Impact of CYP2C19 Phenotype and Drug-Drug Interactions on Voriconazole Concentration in Pediatric Patients

ABSTRACT

Voriconazole (VRC), a first-line agent for the treatment of invasive fungal infections, is mainly metabolized by human cytochrome P450 (CYP) 2C19. In this study, a retrospective analysis was performed to investigate the key factors that influence the plasma trough concentration (C_min) of VRC, and an appropriate dosing regimen for pediatric patients was drafted subsequently. Overall, factors such as age, CYP2C19 phenotype, and combination medication with proton pump inhibitors accounted for 23.4% of variability in dose-normalized C_min values of VRC by a multiple linear regression analysis. Dose-normalized C_min values in the poor metabolizers (PMs) and intermediate metabolizers (IMs) were significantly higher than those in extensive metabolizers (EMs) (P < 0.001). To achieve therapeutic C_min for CYP2C19 ultrarapid metabolizers (UMs) or EMs, patients aged no more than 12 and more than 12 years required doses of 6.53 ± 2.08 and 3.95 ± 0.85 mg/kg of body weight twice daily (P = 0.007). For CYP2C19 PMs or IMs, patients aged under 12 and over 12 years required doses of 5.75 ± 1.73 and 4.23 ± 0.76 mg/kg twice daily, respectively (P = 0.019). Furthermore, coadministration of rifamycin sodium or omeprazole exhibited significant effects on VRC C_min. Taken together, it is necessary to pay attention to the impact of CYP2C19 phenotype and drug-drug interactions to achieve optimal therapy.

INTRODUCTION

Immunocompromised children are highly susceptible to invasive fungal infections (1). VRC is a triazole antifungal agent recommended as the first choice for treating or preventing invasive aspergillosis (2). Several studies have found that the C_min of VRC is associated with its efficacy and safety (3, 4). VRC exhibits nonlinear pharmacokinetics, resulting in highly variable C_min (5, 6). In children, VRC dosing has been particularly challenging due to decreased oral bioavailability, higher elimination capacity, and larger inter- and intrapatient variability compared with adults (7). There is an agreement that children require higher and personalized doses to attain therapeutic C_min (8). Multiple studies have recommended initial dose ranges of 7 to 9 mg/kg of body weight twice daily in Caucasian children aged 2 to 12 years (9, 10). In China, it suggested that 5 to 7 mg/kg is adequate in children aged <2 years of Asian origin (11). However, VRC has been commonly prescribed for children ≤12 years with a low maintenance dose of 6 mg/kg in our center. Until now, the optimal dose in pediatric patients has not been fully established.

Apart from age, C_min of VRC can also be affected by genetic polymorphism of CYP2C19. The allele frequencies of CYP2C19*2 and CYP2C19*3 were 38.6% and 5.2% in Han Chinese populations, respectively, which were 2- to 3-fold higher than those in Europeans (12). The frequencies of PM on CYP2C19 are estimated to be 15.8% in patients of Asian origin compared with 2.2% in Caucasians and 13.7% in Chinese subjects (13). The racial differences imply that Chinese people have a high risk of adverse effects because of elevated C_min. However, limited data are available describing the correlation between CYP2C19 genetic variants and VRC C_min in pediatric patients (14, 15).

On the other hand, concomitant medication with drug-metabolizing enzyme inducers like rifampin is associated with decreased C_min of VRC. While the magnitude of this interaction appears to be less than that observed with rifampin, the coadministration of VRC with rifamycin sodium was more common. The cases reported that C_min of VRC could be maintained within the therapeutic range when coadministered with rifabutin, so that rifabutin could be a substitute for rifampin (16). Whether the use of rifamycin sodium has a clinically significant effect on VRC still requires more evidence. Proton pump inhibitors (PPIs) are able to influence VRC metabolism by competition inhibition against CYP2C19. The magnitude of the interaction with PPIs was greatly influenced by both the type and the dosage (17). To our knowledge, there have been few reports regarding the interaction of VRC and other PPIs other than omeprazole, such as pantoprazole, lansoprazole, and rabeprazole. The clinical implication of this drug interaction needs to be explored further in pediatric patients.

For these goals, this retrospective analysis aimed to investigate key factors that affected VRC C_min in Chinese pediatric patients, with particular attention to the CYP2C19 phenotype and drug-drug interactions.

