Variability of Voriconazole Plasma Concentrations after Allogeneic Hematopoietic Stem Cell Transplantation: Impact of Cytochrome P450 Polymorphisms and Comedications on Initial and Subsequent Trough Levels

Abstract

Voriconazole (VRC) plasma trough concentrations (C_min) are highly variable, and this could affect treatment efficacy and safety in patients undergoing allogeneic hematopoietic stem cell transplantation (AHSCT). We aimed to describe the intra- and interindividual variation of VRC C_min throughout the course of VRC therapy and to identify the determinants of this variation. Clinical data, medications, and VRC C_min (n = 308) of 33 AHSCT patients were retrospectively collected. Cytochrome P450 (CYP450) genotypes of CYP2C19, CYP3A4, and CYP3A5 patients were retrospectively determined before allografting, and a combined genetic score was calculated for each patient. The higher the genetic score, the faster the metabolism of the patient. The VRC C_min inter- and intraindividual coefficients of variation were 84% and 68%, respectively. The VRC dose (D) was correlated to VRC C_min (r = 0.412, P < 0.0001) only for oral administration. The administration route and the genetic score significantly affected the initial VRC C_min. Considering oral therapy, patients with a genetic score of <2 had higher initial VRC C_min/D than patients with a genetic score of >2 (P = 0.009). Subsequent VRC C_min remained influenced by the genetic score (P = 0.004) but were also affected by pump proton inhibitor comedication (P < 0.0001). The high variability of VRC C_min in AHSCT patients is partially explained by the route of administration, treatment with pump proton inhibitors, and the combined genetic score. This study suggests the interest in combined genetic score determination to individualize a priori the VRC dose and underlines the need for longitudinal therapeutic drug monitoring to adapt subsequent doses to maintain the VRC C_min within the therapeutic range.

INTRODUCTION

Recipients of allogeneic hematopoietic stem cell transplants (AHSCT) are at high risk of developing invasive fungal infections (IFIs), in particular invasive aspergillosis (IA). Voriconazole (VRC) is a broad-spectrum triazole antifungal used as the first-line treatment of IA. Despite adequate care, mortality due to IA remains very high, reaching 56% in AHSCT patients (1). One possible explanation for this high rate of treatment failure could be insufficient exposure to treatment. Indeed, several studies performed in heterogeneous cohorts of patients with underlying hematological malignancy, solid organ transplantation, surgery, or various chronic diseases have suggested an association between low plasma trough concentrations (C_min) and treatment failure. For example, the lack of response to VRC therapy was more frequent in patients with a VRC C_min of ≤1 mg/liter (46%) than in those with a VRC C_min of >1 mg/liter (12%) (2). Similarly, VRC C_min in patients failing to respond to VRC therapy were lower than those in successfully treated patients (3). Finally, a median VRC C_min of >2.2 mg/liter was found to be a strong predictor of microbiological or clinical success (4). Thus, a therapeutic target of between 1 and 4 to 6 mg/liter for VRC C_min has recently been proposed by the British Society for Medical Mycology (5). However, the VRC C_min is frequently below this efficacy threshold in patients suffering from hematologic malignancies (6, 7), notably in AHSCT patients (8).

