Inflammation‐driven variability in drug metabolism: Insights from voriconazole treatment of HSCT recipients

Abstract

Aims

Voriconazole is commonly used to prevent fungal infections after haematopoietic stem cell transplantation (HSCT). Although its metabolism is influenced by CYP2C19 genetics and inflammation, their combined effect is rarely considered in clinical practice, and integrated analyses remain limited.

Methods

We retrospectively analysed how inflammation and CYP2C19‐predicted drug‐metaboliser phenotypes affect voriconazole exposure and therapeutic range attainment in 126 HSCT patients. C‐reactive protein (CRP) ≥ 10 mg/L defined inflammation. A linear mixed model (LMM) assessed associations between dose‐corrected voriconazole concentrations, inflammation and CYP2C19‐predicted drug‐metaboliser phenotype.

Results

Dose‐corrected voriconazole trough concentrations were associated with CYP2C19‐predicted drug‐metaboliser phenotype (LMM p = <2e‐16, effect size of CYP2C19‐phenotype: β = −5.99e‐02‐mg/L per mg voriconazole, standard error (SE) = 2.75e‐02‐mg/L per mg voriconazole, p = 0.0315; effect size of CRP: β = 1.54e‐03‐mg/L per mg voriconazole, SE = 2.79E‐04‐mg/L per mg voriconazole, p = 5.49e‐08) and were higher during inflammation. Inflammation increased supra‐therapeutic concentrations and reduced subtherapeutic levels (p = 0.0080). This effect was most pronounced for intermediate metabolisers (n = 18 with and n = 31 without inflammation) and rapid metabolisers (n = 20 with and n = 28 without inflammation), with supra‐therapeutic concentration of 33% vs. 3%, and 15% vs. 0%, respectively (p = 0.02). Finally, in a longitudinal subset (n = 25), concentrations tracked inflammatory status across prior‐, during and post‐inflammation timepoints.

Conclusions

This study demonstrates inflammation and CYP2C19 genotype jointly influence voriconazole exposure and target attainment in clinical practice. These findings support intensified therapeutic drug monitoring with concurrent CRP assessment, particularly for intermediate and rapid metabolisers, to reduce the risk of supra‐therapeutic voriconazole concentrations during inflammation.

1. What is already known about this subject

• Voriconazole is primarily metabolized by CYP2C19.

• During inflammation CYP2C19 activity is downregulated, and dosing algorithms including c‐reactive protein (CRP) have been proposed.

• Genetic variation in CYP2C19 influences voriconazole metabolism. However, the association between CYP2C19 genotype and voriconazole exposure is lost during inflammation.

What this study adds

• This study demonstrates the combined clinical relevance of inflammation and CYP2C19 genotype for voriconazole exposure and target attainment.

• Inflammation increased voriconazole exposure in all CYP2C19‐predicted drug‐metaboliser phenotypes, resulting in concentrations outside the therapeutic window.

• During inflammation the proportion of supra‐ and subtherapeutic concentrations increased and decreased, respectively. This effect appeared most pronounced for intermediate metabolisers (IMs) and rapid metabolisers (RMs)

• Intensified TDM with concurrent CRP monitoring, particularly in IMs and RMs, may help prevent supra‐therapeutic voriconazole concentrations during inflammation.

1. INTRODUCTION

Voriconazole is used both as a treatment and for prophylaxis of invasive fungal infections in patients receiving a haematopoietic stem cell transplantation (HSCT). The drug is characterized by a narrow therapeutic window (range 1–6 mg/L). ¹ Subtherapeutic trough concentrations result in an increased risk of fungal infections, while supra‐therapeutic concentrations place the patient at risk for drug‐related toxicity such as liver damage, hallucinations and visual changes. ¹ , ² , ³ Consequently, to attain voriconazole trough concentrations, in clinical practice, therapeutic drug monitoring (TDM) is used. ¹ , ³ , ⁴

Voriconazole is metabolized primarily through CYP2C19, and to a lesser extent by CYP3A4, CYP2C9, and different flavin‐containing monooxygenases. ⁵ , ⁶ Moreover, voriconazole metabolism is known to exhibit non‐linear kinetics and is subject to autoinhibition. ⁷ Multiple genetic variants in CYP2C19 are known to influence enzyme activity. In clinical practice, different CYP2C19‐predicted drug‐metaboliser phenotypes can be distinguished including poor metabolisers (PM; two nonfunctional alleles, e.g., *2 and *3), intermediate metabolisers (IM; one nonfunctional allele), normal metabolisers (NM; two functional alleles), rapid metabolisers (RM; one increased function allele, e.g., *17) and ultrarapid metabolisers (UM; two increased function alleles, e.g., *17/*17). ⁸ , ⁹