RESULTS

Characteristics of patients.

A total of 108 Chinese pediatric patients (39 females and 69 males) were divided in 3 groups according to age (<6 years, 6 to 12 years, and >12 years) (Fig. 1). The characteristics are summarized in Table 1, and there were no significant differences among the three age groups in any of the covariates. Overall, the median age of total patients was 10 years old (range, 0.67 to 17 years), and 61.11% of patients were younger than 12 years old. The median body weight during VRC treatment was 32.35 kg (range, 7.00 to 84.00 kg). Hematologic malignancies and HSCT were the most common underlying conditions in patients receiving VRC. VRC was administered either targeted therapeutically (n = 27), diagnostic-driven therapeutically (n = 41), empirically (n = 27), or prophylactically (n = 13). CYP2C19 phenotype classifications were UM (n = 1), EM (n = 21), IM (n = 26), and PM (n = 8).

FIG 1.

Flow chart of the study.

TABLE 1.

Characteristics of patients

Parameter	Value for:				P value
	All patients	<6 yr	6–12 yr	>12 yr
No. of patients	108	28	38	42
Age (yr)	10.50 (0.67–18.00)	3.00 (0.67–5.00)	9.00 (6.00–12.00)	16.00 (13.00–17.00)
Body wt (kg)	32.35 (7.00–84.00)	14.50 (7.00–24.20)	29.00 (17.40–46.00)	55.00 (33.60–84.00)
Gender [no. (%)]					0.630
Female	39 (36.11)	9	16	14
Male	69 (63.89)	19	22	28
Underlying disease [no. (%)]					0.054a
Acute lymphoblastic	27 (25.00)	10	8	9
Acute myeloid leukemia	24 (22.22)	6	13	5
Acute myelocytic leukemia	1 (0.93)	1	0	0
HSCT	24 (22.22)	3	11	10
Aplastic anemia	14 (12.96)	5	5	4
Others	18 (16.67)	3	1	14
Treatment indication [no. (%)]					0.484b
Prophylaxis	13 (12.04)	5	0	8
Empirical	27 (25.00)	5	17	5
Diagnostic-driven therapy	41 (37.96)	9	15	17
Targeted therapy	27 (25.00)	9	6	12
Metabolizer phenotype [no. (%)]					0.523c
UM	1 (0.93)	0	1	0
EM	21 (19.44)	4	9	8
IM	26 (24.07)	7	10	9
PM	8 (7.41)	3	1	4
Unknown	52 (48.15)	14	17	21

^aAcute myelocytic leukemia, HSCT, aplastic anemia, and others groups were combined for statistical analysis. HSCT, hematopoietic stem cell transplantation.

^bProphylaxis and empirical groups were combined for statistical analysis.

^cUM and EM groups were combined for statistical analysis, and IM and PM groups were combined for statistical analysis.

Analysis of VRC C_min.

A total of 348 samples were obtained with a median (range) C_min of 1.26 μg/ml (<0.05 to 11.24). A median of 2 samples (range, 1 to 16) was drawn per patient, of which 63.79% were therapeutic. Figure 2 shows the distribution of measured VRC C_min and their respective doses. There was a significant difference in the number of VRC C_min within the therapeutic range between patients ≤12 years and >12 years (106/187 versus 116/161, χ² = 8.843, P = 0.003). A higher proportion of patients had subtherapeutic C_min in patients ≤12 years compared with patients >12 years, which means younger children are at risk for treatment failure (62/187 versus 29/161, χ² = 10.273, P = 0.001). No predictable relationships between dose and C_min was observed in pediatric patients (n = 348, coefficient = −0.056, P = 0.295).

FIG 2.

Scatterplot of the VRC C_min versus the VRC dose. The green square, blue triangles, and orange circles represent data of children aged <6 years, 6 to 12 years, and >12 years, respectively. The area between the horizontal lines indicates the therapeutic range (0.5 to 5.0 μg/ml).

Factors influencing dose-normalized VRC C_min.

The concentrations of VRC in pediatric patients were affected by multiple factors, including age, CYP2C19 genetic polymorphism, drug interactions, and so on (Table 2). Multivariate analysis by linear regression was performed to identify the factors associated with VRC C_min corrected for dose (Table 3). No significant effect was detected of routes of administration, glutamyl transpeptidase, or AKP on VRC C_min corrected for dose, but age, CYP2C19 phenotype, and PPI comedication did have a significant impact.