The VRC C_min exhibits large inter- and intraindividual variabilities (4, 9, 10) that could be related in part to nonlinear pharmacokinetics, metabolization via cytochrome P450 (CYP450), and drug-drug interactions (3, 11). Younger age (3, 6, 12), oral administration of VRC (3), and concomitant medication with enzyme inducers like phenytoin, rifampin, or glucocorticoids (3, 13) are associated with decreased VRC C_min. In this context, therapeutic drug monitoring (TDM) of VRC is of particular interest. A prospective randomized study has recently demonstrated the interest in VRC TDM, since an improvement of treatment response and a decrease of drug discontinuation due to adverse effects was obtained in patients benefiting from TDM (14). Nevertheless, TDM allows dose adjustment a posteriori, whereas it seems crucial to obtain an adequate VRC C_min as soon as possible after the start of treatment (4, 15). Indeed, an initial VRC C_min of ≤0.35 mg/liter appeared to be a strong predictor of mortality (4). Since VRC is extensively metabolized by CYP450 (16), it could be hypothesized that genetics affecting CYP450 have also a strong influence on initial VRC C_min, in addition to drug-drug interactions (3, 11, 13) and route of administration (3). Thus, identification of relevant genetic polymorphisms could be useful when proposing a priori dose adjustment and to reach an adequate VRC C_min from the first determination. The CYPs 2C19, 3A4, 3A5, and to a lesser extent, 2C9, are involved in VRC metabolism (16–18); all of these CYPs exhibit a number of clinically relevant polymorphisms. Genetic studies focusing on the CYP2C19 genotype have already demonstrated a link between the gain-of-function allele *17 and an insufficient VRC C_min (19) or between alleles *2 and *3 and an increased VRC C_min (20). The impact of the CYP2C19 genotype on VRC pharmacokinetics was later confirmed by several population pharmacokinetic studies (13, 21). Moreover, the presence of a CYP2C19 polymorphism has been associated with a higher frequency of out-of-range VRC C_min in lung transplant cystic fibrosis patients (22). Conversely, the variability of CYP3A activity on VRC pharmacokinetics has been less studied. In healthy subjects, cotreatment with a CYP3A inhibitor increased VRC exposure, particularly in poor metabolizers of CYP2C19 (23, 24), and CYP3A5 variants did not influence VRC single-dose pharmacokinetics (25). More recently, the T allele of the rs4646437 polymorphism located in an intronic area of the CYP3A4 was associated with higher VRC levels in Chinese patients (26). However, the impact of the new variant CYP3A4*22 (conferring decreased CYP3A4 activity) (27) on VRC pharmacokinetics has never been investigated.

The intraindividual variability of VRC C_min is also considerable (4, 9, 10), requiring regular VRC C_min determinations throughout the duration of VRC treatment (5). Some authors reported no correlation between the first and subsequent VRC C_min (9), whereas others showed that initial and subsequent VRC C_min were correlated only when the initial VRC C_min was greater than 2 mg/liter (10).

This retrospective study aimed to describe the variations of VRC C_min throughout the course of VRC therapy in a cohort of AHSCT patients and to identify the determinants of the variability in the first and subsequent VRC C_min.

MATERIALS AND METHODS

Patients.

This study was conducted in Grenoble University Hospital Center, France. Adult (>18 years old) patients who had received an AHSCT, were treated with VRC, and underwent TDM of VRC between January 2011 and July 2013 were eligible. VRC C_min determined before the allograft or before VRC pharmacokinetics reached a steady state and patients with less than three VRC C_min determinations were excluded. After treatment initiation (without loading dose) or dose adjustment, a 3-day period was considered necessary to obtain a VRC C_min steady state (23). Demographic, biological (transaminase levels), clinical data, records concerning VRC therapy (C_min, daily dose, and route of administration), and concomitant medications were retrospectively collected. All patients gave written informed consent for genetic analysis, sample collection, and use of their data. This retrospective study was performed on residual samples stored in a biological sample collection (DRC-2013-1983) and was approved by the regional Ethics Committee.

Classification of invasive aspergillosis.

IFIs were classified as possible, probable, or proven according to guidelines from the European Organization for Research and Treatment of Cancer/European Invasive Infections Cooperative Group and the criteria of the National Institute of Allergy and Infectious Diseases-Mycoses Study Group (EORTC/MSG) (28).

Measurement of plasma VRC trough concentration.

The plasma VRC trough concentration was determined on samples handled just before the subsequent VRC administration. VRC C_min were measured by a validated liquid chromatography-tandem mass spectrometry method (29). Briefly, after protein precipitation, samples were injected into a 2-dimensional chromatographic system. In the first step, samples were cleaned in a perfusion chromatography column before being eluted and transferred to an analytical column. Finally, compounds were detected by tandem mass spectrometry. The plasma drug standard curve ranged from 0.1 to 20 mg/liter. The therapeutic range was between 1 and 5 mg/liter.

Genotyping.

Genotyping was performed retrospectively on residual samples from routine biological analyses collected before the allograft. DNA was extracted from white blood cells using Macherey-Nagel NucleoSpin Blood L kit (Macherey-Nagel, Hoerd, France). The quality and quantity of DNA were checked with the NanoDrop 2000 spectrophotometer (Thermo Scientific, Illkirch, France).

CYP2C19 genotyping.

CYP2C19 genotyping was performed by direct sequencing after DNA amplification using specific primers for the *2, *3 (associated with low CYP2C19 activity), and *17 (associated with increased CYP2C19 activity) alleles (see Table S1 in the supplemental material). The presence of the wild-type allele CYP2C19*1 was inferred in the absence of CYP2C19*2, CYP2C19*3, and CYP2C19*17. Patients were classified as ultrarapid (URM), extensive (EM), or intermediate (IM) metabolizers according to the established genotype-phenotype relationships proposed by Mega and colleagues (see Table S2 in the supplemental material) (30).