Voriconazole has widely accepted therapeutic target levels. ¹ Following a loading dose, the standard dose of voriconazole is 200 mg b.i.d. followed by TDM to achieve target attainment. In the prophylactic treatment setting, it has been shown that more CYP2C19 RMs and UMs attain target concentrations with a dose of 300‐mg voriconazole b.i.d. ¹⁰ , ¹¹ Indeed, available dosing guidelines by the Clinical Pharmacogenetics Implementation Consortium (CPIC) and Dutch Pharmacogenetics Working Group (DPWG) recommend a higher dose (300 mg 2 b.i.d.) or switch to an alternative treatment for UMs, and a lower dose (100 mg 2 b.i.d.) for PMs. ⁸ , ⁹ , ¹² However, in a post hoc exploratory analysis of a prospective trial of patients (n = 189) with invasive aspergillosis, no association between CYP2C19 genotype and required voriconazole dose was observed. ¹³ , ¹⁴

In addition to CYP2C19 genotype, inflammation has been shown to impair voriconazole metabolism. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ It has been shown that proinflammatory cytokines reduce the expression of CYP2C19 and CYP3A4. ¹⁹ , ²⁰ Particularly, interleukin (IL)‐6 has been linked to the downregulation of CYP2C19 and CYP3A4. ²¹ In patients with coronavirus disease 2019 (COVID‐19)‐associated aspergillosis (CAPA) and invasive pulmonary aspergillosis an increase in supra‐therapeutic voriconazole levels was observed. ²² However, in clinical practice, IL‐6 is not routinely monitored, and most recent studies of the variability in voriconazole pharmacokinetics included c‐reactive protein (CRP) as a covariate. ¹³ , ¹⁷ , ²³ , ²⁴ , ²⁵

While there are studies that investigated the combined effect of inflammation and CYP2C19 genotype on voriconazole exposure, the combined effect of these two factors is not routinely considered in clinical practice, and a comprehensive and integrated analysis of inflammation and CYP2C19predicted drug‐metaboliser phenotypes is lacking. ²³ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰

Here, we retrospectively analysed the combined effect of inflammation and CYP2C19‐predicted drug‐metaboliser phenotypes on voriconazole exposure and therapeutic range attainment in a cohort of real‐world HSCT patients.

2. METHODS

2.1. Study design and setting

A retrospective analysis of data from the Biobank Hematologische Ziekten of the Leiden University Medical Center (LUMC) was performed. This biobank contains samples from patients treated at the department of Haematology, who provided written informed consented for research with data from their medical records and biological samples. Heparinized venous blood or DNA was collected from the biobank, and clinical data were extracted from the medical record for patients who (1) underwent a HSCT, (2) received voriconazole between February 2015 and February 2021 and (3) had at least one voriconazole trough concentration available. Only patients with a DNA sample collected before HSCT were included to ensure that the CYP2C19 genotype of the HSCT recipients liver was determined and not that of the HSCT donor. ³¹ The study was approved by the Biobank Hematologische Ziekten and the Medical Ethics Committee Leiden The Hague Delft (METC LDD) (B21.044).

2.2. Data synthesis

Data were collected as follows. Data from the biobank were available from 2015 until a patient discontinued treatment in the LUMC, died or reached the moment of data extraction (February 2021). Data regarding voriconazole treatment (dose, route of administration and duration), voriconazole trough concentrations, CRP concentrations, date of HSCT, weight, height, age and prescriptions of concomitant medications influencing (substrate, inhibitor or inducer) CYP2C19 and CYP3A4 as defined in the Flockhart table  ³² were extracted from medical records. The HSCT date was used as the starting point of analysis; all other data points were calculated relative to this timepoint.

Data of patients who received concomitant treatment with strong (a ≥ 5‐fold increase in the plasma area under the curve) CYP2C19 or CYP3A4 inhibitors or inducers were excluded from the analysis. ³³ To avoid overrepresenting patients with multiple measurements, a single voriconazole trough concentration was selected per patient per inflammation episode. We selected the measurement in presence and absence of inflammation based on the following criteria: (1) CRP concentration available within ±7 days of the voriconazole trough concentration; (2) intravenous (IV) data were preferred over oral data; and (3) the latest available data point on a particular dose was selected. Subtherapeutic voriconazole concentrations were defined as below 1 mg/L. ¹ Supra‐therapeutic concentrations were defined as over 6 mg/L. ¹ Dose‐normalized voriconazole concentrations were calculated by dividing the trough concentrations by the total prescribed daily voriconazole dose.

