Population pharmacokinetics of voriconazole and CYP2C19 polymorphisms for optimizing dosing regimens in renal transplant recipients

Abstract

Aims

The aims of the present study were to characterize the pharmacokinetics of voriconazole in renal transplant recipients and to identify factors significantly affecting pharmacokinetic parameters. We also aimed to explore the optimal dosing regimens for patients who developed invasive fungal infections.

Methods

A total of 105 patients (342 concentrations) were included prospectively in a population pharmacokinetic analysis. Nonlinear mixed‐effects models were developed using Phoenix NLME software. Dosing simulations were performed based on the final model.

Results

A one‐compartment model with first‐order absorption and elimination was used to characterize voriconazole pharmacokinetics. Population estimates of clearance, volume of distribution and oral bioavailability were 2.88 l·h⁻¹, 169.3 l and 58%, respectively. The allele frequencies of cytochrome P450 gene (CYP) 2C19*2, *3 and *17 variants were 29.2%, 5.2% and 0.5%, respectively. CYP2C19 genotype had a significant effect on the clearance. Voriconazole trough concentrations in poor metabolizers were significantly higher than in intermediate metabolizers and extensive metabolizers alike. The volume of distribution increased with increased body weight. The oral bioavailability was substantially lower within 1 month after transplantation but increased with postoperative time. Dosing simulations indicated that during the early postoperative period, poor metabolizers could be treated with 150 mg intravenously or 250 mg orally twice daily; intermediate metabolizers with 200 mg intravenously or 350 mg orally twice daily; and extensive metabolizers with 300 mg intravenously twice daily.

Conclusions

Using a combination of CYP2C19 genotype and postoperative time to determine the initial voriconazole dosing regimens followed by therapeutic drug monitoring could help to advance individualized treatment in renal transplantation patients with invasive fungal infections.

What is Already Known about this Subject

• Invasive fungal infection is a devastating disease in patients who have undergone kidney transplantation.

• Voriconazole, with a narrow therapeutic range, exhibits significant interindividual variability.

• Data on the pharmacokinetics and dosage optimization of voriconazole in renal transplant recipients are still very limited.

What this Study Adds

• A population pharmacokinetic model of voriconazole in renal transplant recipients was developed. The cytochrome P450 gene (CYP2) C19 genotype, postoperative time and body weight were identified significantly to affect the pharmacokinetic parameters of voriconazole.

• Model‐based simulations stratified by CYP2C19 phenotype and postoperative time were performed for dosage optimization.

Introduction

The consequence of long‐term administration of potent immunosuppressive agents and broad‐spectrum antibiotics is the development of invasive fungal infections (IFIs). These are one of the leading causes of early death in renal transplant recipients (RTRs). In particular, invasive pulmonary aspergillosis is a devastating disease, with a 6‐week mortality of 31% in postrenal transplantation patients 1.

Voriconazole (VRC) is a second‐generation triazole with potent broad‐spectrum antifungal activity. In 2016, the Infectious Diseases Society of America guidelines recommended VRC as the agent of choice for invasive aspergillosis, and an alternative therapy for candidaemia. It can also be used for prophylaxis against invasive aspergillosis in high‐risk patients 2, 3. However, VRC has numerous adverse events, such as neurotoxicity, hepatotoxicity and visual disturbances 4. Moreover, a narrow therapeutic range, significant interindividual variability and nonlinear pharmacokinetics further complicate its optimal use 5, 6, which could affect the prognosis of RTRs with IFIs. Therefore, it is important to optimize the VRC therapy, to improve clinical outcomes for RTRs with IFIs.

A growing number of studies have shown that VRC exhibits significant exposure–response relationships for efficacy and toxicity; the target trough concentrations (C_min) have been identified 7, 8, 9. Hence, therapeutic drug monitoring (TDM) is now widely recommended to optimize clinical outcomes 2, 10, 11, 12. VRC is known to be metabolized principally by cytochrome P450 (CYP) 2C19 and, to a lesser extent, by CYP3A4 and CYP2C9 enzymes 13. http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=262#1328 polymorphisms are believed to be a crucial determinant of the extensive pharmacokinetic variability of VRC 14, 15. It has been suggested that the optimization of VRC therapy should be based on CYP2C19 genotype 5, 8, 16, 17. Other sources of variability include age, liver function, comedications and inflammation 6, 13, 18. Population pharmacokinetic (PPK) analyses are widely used to characterize the pharmacokinetic profiles of VRC, as well as recommend doses in clinical practice based on the model simulations. At present, published data on VRC focus mainly on haematopoietic stem cell transplantation, liver or lung transplantation, and patients in the intensive care unit. However, information on the pharmacokinetics and dosage optimization of VRC in RTRs is still very limited.