CYP3A genotyping.

The CYP3A4*22 allele (associated with low CYP3A4 activity) and the CYP3A5*1 allele (associated with CYP3A5 expression) were studied using the TaqMan allelic discrimination assay (Life Technologies, Illkirch, France). Patients were classified as intermediate metabolizers (IM) if they had the *22 allele and extensive metabolizers (EM) if they did not.

Determination of genetic score.

To evaluate the impact of both CYP2C19 and CYP3A genotypes, a combined genetic score was calculated for each patient.

First, a specific genetic score was separately attributed to each CYP2C19 and CYP3A genotype cluster, as previously proposed by Goutelle et al. (31) and Moes et al. (32), and both specific scores were then added to obtain the combined genetic score (see Table 2). Briefly, the specific genetic score for each cytochrome was determined as follows: the absence of any polymorphism (genotype *1/*1 for CYP2C19, absence of CYP3A4*22 and presence of CYP3A5*3/*3 for CYP3A) conferred a score equal to 1, whereas the presence of a gain- or a loss-of-function allele inferred, respectively, an increase or a decrease in the genetic score. The fluctuation of the genetic score was ±0.5 when the gain- or the loss-of-function allele was heterozygous or ±1 when the gain- or the loss-of-function allele was homozygous.

TABLE 2.

Calculated combined genetic scores according to CYP2C19 and CYP3A genotypes

^aThe 3 genetic score categories are indicated as follows: a genetic score of <2 is indicated by light gray shading, a genetic score equal to 2 is indicated by medium gray shading, and a genetic score of >2 is indicated by dark gray shading.

The combined genetic score was calculated to reflect the hepatic clearance (Cl_H) of VRC expressed as follows: Cl_H = Cl VRC N-oxidation + Cl VRC hydroxylation (33). VRC N-oxidation depends on CYP2C19 and CYP3A activities, which exhibited similar K_m values in an in vitro study (17), while VRC hydroxylation depends on only the activity of CYP3A isoforms (17). Thus, the same weight was attributed to CYP2C19 and to CYP3A4.

The combined genetic score was expressed in arbitrary units. Patients were classified into three categories of combined genetic score: <2, 2, and >2. A combined genetic score equal to 2 corresponds to normal CYP2C19 and CYP3A activities (or the association of one gain-of-function allele and one loss-of-function allele), whereas a combined genetic score of <2 or >2 reflects diminished or increased CYP2C19 and CYP3A activities, respectively.

Data analysis.

Statistical analyses were performed using Statview software, version 5.0 (SAS Institute, Cary, NC). Quantitative data are expressed as median and 10th and 90th percentiles. Statistical analyses were performed separately on the initial VRC C_min and on the subsequent VRC C_min. Comparisons between groups (male versus female, intravenous [i.v.] therapy versus oral therapy, CYP2C19 and CYP3A phenotypes, and categories of genetic score) were performed with nonparametric tests (Mann-Whitney test to compare 2 groups or Kruskal-Wallis test followed by Bonferroni adjusted t test to compare at least 3 groups). A Wilcoxon rank test was used for the comparison of pairwise series. As the numbers of VRC C_min determinations varied among patients, the frequency of a VRC C_min of ≤1 mg/liter was calculated for each patient. The percentages of VRC C_min of ≤1 mg/liter according to the combined genetic score groups were compared using Fisher's test at the first VRC C_min determination and using the Mann-Whitney test for subsequent VRC C_min. A Spearman test was used to study the correlation between two quantitative variables. The Hardy-Weinberg equilibrium was tested for each polymorphism by the online method of Rodriguez et al. (34). Univariate and multivariate linear regression analyses were used to identify the determinants contributing to the variability of VRC C_min. A P value of <0.05 was considered statistically significant.

RESULTS

Population characteristics.

Thirty-three AHSCT patients were included in our study. Table 1 summarizes their demographic characteristics, hematological diagnoses, and conditioning regimens. All patients received VRC therapy for secondary prophylaxis of IFI. The classification of IFI is detailed in Table 1, but for 2 patients, IFI could not be retrospectively classified.

TABLE 1.