CRP concentrations were utilized as a proxy marker for inflammation. A CRP concentration ≥10 mg/L was considered positive for inflammation, whereas a CRP concentration <10 mg/L or the absence of a CRP measurement was classified as no inflammation. ³⁴ , ³⁵ , ³⁶ Dose‐corrected voriconazole concentrations were categorized in absence and presence of inflammation based on CRP concentrations. If available, a voriconazole trough concentration in absence and presence of inflammation was selected for each patient.

2.3. Analytical procedures

Voriconazole concentrations were quantified in the ISO15189‐certified laboratory of the Department of Clinical Pharmacy and Toxicology of the LUMC, using an UPLC/MS–MS assay validated according to European Medicine Agency guidelines for bioanalytical method validation. ³⁷ From July 2020 onwards, an In Vitro Diagnostics Regulation certified LC–MS/MS assay (Chromsystems Instruments & Chemicals GmbH, Gräfelfing, Germany) was used as described previously. ³⁸ The methods were cross‐validated with good results (<15% deviation). Both methods are described in more detail in Method S1.

CRP was measured using a validated antigen–antibody complex with turbidimetry method (Roche COBAS 8000 Modular ‐C702, Roche Nederland B.V., Woerden, The Netherlands) at the ISO15189‐certified laboratory of the Department of Clinical Chemistry. ³⁹ The lower limit of quantification was 0.3 mg/L.

2.4. CYP2C19 genotyping

CYP2C19 genotyping was performed at the ISO15189‐certified laboratory of the Department of Clinical Pharmacy and Toxicology of the LUMC. First, DNA was isolated with QIAsymphony SO system (QIAGEN Benelux B.V., Venlo, The Netherlands), by eluting 400‐μL blood sample in 200‐μL elution buffer (ATE) from the DSP DNA Midi kit (QIAGEN Benelux B.V., Venlo, The Netherlands). Briefly, CYP2C19*2 rs4244285, CYP2C19*3 rs4986893 and CYP2C19*17 rs12248560 were genotyped using predesigned TaqMan® genotyping assays and analysed on a ViiA7® Real‐time PCR system or Quantstudio 12K Flex Real‐Time PCR system (all from Thermo Fisher Scientific, Bleiswijk, The Netherlands) using a validated method. Genotypes were translated into CYP2C19predicted drug‐metaboliser phenotypes based on CPIC guidelines (details available in Table S1). ⁹

2.5. Data analyses

Three analyses were performed. First, to investigate the influence of inflammation and CYP2C19 genotype on voriconazole metabolism, dose‐corrected trough concentrations in presence and absence of inflammation were compared between the different CYP2C19‐predicted drug‐metaboliser phenotypes. Second, the effect of inflammation on voriconazole therapeutic range attainment was investigated by comparing the frequency of voriconazole trough concentrations within (1–6 mg/L), below (<1 mg/L) or above (>6 mg/L) the therapeutic window between the different CYP2C19‐predicted drug‐metaboliser phenotypes. Third, to minimize the effect of inter‐patient variability, a subgroup analysis was performed of patients with complete longitudinal follow‐up including data prior‐, during and post‐inflammation.

2.6. Statistical methods

Continuous variables were summarized as the median and range, or mean and standard deviation, depending on the distribution of the data. Predicted drug‐metaboliser phenotype and inflammation groups were summarized as count and percentage. Not normally distributed values were log transformed and visually checked for improved normality. CRP levels are routinely quantified as part of standard diagnostic procedures if inflammation is suspected in patients admitted to LUMC. Therefore we assumed no clinically relevant inflammation was present if a voriconazole level was missing a paired CRP value, and CRP concentration was set at 1 mg/L accordingly. Missing weight and height were replaced by group mean values. Genotype predicted drug‐metaboliser phenotype was included as ordinal variable. For the first analysis, a paired Student's t‐test was conducted to compare the means of the differences in absence and presence of inflammation for each phenotype. Next, a linear mixed model (LMM) was employed to analyse the relationship between dose‐corrected voriconazole concentrations and inflammation and CYP2C19‐predicted drug‐metaboliser phenotype. Fixed effects included CRP concentrations, phenotype and patient identifier. Random patient intercept and slope were included to account for the repeated measurements within patients. All parameters were checked for linearity and transformed appropriately. The LMM model was fitted using the maximum likelihood estimation method. Model diagnostics, including residual plots and normality checks, were conducted to ensure the appropriateness of model assumptions. In the second analysis, a chi‐square test of independence was used to test the association between CYP2C19‐predicted drug‐metaboliser phenotype and target attainment. Data were aggregated by adding individual counts across groups to form a contingency table suitable for analysis. Descriptive group statistics were used to describe patient cases in the third analysis. All statistical analyses were performed using Rstudio (version 4.3.3). A p‐value ≤0.05 was considered statistically significant.