The aims of the present observational study were to characterize the pharmacokinetics of VRC in RTRs and to identify factors significantly affecting pharmacokinetic parameters by using PPK analysis, and to recommend optimal dosing regimens for patients who have developed IFIs, in specific clinical scenarios based on the model simulations.

Methods

Patients and data collection

A single‐centre prospective clinical study was conducted from March 2016 to January 2017 in the Department of Urological Organ Transplantation of the Second Xiangya Hospital, Central South University. The study was approved by the Ethics Research Committee of the Second Xiangya Hospital, Central South University (yxlb‐lcys‐201 501). A signed informed consent to participate in the study was obtained from all patients. All RTRs receiving intravenous or oral VRC for the prevention or treatment of IFIs during hospitalization were eligible to enrol in the study. Routine TDM and CYP2C19 genotyping were performed. The initial dose and route of administration were determined based on the VRC manufacture package insert. The subsequent dose was adjusted by the surgeons according to clinical response and TDM results. All patients received tacrolimus or cyclosporine as their primary immunosuppressive agent. The exclusion criteria were as follows: (i) age <18 years; (ii) absence of VRC plasma concentration or CYP2C19 genotyping; (iii) concomitant drugs known to have a large effect on VRC pharmacokinetics (e.g. rifampin, a strong inducer of CYP2C19); and (iv) incomplete important dosing information or clinical data.

Patients' medical records were reviewed individually using a standardized data collection form. From the start of VRC therapy, researchers tracked and accurately recorded the dosing information (indication for therapy, route of administration and dose, administration time and sampling time) and the concomitant drugs taken during VRC therapy [proton pump inhibitors (PPIs), including omeprazole, esomeprazole, pantoprazole, lansoprazole and ilaprazole, and glucocorticoid (methylprednisolone)]. In addition, demographic data (age, gender and weight), laboratory test results (blood, liver, and kidney function index) and the time at which transplantation was completed were collected and blood samples were obtained to analyse for CYP2C19 alleles (CYP2C19*2, CYP2C19*3, and CYP2C19*17).

Blood sampling and analytical assays

Blood samples were taken at 0.5, 1, 1.5, 2, 4, 6, 8 and 12 h (every patient had at least 2–3 samplings) after receiving the first intravenous or oral VRC dose. Plasma concentrations were considered to be at steady state on day 2 (or later) following loading doses, or on day 5 (or later) of treatment without loading doses. At steady state, C_min samples were collected 30 min before the next dose. As described in our previous paper 19, VRC plasma concentrations were measured by automatic two‐dimensional high‐performance liquid chromatography (Demeter Instrument Co., Ltd., Hunan, China). The first dimensional chromatographic column was an ASTON FRO C₁₈ (100 mm × 3.0 mm, 5 μm, ANAX) column, and the second dimensional chromatographic column was an ASTON HD C₁₈ (150 mm × 4.6 mm, 5 μm, ANAX) column.

DNA purification and CYP2C19 genotyping

Blood samples (1–3 ml) for genotype detection were obtained from every enrolled patient. DNA was purified using the E.Z.N.A^® SQ Blood DNA Kit II (Omega Bio‐Tek, Inc., Norcross, GA, USA) method. CYP2C19 genotyping was carried out by using Sanger dideoxy DNA sequencing method with the ABI 3730XL DNA Analyzer (ABICo.; BioSune Biotechnology Co., Ltd, Shanghai, China).

CYP2C19 phenotypes were classified into five categories: ultrarapid metabolizer (UM, CYP2C19*17/*17), rapid metabolizer (RM, CYP2C19*1/*17), extensive metabolizer (EM, CYP2C19*1/*1), intermediate metabolizer (IM, CYP2C19*1/*2, CYP2C19*1/*3, CYP2C19*2/*17) and poor metabolizer (PM, CYP2C19*2/*2, CYP2C19*2/*3, CYP2C19*3/*3) 20.

PPK analysis

A nonlinear mixed‐effects model (NONMEM) was developed using the first‐order conditional estimation‐extended least‐squares (FOCE ELS) method and performed using Phoenix NLME software (Version 7.0, Pharsight, A Certara Company, Princeton, NJ, USA).

Structural model

One‐ or two‐compartment models with first‐order oral absorption and linear or nonlinear (Michaelis–Menten) elimination after the intravenous or oral administration were investigated to determine the optimal structural model. The clearance (CL), volume of distribution (V) and oral bioavailability (F) of VRC were characterized and estimated. The absorption rate constant was fixed at 1.1 h⁻¹ based on the literature report 21.