Demographic and clinical characteristics of 33 allogeneic hematopoietic stem cell transplantation patients who experienced therapeutic drug monitoring of voriconazole

Characteristic	Value for characteristic
Demographics
Age, yra	52.2 (35.1–60.1)
Male, no. (%)	20 (61)
BMI, kg/m2a	22.5 (16.6–31.4)
Classification of IFI, no. (%)b
Proven	2 (6)
Probable	15 (46)
Possible	14 (42)
NDc	2 (6)
Hematological diagnoses, no. (%)
Acute myeloid leukemia	22 (67)
Acute lymphoblastic leukemia	5 (15)
Non-Hodgkin lymphoma	3 (9)
Chronic lymphocytic leukemia	2 (6)
Myelodysplastic syndrome	1 (3)
Conditioning regimens, no. (%)
Myeloablative	8 (24)
Non-myeloablative	25 (76)
Matching, no. (%)
Yes	22 (67)
No	11 (33)

^aThese data are expressed as median (10th to 90th percentiles).

^bIFI, invasive fungal infection. All patients were treated for invasive pulmonary aspergillosis, except one patient who had invasive Candida infection.

^cND, not determined.

CYP450 polymorphisms.

Genotyping was performed for 88% of patients (n = 29) (blood samples before the allograft were not available for the other patients). The proportions of the CYP2C19 and -3A genotypes are described in Tables S2 and S3 in the supplemental material. For CYP2C19, 37% (n = 11), 34% (n = 10), and 28% (n = 8) of patients were classified as URM, EM, and IM, respectively. All patients classified as IM for CYP2C19 expressed the *2 allele, while no *3 allele was detected in our cohort. The CYP3A4*22 polymorphism was detected in 24% (n = 7) of patients, and a single patient expressed CYP3A5. The Hardy-Weinberg equilibrium was respected for each allele (CYP2C19*2, χ² = 0.74, P = 0.39; CYP2C19*17, χ² = 0.34, P = 0.56; CYP3A4*22, χ² = 0.55, P = 0.46), and no linkage disequilibrium was detected between the CYP3A4 and CYP2C19 genotypes (P = 0.46). Concerning the combined genetic score, 34% of patients (n = 10) had a genetic score of <2, 34% (n = 10) had a genetic score equal to 2, and 31% (n = 9) had a genetic score of >2 (Table 2).

VRC therapeutic drug monitoring.

A total of 308 VRC C_min were analyzed (Table 3), representing a median of 9 dosages per patient (range, 3 to 25 dosages) for a median follow-up duration of VRC C_min TDM of 94 days (range, 14 to 269 days). The median delay between two VRC C_min determinations was 7 days (range, 4 to 21 days). The descriptions of VRC C_min, proportions of oral VRC therapy, and comedications are detailed in Table 3. The interindividual coefficient of variation (CV) was 84%, whereas the median intraindividual CV was 68% (range, 14 to 185%). Forty-two percent of VRC C_min were out of the therapeutic range, and 85% (n = 28) of patients had at least one inadequate VRC C_min during their course of treatment. Figure 1a illustrates this high intra- and interindividual variability.

TABLE 3.

Voriconazole plasma trough concentrations and concomitant treatments

Parametera	Result for VRC Cmin determinations
	All (n = 308)	Initial (n = 33)	Subsequent (n = 275)
VRC Cmin, mg/liter
Mean	1.6	1.9	1.6
Median	1.3	1.4	1.3
Range	<0.1–9.7	<0.1–9.7	<0.1–8.1
VRC Cmin level, %
≤1 mg/liter	40	33	36
>5 mg/liter	2	6	1
Frequency of VRC oral therapy, %b	82	58	85
Determination during hospitalization, %	29	64	25
VRC dosing, mg/dayc	400 (400–600)	400 (400–617)	400 (400–600)
Concomitant treatment, %
Cyclosporine/tacrolimus/otherd	47/13/40	73/9/18	44/14/42
Enzyme inhibitor(s)e	41	15	44
Glucocorticoid(s)f	29	18	30
Esomeprazole/pantoprazole/rabeprazole/no PPI	40/55/3/2	42/49/3/6	40/55/3/2

^aVRC, voriconazole; C_min, trough concentration; PPI, pump proton inhibitor.

^bOnly the tablet form was used for oral VRC therapy.

^cThese data are expressed as median (10th to 90th percentiles).

^d“Other” included immunosuppressive therapy with mycophenolate or everolimus or the absence of immunosuppressive therapy.

^eEnzyme inhibitors included dihydropyridines, macrolides, verapamil, amiodarone, amitriptyline, citalopram, and metronidazole (all of these drugs are CYP2C19 and/or CYP3A inhibitors).

