Recipient Versus Donor CYP3A4/5 Genotypes in Adult Liver Transplant Recipients to Achieve Therapeutic Tacrolimus Concentrations Early Post‐Transplant

ABSTRACT

Tacrolimus is an immunosuppressive agent with difficult dosing due to a narrow therapeutic index and large interpatient pharmacokinetic variability, for which CYP3A5 variation plays a role. Tacrolimus/CYP3A5 pharmacogenetic guidelines exclude liver transplant patients with a donor/recipient CYP3A5 mismatch. We sought to determine the influence of donor vs. recipient CYP3A5 genotype early post‐transplant and evaluate tacrolimus dosing strategies in liver transplant recipients with various recipient/donor CYP3A5 genotype combinations. This was a single‐center, retrospective analysis of 58 prospectively enrolled adult liver transplant patients prescribed tacrolimus post‐transplant, with donor and recipient CYP3A5 genotype data, which were categorized as recipient expresser (RE) or nonexpresser (RN) and donor expresser (DE) or nonexpresser (DN). Patients were stratified as REDE, REDN, RNDN, or RNDE with comparisons across groups. Prediction error (PE) based on predicted versus actual therapeutic dose was calculated for eight published dosing methods. Donor/recipient genotype was mismatched in 41% of liver transplant recipients. Weight‐adjusted total daily dose for first therapeutic tacrolimus concentration was significantly different across CYP3A5 combinations, with recipient expressors requiring the highest doses (REDE 0.15 mg/kg/day, REDN 0.19 mg/kg/day, RNDE 0.12 mg/kg/day, RNDN 0.09 mg/kg/day, p = 0.006). Of the eight dosing methods analyzed, two performed with < 2% PE overall, and ≤ 20% PE for all four genotype combinations: one included recipient genotype only; the other included donor and recipient genotype. Collectively, our findings suggest recipient CYP3A5 genotype may be more important than donor genotype in the immediate post‐liver transplant period for predicting optimal tacrolimus dosing to achieve therapeutic trough concentrations.

Study Highlights.

• What is the current knowledge on the topic?

  • ○
    The relative influence of donor and recipient genotypes on tacrolimus dosing immediately post‐liver transplant remains unclear. In the context of liver transplant recipients, CPIC guidelines indicate the recommendations only apply to those in which donor and recipient genotypes are identical, which means both donor and recipient genotypes are needed and there is no guidance in the case of donor/recipient genotype mismatch.
• What question did this study address?

  • ○
    This study aimed to determine the importance of donor versus recipient CYP3A5 genotype and an optimal tacrolimus dosing strategy to achieve first therapeutic tacrolimus trough concentration in liver transplant recipients with various recipient/donor CYP3A5 genotype combinations.
• What does this study add to our knowledge?

  • ○
    The data suggest recipient genotype is more important than donor genotype in the early post‐transplant tacrolimus dosing period; thus, CPIC recommendations for starting dose could be expanded to all liver transplant recipients where their CYP3A5 genotype is known, with 1.5–2 times the starting dose for Normal Metabolizer and Intermediate Metabolizer phenotypes. This strategy may be further personalized by incorporating factors included in model‐based dosing (e.g., CYP3A4 inhibitor use, age, hematocrit) to ensure tacrolimus dosing that will lead to the faster achievement of goal therapeutic trough concentrations to prevent graft rejection.
• How might this change clinical pharmacology or translational science?

  • ○
    This study suggests that the implementation of pharmacogenetics in the immediate post‐liver transplant setting may be facilitated by the use of a dosing algorithm that requires recipient genotype only, which bypasses the logistical challenges of obtaining the donor liver sample for genotype determination.

1. Introduction

1.1. Background

Tacrolimus is a calcineurin‐inhibitor immunosuppressive agent that is Food and Drug Administration (FDA)‐approved for prophylaxis of organ rejection in adult and pediatric patients receiving allogeneic kidney, liver, and heart transplants, in combination with other immunosuppressants [1]. Tacrolimus exhibits a narrow therapeutic index and large interpatient pharmacokinetic (PK) variability [2]. Currently, therapeutic drug monitoring (TDM) strategies are used to guide individual tacrolimus dose adjustments, as overexposure may lead to adverse effects (e.g., nephrotoxicity, neurotoxicity, type 2 diabetes mellitus, immunosuppression‐associated opportunistic infections, cardiovascular toxicities, malignancies), while underexposure increases the risk of graft rejection due to insufficient immunosuppression [3, 4, 5]. Extensive pre‐systemic metabolism by the CYP3A enzymes in the gut wall results in poor oral bioavailability [6]. In addition to its poor and variable oral bioavailability, the large interpatient PK variability exhibited by tacrolimus is in part due to genetic variations altering the functionality of CYP3A4 and CYP3A5, which are primarily responsible for its metabolism/clearance [7, 8].