Statistical model

The interindividual variability in VRC pharmacokinetic parameters was described using exponential models: P_ij = P_pop × exp(η_ij), where P_ij is the jth pharmacokinetic parameter estimation of the ith individual, P_pop is the population typical value of the jth parameter and η_ij is a random variable distributed with a mean of zero and a variance of ω². Residual variability was evaluated by comparing the following models:

Additive error model:

C_obs = C_pred + ε

Proportional error model:

C_obs = C_pred × (1 + ε)

Combined error model:

C_obs = C_pred × (1 + ε) + ε’

Exponential error model:

C_obs = C_pred × exp(ε)

In the above four equations, C_obs and C_pred are the observed and predicted concentrations, and ε and ε’ are random variables distributed with a mean of zero and variances of σ² and σ’², respectively. The optimal residual variability model was determined by balancing the objective function value (OFV), percentage coefficient of variation (CV%) and RetCode value.

Covariate model

The correlations between the pharmacokinetic parameters and the covariates were preliminarily inspected by the linear plots (continuous variables) and the box plots (categorical variables). Subsequently, 17 covariates were screened by forward addition followed by backward elimination steps (stepwise method) to establish the full model and the final model, including gender; age; body weight (WT); white blood cell (WBC) count; haemoglobin (HGB) level; platelet (PLT) count; alanine aminotransferase (ALT), aspartate aminotransferase (AST), albumin (ALB), total bilirubin (TBIL), direct bilirubin (DBIL) and serum creatinine (SCr) levels; CYP2C19 phenotype (PM, IM and EM); and postoperative time (POT), as well as concomitant medications including lansoprazole, ilaprazole and methylprednisolone.

A significant covariate was retained in the final model when the following criteria were met: (i) a decrease in OFV >3.84 (P < 0.05) was included in the forward addition steps and an increase in OFV >10.83 (P < 0.001) was significant in the backward elimination steps (approximate to χ² distribution, χ² _0.05,1 = 3.84; χ² _0.001,1 = 10.83); (ii) clinical plausibility for the parameter–covariate relationship; and (iii) the 95% confidence interval (CI) for the parameter estimates did not include zero.

Model evaluation and validation

Goodness‐of‐fit (GOF) plots were used to evaluate the adequacy of fitting. The stability of the final model and the precision of parameter estimation were assessed by the bootstrap method. One thousand resamples from the original data were performed. The parameters (median and 95% CI) obtained from the bootstrap analysis were compared with the estimates of the final model.

Model‐based simulations

In order to recommend optimal dosing regimens for RTRs with IFIs in specific clinical scenarios, various dosing simulations based on the established final model were performed using Phoenix NLME Version 7.0. The parameter estimates of VRC CL, V and oral F, as well as their interindividual variability, were used to simulate steady‐state C_min with 1000 replicates following different intravenous or oral maintenance doses. A VRC C_min of at least 2 μg·ml⁻¹ for critically ill patients with a poor prognosis was recommended by the British Society for Medical Mycology 11. Considering that IFI is a devastating infection for RTRs, the lower end of the therapeutic C_min range was targeted at 2 μg·ml⁻¹ for dosing simulations. The supratherapeutic threshold was defined at 6 μg·ml⁻¹, based on the recent report 22.

Statistical analysis

The Kruskal–Wallis test and Dunn–Bonferroni post hoc test were applied to check the differences in VRC C_min among CYP2C19 phenotype groups, as well as among different periods after the renal transplantation. P < 0.05 was considered statistically significant. Statistical analysis was performed using SPSS version 22.0 (IBM Corporation, Armonk, NY, USA).

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 23, and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 24.

Results

Patients' demographics and dose characteristics

A total of 129 patients were initially enrolled in the study; 23 patients were excluded and 106 patients included, with 345 VRC concentrations obtained. C_min was measured in 201 samples [median of two per patient (range 1–8)] from 93 patients. The median [interquartile range (IQR)] of the 201 C_min values was 2.32 (2.01) μg·ml⁻¹. Even though TDM was performed, a broad variability in VRC C_min was still observed – i.e. 15.4% of C_min values were lower than 1.0 μg·ml⁻¹, and 9.5% were higher than 4.5 μg·ml⁻¹. The subjects were divided into four groups according to their CYP2C19 phenotype [12 PMs (*2/*2, *2/*3), 49 IMs (*1/*2, *1/*3), 44 EMs (*1/*1) and one RM (*1/*17)]. No UM patients were found. The allele frequencies of CYP2C19*2, *3 and *17 variants were 29.2, 5.2 and 0.5%, respectively. As only one RM patient (with 3 C_min) was observed, the RM group was not included in the analyses. Patients' demographics and clinical characteristics are summarized in Table 1.

Table 1.