TABLE 2.

Univariate analysis of factors influencing dose normalized plasma VRC concentration

Covariate	Coeffcient	P valuea
Age	0.380	<0.001
Body wt	0.393	<0.001
Gender
Female		0.175
Male
CYP2C19 phenotype
UM/EM		<0.001
IM/PM
PPIs
Not concomitantly administered		0.016
Concomitantly administered
Routes of administration
Injection		0.133
Oral
ALT	−0.111	0.143
AST	−0.071	0.350
Glutamyl transpeptidase	0.180	0.017
AKP	−0.163	0.031
Bilirubin	0.116	0.138

^aThe boldface values indicate significant differences (P < 0.05).

TABLE 3.

Multiple linear regression analysis for dose-corrected VRC C_min

Covariate	Coeffcient	95% CI	P value	R 2
Age	0.437	0.029, 0.056	<0.001	0.234
CYP2C19 phenotype
UM/EM	0.192	0.058, 0.337	0.006
IM/PM
PPIs
Not concomitantly administered	−0.182	−0.309, −0.043	0.01
Concomitantly administered
Routes of administration
Injection	0.107	−0.042, 0.296	0.14
Oral

Impact of CYP2C19 phenotype on dose-normalized C_min.

A scatterplot of the relationships between the CYP2C19 phenotype and C_min corrected for dose is shown in Fig. 3. CYP2C19 IMs (median, 0.31 μg/ml/mg/kg; range, 0.01 to 2.60; P < 0.001) and PMs (median, 0.48 μg/ml/mg/kg; range, 0.04 to 1.75; P < 0.001) had higher dose-normalized C_min than EMs (median, 0.11 μg/ml/mg/kg; range, 0.01 to 2.02). CYP2C19 UMs did not have significantly lower dose-normalized C_min (median, 0.09 μg/ml/mg/kg; range, 0.04 to 0.60; P = 1.000) than EMs. There was no difference in the dose-normalized C_min between the CYP2C19 PMs and IMs (P = 0.155).

FIG 3.

CYP2C19 phenotype and dose-normalized C_min. Horizontal bars represent median values and interquartile range of dose-normalized C_min.

Maintenance dose to achieve the target range.

Patients whose C_min were within therapeutic range were divided into four groups according to age and phenotype, and the average maintenance dose required to achieve therapeutic C_min was inversely related to age (Fig. 4). Patients with CYP2C19 UM or EM phenotype of  ≤12 and >12 years required mean ± standard deviation (SD) doses of 6.53 ± 2.08 and 3.95 ± 0.85 mg/kg twice daily, respectively (P = 0.007). With CYP2C19 PM or IM phenotype, patients ≤12 and >12 years required doses of 5.75 ± 1.73 and 4.23 ± 0.76 mg/kg twice daily, respectively (P = 0.019), to achieve therapeutic C_min. In children ≤12 and >12 years, there was no difference in the EM/UM and PM/IM groups required to achieve therapeutic C_min (≤12 years, 6.53 ± 2.08 mg/kg versus 5.75 ± 1.73 mg/kg, P = 0.274; >12 years, 3.95 ± 0.85 mg/kg versus 4.23 ± 0.76 mg/kg, P = 0.505). Before genotyping of CYP2C19, the previous pediatric dosing regimen was explored according to therapeutic drug monitoring and clinical experience, leading to attainment of target concentrations of 56.68% in children ≤12 years. No significant differences between the doses recommended for each CYP2C19 phenotype/age group and the previous dose were observed.

FIG 4.

Average dose required to achieve target VRC C_min (0.5 to 5 μg/ml) categorized by age and CYP2C19 phenotype. Horizontal bars represent mean ± SD values of VRC C_min.

Influence of drug interactions on VRC C_min.