1.2. Genetic Influence on Inter‐Patient Tacrolimus PK Variability

The most investigated genes affecting tacrolimus PK following all transplantation types are CYP3A4 and CYP3A5, with the strongest evidence associated with CYP3A5. Known CYP3A5 variant alleles that encode a nonfunctional protein and therefore loss of CYP3A5 enzyme activity are CYP3A5*3 (rs776746), *6 (rs10264272), and *7 (rs41303343). The Association for Molecular Pathology (AMP) Clinical PGx Working Group has classified these alleles as tier 1 recommended variants to test due to their well‐characterized effect on functional enzyme activity and a prevalence of > 1% in at least one ancestral population [8]. Absence of functional CYP3A5 enzyme (poor metabolizer, PM) is the most common phenotype in many populations. Specifically, populations of European and East Asian ancestry, ~85% and 56%, respectively, lack functional protein due to being homozygous for CYP3A5 loss of function alleles, whereas only ~30% of those with African ancestry lack functional CYP3A5 enzyme based on genotype [8]. Current phenotype terminology refers to CYP3A5*3/*3, *6/*6, *7/*7, *3/*6, *3/*7, *6/*7 genotypes as PM or nonexpresser, *1/*1 genotype as normal metabolizer (NM) or expresser, and *1/*3, *1/*6, *1/*7 genotypes as intermediate metabolizer (IM) or expresser.

For those expressing the CYP3A5 enzyme (i.e., CYP3A5 NM or IM), Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines recommend increasing the tacrolimus starting dose by 1.5–2 times the recommended initial total daily dose (TDD), with subsequent dosing adjustments based on TDM and clinical factors [9]. In the context of liver transplant patients, the CPIC guidelines indicate the recommendations only apply to those in which donor and recipient genotypes are identical. Given that donor genotype would rarely be known, this essentially eliminates liver transplant recipients from potential benefits of CYP3A5 pharmacogenetics‐guided dosing.

CYP3A4 is the most abundantly expressed CYP enzyme and accounts for approximately 30 to 40% of the total CYP content in human adult liver and small intestine [10]. The CYP3A4*22 (rs35599367) allele was recently classified as a tier 1 recommended variant by the AMP Clinical PGx Working Group [8] and is predicted to be included in the next CPIC PGx‐guided tacrolimus dosing guideline update. The CYP3A4*22 decreased function variant has been shown to significantly influence CYP3A4 hepatic expression, and individuals carrying this variant are predicted to be at increased risk of tacrolimus overexposure requiring lower doses compared to the CYP3A4*1/*1 genotype [11].

1.3. Recipient and Donor Combinatorial Genotype Effect on Tacrolimus PK

In the setting of liver transplantation, assessment of tacrolimus PK becomes even more complex as it is necessary to consider the genotype of both the donor liver graft and the recipient (i.e., CYP3A enzymes responsible for gut metabolism), as they may differ. To date, the relative influence of donor and recipient genotypes on tacrolimus dosing post‐liver transplant remains unclear. Additionally, it is possible the impact of donor versus recipient genotype may be dynamic and change over time. Although some studies show that both recipient and donor genotypes affect dose‐adjusted trough concentrations from the 1st week post‐transplant, others show that donor genotype does not begin to play a role until the 2nd week or even 6 months following transplantation [12, 13, 14].

1.4. Published Model‐Predicted Tacrolimus Dosing Methods

Existing PGx‐guidelines recommend empirically increasing tacrolimus dose based on genotype‐predicted CYP3A5 expression immediately post‐transplantation; however, this strategy may be further personalized by incorporating factors included in model‐based dosing to ensure tacrolimus dosing that more rapidly achieves goal therapeutic trough concentrations to prevent graft rejection. Despite the use of TDM, it may still take up to 3 weeks post‐transplant for a patient to reach the target concentration range due to both genetic and clinical factors. Therefore, dosing algorithms have been developed to reduce the incidence of tacrolimus concentrations that fall outside of the target range, with a goal of improving clinical outcomes (e.g., prevent organ rejection, reduce risk of adverse effects) [15]. Starting dose algorithms have been developed to aid prescribers in individualizing a patient's dose requirement based on clinical and/or genetic factors rather than the standard body weight (kg)‐based dose.