Demographic and clinical data

Characteristic	Value [Mean ± SD (median, range)]
Gender (male/female) (n)	84/21
Age (years)	36 ± 9 (36, 18–58)
Weight (kg)	56.9 ± 10.5 (56.1, 38.9–87.5)
WBC (109·l−1)	7.33 ± 3.96 (6.96, 0.06–22.21)
HGB (g·l−1)	103 ± 24 (99, 53–175)
PLT (109·l−1)	186 ± 71 (176, 41–505)
ALT (U·l−1)	27.2 ± 48.0 (15.3, 3.3–704.4)
AST (U·l−1)	24.6 ± 30.2 (18.1, 3.7–433.4)
ALB (g·l−1)	33.2 ± 3.6 (33.3, 23.7–51.7)
TBIL (μmol·l−1)	8.5 ± 5.1 (7.1, 2.0–34.5)
DBIL (μmol·l−1)	3.7 ± 3.5 (2.7, 0.2–25.5)
SCr (μmol·l−1)	168.6 ± 108.7 (144.7, 51.7–1328.0)
RM; EM; IM; PM [n (%) of patients] a	1; 44; 49; 12 (0.9; 41.5; 46.2; 11.3)
Postoperative time [n (%) of patients]
≤1 month	33 (31.4)
1–6 months	35 (33.3)
6–12 months	22 (21.0)
> 1 year	15 (14.3)
Indication for therapy [n (%) of patients]
Antifungal prophylaxis	29 (27.6)
Empirical therapy	76 (72.4)
Route of administration [n (%) of patients]
Oral	28 (26.7)
Switch from intravenous to oral	77 (73.3)
Concomitant medication [n (%) of patients]
Omeprazole	7 (6.7)
Esomeprazole	9 (8.6)
Pantoprazole	6 (5.7)
Lansoprazole	30 (28.6)
Ilaprazole	25 (23.8)
Methylprednisolone	81 (77.1)

ALT, alanine aminotransferase; ALB, albumin; AST, aspartate aminotransferase; DBIL, direct bilirubin; EM, extensive metabolizer; HGB, haemoglobin; IM, intermediate metabolizer; PLT, platelets; PM, poor metabolizer; RM, rapid metabolizer; SCr, serum creatinine; SD, standard deviation; TBIL, total bilirubin; WBC, white blood cell

^aData are available for 106 patients. As only one patient was observed, the RM group was not included in the population pharmacokinetic analyses

The differences in VRC C_min among CYP2C19 phenotype groups or in different periods after the renal transplantation were tested statistically in 92 patients. As Table 2 and Figure 1A show, the C_min of PMs was significantly higher than that of both IMs (P = 0.015) and EMs (P < 0.001) but there was no significant difference in the C_min between IMs and EMs (P = 0.228). In addition, the C_min of patients at the early postoperative stage (POT within 1 month) was significantly lower compared with those at 1–6 months, 6–12 months and over 1 year after the operation (P = 0.006). No significant differences in the C_min among the latter three POT groups were found (Figure 1B).

Table 2.

Values (median ± interquartile range) of C_min in different CYP2C19 gene status in 93 RTRs

	CYP2C19 phenotype
	RM (n = 1) a	EM (n = 41)	IM (n = 40)	PM (n = 11)
CYP2C19 genotype (n)	*1/*17 (1)	*1/*1 (41)	*1/*2 (34)	*2/*2 (7)
			*1/*3 (6)	*2/*3 (4)
Number of Cmin (n)	3	97	81	20
Cmin (μg·ml−1) b	1.90 ± 0.30	2.19 ± 2.01	2.32 ± 2.05	3.86 ± 2.96

C_min, trough concentration; CYP2C19, gene encoding cytochrome P450 2C19; EM, extensive metabolizer; IM, intermediate metabolizer; PM, poor metabolizer; RM, rapid metabolizer; RTR, renal transplant recipient

^aThe RM group was not included in the statistical analyses

^bKruskal–Wallis test, P < 0.001

Figure 1.

Distribution of voriconazole (VRC) trough concentrations (C_min) among cytochrome P450 (CYP) 2C19 phenotype groups (A) and at various periods after the renal transplantation (B). The Dunn–Bonferroni correction was made for pairwise comparisons following the Kruskal–Wallis test. Data are expressed as the median ± interquartile range

PPK analysis

The PPK analysis was based on 342 VRC plasma concentrations from 105 patients. A one‐compartment model with first‐order absorption and elimination was sufficient to characterize VRC pharmacokinetics. The interindividual variability and the residual variability were described by the exponential model and the additive error model, respectively.

