 (2.3–7.1)
Chen- Initial	Weight, CYP3A5 genotype, post-transplantation day, presence of Wuzhi capsule	9.2 (4.2-NC)	6.5 (2.3–11.2)#	7.0 (2.9–14.7)
Chen – therapeutic	Weight, CYP3A5 genotype, post-transplantation day, presence of Wuzhi capsule	6.0 (3.3 -NC)	4.4 (1.6–10.2)	4.4 (1.7–10.0)

Dose determined for a target trough of 10 ng/mL = AUC 185 ng/dL

^*comparison of the overall dose between models and the standard therapeutic dose- Friedman test

^#p<0.05, comparing CYP3A5 expresser and CYP3A5 non-expresser for the same model

NC: Not calculable; protocol-initial: weight-based dosing per institutional dosing protocol; protocol-therapeutic: dose at first steady-state, therapeutic trough concentration per clinician guided dose adjustments; Chen-initial: dose prediction for the first day of tacrolimus administration post-transplant; Chen-therapeutic: dose prediction for the post-operative day that the first steady-state therapeutic trough was observed.

A significant difference was found between the model-predicted doses and the first therapeutic dose (p<0.001). Post-hoc comparisons revealed that all model-predicted doses, except the Chen et al. therapeutic dose and the Andrews et al. dose, differed significantly from each other (p<0.001). The model-predicted doses were lower than the actual therapeutic doses. Figure 1 illustrates the linear fit plots of the observed actual therapeutic dose to the model-predicted dose on a weight basis. The model-predicted doses performed better than the initial weight-based protocol recommended dose (0.2 mg/kg). The similar slopes of these lines with a shift to the left from unity (actual therapeutic dose) illustrate the prediction bias. The Chen et al. model-predicted initial dose was less biased than the dose predicted by other models for kidney transplant recipients.

Figure 1.

Linear fit plots of the observed therapeutic dose to the model predicted doses in mg/kg with the reference initial dose (0.2 mg/kg) and reference line of unity (observed therapeutic dose) for kidney transplant patients.

Heart Transplant

The median (5^th–95^th percentile) initial tacrolimus dose for heart transplant recipients was 4 mg (1.0–6.1 mg), which did not differ between the CYP3A5 expressers and non-expressers. The median therapeutic dose for all heart transplant recipients was 6 mg (1.0–18 mg), which was numerically, but not significantly, higher in CYP3A5 expressers than non-expressers (Table 4). There were significant differences among the initial, therapeutic, and popPK model-predicted tacrolimus doses (P <0.001). In post-hoc analysis, all respective model-predicted doses differed significantly from the actual initial dose.

Table 4.

Comparison of the alternate population pharmacokinetic model-predicted and institutional weight-based dosing protocol- initial doses to the first observed therapeutic dose for patient status post heart transplantation by the CYP3A5 phenotype

Model – Dose Type	Factors	Median (5th–95th Percentile) Daily Dose (mg)			P-value *
		CYP3A5 Expresser N=6	CYP3A5 non-Expresser N=32	Overall N=38
Protocol – Initial	Weight	4 (1.6-NC)	3.5 (0.9–6.7)	4 (1.0–6.1)
Protocol – Therapeutic	TDM	9 (3-NC)	5 (0.9–15.6)	6 (1.0–18)	p < 0.001
Andrews	Weight, CYP3A5 genotype	8.2 (4.3-NC)	5.1 (2.7–7.4)#	5.5 (2.9–8.5)
Zhao	Weight, CYP3A5 genotype, Hct	6.6 (3.3-NC)	3.7 (2.3–5.3)#	3.9 (2.3–7.1)
Chen- Initial	Weight, CYP3A5 genotype, POD, Wuzhi capsule	11.7 (5.1-NC)	6.8 (2.7–11.1)#	7.0 (2.8–12.7)
Chen – therapeutic	Weight, CYP3A5 genotype, POD, Wuzhi capsule	9.3 (3.7-NC)	6.0 (2.0–10.2)	6.4 (2.4–11.3)

Dose determined for a target trough of 10 ng/mL = AUC 185 ng/dL

^*comparison of the overall dose between models and the standard therapeutic dose- Friedman test

^#p<0.05, comparing CYP3A5 expresser and CYP3A5 non-expresser for the same model

NC: Not calculable; protocol-initial: weight-based dosing per institutional dosing protocol; protocol-therapeutic: dose at first steady-state, therapeutic trough concentration per clinician guided dose adjustments; Chen-initial: dose prediction for the first day of tacrolimus administration post-transplant; Chen-therapeutic: dose prediction for the post-operative day that the first steady-state therapeutic trough was observed.