While numerous dosing strategies developed to guide tacrolimus dosing post‐kidney transplant exist, optimal dosing strategies in the setting of liver transplant, which consider long‐term PK effects and changes in tacrolimus dosing based on discordant recipient and donor phenotype, are still needed. To identify which of these approaches most accurately estimates the initial dose required in liver transplant recipients to achieve target therapeutic concentrations, we tested the existing dosing strategies from a variety of published approaches in liver and kidney transplant in various recipient/donor CYP3A5 genotype combinations.

2. Materials and Methods

2.1. Study Design and Patient Population

This was a single‐center, retrospective analysis of a prospectively enrolled study population of 58 adult patients who underwent orthotopic liver or liver‐kidney transplantation from deceased donors and received tacrolimus post‐transplant. Tacrolimus trough blood concentration and dose data were documented prospectively as part of clinical care and retrieved from the electronic health record for use in this analysis. The study protocol was approved by the University of Florida (UF) Institutional Review Board (IRB201800053) [16]. Written informed consent was obtained from all transplant recipients prior to enrollment.

2.2. DNA Isolation and Genotyping of CYP3A4 and CYP3A5 Polymorphisms

Liver biopsies were obtained from both donor and native livers for each liver transplant patient. Tissue samples were formalin‐fixed and paraffin‐embedded (FFPE). DNA was extracted from tissue scrolls with the use of GeneRead DNA FFPE Kits (Qiagen, Hilden, Germany). The following SNPs from both donor and recipient were genotyped in the University of Florida Center for Pharmacogenomics using a TaqMan genotyping assay: CYP3A5*3 (rs776746), *6 (rs10264272), *7 (rs41303343); and CYP3A4*22 (rs35599367) [17], with those lacking variability at any of these sites labeled as *1. As is standard in the literature, those with two *1 alleles were designated as NM, those with one *1 allele were IM, and those with zero *1 alleles were PM. We defined NM and IM as expressors and PM as non‐expressors.

2.3. Drug Administration and TDM Per UF Health Protocol Following Liver Transplantation

All study patients were started on the standard‐of‐care (SOC) medication regimen per established center protocol: a three‐drug combination of tacrolimus (Prograf) capsule 0.05 mg/kg administered orally (PO) or sublingually (SL) every 12 h starting on the day of transplantation, corticosteroid, and mycophenolate. Subsequent tacrolimus dosing was based on TDM and patient‐specific clinical factors determined by a transplant‐specialized clinical pharmacist, together with the on‐call transplant surgeon. Additionally, a dose conversion of 2:1 was used per hospital protocol when converting tacrolimus PO to SL formulation, based on limited PK evidence [18, 19]. Tacrolimus trough levels were measured clinically before the patient's morning dose via liquid chromatography/tandem mass spectrometry (LC/MS–MS). The standard target trough range was 8–10 ng/mL for the first month after transplant. A therapeutic trough concentration is defined as two consecutive concentrations at least 48 h apart in the therapeutic range without any changes in tacrolimus dose.

2.4. Data Collection

Liver transplant recipient and donor demographic information, and recipient laboratory data were obtained from electronic health records (EHR). Most data were collected for the duration of inpatient hospitalization using a standardized data collection form. Demographic data, including age, sex, race/ethnicity, weight, height, body mass index (BMI), and preoperative diagnosis were collected. Past medical history (including etiology of liver, kidney, heart, and lung disease), tacrolimus daily doses (mg), blood tacrolimus trough levels (ng/mL), dates (of transplant operation and of subsequent labs/medication/care records during hospital stay), post‐operative day (POD), hematocrit (HCT), total bilirubin (TBILI), both donor and recipient CYP3A5 and CYP3A4 genotypes, concomitant medications (drugs given including time/date/dose) and events including rejection and adverse drug effects were collected from the EHR during the inpatient stay, up to discharge. Further information regarding data collection is provided in the Data S1.