^fThe glucocorticoids used were prednisone, prednisolone, and methylprednisolone.

FIG 1.

Variability of voriconazole plasma trough concentrations. (a) Serial voriconazole plasma trough concentrations determined in 33 post-AHSCT patients. All patients had at least 3 voriconazole concentration determinations, including several similar voriconazole concentrations for patients 7 and 23. The area shaded in gray represents the therapeutic range. (b) Relationships between voriconazole dose (expressed as mg per day) and plasma trough concentration for both routes of administration. Data are presented as interquartile range (boxes), data range (whiskers), and median (horizontal line). Numbers are indicated above boxes.

During the course of treatment of the 33 patients, 30 VRC dose adjustments for 16 patients were performed, representing 11% of 275 follow-up VRC C_min for 48% of patients. Seventy percent (n = 21) of these dose modifications were realized without changing the VRC route of administration, while 27% (n = 8) resulted from an i.v.-to-oral switch, and 3% (n = 1) were discontinuations of VRC therapy due to an overdose.

Figure 1b shows the relationships between VRC daily dose (expressed in mg per day) and VRC C_min for both routes of administration. Considering oral administration, the VRC C_min presented a CV of 72% and was weakly correlated to the VRC daily dose (r = 0.412, P < 0.0001), while i.v. administration led to a VRC C_min CV of 101%, and no correlation was found with the daily dose (r = 0.096, P = 0.5).

Initial VRC trough concentrations.

The description of the initial VRC C_min is detailed in Table 3. Thirty-nine percent of initial VRC C_min were out of the therapeutic range, with the majority under the lower therapeutic limit: 33% were ≤1 mg/liter, and 6% were >5 mg/liter. VRC i.v. administration was associated with a higher initial VRC C_min compared to oral treatment (i.v., 1.7 mg/liter [range, 0.3 to 5.9 mg/liter], versus oral, 1.0 mg/liter [range, 0.1 to 2.9 mg/liter]; P = 0.03). Sex, age, body mass index (BMI), and proton pump inhibitor (PPI) treatment or enzymatic inhibitor treatment as well as transaminase levels had no impact on the initial VRC C_min (data not shown).

Influence of CYP450 polymorphisms on initial VRC trough concentrations.

The influence of CYP450 polymorphisms was studied on the initial VRC C_min adjusted on VRC dose (VRC C_min/D) to overcome the influence of VRC dose. After oral administration, the CYP2C19 phenotype affected the initial VRC C_min/D (Kruskal Wallis test, P = 0.04), with a higher initial VRC C_min/D in IM compared to URM patients (post hoc test, P = 0.007) (Fig. 2a). Similarly, the presence of the *22 allele for CYP3A4 was associated with a statistically significant increased VRC C_min/D (Fig. 2c). When patients were stratified on CYP2C19 phenotype, expression of CYP3A4*22 was associated with increased initial VRC C_min/D whatever the CYP2C19 phenotype (see Fig. S1 in the supplemental material). In univariate analysis, the combined genetic score was significantly correlated to VRC C_min/D (r = −0.748; P = 0.002). As shown in Fig. 2, a genetic score of >2 was associated with a lower VRC C_min/D compared to a genetic score of <2 (post hoc test, P = 0.009) (Fig. 2e).

FIG 2.

Influence of CYP450 polymorphisms on initial voriconazole plasma trough concentrations adjusted on the dose (VRC C_min/D). (a and b) Influence of CYP2C19 phenotypes on initial VRC C_min/D ratio obtained during VRC oral treatment (PO) (a) and on all initial VRC C_min/D ratios (oral and i.v. [PO + IV]) (b). (c and d) Influence of CYP3A phenotypes on the initial VRC C_min/D ratio obtained during VRC oral treatment (c) and on all initial VRC C_min/D ratios (d). (e and f) Influence of the combined genetic score on the initial VRC C_min/D ratio obtained during VRC oral treatment (e) and on all initial VRC C_min/D ratios (f). Data are presented as interquartile range (boxes), data range (whiskers), and median (horizontal line). Only the significant P value obtained for the post hoc Bonferroni test (<0.0167) is shown. Abbreviations: VRC, voriconazole; CYP, cytochrome; PO, per os; IV, intravenous; URM, ultrarapid metabolizer; EM, extensive metabolizer; IM, intermediate metabolizer.