2.7. Targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/22. ⁴⁰

3. RESULTS

Out of 397 voriconazole treated patients recorded in the biobank, 126 patients matched inclusion criteria; HSCT, treated with voriconazole and availability of a pre‐transplantation sample for DNA analysis (Figure 1). A total of 526 voriconazole trough concentrations were available, along with selected covariates, including voriconazole dose (with start and end dates of prescription), CYP2C19‐predicted drug‐metaboliser phenotype, age, date of HSCT, body weight, height, administration route, available CRP measurements ±7 days of voriconazole trough level and concomitant medications influencing CYP2C19 and CYP3A4. Out of the 126 patients, 99 patients had more than one voriconazole measurement available. Missing CRP values were defaulted to 1 mg/L for 415 (79%) voriconazole level measurements. Missing weight (8%) and height (4%) were replaced with the group mean. At the first voriconazole measurement, the median age was 57.5 years and 67% of the patients was male (Table 1). Median time since HSCT was 4.5 years. Follow‐up was 1 to 7 years, during which patients had one to 26 voriconazole trough concentration measurements. Of the patients, 99 (79%) received concomitant treatment with CYP2C19 or CYP3A4 substrates or weak or moderate inhibitors (Table 1). Five measurements of two patients receiving concomitant treatment with the strong CYP3A4 inhibitor clarithromycin were excluded. All of the 99 patients on concurrent medication received the CYP2C19 and CYP3A4 substrate pantoprazole. No patients received strong CYP2C19 inhibitors. Other concomitant CYP2C19 or CYP3A4 substrates were prescribed with a low frequency, and most were weak CYP3A4 inhibitors (Table S2). The CYP2C19‐predicted drug‐metaboliser phenotype distribution was three PMs (2%), 41 IMs (33%), 38 NMs (30%), 38 RMs (30%) and six UMs (5%) (Table 1), as expected for patients of European descent.

FIGURE 1.

Extraction and synthesis of the study data for the different analysis. HSCT: haematopoietic stem cell transplantation. * patients can be in both groups.

TABLE 1.

Patient characteristics (n = 126) at first voriconazole trough level.

Variable	Outcome
Male n (%)	85 (67)
Age (years) median (range)	57.5 (19–76)
Weight (kg) mean ± SD	78 ± 15
Height (cm) mean ± SD	177 ± 11
CYP2C19‐predicted drug‐metaboliser phenotype n (%)
PM	3 (2)
IM	41 (33)
NM	38 (30)
RM	38 (30)
UM	6 (5)
Cotreatment with CYP2C19 and/or CYP3A4 substrate, weak/moderate inhibitor or inducer n (%)	99 (79)
Number of voriconazole measurements per patient (n)	1–26
Patients with (at least one) measurement with I.V. voriconazole dosing n (%)	50 (40)
Time between voriconazole and CRP level (days) median (range)	0 (−6–7)

Abbreviation: I.V., intravenous.

First, we compared voriconazole trough concentrations in absence and presence of inflammation at the time of voriconazole trough concentration measurement to study the effect of inflammation and CYP2C19‐predicted drug‐metaboliser phenotype on voriconazole exposure. Voriconazole trough concentrations of 98 occasions in absence of inflammation and 55 occasions in presence of inflammation were available (Figure 1, Table S3). Dose‐corrected voriconazole trough concentrations correlated with CYP2C19‐predicted drug‐metaboliser phenotype (LMM p = <2e‐16, effect size of phenotype: β = −5.99e‐02‐mg/L per mg voriconazole, standard error (SE) = 2.75e‐02‐mg/L per mg voriconazole, p = 0.0315; effect size of CRP: β = 1.54e‐03‐mg/L per mg voriconazole, SE = 2.79E‐04‐mg/L per mg voriconazole, p = 5.49e‐08) and were higher during inflammation compared to the noninflamed state (Table 2) (Figure 2). These results show influence of both CRP level and CYP2C19 phenotype on voriconazole trough concentration. No statistically significant effect of inflammation on dose‐corrected voriconazole trough concentrations were observed in PMs (n = 3) and UMs (n = 6).