Medicines for possible drug interaction with VRC were identified. More than 59.26% of patients (64/108) had concomitant medications with potential drug interaction with VRC: rifamycin sodium (n = 17), rifampin (n = 1), fluconazole (n = 1), omeprazole (n = 32), pantoprazole (n = 39), lansoprazole (n = 16), and rabeprazole (n = 4). According to clinical experience, when patients received concomitant rifamycin sodium, VRC C_min were obviously low until 14 days after discontinuation. As shown in Fig. 5a, there was a statistically significant difference in dose-normalized VRC C_min between the groups coadministered with or without rifamycin sodium (P < 0.001). Moreover, the patients receiving PPIs had a higher dose-normalized VRC C_min (P = 0.038). Of the five PPIs, omeprazole was shown to have a significant effect on VRC concentration (Fig. 5b) (P <  0.001) compared with those with no PPIs, whereas pantoprazole (P = 0.870), lansoprazole (P = 0.828), and rabeprazole (P = 0.208) were not found to influence VRC C_min significantly. VRC C_min was significantly lower in patients treated with pantoprazole (P = 0.006) and lansoprazole (P = 0.023) versus those treated with omeprazole. There were no significant differences in VRC C_min of patients with omeprazole and rabeprazole (P = 0.711).

FIG 5.

(a) Effect of rifamycin sodium on dose-normalized C_min. (b) Effect of omeprazole on dose-normalized C_min. Horizontal bars represent median values of dose-normalized C_min.

Relationship between VRC C_min and treatment response.

Outcome data were analyzed in patients with targeted antifungal therapy. Data were available for 27 courses. Twenty-one had a successful outcome (21/27; 77.78%), while six were classified as failure (6/27; 22.22%). Univariate analysis did not reveal any statistically significant association between mean C_min values and treatment response (P = 0.369). The incidence of treatment success in all patients was 88.89% (96/108). No association between C_min values and treatment response was found (P = 0.158), although 20.37% (22/108) of patients did not attain mean C_min of ≥0.5 μg/ml.

According to the CTCAE criteria, 18 patients (16.67%) exhibited hepatotoxicity 14 days (range, 2 to 26 days) after starting VRC therapy. Hepatotoxicity was more common in patients with higher C_min (1.73 μg/ml versus 1.04 μg/ml, P = 0.013) (Fig. 6). Logistic regression analysis showed that the probability of hepatotoxicity increased by 16.2% while C_min increased by 1 μg/ml. When C_min is higher than 2.4 μg/ml, the incidence of hepatotoxicity is significantly increased (χ² = 3.938, P = 0.047). Other adverse drug reactions were observed in four pediatric patients (3.7%) during VRC treatment, including skin rashes in one case, myocardial injury in two cases, and neutropenia in one case. No ophthalmological disorders were reported, likely due to the young age of the patients.

FIG 6.

Hepatotoxicity and VRC C_min. Horizontal bars represent median, minimum, and maximum values of VRC C_min.

DISCUSSION

For children, maintaining VRC C_min within a target range is clinically important in order to obtain an optimal therapeutic response. We found high interindividual variability in VRC C_min that cannot be explained by a dose-concentration relationship. A large fraction of pediatric patients fails to achieve therapeutic C_min, with a consequently increased risk of therapeutic failure. This study demonstrated that serum concentrations of VRC were affected by age, CYP2C19 phenotype, and coadministration of PPIs (Table 3).

As has been shown previously in adults (18, 19), results of the current study indicate that VRC C_min in pediatric patients is correlated with CYP2C19 polymorphisms. CYP2C19 IMs and PMs had higher dose-normalized C_min than EMs, and there was no difference in the dose-normalized C_min between the CYP2C19 PMs and IMs (Fig. 3). In addition, the incidence of IM/PM among Asian populations, about 60.71% in our data, is higher than that for Caucasians. Therefore, it remains to be considered whether the recommended dose (7 to 9 mg/kg intravenously every 12 h) of previous studies conducted in western nations should be used in China. Our findings suggest that, in the younger age group, the recommended dose should be increased. For CYP2C19 UMs or EMs, most children ≤12 and >12 years should receive maintaining doses of 6.53 and 3.95 mg/kg twice daily, respectively; with CYP2C19 PM or IM phenotype, patients ≤12 and >12 years, respectively, required doses of 5.75 and 4.23 mg/kg twice daily. Lower VRC doses were required to achieve the target C_min among older children, which was consistent with one previous study (20).

