Effects of antifungal stewardship using therapeutic drug monitoring in voriconazole therapy on the prevention and control of hepatotoxicity and visual symptoms: A multicentre study conducted in Japan

Summary Background Hepatotoxicity and visual symptoms are common adverse effects (AEs) of voriconazole therapy. Objective To retrospectively evaluate the effects of treatment modification based on therapeutic drug monitoring on AEs in patients undergoing voriconazole therapy. Methods The target voriconazole trough concentration (C min) was 1‐5 µg/mL. Receiver operating characteristic curves were used to determine C min cut‐offs for AEs. Results A total of 401 patients were included. Among 108 patients with high initial C min, voriconazole was discontinued in 32 and the dose was reduced in 71. Among 44 patients with low initial C min, voriconazole was discontinued in 4 and the dose was increased in 19. Hepatotoxicity occurred in 6.0% of patients, after a median of 10 days. Visual symptoms were evident in 9.5% of patients after a median of 4 days. Initial C min was significantly associated with visual symptoms but not hepatotoxicity, which suggested the effect of treatment modification on hepatotoxicity. However, both hepatotoxicity and visual symptoms were significantly correlated with C min at the onset of AEs, and the C min cut‐offs were 3.5 μg/mL for hepatotoxicity and 4.2 μg/mL for visual symptoms. Voriconazole was discontinued after the occurrence of AEs in 62.5% of patients with hepatotoxicity but only 26.3% of patients with visual symptoms. With dose adjustment, treatment was completed in 8/9 patients with hepatotoxicity and 27/28 patients with visual symptoms. Conclusions A significant preventive effect was demonstrated on hepatotoxicity, but not on visual symptoms because of earlier occurrence. With treatment modification after the occurrence of AEs, most patients completed therapy.


| INTRODUC TI ON
Voriconazole is an antifungal agent that is used to treat invasive aspergillosis and invasive candidiasis. 1,2 It is associated with adverse side effects including visual symptoms, neurological disorders and hepatotoxicity. 3 Voriconazole TDM significantly reduced drug discontinuation due to adverse effects, and a higher proportion of patients achieved a clinical response with TDM compared to the non-TDM group. This is particularly important for Asian populations. Voriconazole is metabolised primarily via CYP2C19, and to a lesser extent CYP3A4 and CYP2C9. 13 Allelic polymorphism of CYP2C19 has been reported, and CYP2C19 non-wild-type alleles are generally found in 60%-70% of people in Asian populations but only 30% of Caucasian and African Americans. 14 Voriconazole concentrations can be as much as four times higher in individuals with non-wild-type alleles than in those with wild-type alleles. 15,16 These observations imply that Japanese individuals taking voriconazole are at a comparatively high risk of hepatotoxicity because of their genetic background. Jin et al 17 conducted a systematic review and meta-analysis to determine the optimal voriconazole trough concentration (C min ) and reported that C min > 3 μg/mL was associated with increased hepatotoxicity in Asians but not in other races.
In many institutions, monitoring serum voriconazole concentrations has become routine, and antimicrobial stewardship programmes incorporate TDM. Notably however, few studies have assessed the effects of antifungal stewardship using voriconazole TDM in real clinical settings. The aim of the current study was to evaluate the influence of treatment modification of voriconazole based on initial C min on the prevention of adverse effects, and the capacity of treatment modification after the occurrence of adverse effects to enable continued effective voriconazole therapy in Japanese patients.

| Setting
Patients aged ≥18 years who were receiving voriconazole were being monitored by an antimicrobial stewardship team and had had their voriconazole C min measured at least once during therapy was eligible for inclusion. Exclusion criteria were prophylactic use of voriconazole and unconsciousness precluding the determination of visual symptoms. This retrospective study was conducted at five hospitals in Japan between April 2015 and March 2018. Medical records were individually reviewed at each study site using a standardised data collection template to collect demographic information and clinical data on adverse events, as well as voriconazole dosing information. Hepatotoxicity was evaluated using laboratory data obtained at least once a week or at the time of TDM. Proven invasive fungal infection and probable/possible fungal infection were diagnosed in accordance with previously reported criteria. [18][19][20]