TABLE 2.

The dose‐corrected voriconazole trough concentrations for the different CYP2C19‐predicted drug‐metaboliser phenotypes in the absence (CRP < 10 mg/L) and presence of inflammation (CRP ≥ 10 mg/L). Patients are represented only once in both inflammation groups (absence and presence).

CYP2C19‐predicted drug‐metaboliser phenotype	Inflammation a	Number of patients	Voriconazole trough concentration (mg/L)	Dose‐corrected voriconazole trough concentration (mg/L per mg voriconazole) median (range)	p‐Value	Daily voriconazole dose (mg) median (range)
PM	No	3	3.6 (3.4–5.1)	0.0051 (0.0045–0.0085)	0.74	800 (400–1000) *
	Yes	1	5.2	0.0065		800 *
IM	No	31	1.8 (0.10–7.5)	0.0035 (0.0025–0.019)	0.016	400 (400–1200)
	Yes	18	3.5 (0.20–9.7)	0.0083 (0.0050–0.024)		400 (300–800)
NM	No	30	2.2 (0.30–7.5)	0.0048 (0.00075–0.017)	0.034	400 (400–1200)
	Yes	15	3.0 (0.20–7.9)	0.0064 (0.00036–0.020)		400 (250–900)
RM	No	28	1.7 (0.20–5.5)	0.0031 (0.00044–0.011)	0.012	400 (400–1050)
	Yes	20	2.7 (0.30–9.6)	0.0061 (0.00075–0.019)		400 (400–680)
UM	No	6	2.4 (0.40–15)	0.0059 (0.00033–0.013)	0.51	480 (400–1200)
	Yes	1	7	0.0047		1500

Abbreviations: CRP, C‐reactive protein; IM, intermediate metaboliser; NM, normal metaboliser; PM, poor metaboliser; RM, rapid metaboliser; UM, ultrarapid metaboliser.

^a Inflammation is defined as ‘Yes’ if CRP ≥ 10 mg/L or ‘No’ if CRP < 10 mg/L.

^*two patients received high intravenous voriconazole doses.

FIGURE 2.

Effect of inflammation on dose‐corrected voriconazole trough concentrations (n = 126) stratified by CYP2C19‐predicted drug‐metaboliser phenotype (PM: 3 patients, three measurements without inflammation, one measurement with inflammation, IM: 41 patients, 31 measurements without inflammation, 18 measurements with inflammation, NM: 37 patients, 30 measurements without inflammation, 15 measurements with inflammation, RM: 38 patients, 28 measurements without inflammation, 20 measurements with inflammation, UM: seven patients, six measurements without inflammation and one measurement with inflammation). Inflammation was defined as c‐reactive protein concentrations ≥10 mg/L (n = 55). PM (black): poor metaboliser, IM (blue): intermediate metaboliser, NM (green): normal metaboliser, RM (yellow): rapid metaboliser, UM (red): ultrarapid metaboliser.

Second, the effect of inflammation on voriconazole target attainment was investigated in the different CYP2C19‐predicted drug‐metaboliser phenotypes in the same 126 patients. Overall, in the absence of inflammation 79 (74%) concentrations were within therapeutic range vs. 37 (71%) during inflammation (not significant (NS)). During inflammation, the distribution of concentrations within the target range per CYP2C19‐predicted drug‐metaboliser phenotype group differed ( $\chi$ ² test p = 0.008). The percentage of patients with subtherapeutic concentrations (n = 7, 13%) was lower during inflammation compared to no inflammation (n = 15, 15%) ( $\chi$ ² test p = 0.02) (Figure 3). Importantly, in accordance with the observed increase in voriconazole trough concentrations during inflammation, the overall proportion of patients with supra‐therapeutic concentrations increased during inflammation compared to no inflammation (n = 11, 20% vs. n = 4, 4%) ( $\chi$ ² test p = 0.02). In particular, IMs showed a substantial increase in the proportion of supra‐therapeutic concentrations with one (3%) in the absence of inflammation compared to 6 (33%) during inflammation ( $\chi$ ² test p = 0.02). In RM patients, inflammation was also associated with an increased likelihood of supra‐therapeutic voriconazole concentrations (n = 3, 15%), whereas under noninflammatory conditions only subtherapeutic (n = 6, 21%) and therapeutic (n = 22, 79%) levels were observed. Interestingly, inflammation did not affect target attainment in NMs (Figure 3).