Studying all initial VRC C_min independently of the VRC route of administration provided similar results: a trend toward an influence of the CYP2C19 phenotype (Kruskal-Wallis test, P = 0.08) (Fig. 2b) and the *22 allele (Mann-Whitney test, P = 0.05) (Fig. 2d) as well as the combined genetic score (Kruskal-Wallis test, P = 0.05) (Fig. 2f) on the first VRC C_min/D.

None of the patients having a combined genetic score of <2 presented an initial VRC C_min of ≤1 mg/liter, while 47% of the patients having a combined genetic score of ≥2 had an initial VRC C_min of ≤1 mg/liter (Fisher's test, P = 0.01).

In a multiple linear regression model integrating the VRC route of administration, initial VRC dose, and combined genetic score, the combined genetic score remained an independent predictor of initial VRC C_min (r² = 0.33; P = 0.04).

Subsequent VRC trough concentrations.

Longitudinal TDM of VRC C_min is detailed in Table 3. Thirty-seven percent of the subsequent VRC C_min were outside the therapeutic range, with 36% at ≤1 mg/liter and 1% at >5 mg/liter. The route of administration had no impact on follow-up VRC C_min (i.v., 0.9 mg/liter [range, 0.3 to 4.8 mg/liter], versus oral, 1.4 mg/liter [range, 0.2 to 3.3 mg/liter]; P = 0.9). As observed for initial VRC C_min, age, BMI, and transaminase levels were not related to subsequent VRC C_min (data not shown).

Influence of CYP450 polymorphisms on subsequent VRC trough concentrations.

The combined genetic score influenced the subsequent VRC C_min or subsequent VRC C_min/D (Fig. 3a and b) (Kruskal-Wallis test, P = 0.004 and P < 0.0001, respectively). The subsequent VRC C_min of patients having a combined genetic score of >2 were lower than those of patients with a combined genetic score of <2 (post hoc test, P = 0.002).

FIG 3.

Influence of the genetic score on follow-up voriconazole plasma trough concentrations. Shown are the follow-up VRC C_min (a) or C_min/D ratio (b) according to the combined genetic score. Data are presented as interquartile range (boxes), data range (whiskers), and median (horizontal line), and only the significant P value obtained for the post hoc Bonferroni test (<0.0167 for three groups) is shown. (c) Temporal evolution of VRC trough concentrations throughout longitudinal TDM according to the combined genetic score. The area shaded in gray represents the therapeutic range. Solid circles, triangles, and squares represent the mean (± standard deviation) VRC C_min for patients having genetic scores of <2, equal to 2, and >2, respectively. The asterisk indicates a P value of <0.05 for the Kruskal-Wallis test. Second, third, fourth, and fifth determinations of VRC C_min were obtained after median periods of 7 (range, 2.8 to 28), 14 (range, 7.0 to 36), 35 (range, 11 to 59), and 53 (range, 20 to 87) days after the first determination.

Figure 3c shows the evolution of VRC C_min during repeated TDM in patients stratified according to the combined genetic score. As previously shown in Fig. 2c, the initial VRC C_min was significantly affected by the combined genetic score (Kruskal-Wallis test, P = 0.03), while at each subsequent TDM VRC C_min measurement, the VRC C_min was not significantly different between the combined genetic score categories.

The frequency of VRC C_min of ≤1 mg/liter was higher (P = 0.05) in patients having a combined genetic score of ≥2 (50% [range, 0 to 100%]) than in those with a combined genetic score of <2 (21% [range, 0 to 56%]).

Influence of diarrhea on subsequent VRC trough concentrations.

Clinical events contemporary with VRC C_min measurements were available for 77% of subsequent VRC C_min (n = 212). Since diarrhea could impact VRC C_min by reducing VRC absorption, the effect of diarrhea was investigated during per os VRC therapy. Among the 235 VRC C_min obtained during oral VRC treatment, clinical events were available for 193 VRC C_min. Eleven episodes of diarrhea were reported by 9 different patients (i.e., a frequency of 6%). VRC C_min was not influenced by diarrhea (VRC C_min with and without diarrhea, 1.6 mg/liter [range, 0.3 to 3.8 mg/liter] versus 1.3 mg/liter [range, 0.2 to 2.9 mg/liter]; P = 0.2). The influence of gastrointestinal graft versus host disease (GVHD) or mucositis could not be tested in our cohort because only two episodes of gastrointestinal GVHD and one of mucositis were reported during the course of VRC oral treatment.

Influence of concomitant medication on subsequent VRC trough concentrations.