Doses based on age and CYP2C19 phenotype should theoretically improve attainment of target concentrations, but there were no significant differences in the required dose and the previous dose. Our data cannot provide the guidance for optimizing VRC dosing based on the phenotype. The most likely cause is the small sample size for each phenotype group in this retrospective study, but it was suggested that phenotype should be identified before dosing. There was a significant difference in C_min between different phenotype groups. CYP2C19 phenotype to predict C_min may be a strategy for reducing the risk of subtherapeutic and supratherapeutic C_min.

Rifamycin sodium may upregulate CYP2C19 expression, thereby resulting in decreased VRC C_min, even lower than the lower limit of quantification of 0.05 μg/ml. The findings suggest that rifamycin sodium’s inductive effects were notable for at least 14 days. Patients could benefit from the selection of other drugs instead of rifamycin sodium to avoid the risk of subtherapeutic VRC C_min and therapeutic failure. This retrospective study is the first to assess the influence of comedications with rifamycin sodium in children.

Notably, out of the five most frequently recorded concomitant PPIs, omeprazole promoted higher dose-normalized VRC C_min. In contrast, pantoprazole, lansoprazole, and rabeprazole showed no significant effects on VRC C_min. A multicenter prospective study in adults has demonstrated increased VRC exposure with omeprazole due to an inhibition of CYP2C19 (21). Our data support differences in the degrees of interaction observed with different PPIs. This observation is consistent with the results of a physiologically based pharmacokinetic model that pharmacokinetic values of VRC increased to various degrees when administered PPIs (22). There was a statistically significant association between increased VRC C_min and omeprazole use but not with rabeprazole, lansoprazole, or pantoprazole use, suggesting a greater CYP2C19 enzyme inhibitory effect of omeprazole. That should be helpful for the choice of PPIs for pediatric patients treated with VRC. There were no significant differences in VRC C_min of patients with omeprazole and rabeprazole, which may be due to the low number of patients with rabeprazole.

We found no consistent correlation between VRC mean C_min values and treatment response, similar to findings of Liu et al. (11). A large portion of patients used VRC as diagnostic-driven therapy, empirical therapy, or prophylaxis. They showed clinical improvement regardless of C_min, which might explain why VRC did not show significant improvement in treatment success rates. Second, antifungal combination therapy with echinocandins or amphotericin B may confound the assessment of an exposure-response relationship of VRC.

Hepatotoxicity was the most frequent adverse event, and the incidence (16.67%) is comparable to previously reported rates (14.3% to 17.5%) (23, 24). A C_min of >2.4 μg/ml is associated with increased hepatotoxicity, which is considerably lower than that described in previous studies (25). The relationship between VRC C_min and hepatotoxicity may follow a different profile among pediatric patients. A meta-analysis indicated that the optimal trough concentration for increasing clinical success and minimizing hepatotoxicity during VRC therapy for Asian children populations is 1.0 to 3.0 μg/ml (26). Therefore, a lowered upper limit of the target C_min should be considered for Asian pediatric patients compared with adults.

However, some limitations are present in this study. This was a retrospective analysis of VRC drug monitoring from a single tertiary center in China. We failed to find a positive relationship between VRC C_min and treatment response. Despite these limitations, this study supports several conclusions regarding clinical application of VRC in pediatric patients: (i) the CYP2C19 genotype is likely one of the factors contributing to the high variability of VRC C_min in pediatric patients, and (ii) coadministration of rifamycin sodium or omeprazole may interact with VRC, resulting in significantly lower or higher plasma C_min.

As a result, we suggest that CYP2C19 phenotype helps in explaining lower VRC dosage required in Chinese populations, considering a higher portion of poor CYP2C19 metabolizers than those for Caucasian or African-American populations (27). A careful selection of the drugs that are concomitantly administered with VRC should be conducted to obtain an optimal clinical outcome. A prospective study with larger patient numbers is warranted to confirm the results of the present study, which showed the association between VRC C_min and CYP2C19 phenotype in pediatric patients.

MATERIALS AND METHODS

Patients and data collection.