FIGURE 3.

Voriconazole range attainment in absence and presence of inflammation. The left y‐axis shows the percentage of patients with a voriconazole measurement within range; subtherapeutic (blue), therapeutic (green) and supra‐therapeutic (purple). The right y‐axis shows the voriconazole dose. Dot size represents the number of observations. Patients are represented only once in both inflammation groups (absence and presence).

Finally, to exclude the effect of inter‐patient variability we investigated voriconazole concentrations in a subset of 25 patients with longitudinal follow‐up data prior‐, during and post‐inflammation. Voriconazole concentrations changed in parallel to inflammatory status (Figure 4: 1 selected case example per CYP2C19‐predicted drug‐metaboliser phenotype, all 25 patients in Figure S1). Interestingly, in six out of the eight patients with supra‐therapeutic voriconazole levels (4× IM, 1× RM, 1× UM) observed concomitant CRP concentrations were considerably above the predefined threshold of ≥10 mg/L and ranged from 62 to 230 mg/L.

FIGURE 4.

Selected case example per CYP2C19‐predicted drug‐metaboliser phenotype illustrating the effect of inflammation on voriconazole pharmacokinetics. For each CYP2C19‐predicted drug‐metaboliser phenotype one case example was selected. Results of all 25 patients are available in Figure S1. Panel A: CYP2C19‐predicted intermediate metaboliser, Panel B: CYP2C19‐predicted normal metaboliser, Panel C: CYP2C19‐predicted rapid metaboliser and Panel D: CYP2C19‐predicted ultrarapid metaboliser. Day 0 is defined as the date of transplantation, voriconazole (•), CRP (•), the red and blue shades indicate the therapeutic ranges of voriconazole and normal CRP concentrations, respectively. CRP = C‐reactive protein.

4. DISCUSSION

We investigated the effect of CYP2C19‐predicted drug‐metaboliser phenotypes in HSCT patients receiving voriconazole in absence and presence of inflammation. Our data confirm that inflammation and CYP2C19 genotype are independently associated with voriconazole concentration. Furthermore, we showed that both influenced voriconazole target attainment.

To the best of our knowledge, we are the first to investigate the effect of all types of inflammation on the occurrence of sub‐ and supra‐therapeutic voriconazole concentrations. During inflammation, an increase in the frequency of supra‐therapeutic concentrations of voriconazole was observed, while the percentage of subtherapeutic concentrations decreased, despite the application of TDM. This observation confirms the impact of inflammation previously reported. ²³ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ Moreover, in the subset of 25 patients with data available prior‐ during and post‐inflammation, the majority of patients with supra‐therapeutic voriconazole concentrations had a CRP of over 60 mg/L (62–230 mg/L). Based on available clinical guidelines, we applied a predefined CRP cut‐off of ≥10 mg/L for inflammation. ³⁴ , ³⁵ , ³⁶ Interestingly, and in line with our findings, a study in liver transplant recipients receiving tacrolimus concluded that patients with CRP concentrations >49 mg/L were 2.6 times more at risk for supra‐therapeutic tacrolimus exposure. ⁴¹ Tacrolimus is extensively metabolized by CYP3A4/5, which like CYP2C19, is highly susceptible to inflammation‐induced downregulation. ²¹ , ⁴² For voriconazole, this was also studied previously. In a study investigating the correlation of pre‐albumin, CRP and voriconazole overexposure (C_min > 5.0 mg/L), 70% of 123 patients treated with voriconazole had adverse reactions when CRP > 102 mg/L. ⁴³ In a case–control study of haematological patients who benefited from voriconazole TDM, the sole independent risk factor for voriconazole overdose was a CRP level >96 mg/L. ¹⁸ Future studies should investigate if setting a CRP threshold higher than ≥10 mg/L would be more appropriate. Our data clearly indicate the importance to monitor inflammation during voriconazole treatment and emphasize the need to tailor voriconazole dose and potentially TDM application during inflammation.