The recipient's weight (kg), tacrolimus daily dose (mg/day) and first therapeutic trough concentration (i.e., within range of 8–10 ng/mL, with concentrations up to 11 ng/mL included as dose revisions were rare in the 8–11 ng/mL range) were recorded. The daily tacrolimus dose was noted at the time of the first therapeutic trough concentration, and the weight‐adjusted tacrolimus dose was calculated using daily tacrolimus dose/weight (mg/kg/day). The dose‐adjusted tacrolimus trough concentration [C/D ratio, (ng/mL)/(mg/kg/day)] was calculated by dividing the measured concentration by the corresponding daily weight‐adjusted tacrolimus dose.

2.5. Dosing Algorithms Tested

We compared five liver transplant dosing algorithms [20, 21, 22, 23, 24], and three kidney transplant dosing algorithms [25, 26, 27] that included CYP3A5 genotype and various other factors including total bilirubin, hematocrit, body surface area, age, CYP3A4*22 variant expression, drug interactions, and post‐op day, as shown in Table 1. Details for the dosing equations algorithms/equations are reported in the Data S1. For the three kidney transplant dosing equations, we did two sets of calculations: one using the recipient genotype and one using the donor genotype.

TABLE 1.

Dosing algorithms analyzed and covariates included in each algorithm.

Author (year)	N	Transplanted organ	Body weight (kg) a	CYP3A5 donor genotype	CYP3A5 recipient genotype	CYP3A4 inhibitor	POD	Other covariates
Gerard et al. (2014) [20]	66	Liver	X	X				HCT
Shao et al. (2020) [21]	43	Liver		X	X
Ji et al. (2018) [22]	58	Liver		X	X		Up to POD 14	Tacrolimus goal trough (10 vs. 15 ng/mL)
Csikány et al. (2021) [23]	112	Liver	X	X				Recipient CYP3A4*22 (if donor CYP3A5 PM)
Cai et al. (2022) [24]	176	Liver			X	X		TBILI HCT
Andrews et al. (2019) [25]	337	Kidney			X			Age BSA Recipient CYP3A4*22
Srinivas et al. (2021) [27]	156	Kidney	X		X			Desired tacrolimus level on Day 6
Passey et al. (2011) [26]	681	Kidney			X	X	Up to POD 180	Age Tacrolimus goal trough

Abbreviations: BSA, body surface area; HCT, hematocrit; POD, post‐op day; TBILI, total bilirubin.

^a Algorithms that reported dose as mg/day were not marked as including body weight (kg).

2.6. Study Outcomes and Statistical Analysis

The primary aim of this study was to compare the prediction error (PE) of model‐based dosing methods relative to the observed first therapeutic tacrolimus dose to determine an optimal dosing method in the early period after liver transplantation, including whether dosing methods that included both donor and recipient genotype performed best.

Performance of various dosing strategy's ability to accurately predict tacrolimus doses versus over‐ or under‐estimating dose requirements was calculated using the following equation:

$$\mathrm{PE}\left(\%\right)=\left(D_{\text{pred}}-D_{\mathrm{obs}}\right)/D_{\mathrm{obs}}\times 100\%$$

where D is the dose expressed in mg/day. PE was summarized with mean (95% CI) and median (IQR) and visualized with boxplots. In this study, a predicted tacrolimus dose falling within ±20% of the actual therapeutic dose demonstrated the clinically significant PE and accuracy of the dosing algorithm. A model is predicted to demonstrate high performance when PE is closer to 0%.

The Kruskal Wallis test was performed for non‐parametric analysis followed by post hoc Dunn‐Bonferroni tests for pharmacokinetic parameters with significant between group differences. One‐way ANOVA was performed for parametric analysis between groups. One‐way ANOVA effect size analysis determined an eta squared (η ²) value of 0.29, suggesting a large effect size for Recipient/Donor CYP3A5 expression groups on the pharmacokinetic parameters of tacrolimus daily dose, Tac daily dose, first therapeutic concentration, and dose‐corrected concentration (C/D ratio). A p‐value less than 0.05 was considered statistically significant.

3. Results

Fifty‐eight liver transplant recipients were enrolled, of whom nine were liver‐kidney transplant recipients. Their baseline characteristics (along with data on the donors) are summarized in Table 2. Of these 58 patients, the majority were White (88%), and no transplant recipients were Black, whereas 26% of donors were Black. The median age was 58. Fifty‐six participants were included in the analysis comparing existing dosing algorithms; two participants had adverse effects and were switched to cyclosporine prior to achieving a therapeutic trough concentration on tacrolimus.

