Pharmacokinetic Modeling of Voriconazole To Develop an Alternative Dosing Regimen in Children

ABSTRACT

The pharmacokinetic variability of voriconazole (VCZ) in immunocompromised children is high, and adequate exposure, particularly in the first days of therapy, is uncertain. A population pharmacokinetic model was developed to explore VCZ exposure in plasma after alternative dosing regimens. Concentration data were obtained from a pediatric phase II study. Nonlinear mixed effects modeling was used to develop the model. Monte Carlo simulations were performed to test an array of three-times-daily (TID) intravenous dosing regimens in children 2 to 12 years of age. A two-compartment model with first-order absorption, nonlinear Michaelis-Menten elimination, and allometric scaling best described the data (maximal kinetic velocity for nonlinear Michaelis-Menten clearance [V_max] = 51.5 mg/h/70 kg, central volume of distribution [V₁] = 228 liters/70 kg, intercompartmental clearance [Q] = 21.9 liters/h/70 kg, peripheral volume of distribution [V₂] = 1,430 liters/70 kg, bioavailability [F] = 59.4%, K_m = fixed value of 1.15 mg/liter, absorption rate constant = fixed value of 1.19 h⁻¹). Interindividual variabilities for V_max, V₁, Q, and F were 63.6%, 45.4%, 67%, and 1.34% on a logit scale, respectively, and residual variability was 37.8% (proportional error) and 0.0049 mg/liter (additive error). Monte Carlo simulations of a regimen of 9 mg/kg of body weight TID simulated for 24, 48, and 72 h followed by 8 mg/kg two times daily (BID) resulted in improved early target attainment relative to that with the currently recommended BID dosing regimen but no increased rate of accumulation thereafter. Pharmacokinetic modeling suggests that intravenous TID dosing at 9 mg/kg per dose for up to 3 days may result in a substantially higher percentage of children 2 to 12 years of age with adequate exposure to VCZ early during treatment. Before implementation of this regimen in patients, however, validation of exposure, safety, and tolerability in a carefully designed clinical trial would be needed.

INTRODUCTION

Opportunistic invasive fungal diseases (IFDs) are important causes of morbidity and mortality in immunocompromised children and adolescents (1). Voriconazole is a synthetic expanded-spectrum triazole with broad-spectrum antifungal activity in vitro (2). The compound is available in oral and intravenous formulations and has demonstrated clinical efficacy and safety in adult phase III clinical trials of its use for primary treatment of superficial and invasive candidiasis (3, 4) and invasive aspergillosis (5) and as antifungal prophylaxis in high-risk patients (6, 7). Voriconazole is approved for use for the indications listed above in subjects ≥12 years of age in both the United States and the European Union. The recommended intravenous dosage is 4 mg/kg of body weight two times daily (BID; day 1, 6 mg/kg BID), and the oral dosage is 200 mg BID (day 1, 400 mg BID). While the compound has been approved in the European Union for use in children ≥2 to 11 years of age since 2005, the dose finding in this age group has been difficult, with several revisions in the recommended dosages being made. Currently, an intravenous dose of 8 mg/kg BID (day 1, 9 mg/kg BID) and an oral dose of the suspension of 9 mg/kg BID have been adopted by the European Medicines Agency (EMA) for children ≥2 to 11 years of age and those 12 to 14 years of age weighing <50 kg (8–10).

The pharmacokinetics of voriconazole in children are nonlinear, complex, and still incompletely understood. Apart from a considerable intrasubject variability, there is wide between-subject variability in exposure that may be explained only in part by genetic CYP2C19 polymorphisms (11–15). On the basis of this variability, the significant correlations between exposure and effect (16), and the significant effects of therapeutic drug monitoring (TDM) on treatment responses and adverse effects (17, 18), TDM is advocated in patients with life-threatening IFDs, with the target trough concentration of voriconazole being between 1 and 6 mg/liter, depending upon safety and tolerability (17–20).