Significant differences were found between the model-predicted and therapeutic tacrolimus doses (p<0.001). In post-hoc comparisons, the model-predicted dose by Zhao et al. significantly differed from the actual therapeutic dose (p<0.001); however, the doses predicted by the Andrews et al. and Chen et al. models did not significantly differ from the actual therapeutic dose. Figure 2 illustrates the linear fit plots of the observed therapeutic dose to the model-predicted dose on a weight basis. The model-predicted doses performed better than the initial weight-based protocol recommended dose (0.1 mg/kg) but were imprecise for heart transplant recipients. The dissimilar slopes of these lines that cross the line of unity (actual therapeutic dose) illustrate this bias and imprecision.

Figure 2.

Linear fit plots of the observed therapeutic dose to the model predicted doses in mg/kg with the reference initial dose (0.1 mg/kg) and reference line of unity (observed therapeutic dose) for heart transplant patients.

Discussion

In this study, we evaluated the capacity of three previously published popPK models to predict tacrolimus dose requirements for pediatric kidney and heart transplant recipients. For kidney transplants, the Zhao et al. model predicted lower doses overall, the Andrews et al. model predicted lower doses in CYP3A5 non-expressers, and the Chen et al. model predicted higher doses for CYP3A5 expressers compared to the currently utilized weight-based dosing. All model-predicted doses for kidney transplant recipients were consistently lower than the actual therapeutic dose, although the Chen et al. initial dose prediction was closer to the actual therapeutic dose than the other models. This finding suggests that the Chen et al. model may improve initial dose selection for pediatric renal transplants compared to the current weight-based dosing strategy. For heart transplant recipients, the Zhao et al. model predicted similar doses overall, while the Andrews et al. and Chen et al. models predicted higher initial doses than the weight-based dosing strategy. For heart transplant recipients, the model-predicted doses of Andrews et al. and Chen et al. were not significantly different from the therapeutic dose, although the model-predicted dose by Andrews et al. was slightly lower and that by Chen et al. was slightly higher than the actual dose. These findings suggest that the addition of clinical variables beyond the CYP3A5 genotype does not significantly improve therapeutic dose prediction for pediatric heart transplant recipients.

Although many tacrolimus dosing models or other personalized dosing strategies have been developed, few have been evaluated for their impact on improving initial tacrolimus dose selection. Andrews et al. assessed a previously published popPK-based dosing strategy in pediatric kidney transplant recipients. In this study, the initial tacrolimus dose was based on the equation: Dose (mg/day) = 209 ng*h/mL × 54.9 × (weight/70)^0.75 × (1.8 if CYP3A5*1) × (0.74, if living donor)/1000. Investigators discontinued this prospective study at the time of interim analysis due to poor dose prediction, particularly in CYP3A5 expressers with deceased donors, and the concern for tacrolimus over-exposure with this model.²² We found that their revised model tended to under-predict the actual therapeutic tacrolimus dose for our kidney transplant recipients, suggesting that the revised model may be less likely to result in significant tacrolimus over-exposure than the prior model. However, the model-predicted doses by both the Andrews et al. and Zhao et al. models were approximately 50% lower than the actual therapeutic dosing in this study, suggesting a limited impact on improving the time to achieve a therapeutic concentration. The dose predictions of Chen et al. model at the time of the initial dose and the first therapeutic dose were also lower than the actual therapeutic dose, although the initial dose prediction by this model was the closest to the actual therapeutic dose , suggesting that this dosing strategy could help achieve therapeutic concentrations more quickly without significant risk for supra-therapeutic dosing.