The full model contained gender and WT as significant covariates for V; CYP2C19 phenotype as a significant covariate for CL; and POT and TBIL as significant covariates for F. Following the backward elimination steps, the significant covariates – namely, WT, CYP2C19 phenotype and POT – were incorporated into the final model: V(l) = 169.27 × (WT/56.1)^1.30 × exp(η_V); CL (l·h⁻¹) = 2.88 × exp(PM = 0) × exp[0.45 × (IM = 1)] × exp[0.80 × (EM = 1)] × exp(η_CL); F (%) = 58 × exp(POT₁ = 0) × exp[0.43 × (POT₂ = 1)] × exp[0.57 × (POT₃ = 1)] × exp[0.57 × (POT₄ = 1)] × exp(η_F), in which POT₁, POT₂, POT₃ and POT₄ stand for postoperative time ≤1 month, 1–6 months, 6–12 months and >1 year, respectively. The population parameter estimates (including V, CL, F and the interindividual variability and residual variability) of the base model and the final model are presented in Table 3.

Table 3.

Population parameter estimates of the base model and the final model

Parameter	Base model	Final modela	Bootstrap
	Estimate	Estimate	Median	95% CI
θV	176.97	169.27	169.70	145.33, 197.08
θCL	5.30	2.88	2.89	2.07, 3.95
θF	0.85	0.58	0.58	0.43, 0.76
θ1	–	1.30	1.28	0.53, 2.11
θ2	–	0.45	0.45	0.19, 0.73
θ3	–	0.80	0.79	0.51, 1.12
θ4	–	0.43	0.43	0.05, 0.80
θ5	–	0.57	0.56	0.28, 0.85
θ6	–	0.57	0.57	0.31, 0.85
ωV	0.43	0.39	0.39	–
ωCL	0.49	0.42	0.42	–
ωF	0.35	0.22	0.21	–
σ	0.58	0.57	0.56	–

CI, confidence interval; CL, clearance; EM, extensive metabolizer; F, oral bioavailability; IM, intermediate metabolizer; PM, poor metabolizer; POT, postoperative time; V, volume of distribution

^aThe final model: V(l) = θ_V × (WT/56.1)^θ₁ × exp(η_V); CL (l·h⁻¹) = θ_CL × exp(PM = 0) × exp[θ₂ × (IM = 1)] × exp[θ₃× (EM = 1)] × exp(η_CL); F = θ_F × exp(POT₁ = 0) × exp[θ₄ × (POT₂ = 1)] × exp[θ₅× (POT₃ = 1)] × exp[θ₆× (POT₄ = 1)] × exp(η_F); θ_V, θ_CL, θ_F: typical population values of pharmacokinetic parameters; θ₁, θ₂, θ₃, θ₄, θ₅, θ₆: correction factors of parameters; η_V, η_CL, η_F: interindividual variation of parameters; POT₁, POT₂, POT₃, and POT₄: postoperative time ≤1 month, 1–6 months, 6–12 months, and >1 year, respectively

Model evaluation and validation

The OFV decreased by 53.42 in the final model compared with the base model, indicating that the incorporated covariates WT, CYP2C19 phenotype and POT contributed to substantial model improvement. The population‐predicted concentrations (PRED) strongly deviated from the observed VRC concentrations in the base model but the PRED agreed with the detected values (DV) in the final model in the scatter plots of DV vs. PRED (Figure 2A1, A2). In addition, the conditional weight residuals (CWRES) in the final model were more uniformly distributed within the accepted range (y = ± 2). By contrast, the two average CWRES trend lines of the base model slightly extended outward at the end (Figure 2B1, B2). In brief, the final model was significantly improved in terms of the GOF and allowed more accurate prediction of VRC levels.

Figure 2.

The goodness‐of‐fit plots of the base model (left) and the final model (right): detected concentrations vs. population‐predicted concentrations (DV‐PRED) scatterplots (A1, A2) and condition weighted residuals vs. population‐predicted concentrations (CWRES‐PRED) scatterplots (B1, B2). Both the blue and red lines are CWRES trend lines reflecting the trend of the residual distribution, where the blue line is obtained by locally weighted regression (LOESS), and the red line is obtained by taking the absolute value and its mirror image

Bootstrap method, based on data resampling technique, is recommended for the internal validation of the PPK model. In the bootstrap for the final model, all the 1000 replications ran successfully. The population parameter estimates were close to the median values from bootstrapping analysis and fell within 95% CIs (Table 3), suggesting that the final model was robust and accurate.