TABLE 2.

Baseline characteristics of the analyzed study population.

Baseline characteristic	N = 58
Gender
Male	34 (58.6%)
Female	24 (41.4%)
Age (years)	62 (54–67)
Race
White/non‐Hispanic	51 (87.9%)
White/Hispanic	6 (10.3%)
Asian	1 (1.7%)
Disease etiology
AIH	4 (6.9%)
ETOH	19 (32.8%)
HCV	8 (13.8%)
NASH	15 (25.9%)
Other	12 (20.7%)
SLKT	9 (15.5%)
Redo OLT	2 (3.4%)
HCC	11 (19%)
MELD score	25 (22–33)
LOS (days)	12 (9–18)
Fluconazole use post‐LT	37 (63.8%)
Body weight (kg)	80.9 (70.6–98.8)
BSA (m2)	1.9 (1.8–2.2)
Total bilirubin (mg/dL)	2.35 (1.3–3.9)
Hematocrit (%)	26.8 (23.8–29.7)
Donor information
Donor race
White/non‐Hispanic	35 (60.3%)
White/Hispanic	7 (12.1%)
Black	15 (25.9%)
Other	1 (1.7%)
Donor cause of death
Anoxia	18 (31%)
CVA	21 (36.2%)
Trauma	19 (32.8%)
Recipient/donor CYP3A5 expresser status
RExpresserDExpresser (REDE)	4 (6.9%)
IM/IM	1
IM/NM	1
NM/IM	2
RExpresserDNonexpresser (REDN)	6 (10.3%)
IM/PM	6
RNonexpresserDExpresser (RNDE)	18 (31%)
PM/IM	14
PM/NM	4
RNonexpresserDNonexpresser (RNDN)	30 (51.7%)
PM/PM	30
Recipient/donor CYP3A4*22 carrier status
Yes/No	6 (10.3%)
Yes/Yes	1 (1.7%)
No/Yes	6 (10.3%)
No/No	45 (77.6%)

Note: Presented as frequency N (%) or median (interquartile range: Q1–Q3).

Abbreviations: AIH, autoimmune hepatitis; BSA, body surface area; CVA, cerebral vascular accident; D, donor; ETOH, alcohol‐related; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; IM, intermediate metabolizer; LOS, length of stay post‐LT to discharge; LT, liver transplant; MELD, model for end‐stage liver disease; NASH, nonalcoholic steatohepatitis; NM, normal metabolizer; OLT, orthotopic liver transplant; PM, poor metabolizer; R, recipient; SLKT, simultaneous liver‐kidney transplant.

Participants were stratified into four groups based on recipient and donor CYP3A5 genotype‐based expression status (i.e., expresser vs. nonexpresser): (1) CYP3A5 IM/NM recipient of CYP3A5 IM/NM donor graft (REDE, n = 4); (2) CYP3A5 IM/NM recipient of CYP3A5 PM donor graft (REDN, n = 6); (3) CYP3A5 PM recipient of CYP3A5 IM/NM donor graft (RNDE, n = 18); (4) CYP3A5 PM recipient of CYP3A5 PM donor graft (RNDN, n = 30). Thus, 41% of liver transplant recipients received a liver whose CYP3A5 genotype‐inferred phenotype was discordant from theirs.

3.1. Tacrolimus PK Findings Based on Recipient/Donor CYP3A5 Phenotype Combinations

TDD, weight‐adjusted TDD (mg/kg/day) and the C/D ratio associated with the first therapeutic tacrolimus trough by recipient/donor CYP3A5 genotype combinations are shown in Table 3, and weight‐adjusted TDD data are also represented in Figure 1. As shown in Table 3, doses were significantly different across groups, with REs requiring the highest doses, which were significantly different from those for whom both recipient and donor were nonexpressers. Of note, RE mean weight‐adjusted TDD values (REDE = 0.15 mg/kg/day; REDN = 0.19 mg/kg/day) closely aligned with CPIC guideline recommendations to increase the dose 1.5–2‐fold in expressers relative to the standard recommended starting dose of 0.1 mg/kg/day. Weight‐adjusted TDD of those with liver‐kidney transplants was similar to the recipient/donor group into which they were included. Specifically, six were RNDE, two were RNDN, and one REDN, with median weight‐adjusted TDD of 0.11 mg/kg/day, 0.08 mg/kg/day, respectively, and 0.06 mg/kg/day for the one REDN participant. There were no significant differences in the time to first therapeutic concentration, though there were more REs than RNs, and the median 3‐day difference in time to first therapeutic concentration between RNDN and REDE might be considered clinically relevant. As expected, there were no differences in the first therapeutic trough concentration, but there were marked differences between the C/D ratio.