| Dosage adjustment and therapeutic drug monitoring
An adequate voriconazole dose was defined as a loading dose of 5-6 mg/kg twice daily followed by a maintenance dose of 3-4 mg/ kg twice daily. The maintenance dose was decreased to 1.5-2 mg/ kg in patients with liver dysfunction (Child-Pugh A-C). 21 Because the dose was calculated on the basis of body weight, dose rounding within 10% of the recommended dose was considered as appropriate. Optimal timing of TDM was defined as 4-10 days after the start of therapy. 21 Voriconazole concentrations were measured using high-performance liquid chromatography. The lower limit of quantification of the assay was 0.1 μg/mL; therefore, <0.1 μg/mL was recorded as zero for the purposes of data analysis. The target voriconazole C min was set at 1-5 µg/mL in the current study (≥1-2 µg/mL for efficacy and <4-5 µg/mL to prevent adverse effects 21 ). Relationships between well-known concentration-dependent adverse effects including hepatic dysfunction and visual symptoms and C min (initial C min and C min at the onset of adverse effects) were analysed.

Funding information
This research was supported by the Agency for Medical Research and Development (grant number JP18fk0108045). The funders had no role in the study design, data collection or analysis, the decision to publish, or preparation of the manuscript.

Conclusions:
A significant preventive effect was demonstrated on hepatotoxicity, but not on visual symptoms because of earlier occurrence. With treatment modification after the occurrence of AEs, most patients completed therapy.

K E Y W O R D S
antifungal stewardship, hepatotoxicity, therapeutic drug monitoring, visual symptoms, voriconazole

| Adverse effects
Elevations in liver function test results including alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase, gamma-glutamyl transpeptidase and bilirubin were recorded, and adverse events were graded in accordance with the Common Terminology Criteria for Adverse Events version 5.0. 22 Hepatic dysfunction was defined as AST or ALT levels at or above three times the upper limit of normal. If the AST or ALT baseline was abnormal, hepatic dysfunction was defined as AST or ALT at or above three times the baseline. Any visual symptoms during therapy including changes in colour perception, blurred vision, bright spots, wavy lines and photophobia were considered to be adverse effects caused by voriconazole. Visual hallucinations are usually classified as a symptom of neurotoxicity, but because clearly differentiating them from visual symptoms can be challenging, they were included as visual symptoms in the current study.

| Statistical methods
The data were expressed as medians and interquartile ranges (IQRs).
The chi-squared test was used to analyse categorical data, and the paired t test was used to analyse continuous data. Statistical analysis was performed with spss ver. 24 (SPSS Inc, Chicago, IL, USA). P < .05 was deemed to indicate statistical significance. Cut-off values were the maximum area under the curve (AUC) as determined via a receiver operating characteristic (ROC) curve.

| Patient characteristics
Overall, data from 583 patients were reported. Of these, 182 in whom voriconazole use was prophylactic were excluded from the study, and 401 in whom voriconazole was used for treatment were included. The median duration (IQR) of follow-up was 48 days

| Dosage adjustment and timing of initial therapeutic drug monitoring
Loading doses were administered to 65.8% of patients. The median dose (IQR) on the initial day of treatment was 5.9 mg/kg twice daily (5.4-6.1) in patients who received a loading dose. The median maintenance dose was 3.8 mg/kg twice daily (IQR 3.2-4.1) and the rate of adherence to the standard dose was 75.1%. Low adherence to a reduced dose (3/13 patients, 23.1%) was observed in patients with liver cirrhosis. The median day of TDM after the start of therapy was 6 (IQR 5-7), and the rate of adherence to adequate timing was 88.5%. TDM was performed a median (IQR) of two times (1-4) for
In 26 patients with sequential therapy (intravenous to oral administration and oral to intravenous administration) whose dose was not altered when the administration route was changed, C min was significantly lower in patients who were treated orally than in and pregabalin, and worsening of liver function, which might influence the voriconazole concentration was not observed ( Figure S1).  In total, 87.3% of patients who underwent dose adjustment subsequently achieved a C min within the target range (Table 3).