Dosing simulations for optimal VRC doses

The probability of attaining the VRC target C_min with 150–300 mg twice daily intravenous (2‐h infusion) dosing regimens stratified by CYP2C19 phenotype is shown in Table 4. According to the results, 150 mg intravenously twice daily was adequate for the PM patients to reach the VRC therapeutic range (90.9% of patients would reach the lower target C_min of 2 μg·ml⁻¹ and 6.3% of patients would exceed the upper target of 6 μg·ml⁻¹). For IM patients receiving VRC 200 mg intravenously twice daily, 81.5% of patients would reach the lower target of the therapeutic range and 3.5% would exceed the upper target. The EM patients were predicted to require a higher dose, with 300 mg intravenously twice daily; this would result in 80.3% of patients attaining the lower target, meeting their therapeutic need.

Table 4.

Probability of achieving VRC target C_min with 150–300 mg twice daily intravenous (2‐h infusion) dosing regimens in different CYP2C19 phenotype groups

Cmin (μg·ml−1)	Probability (%), stratified by CYP2C19 phenotype and intravenous dosing regimen
	PM			IM			EM
	150 mg	200 mg	250 mg	150 mg	200 mg	250 mg	150 mg	200 mg	250 mg	300 mg
1.5	97.1	99.2	99.6	81.5	92.5	96.6	53.8	74.3	84.8	90.9
2	90.9	97.0	98.7	61.2	81.5	90.6	29.8	54.0	69.8	80.3
4	34.7	64.6	82.3	7.1	24.2	44.6	1.2	6.5	17.1	29.5
5	15.4	41.5	64.7	1.8	9.7	24.3	0.2	1.7	6.4	14.2
6	6.3	23.9	46.3	0.4	3.5	11.7	0.0	0.5	2.3	6.4

The percentages represent the probabilities of obtaining C_min above the reported targets. Probability results are based on dosing simulations of VRC steady‐state pharmacokinetics in different dosing regimens. C_min, trough concentration; CYP2C19, gene encoding cytochrome P450 2C19; EM, extensive metabolizer; IM, intermediate metabolizer; PM, poor metabolizer; VRC, voriconazole

The oral dosing simulations with 150–500 mg twice daily in different CYP2C19 phenotype groups (Table 5) indicated that within 1 month after transplantation, for the PM patients, 250 mg orally twice daily was adequate to reach the VRC therapeutic range (88.0% of patients would reach the lower target and 6.6% of patients would exceed the upper target). For the IM patients, about 80% of patients receiving a VRC dose of 350 mg orally twice daily would reach the lower target of therapeutic range and 6.1% would exceed the upper target.

Table 5.

Probability of achieving VRC target C_min with 150–500 mg twice daily oral dosing regimens in different CYP2C19 phenotype groups (postoperative time within 1 month)

Cmin (μg·ml−1)	Probability (%), stratified by CYP2C19 phenotype and oral dosing regimen
	PM				IM				EM
	150 mg	200 mg	250 mg	300 mg	250 mg	300 mg	350 mg	400 mg	350 mg	400 mg	450 mg	500 mg
1.5	73.9	89.1	96.3	97.9	76.7	85.8	91.5	94.4	73.0	81.0	84.6	89.2
2	47.4	73.5	88.0	93.4	56.0	70.3	80.2	86.0	53.6	63.8	70.8	77.7
4	3.9	15.8	31.9	48.1	8.5	16.8	26.9	37.8	9.0	15.4	20.9	29.3
5	0.9	6.1	14.6	28.1	3.0	7.0	13.4	21.1	3.3	6.4	9.8	16.0
6	0.4	2.5	6.6	15.6	0.9	2.7	6.1	11.4	1.3	2.6	4.7	8.6

EM, extensive metabolizer; IM, intermediate metabolizer; PM, poor metabolizer; VRC, voriconazole

Discussion

The pharmacokinetic data on VRC in RTRs are limited. In the current study, a PPK model of VRC in RTRs was developed. CYP2C19 genotype, POT and WT were identified to have a significant effect on the pharmacokinetic parameters of VRC. Furthermore, model‐based simulations stratified by CYP2C19 phenotype and POT were performed for dosage optimization.