Although the popPK models assessed in this study were developed for kidney transplant recipients, as no popPK models that include the CYP3A5 genotype have been developed for pediatric heart transplant,^16,26 we evaluated their ability to predict doses in this population. The Andrews et al. model appeared to perform the best in this population, with initial dose predictions closest to the actual therapeutic dose, without being higher than the actual dose. This finding supports the hypothesis that the CYP3A5 genotype remains a significant contributor to tacrolimus pharmacokinetics in heart transplant recipients. A popPK model for pediatric heart transplant recipients that did not include CYP3A5 identified concomitant fluconazole and creatinine clearance as covariates affecting tacrolimus CL/F,¹⁶ while a popPK study with adult heart transplant recipients revealed that the CYP3A5 genotype and concomitant azole medications significantly impact tacrolimus clearance.²⁶ Patients who received concomitant CYP3A inhibitors were excluded from our analysis; however, our findings suggest that transplant-specific models might be needed for improved tacrolimus dose selection. This notion is supported by a recent study that evaluated the tacrolimus dose predictions of adult popPK models developed in multiple transplant types for adult heart transplant recipients. This study concluded that the models did not adequately predict tacrolimus dosing for this population and should not be used for clinical dosing prediction.²⁷

The models evaluated in this study tended to under-predict the actual therapeutic dose. This finding suggests that the CL/F estimates in the models are lower than the actual patient clearance and would have a lower likelihood of resulting in supratherapeutic exposure. Owing to the risk of neurotoxicity and nephrotoxicity with acute supra-therapeutic concentrations, a slight under-prediction of dose requirements may be preferred clinically relative to an overestimation.

The clinical impact of personalized tacrolimus dosing in pediatric solid organ transplant recipients has only been evaluated in one study. Min et al. evaluated a genotype and age-based dosing strategy in pediatric heart, liver, and kidney transplant recipients.¹⁹ In this study, CYP3A5 expressers received 50% higher doses than non-expressers and patients ≤ 6 years had an additional 25% dose increase, capped at 5 mg/dose. Compared with standard weight-based dosing, the personalized dosing strategy was associated with a decreased time to achieve a therapeutic tacrolimus concentration and a reduction in out-of-range concentrations. Of all the models assessed in this study, the Chen et al. model-predicted initial doses were most similar to the dosing strategy used in this prospective clinical trial, suggesting that the application of this model in pediatric kidney transplant recipients could also help reduce time to achieve therapeutic concentrations to improve medication safety and efficacy in this population. The similarity in dose predictions also supports pediatric patients who are <6 years old may require additional dosing adjustments beyond the currently recommended 1.5–2x dose increase in CYP3A5 expressers from the Clinical Pharmacogenetics Implementation Consortium guidelines.²⁵ However, the dosing used in the Min et al. study was higher than the patient-predicted doses for our heart transplant recipients. This finding suggests that organ-specific considerations may be muted in multi-transplant dosing models. Although the personalized dosing strategy assessed by Min et al. reduced out-of-range concentrations, the study did not evaluate this outcome individually by transplant type. Therefore, it remains unclear whether the dose requirements and tacrolimus exposure vary among transplant types.

Patients younger than 2 years old were excluded from our analyses owing to their exclusion from the popPK model development.¹⁵ Substantial changes in CYP3A expression occur over the first year of life. CYP3A7 is the predominant enzyme in newborn hepatic tissue, although expression wanes over the first year of life. Furthermore, during this time, CYP3A4 expression increases, while CYP3A5 expression appears to be constant.²⁸ This natural ontogeny could contribute to additional variability in tacrolimus dosing requirements in children younger than 2 years old and further study is needed within this population to improve tacrolimus dose selection. Model-based dosing might be inappropriate for any patient receiving concomitant CYP3A inhibitors at the time of tacrolimus initiation, and a further study is needed to understand how to account for this covariate. The models also only apply to the immediate-release formulation of tacrolimus and might not be accurate for extended-release formulations. Our findings are limited by the small patient sample size and single-center design, and differences in institutional practices for post-transplant management could further influence the tacrolimus pharmacokinetic parameters.

Conclusion

Individualization of tacrolimus dosing has the potential to improve medication efficacy and decrease toxicity by reducing the time required to meet and maintain therapeutic tacrolimus concentrations and decreasing the risk of out-of-range concentrations. Our findings demonstrate that the current weight-based standard dosing is suboptimal for most pediatric transplant recipients, and the Chen et al. model-predicted initial doses more closely correlated with actual therapeutic doses for kidney transplant recipients. The models developed for kidney transplant patients do not appear to translate well to heart transplant patients, which suggests that model-based tacrolimus dosing should be tailored to transplant type. A prospective application of validated popPK models should be evaluated in future studies to assess their impact on clinical patient outcomes.