A retrospective observational study of pediatric patients below the age of 18 who were subjected to VRC C_min monitoring between May 2017 and October 2020 was performed (Fig. 1). Clinical and demographic data were collected from the electronic medical records in The First Affiliated Hospital of Zhengzhou University (Zhengzhou, China). This study was approved by the Zhengzhou University Medical Research and Ethics Committee (number 2020-KY-0355-002). Demographic characteristics, drug administration history, outcomes of therapy, adverse events, and laboratory test results, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), glutamyl transpeptidase, alkaline phosphatase (AKP), and bilirubin, were also collected. In addition, rifampin and rifamycin sodium can significantly influence VRC C_min even for 2 weeks after discontinuation (28, 29). Therefore, the biological data were collected on the day C_min was measured, and potential drug interactions with rifampin or rifamycin sodium were collected during the 14 days before C_min measurements. Only patients with a VRC treatment regimen of ≥14 days were used to assess clinical response.

Diagnostic criteria.

Diagnosis and treatment of invasive fungal infections was classified according to the definitions of the European Organization for Research and Treatment of Cancer/Mycoses Study Group (EORTC/MSG). Treatment response was assessed by clinical, radiological, and microbiological criteria. Outcome for targeted therapy was classified as a success when two of the following criteria were met: clinical improvement, radiologic improvement, or decreasing mycologic burden. Otherwise, the course was defined as a failure. Diagnostic-driven, empirical therapy and prophylaxis were classified as a success if the course was completed without breakthrough fungal infection (30). Adverse drug reactions and their relationship with VRC were based on the Common Terminology Criteria for Adverse Events (CTCAE) defined by the National Cancer Institute.

Therapeutic drug monitoring of VRC.

VRC C_min between 0.5 and 5 μg/ml were considered therapeutic. Only patients having achieved steady state were included in the analysis, and for patients who received a loading dose, C_min measurements taken on day 3 of dosing or later were included. In patients who did not receive a loading dose, C_min measurements taken on day 5 of dosing or later were included. Dose adjustments were taken according to the guidelines (31).

C_min was measured by an ultra-high-performance liquid chromatograph-tandem mass spectrometry method (UPLC-MS/MS; Waters, United States). Following protein precipitation, the analyte and internal standard voriconzole-d3 (TRC; purity was 98%) were separated on a C₁₈ column (Waters Acquity UPLC BEH; 2.1 mm by 50 mm, 1.7 μm). The electrospray ion source (ESI) was employed as the ion source. MS acquisition was performed in multiple reaction monitoring mode, and the mass transitions were 350.0→127.0 for VRC and 353.0→127.0 for internal standard. This method was developed and validated according to FDA guidelines. The analytical range was 0.05 to 10.00 μg/ml. The intra- and interassay relative standard deviations (RSD) were 3.43 to 6.18%, 4.47 to 4.55%, and 2.54 to 3.28% for low, medium, and high levels of internal quality controls, respectively. The intra- and interassay relative deviation (RE) was ≤4.66%. The accepted inter/intraday precision and accuracy were less than 20%.

CYP2C19 phenotype assignment.

Genotyping was performed at the Henan Key Laboratory of Precision Clinical Pharmacy, Zhengzhou University, using the sequencing-by-hybridization array on a Fascan 48E (Xi'an Tianlong Technology Co.). Patients were classified into categories of metabolizer phenotypes with the use of the established common-consensus star allele nomenclature. Thus, carriers of a single CYP2C19 *17 allele (*1/*17) and *17 homozygotes were classified as UMs; patients without a CYP2C19 *2, *3, or *17 allele (*1/*1) were classified as EMs; those with one CYP2C19*2 or *3 allele (*1/*2 or *1/*3) and patients with one CYP2C19 *17 allele and one loss-of-function allele (*2/*17 or *3/*17) were classified as IMs; and those with two CYP2C19 *2 or *3 alleles (*2/*2, *2/*3, or *3/*3) were classified as PMs (32).

Data analysis.

IBM SPSS Statistics software v.19.0 (IBM Corp., Armonk, NY) was used for statistical analysis. The χ² test was used to compare patient characteristics between age groups. Univariate analysis of factors influencing dose-normalized plasma VRC concentration was conducted using the Spearman correlation analysis or Mann-Whitney U. Variables with P values of <0.05 by univariate analysis were entered into the multivariable model. The correlation between covariates (age, CYP2C19 phenotype, interacting drugs, and routes of administration) and VRC C_min corrected for dose were examined using multiple linear regression analysis. The association of C_min with efficacy and safety was analyzed by logistic regression. Continuous variables were compared by t test or nonparametric Mann-Whitney U test. A P value of <0.05 was considered statistically significant.