The population estimate of CL (2.88 l·h⁻¹) in the present study is similar to the reported value for lung transplant recipients (3.45 l·h⁻¹) 25 but lower than that of other patients with IFIs 9, 15, 21. The lower CL possibly results from the unrecovered renal function of the RTRs, with a lower baseline creatinine CL (mean ± standard deviation) of (47.3 ± 23.4) ml·min⁻¹. The CL was significantly affected by CYP2C19 genotype (EMs > IMs > PMs). The empirical Bayesian estimates from the final model showed that the median CL of the EMs, IMs and PMs was 6.67, 4.51 and 2.86 l·h⁻¹, respectively, indicating that CYP2C19 polymorphisms modulate CYP2C19 enzyme activity, and as a result affect the metabolism of VRC in different patients 8, 26. Similarly, many studies also showed CYP2C19 polymorphisms have an extensive effect on VRC C_min. Comparable with previous findings 8, 16, 17, C_min was increased in the PMs as compared with that in both the IMs and the EMs in the present study. Moreover, the distribution of CYP2C19 varies in different ethnic populations. As the CYP2C19 frequency table 27 shows, CYP2C19*2 and *3 are seen frequently among East Asians (29.0 and 8.3%, respectively), but the frequency of CYP2C19*17 in this population is less than 2%. By contrast, the prevalence of CYP2C19*3 in Caucasians is relatively low (<1%), but that of CYP2C19*17 is high, at 21.3%. The allele frequencies of CYP2C19*2, *3 and *17 variants (29.2, 5.2 and 0.5%, respectively) in our study were roughly in line with the data for East Asians mentioned above, indicating that Chinese people are susceptible to adverse reactions due to slow metabolism. Considering that CYP2C19 polymorphisms contribute largely to the wide interindividual variability of VRC exposure, refining the initial dosing regimen based on CYP2C19 genotype is recommended, to optimize clinical outcomes. This is supported by a recent study showing that genotype‐directed dosing allows paediatric patients to reach the target range much faster than the traditional dosing regimens 28.

According to empirical Bayesian estimates from the final model, the VRC oral F for RTRs within 1 month after the operation was 57.6% (median), which is substantially lower than in healthy volunteers (96%) 29 as well as other nontransplant populations 6, 15, 30. However, the F value increased with the POT and reached 89% in patients 1–6 months after transplantation, and slightly elevated but became stable 6 months after transplantation. A significantly lower F (45.9%) during the early postoperative stage has also been reported in the lung transplant recipients by Han et al. 25. In their study, the F value significantly increased with the POT and reached the maximum levels within 1 week. The lower F may be due to gastrointestinal dysfunction after the transplant surgery, whereas the recovery of gastrointestinal absorption over the POT contributes to the increased F.

In consideration of the substantially lower F, higher oral doses are required for RTRs during the early postoperative period. Oral doses of 250 mg and 350 mg twice daily are needed for the PM patients and the IM patients, respectively. However, the EM patients who are at higher risk of therapy failure, require an intravenous dose of 300 mg twice daily to ensure complete absorption to sustain the target therapeutic range. This recommendation is based on the results of the dosing simulation at a high oral dose of 500 mg twice daily. Simulation results showed that the EM patients had an accepted probability (8.6%) of exceeding the upper target C_min but the probability of obtaining the lower target was 77.7%. Additionally, as the oral dosage form is less invasive and cheaper compared with intravenous dosage, oral/intravenous interchange programmes should be considered as soon as possible when all factors affecting VRC pharmacokinetics have been taken into account 31. Similarly, our results suggest that switching from intravenous administration to therapeutically equivalent oral doses may be feasible for the patients 1 month after transplantation if patients are clinically stable.

WT‐adjusted dosing in adults is not supported as several studies have shown no association between WT and VRC pharmacokinetics 6, 8, 21. The V increased with increased WT in our study, which is consistent with the result of Han et al. 25. The latter authors found that the variability of VRC pharmacokinetic profiles did not improve with a WT‐adjusted dose as compared with a fixed dose. In addition, high VRC exposure was observed in obese patients dosed according to their actual WT 32, 33. Therefore, dosing based simply on the actual WT may not be feasible, especially for obese patients. An adjusted WT dosing strategy has been proposed 32, 33, 34 but needs further study.

In recent years, CYP2C19 inhibitor PPIs have been confirmed to affect VRC pharmacokinetics, but the degree depends on the type of PPI 35, 36. The effect of glucocorticoids on VRC exposure remains controversial. Some studies have found that they reduced the VRC concentrations, whereas others have made contradictory findings 35, 37, which may have been due to different types and doses of glucocorticoids. Similarly to our retrospective study 19, neither PPIs (including lansoprazole and ilaprazole) nor glucocorticoids (methylprednisolone) seemed to affect the VRC pharmacokinetic parameters in the current study. It should be mentioned that omeprazole, pantoprazole and esomeprazole were not included in the PPK analyses owing to limited sample size. As concomitant use of PPIs/glucocorticoids and VRC is relatively common in RTRs, surgeons should be aware of the potential impact on VRC concentrations. In addition, the present study did not evaluate the drug–drug interactions (DDIs) between VRC and immunosuppressants. However, VRC has been identified to increase the exposure to coadministered cyclosporine, tacrolimus, sirolimus and everolimus 38. The strong inhibition of CYP3A4 by VRC in the small intestine and liver is likely to be the main mechanism of these DDIs. The magnitude of the interaction between VRC and tacrolimus was found to be affected by CYP3A5 polymorphisms 39 as well as by CYP2C19 polymorphisms that result in different VRC exposure 39, 40. It was shown that VRC could increase intracellular cyclosporine concentrations via inhibiting the P‐glycoprotein‐mediated transport of cyclosporine 41. It remains difficult to formulate general dosing recommendations owing to the wide variability in the effect of VRC on cyclosporine/tacrolimus concentrations. Further studies are required to quantify the effects of VRC on immunosuppressants and recommend adequate dosing adjustment strategies for RTRs.