The largest effect of inflammation on voriconazole concentration was observed in CYP2C19 IM and RM patients with an increase in supra‐therapeutic voriconazole concentrations from 3% to 33% and 0% to 15%, respectively. This effect might appear contradictory with IMs having decreased CYP2C19 activity, while RMs have an increased CYP2C19 activity. However, many of the IMs were treated with 200 mg b.i.d., a dose also used in CYP2C19 NMs. We hypothesize that during inflammation, the reduced CYP2C19 enzyme activity in IMs due to loss‐of‐function variants is further reduced, and the risk for developing supra‐therapeutic concentrations is elevated compared to NMs who have an a priori normal function of CYP2C19.

By contrast, many RMs do not attain target concentrations when treated with voriconazole 200 mg b.i.d. in the absence of inflammation and are typically prescribed a higher dose to attain target levels. During inflammation, CYP2C19 activity is reduced, and RMs are at increased risk for developing supra‐therapeutic concentrations given their a priori higher voriconazole dose. Indeed, data from our previous study show that during inflammation 50% of the supra‐therapeutic voriconazole concentrations belonged to CYP2C19 RMs. ²⁶

Plausibly due to the small sample sizes of the PM (n = 3) and UM (n = 6) groups, no statistically significant effects could be observed. Future studies involving larger, well‐characterized cohorts are needed to elucidate the effects of these metaboliser phenotypes.

A considerable proportion of patients receiving voriconazole acquire an infection. The absolute risk of fungal infection after liver transplantation varies regionally and between institutions. For example, in a cohort of voriconazole treated liver transplant patients in Iran (n = 104) over 50% experienced an invasive fungal infection. ⁴⁴ While this is considerably lower after HSCT, a recent meta‐analysis showed 7% of pooled incidence of proven/probable breakthrough invasive fungal infections in 991 patients from published studies. ⁴⁵ In addition, the positioning of voriconazole compared to other antifungal drugs like micafungin and fluconazole varies. However, breakthrough infections were found to have high mortality rates (20%–50% have been reported). ⁴⁶ , ⁴⁷ The success of voriconazole treatment depends on multiple factors, including regional infection risk and patient compliance, alongside CYP2C19 genotype and inflammation status. Potentially, it could be a useful strategy to determine CYP2C19 genotype and apply more intensive TDM and CRP follow‐up in CYP2C19 IM and RM to prevent supra‐therapeutic voriconazole concentrations and associated adverse events such as liver toxicity and neurological symptoms. ¹ , ² However, CYP2C19 phenotype or phenoconversion could probably not have prevented those adverse events.