TABLE 3.

Observed tacrolimus pharmacokinetic parameters at first therapeutic tacrolimus dose and concentration.

PK parameter	R/D group (n = 56) a	Mean	Median	IQR (Q1–Q3)	Overall sig. between groups (p)
Weight adjusted TDD (mg/kg/day)	RNDN	0.09	0.08	0.06–0.12	0.006
	RNDE	0.12	0.11	0.09–0.16
	REDN	0.19	0.18	0.15–0.25
	REDE	0.15	0.14	0.1–0.18
POD at first therapeutic tacrolimus trough concentration (days)	RNDN	5.5	4.5	3–6	0.161
	RNDE	7.1	5	3.3–9.5
	REDN	6.3	6	5.3–7.5
	REDE	8.8	7.5	6.3–10
TDD (mg/day)	RNDN	7.7	6.5	6–10	0.005
	RNDE	9.6	8	6.6–11
	REDN	13.7	14.8	13.4–15
	REDE	12.9	13	11.4–14.5
First therapeutic tacrolimus trough concentration (ng/mL)	RNDN	9.5	9.5	9–10.1	0.338
	RNDE	9.5	9.3	8.8–10.2
	REDN	9	8.7	8.2–9.7
	REDE	8.9	8.4	8.2–9.1
C/D ratio [(ng/mL)/(mg/kg/day)]	RNDN	1.47	1.43	0.94–1.73	0.003
	RNDE	1.11	1.08	0.9–1.41
	REDN	0.77	0.64	0.58–0.74
	REDE	0.7	0.67	0.65–0.75

Abbreviations: C/D, tacrolimus concentration dose ratio; IQR (Q1–Q3), interquartile range (quartile 1–quartile 3); PK, pharmacokinetic; POD, post‐op day; R/D, recipient/donor; REDE, recipient and donor expresser; REDN, recipient expresser and donor nonexpresser; RNDE, recipient nonexpresser and donor expresser; RNDN, recipient and donor nonexpresser; TDD, total daily dose resulting in first therapeutic tacrolimus trough concentration.

^a Two patients within the RNDN group were excluded from analysis as they were switched to an alternative immunosuppressant as a result of tacrolimus‐related seizures.

FIGURE 1.

Weight adjusted TDD by recipient/donor CYP3A5 expression groups. Dashed line within boxplots indicates mean weight adjusted TDD; NS, not significant; REDE, recipient and donor expresser; REDN, recipient expresser and donor nonexpresser; RNDE, recipient nonexpresser and donor expresser; RNDN, recipient and donor nonexpresser; TDD, total daily dose resulting in first therapeutic tacrolimus trough concentration.

3.2. Dosing Algorithm Comparisons

Performance of each of the dosing algorithms in the four recipient/donor CYP3A5 phenotype combinations is shown in Figure 2 and Table 4, ordered generally by those that most underestimated to overestimated the dose. The three kidney dosing algorithms are represented twice, where recipient and donor genotype were tested separately, and those are noted with an (R) or (D), respectively. The best performing algorithms were those by Passey et al. [26], where the recipient genotype was included, and Shao et al. [21], which incorporated both recipient and donor genotype. Across all study participants, these two dosing approaches had < 2% predicted error overall, with a PE ≤ 20% error for all four recipient/donor genotype combinations. Mean absolute prediction error (MAPE) was also calculated and reported in Table S2 and Figure S1, where the recipient algorithm by Passey and the Shao algorithm remained the best performing.

FIGURE 2.