Data on concomitant treatments were available for 95% of VRC C_min (n = 259). Ninety-four percent (n = 244) of the subsequent VRC C_min were determined while the patient was on PPI treatment; esomeprazole and pantoprazole were the most frequently used PPIs (Table 3). Each PPI influenced the VRC C_min differently (Kruskal-Wallis test, P < 0.0001) (Fig. 4). Post hoc analysis revealed a higher VRC C_min when the PPI was esomeprazole compared to pantoprazole (P < 0.0001), whereas there was no statistically significant difference between rabeprazole and esomeprazole or pantoprazole. This difference was also observed after stratification of patients according to combined genetic scores (data not shown).

FIG 4.

Influence of pump proton inhibitor treatment on voriconazole plasma trough concentrations. Data are presented as interquartile range (boxes), data range (whiskers), and median (horizontal line).

Treatment with glucocorticoids had no impact on VRC C_min (1.3 mg/liter [range, 0.1 to 3.2 mg/liter] with treatment versus 1.4 mg/liter [range, 0.3 to 3.4 mg/liter] without; P = 0.2). Since a multicenter study recently reported an interaction between glucocorticoids and VRC (3), a pairwise patient analysis was performed to further explore this drug-drug interaction. Patients who benefited from VRC C_min determinations with and without concomitant treatment by corticoids for at least 7 days were selected (n = 10). For each patient, the mean VRC C_min with or without glucocorticoid treatment was calculated. To overcome the influence of the route of administration, only VRC C_min obtained during oral therapy were used. Comparison of VRC C_min in these paired series confirmed the absence of any effect of glucocorticoid treatment on VRC C_min (see Fig. S2 in the supplemental material).

A possible interaction between PPI and steroid use was excluded since the frequencies of steroid use during esomeprazole and pantoprazole treatment were similar (36% of steroid use during esomeprazole treatment versus 29% for pantoprazole; χ² test, P = 0.4).

Immunosuppressive therapies, including calcineurin inhibitors, mycophenolate, or everolimus, had no effect on subsequent VRC C_min. Similarly, the subsequent VRC C_min was not influenced by comedication with various CYP2C19 and/or CYP3A enzyme inhibitors (data not shown).

DISCUSSION

The present study demonstrates that in a homogeneous cohort of AHSCT patients, the VRC C_min exhibits high inter- and intraindividual variability, resulting in almost half of the VRC C_min being out of the therapeutic range. On the whole, this variability could be partially explained by a combined genetic score that takes into account both CYP2C19 and 3A genotypes, the VRC route of administration, and any comedication with PPI.

Early optimal antifungal therapy improves the clinical outcome in patients with IFI (35, 36), and an initial VRC C_min of ≤0.35 mg/liter is a strong predictor of mortality (4). These studies suggested the importance of adjusting the VRC dose right from the first administration to reach the VRC C_min within the therapeutic range early on. Our study demonstrates for the first time that a genetic score combining both CYP2C19 and -3A4 genotypes is an independent predictor of initial VRC C_min. The impact of a CYP2C19 phenotype on the initial VRC C_min was already suggested by a study on a Korean cohort of patients with hematological diseases, without reaching statistical significance (37). In our cohort, CYP2C19 IM patients had a statistically significantly higher initial VRC C_min/D than CYP2C19 URM patients, confirming the importance of CYP2C19 polymorphisms on the variability of the initial VRC C_min. In addition to CYP2C19, VRC is also metabolized through the CYP3A pathway (16), which also exhibits clinically relevant polymorphisms. Our study provides the first demonstration that the CYP3A4*22 polymorphism also significantly influences the VRC C_min in AHSCT patients, since it was associated with a higher initial VRC C_min/D. Recently, the rs4646437 polymorphism of the CYP3A4 gene was also associated with higher VRC levels in Chinese patients (26). These genetic data, combined with several studies focusing on drug-drug interaction (23, 24), suggest the importance of CYP3A4 activity in VRC metabolism.

Since the frequencies of CYP2C19 and CYP3A single nucleotide polymorphisms were independent of one another, it was therefore of interest to calculate a combined genetic score integrating both CYP genotypes. Interestingly, a combined genetic score of <2 was associated with a higher VRC C_min/D compared to a combined genetic score of >2. Moreover, none of the patients having a combined genetic score of <2 presented an initial VRC C_min inferior to the efficacy threshold fixed at 1 mg/liter, while 47% of the patients with a combined genetic score of ≥2 had a subtherapeutic initial VRC C_min. These data suggest that this combined genetic score could help to individualize the VRC dose from the first administration onwards, as recently suggested for CYP2C19 (13, 38), and reduce the risk of out-of-range VRC C_min at the first determination.