| Adverse effects and voriconazole trough concentration
Hepatotoxicity occurred in 24 patients (6.0%), and visual symptoms were reported by 38 patients (9.5%) (photophobia 14, visual hallu- of adverse effects in patients with visual symptoms was significantly higher than that in patients with hepatotoxicity (7.7% vs 3.0%, P = .003).
The ROC curve of initial C min and C min at the onset of adverse effects used to predict adverse effects (C min at last TDM during therapy was used in patients without adverse effects) is shown in Figure 1. Although higher initial C min was associated with visual symptoms (AUC 0.603, cut-off 4.9 μg/mL, OR 3.59, P = .037), there was no significant correlation between hepatotoxicity and initial C min (AUC 0.562, cut-off 3.6 μg/mL, OR 1.67, P = .292). In contrast,

F I G U R E 1
Receiver operating characteristic curve of initial trough concentration (C min ) and C min at the occurrence of adverse effects to predict adverse effects Initial C min C min at the occurrence of adverse effects Hepatotoxicity Visual symptoms (96.4%). One patient in whom visual symptoms did not improve had concomitantly used opioids. In seven patients with visual hallucination, the initial median C min was 4.87 μg/mL. Although voriconazole was continued (the same dose in six patients and dose reduction in one patient), visual hallucination improved in all patients.

| D ISCUSS I ON
Achieving the target C min is important for preventing adverse effects and improving clinical efficacy. 21 Pascual et al 23 reported that median C min was 2.9 μg/mL in patients on ≥8 mg of voriconazole per day, and 1.7 μg/mL in patients on 7 mg/kg per day. Racil et al 24 reported that half of the patients had a median C min of <1.0 μg/mL, and C min was only >5.0 μg/mL in 3.1% of patients. Compared with these previous reports, higher C min was evident in Japanese patients who fulfilled standard dosing criteria and those with TDM obtained at an appropriate time in the current study, a third of patients had high C min , and only 5% had low C min . These results suggest that TDM should be mandatory in Japanese patients who are prone to adverse effects when they exhibit high voriconazole concentrations.
In the present study, there was a clear correlation between adverse effects and C min at the onset of adverse effects, and the on initial C min may prevent adverse effects that could subsequently occur thereafter. In the current study, treatment modification was conducted in almost all patients with high initial C min . After dose adjustment, C min within the target range was achieved in approximately 90% of patients. As a consequence, only 6.0% of patients in the present study had hepatotoxicity, which is much lower than previously reported frequencies. 3,4,26,27 In a multicentre study conducted by Saito et al, 3  In the present study, there was a significant correlation between initial C min and visual symptoms, possibly because of the comparatively earlier occurrence of this adverse effect. TDM was conducted a median of 6 days after the initiation of voriconazole treatment, C min was usually recorded 2 days later, and in total, at least 8 days was required before the results of TDM were observed. Given the comparatively early onset of visual symptoms, treatment modification arising from initial TDM results before the occurrence of adverse effects could not be conducted in approximately 80% of patients who reported visual symptoms. Other authors have also reported that visual symptoms and hallucinations tended to occur during the first week of therapy and that symptoms were reduced or disappeared despite continued therapy in most patients. 8,28,29 As antifungal stewardship for patients with adverse effects, voriconazole discontinuation or dose reduction was more frequently performed in patients with hepatotoxicity than in those who re-  30 and relationships between C min and clinical success were not evaluated. Finally, although no previous studies evaluated the safety of voriconazole in patients with Child-Pugh C severe liver disease, we included seven patients with Child-Pugh C disease.
In conclusion, in the current study the high performance of voriconazole treatment modification based on initial C min by way of an antifungal stewardship programme was confirmed in clinical practice in Japan, and the results of the study suggest that it may promote a low incidence of hepatotoxicity. The chance of optimising voriconazole levels based on the initial TDM result was low in patients who reported visual symptoms, however, because of their comparatively earlier onset. In addition to the above-described preventive effects on hepatotoxicity, modification of therapy after the occurrence of adverse effects had a substantial capacity to facilitate continuation and completion of voriconazole therapy.

ACK N OWLED G M ENTS
We thank Dr Owen Proudfoot from Edanz Group (www.edanz editi ng.com/ac) for editing a draft of this manuscript.

E TH I C A L A PPROVA L
The study was approved by the institutional review boards of Hyogo