Clustered boxplot of percent prediction error by algorithm for predicted dose. Passey (R) and Shao performed the best overall at predicting observed tacrolimus dose; dashed line indicates clinically acceptable prediction error defined as within ±20%; mild outliers: values > 1.5 × interquartile range (IQR) below Q1 or above Q3 are represented by circles (•); extreme outliers: values > 3.0 × IQR below Q1 or above Q3 are represented by asterisks (*); (D), calculated using donor genotype only; (R), calculated using recipient genotype only; REDE, recipient and donor expresser; REDN, recipient expresser and donor nonexpresser; RNDE, recipient nonexpresser and donor expresser; RNDN, recipient and donor nonexpresser.

TABLE 4.

Mean percent prediction error of observed dose versus algorithm‐predicted dose by recipient/donor CYP3A5 expression groups.

R/D group	Ji	Cai (R)	Andre. (R)	Srivan. (R)	Cai (D)	Andre. (D)	Srivan. (D)	Passe. (R)	Shao	Gerard	Passe. (D)	Csikány
RNDN	−46	−16*	−26	−18*	−16*	−25	−18*	8*	7*	27	8*	24
RNDE	−57	−51	−45	−42	−35	−8*	−14*	−20*	5*	−17*	40	85
REDN	−45	−39	−29	−41	−49	−57	−56	9*	−20	10*	−36	−35
REDE	−14*	−48	−17*	−16*	−42	−17*	−32	20*	−15*	35	14*	51
Total	−47	−32	−32	−28	−28	−23	−22	−0.1*	1.8*	11*	13*	39

Note: Asterisk (*) indicates percent prediction error falls within clinically acceptable range of ±20%.

Abbreviations: (D), calculated using donor genotype only; (R), calculated using recipient genotype only; REDE, recipient and donor expresser; REDN, recipient expresser and donor nonexpresser; RNDE, recipient nonexpresser and donor expresser; RNDN, recipient and donor nonexpresser.

4. Discussion

Overall, the data suggest that recipient CYP3A5 genotype may be more important initially than donor genotype for predicting optimal tacrolimus dosing to achieve a therapeutic trough concentration. Our data also support the CPIC guideline recommendation of a 1.5–2‐fold increase from the usual starting dose among CYP3A5 REs. CPIC guidelines currently only apply to: (1) kidney, heart, lung, and hematopoietic stem cell transplant patients, and (2) liver transplant patients in which the donor and recipient genotypes are identical. However, the data from this study suggest their recommendations for starting dose could be expanded to all liver transplant recipients where CYP3A5 is known in the recipient, with 1.5–2 times the starting dose for NM and IM phenotypes.

It is noteworthy that the best performing algorithm in our liver transplant population was one developed for kidney transplant patients, suggesting that the Passey et al. equation [26] may also be used clinically in the liver transplant setting. One note about the Passey et al. algorithm relates to the treatment of CYP3A inhibitors. The Passey algorithm incorporated CYP3A4 inhibitors into the dosing equation, which may have improved the method's ability to predict the actual optimal initial dose requirement for this population. Although the Passey equation specifically incorporates use of CYP3A4 inhibitors diltiazem and verapamil, none of the patients enrolled were on these agents. However, among our study participants, 64% were administered fluconazole, a moderate CYP3A4 inhibitor that may increase tacrolimus concentrations; therefore, we substituted azole antifungals as a factor in the equation.

The other high performing algorithm by Shao et al. [21] was developed in liver transplant recipients and incorporated both donor and recipient genotype. One other algorithm from liver transplant patients that included donor genotype only [20] was within 20% PE overall but performed poorly in RNDN and REDE, while performing much better in those with discordant recipient vs. donor genotypes. The other algorithm by Ji et al. [22] that included both recipient and donor genotype performed the worst of all dosing approaches tested. Two of the dosing algorithms also included CYP3A4*22 [23, 25], but these dosing approaches performed relatively poorly in our study population. It may be that some of the equations overestimated the patient's tacrolimus dose due to the higher goal tacrolimus trough level of 15 ng/mL versus the goal of 10 ng/mL at our institution.

That the logistically simpler algorithm, which included only recipient genotype, outperformed those that required donor genotype with or without recipient genotype may have important implications for facilitating clinical adoption of pharmacogenetic guidance of tacrolimus dosing in the setting of liver transplant. Recipient genotype can be determined as early as the time of waitlisting for transplant; whereas collection of a tissue sample and subsequent determination of donor genotype is both logistically challenging and time‐constrained relative to the first dose of tacrolimus.