Common monitoring parameters for liver function, including ALT, AST, DBIL and TBIL levels, had no significant effect on VRC pharmacokinetics in the present study, which was inconsistent with our previous finding that the CL decreased with elevated AST levels 19. This discrepancy may be attributed to different sample sizes and investigated covariates. Nevertheless, one study reported that decreased liver function resulted in a significant reduction in VRC metabolism 42, so it is essential to monitor VRC concentrations closely in patients with hepatic dysfunction. Furthermore, the interindividual variability of V, CL and F was as high as 39%, 42% and 22%, respectively, indicating that the covariates introduced into the final model explain only part of the variability in VRC pharmacokinetics in RTRs. Other factors, such as diet 29 and inflammation, reflected by C‐reactive protein concentrations 18, 43, were not tested in the present study but may have had an impact on VRC pharmacokinetics. These factors should be explored further, to clarify the remaining variability. Meanwhile, a timely and optimized dosing regimen, based on closely monitored drug concentrations, is crucial to ensure clinical efficacy and safety.

There were several limitations to the present study. Firstly, the RM group was not included in the analyses as only one such patient was observed, and no UM patients were detected. Nevertheless, the CYP2C19*17 allele has been demonstrated to be an important determinant of the CL 5. Studies to determine the effect of *17 carriers on VRC pharmacokinetics are needed in the future. Secondly, the study did not evaluate the relationships between efficacy and toxicity and VRC exposure. Linking the pharmacokinetic results to clinical outcomes, in order to target an appropriate therapeutic range, is critical but challenging for VRC therapy, and further research is warranted. Additionally, given the potential complications and the complex medications that RTRs take during the early postoperative stage, the pharmacokinetic profile of VRC during this period should be further characterized. Despite all the limitations, the current study may provide the theoretical basis for individualized VRC therapy and serve as a good reference for RTRs.

Conclusion

The CL of VRC is strongly affected by CYP2C19 polymorphisms and the F is substantially lower within 1 month after kidney transplantation. Based on the dosing simulations during the early postoperative period, the regimens recommended are: 150 mg intravenously or 250 mg orally twice daily for PM patients; 200 mg intravenously or 350 mg orally twice daily for IM patients; and 300 mg intravenously twice daily for EM patients. With careful monitoring, oral/intravenous interchange may be feasible for patients 1 month post‐transplantation. Using the combination of CYP2C19 genotype and POT to determine the initial VRC dosing regimens, followed by TDM, will help to advance individualized VRC therapy.

Competing Interests

There are no competing interests to declare.

The authors thank Yan‐qin Wu and Xiao‐pei Tong for their assistance in the DNA extraction and detection, and Lin Li, Wen‐ding Hui, Jia‐min Wu, Ying Li, Ping Yang, Zhou‐qi Tang, Shan‐shan Gui, Ni Wu and Wen‐dan Mao for their assistance in the collection of clinical data and blood samples. They are very grateful to Professor Hoan Linh Banh from the University of Albert (Canada) and Chun Wu for reviewing the final manuscript. The study was supported by the Project of New Clinic Techniques of Central South University, China (No. 2015053).

Contributors

X.‐b.L. and Z.‐w.L. participated in the DNA extraction and detection, data collection and manuscript preparation, and conducted the statistical analyses. M.Y. and F.‐h.P. designed the study protocol and participated in the manuscript preparation and editing. X.‐b.X., S.‐j.Y. and G.‐b.L. helped with the medical management of study patients. W.L. performed the statistical analyses and graphic review. F.W. helped with the blood sampling and concentration analysis. D.‐x.X. and P.X. participated in data extraction and patient chart review. B.‐k.Z. managed the study database. All authors read and approved the final manuscript.

Lin, X. , Li, Z. , Yan, M. , Zhang, B. , Liang, W. , Wang, F. , Xu, P. , Xiang, D. , Xie, X. , Yu, S. , Lan, G. , and Peng, F. (2018) Population pharmacokinetics of voriconazole and CYP2C19 polymorphisms for optimizing dosing regimens in renal transplant recipients. Br J Clin Pharmacol, 84: 1587–1597. doi: 10.1111/bcp.13595.