Throughout VRC treatment, repeated adequate sustained VRC C_min measurements, rather than a single VRC C_min determination, have been shown to ensure treatment success (4, 12, 39). However, few studies have focused on longitudinal VRC TDM (4, 10, 12, 22), even though it is recommended to repeat VRC C_min determinations during therapy (5) in view of the large VRC intraindividual variability (4, 7), especially in AHSCT patients (10).

Follow-up VRC C_min remained influenced by the combined genetic score, despite dose adjustment being made during longitudinal TDM. However, this finding could be explained by the low frequency of dose adjustments in our cohort; in fact, VRC dose adjustments were made for only 16% of patients, while 85% of patients had at least one VRC C_min outside the therapeutic range and would be likely to need a VRC dose adjustment. Nevertheless, the genetics-related variability of VRC C_min was slightly reduced during TDM, notably after 5 VRC C_min measurements, although VRC C_min variability remained high.

Concomitant treatments are known to affect VRC C_min, notably drugs able to influence VRC metabolism by competition at the CYP catalytic site, such as PPIs (3, 7), or by induction of protein expression, such as phenytoin (3). In our cohort, PPIs were used widely, and pantoprazole was associated with lower VRC C_min than esomeprazole. Consistent with these findings, an in vitro study on human liver microsomes has demonstrated that pantoprazole has a lower inhibitory potency on CYP2C19 activity than esomeprazole and to a lesser extent rabeprazole (40). Conversely, PPIs exhibited no inhibition of CYP3A activity at concentrations in the same order of magnitude as the plasma levels encountered in vivo (40). Thus, the reduced VRC C_min in patients treated with pantoprazole we observed could be explained by its lower inhibitory potency on CYP2C19 activity. However, the effect of PPIs, especially pantoprazole on the VRC C_min, is far from fully understood since clinical studies investigating VRC pharmacokinetics have demonstrated either increased (3) or decreased (6, 41) VRC C_min in patients treated with pantoprazole compared to those in untreated PPI patients. Since it is well known that PPIs competitively inhibit CYP2C19 (3, 7, 40), taken together, these results suggest that PPI treatment significantly contributes to the variability of VRC C_min.

Another drug-drug interaction, between VRC and glucocorticoids, leading to reduced VRC C_min has been suggested by a single study (3) and has been proposed to be the result of CYP induction by glucocorticoids. However, our data and those from other studies (6, 42) did not support such an interaction, and in the present study, steroid use induced no significant change in VRC C_min both in the overall analysis performed with all VRC C_min and in the pairwise patient analysis. These conflicting results could be explained by the heterogeneity of the studied populations and the type and dose of the glucocorticoid. Indeed, AHSCT patients in our study preferentially received prednisone, prednisolone, and/or methylprednisolone at low and decreasing doses, whereas the effect of glucocorticoids was previously demonstrated in a heterogeneous population, including patients suffering from autoimmune diseases for whom dexamethasone treatment was more frequent and glucocorticoid doses were higher (3).

In our cohort, sex, age, and BMI had no impact on VRC C_min, whereas previous studies have identified age (3, 6, 12) and BMI (3, 6) as factors influencing VRC C_min. However, the patients in the latter studies were much older and had a higher BMI than those of our cohort, which could contribute to explaining these discrepancies.

There are several limitations of our study. First, this study was retrospective, performed in a single center and on a small number of patients: thus, the influence of the combined genetic score on VRC efficacy and safety could not be evaluated. Second, we have arbitrarily chosen to attribute similar weights to CYP2C19 and CYP3A since (i) VRC N-oxidation results from the simultaneous action of CYP2C19 and CYP3A (17, 18), both CYPs exhibiting similar K_m values (17), and (ii) VRC oxidation depends on only CYP3A isoforms (17). However, we acknowledge that the relative contributions of each CYP and relationships between CYP polymorphisms and activities remain to be further defined.

In conclusion, our data confirm the influence of CYP2C19 and provide strong evidence that CYP3A genotypes also influence VRC exposure in AHSCT patients. In this regard, the determination of a genetic score taking into account both CYP genotypes could be of clinical interest to adjust a priori the first VRC doses. Our data provide the groundwork for a multicenter prospective trial to confirm the clinical benefit of VRC first-dose adjustment for the combined genetic score, route of administration, and comedications and adaptation of the following doses according to longitudinal TDM.