As shown in the Data S1, most genotype‐guided tacrolimus dosing methods in the liver transplant setting utilized a dosing chart, whereas those used in the kidney transplant setting were much more comprehensive and personalized with additional clinical factors. Therefore, the creation of a more comprehensive individualized tacrolimus dosing equation versus dosing table for liver transplant recipients may be needed. Du et al. recently developed an artificial neural network (ANN), a model‐based learning algorithm, to identify variables significantly associated with tacrolimus concentration in liver transplant recipients [28]. Aside from recipient (rank 3) and donor CYP3A5 genotype (rank 5), the ANN model identified 15 other variables that influenced tacrolimus concentration, ranked as follows: (1) daily dose; (2) post‐operative day; (4) recipient age; (6) γ‐glutamyl transpeptidase; (7) concomitant use of caspofungin; (8) serum creatinine (SCr); (9) alanine aminotransferase; (10) recipient BW; (11) hematocrit; (12) concomitant use of Wuzhi capsule; (13) total bilirubin; (14) albumin; (15) recipient sex; (16) globulin; and (17) aspartate aminotransferase. Sophisticated dosing approaches that incorporate these variables may represent the future for precision medicine. Of note, the Passey et al. dosing algorithm was found to be the most accurate prediction method in the study and is one of the few algorithms analyzed that account for differences in post‐op day. This confirms that post‐op day (listed as number 2 per Du et al. study) is an important factor to incorporate into a dosing algorithm to predict a therapeutic tacrolimus dose.

There were no differences in time to first therapeutic concentration, which might represent a type II error due to insufficient sample size. However, it is of interest that the two genotype combinations expected to have the most extreme differences in tacrolimus metabolism had a median 3‐day difference in time to first therapeutic concentration (RNDN 4.5 days vs. REDE 7.5 days), consistent with expressors being underdosed when CYP3A5 genotype is not considered with initial dosing of tacrolimus. Other studies would be required to determine the clinical relevance of a 3‐day difference in time to therapeutic concentration, but it could contribute to a longer inpatient stay, with attendant increased costs and the possibility of poor transplant‐related outcomes because of subtherapeutic or supratherapeutic tacrolimus concentration. These data highlight the potential clinical implications of genotyping the recipient prior to transplant and considering that genotype in defining the initial tacrolimus dose.

A limitation of this study is that it is a retrospective analysis of data in a liver transplant recipient population from a single institution, and thus is a relatively small sample size. Therefore, further prospective research is needed to document whether one of the top‐performing algorithms in this study is superior to a standard‐of‐care approach to tacrolimus dosing. Our analysis also does not provide insight into the role of donor genotype after the acute‐post‐transplant period. We acknowledge that some patients were likely not at steady state relative to the dose that was used to define a therapeutic level, and thus the C/D ratio we report may be underestimated relative to the C/D ratio at steady state. Additionally, other data suggest that donor genotype may become more important over time, and thus it is possible that chronic tacrolimus doses are defined by a combination of recipient genotype (for its influence in gut metabolism) and donor genotype (for liver metabolism) [13, 29].

In conclusion, recipient CYP3A5 genotype, regardless of donor CYP3A5 genotype, may be more important in the immediate post‐liver transplant period for predicting optimal tacrolimus dosing to achieve a trough concentration. Our data support the CPIC guideline recommendation of 1.5–2‐fold increase from the usual starting dose among CYP3A5 expressers and suggest expanding the recommendation to liver transplant recipients regardless of donor genotype for initial dosing. Implementation of pharmacogenetics in the setting of liver transplant may be facilitated by use of a dosing algorithm that requires recipient genotype only, which bypasses the logistical challenges of determining the donor genotype.

Author Contributions

E.E., T.Y.L., A.Z., and J.A.J. wrote the manuscript; E.E., J.A.J., and A.Z. designed the research; E.E., S.D., I.A.P., A.D.L., C.W. and T.L. performed the research; E.E. analyzed the data.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: cts70329‐sup‐0001‐Supinfo.docx.

Eken E., Duarte S., Angeli‐Pahim I., et al., “Recipient Versus Donor CYP3A4/5 Genotypes in Adult Liver Transplant Recipients to Achieve Therapeutic Tacrolimus Concentrations Early Post‐Transplant,” Clinical and Translational Science 18, no. 9 (2025): e70329, 10.1111/cts.70329.