Our study has some limitations. First, due to the retrospective nature of our study, data were incomplete and we applied imputation. Consequently, the effect of inflammation was studied at the population level. We defined inflammatory status solely based on CRP as other parameters related to inflammation were not routinely recorded in the EMR. We assumed the absence of inflammation if no CRP data was available, which potentially could have resulted in the incorrect assignment of noninflammatory status. However, in the LUMC it is standard practice to measure CRP if the patient is suspected to have an inflammatory event. Therefore, we estimate the proportion of patients with a missed CRP ≥ 10 mg/L to be extremely low. In addition, we checked the availability of other potential markers for inflammation in the EMR such as body temperature and leukocytes in the EMR. We noticed that if one or more of these markers were available CRP was also available. We applied a relatively wide range of +/−7 days around the voriconazole trough level for the CRP level extraction compared to other studies who have paired CRP and trough levels on the same day or within 48 h. Typically, these studies have been conducted in admitted and often instable patients. ¹⁸ , ⁴⁸ By contrast, many patients in our study were stable and managed in the outpatient setting. CRP measurement were mostly performed during control visits and we consider a 7 day period acceptable as substantial changes in inflammatory status are unlikely within this timeframe in clinically stable patients. Moreover, we primarily focused on acute inflammation, where CRP levels are expected to be >100 mg/L as opposed to patients with chronic inflammatory conditions, such as diabetes, cancer or rheumatoid arthritis that typically exhibit lower CRP levels (10–100 mg/L), reducing the likelihood of misclassification. ⁴⁹ , ⁵⁰ , ⁵¹ Next, liver function can also modulate CYP450 activity with a negative correlation between Child‐Pugh score and liver function. ⁵² , ⁵³ , ⁵⁴ However, the protein abundance of CYP2C19 is relatively stable with stable protein levels in Child‐Pugh score C. ⁵⁵ Moreover, a Child‐Pugh score C is a contraindication for the use of voriconazole. Consequently, the haematologist do not prescribe voriconazole to HSCT patients with a Child‐Pugh score C. Therefore, we consider the potential risk of impact of impaired liver function to our analysis limited. Second, the available data per patient varied considerably, particularly the number of voriconazole measurements which could have resulted in selection bias. To mitigate, we stratified by selecting one voriconazole concentration per patient and inflammation status, and used correction factors in our statistical analyses. Third, concomitant use of CYP2C19 and CYP3A4 inhibitors and inducers can potentially affect voriconazole concentrations. To limit the impact of drug–drug interactions on our outcomes, we recorded the use of CYP2C19/CYP3A4 influencing drugs and excluded patients receiving strong CYP3A4 inhibitors. Moreover, since voriconazole is a strong CYP3A4 inhibitor, the impact of concomitant CYP3A4 modulating drugs is expected to be small. Indeed, with the exception of strong CYP3A4 inhibitors used in HIV treatment, registration and post‐marketing studies have reported no clinically relevant interactions that increase voriconazole concentrations. ¹ , ⁵⁶ Consequently, no specific management is required if a patient receives voriconazole and concomitant treatment with mild or moderate CYP3A4 inhibitors as, for example, ciprofloxacin. We can also not completely exclude the concomitant use of over the counter (OTC) drugs affecting CYP2C19. However, typically stem‐cell transplant recipients are closely monitored in the outpatient setting, and the use of OTC drugs not registered in our EMR is not likely. Our analyses accounted for the concurrent treatment with substrates and weak or moderate inhibitors of CYP2C19 and CYP3A4. We did not correct for concomitant use of other haematology drugs that may cause liver enzyme abnormalities such as the echinocandins. However, the impact of these drugs on CYP2C19 activity appears limited. ⁵⁷ Lastly, only CYP2C19*2, *3 and *17 were analysed. Therefore, presence of rare genetic variants with an effect on CYP2C19 activity cannot be excluded. However, in the Dutch population the three selected variants represent >98% of the CYP2C19 variants. ⁵⁸ To address these limitations, an independent validation of our finding is warranted.

In conclusion, both inflammation and CYP2C19 genotype were found to influence voriconazole exposure. Moreover, our data suggest that the use of intensified TDM and follow‐up CRP measurement to prevent supra‐therapeutic voriconazole concentrations during inflammation could be most beneficial in CYP2C19‐‐predicted IMs and RMs. Future studies should aim to develop dosing algorithms including CYP2C19 genotype and inflammatory status to direct the voriconazole starting maintenance dose.

AUTHOR CONTRIBUTIONS

S.D.K., H‐J.G., P.A.vdB. and J.J.S. designed the research. S.D.K. and E.H.M.vE. performed data collection. S.D.K. and S.B contributed to data analysis and statistical analysis. S.D.K. and J.J.S. wrote the first draft of the manuscript. S.D.K., M.L.M., H‐J.G., S.B., E.H.M.vE., P.A.vdB and J.J.W wrote the final manuscript. All authors read and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Table S1: CYP2C19 genotype to phenotype translation table.

Table S2: Concomitantly prescribed CYP2C19 and/or CYP3A4 substrates, weak/moderate/strong inhibitor or inducers.

Table S3: Measurement characteristics in absence and presence of inflammation (CRP ≥ 10 mg/L). Patients (n = 526 measurements) are represented only once in both inflammation groups.

Figure S1: Effect of inflammation on voriconazole pharmacokinetics for different CYP2C19‐predicted drug metaboliser phenotypes. This figure shows the voriconazole, c‐reactive protein (CRP) concentrations and the voriconazole dose, for (A) CYP2C19‐predicted intermediate metabolisers, (B) CYP2C19‐predicted normal metabolizers, (C) CYP2C19‐predicted rapid metabolisers, (D) CYP2C19‐predicted ultrarapid metabolisers. Day 0 is defined as the date of transplantation, follow‐up time differs for all patients, voriconazole (•), CRP (•), the red and blue shades indicate the therapeutic ranges of voriconazole and normal CRP concentrations, respectively.

Klomp SD, Manson ML, Guchelaar H‐J, et al. Inflammation‐driven variability in drug metabolism: Insights from voriconazole treatment of HSCT recipients. Br J Clin Pharmacol. 2026;92(7):2250‐2259. doi: 10.1002/bcp.70501

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.