However, the large fraction of pediatric patients with inadequate exposure to voriconazole upon standard dosing (21–24) and the time delays in correcting inadequate exposures through TDM contrast with the clinical mandate of early and adequate treatment in patients with life-threatening IFDs (25–27). Using population pharmacokinetic modeling and Monte Carlo-based simulations, we therefore explored alternative approaches to voriconazole dosing in children for their utility to optimize voriconazole exposure, particularly in the critical first days of treatment.

(The results of the population-based pharmacokinetic analysis were presented in preliminary form at the 25th PAGE Meeting, Lisbon, Portugal, 7 to 10 June 2016.)

RESULTS

Study population.

The study population included 24 patients with 225 concentration values available for analysis. Of the 225 samples, 38 had to be excluded for one of the following reasons: missing or unclear time documentation (n = 26) or implausible elevation of plasma concentrations at two subsequent time points without any doses administered in the meantime (n = 12, considered analytical error). Ultimately, 187 blood samples obtained from 23 pediatric patients were considered for building of the population pharmacokinetic model. The demographic parameters for the 23 patients are listed in Table 1, and observed plasma concentration-versus-time plots are depicted in Fig. 1.

TABLE 1.

Dosing regimens and demographic parameters for 23 pediatric patients considered for building of the pharmacokinetic model of voriconazole^a

Parameter	Value(s) for patients:
	≤12 yr of age	>12 yr of age
Intravenous dose (mg/kg BID)	7	4 (day 1, 6)
Oral dose (mg BID)	200	200 (day 1, 400)
No. of males/no. of females	6/6	9/2
Median (range) values for:
Age (yr)	8 (0.5–12)	14 (13–21)
Body wt (kg)	27 (7–44)	56 (39–85)
Body surface area (m2)	0.99 (0.3–1.4)	1.6 (1.3–2.1)
Bilirubin concn on day 1 (mg/dl)	0.3 (0.1–0.9)	0.4 (0.1–0.9)
Creatinine concn on day 1 (mg/dl)	0.3 (0.1–0.6)	0.6 (0.3–0.9)
AST concn on day 1 (U/liter)	50 (35–310)	32 (19–315)
ALT concn on day 1 (U/liter)	49 (14–275)	42 (22–385)
γ-GT concn on day 1 (U/liter)	98 (23–1160)	91 (31–534)
AP concn on day 1 (U/liter)	101 (8–723)	75 (40–151)
CRP concn on day 1 (mg/dl)	0.6 (0.2–18.4)	1.4 (0.1–8.1)
No. of samples for PK analysis	98	89

^aBID, two times daily; AST, aspartate aminotransferase; ALT, alanine aminotransferase; γ-GT, gamma-glutamyltransferase; AP, alkaline phosphatase, CRP, C-reactive protein; PK, pharmacokinetic.

FIG 1.

Observed plasma concentration-versus-time plots of the study population. Circles, concentration data for voriconazole after intravenous dosing; triangles, concentration data after oral dosing.

Population pharmacokinetic model building.

The basic structural model consisted of two compartments. Clearance was tested as the linear, nonlinear Michaelis-Menten type and as the combined linear-nonlinear type. Out of these, the nonlinear Michaelis-Menten clearance best described the data. Use of a combined clearance by addition of a linear clearance term did not significantly improve the model (P > 0.01, difference in the objective function value [OFV] = −0.828). Initially, the compartmental structure and clearance type were determined on the basis of intravenous dosing data only. After addition of the oral dosing data, bioavailability (F) and the rate of absorption were integrated into the model. The final model contains K_m and the absorption rate constant (K_a) fixed to literature values, as it was not possible to obtain reasonable estimates of these parameters due to the small sample size (28). Interindividual variability (IIV) was added to the nonlinear clearance represented by the maximal kinetic velocity for nonlinear Michaelis-Menten clearance (V_max), the central volume of distribution (V₁), intercompartmental clearance (Q), and bioavailability (F). Covariance of 83% was found for IIVs for V_max and V₁.

To test the appropriateness of the classic allometric scaling approach for this pediatric population, the exponents for clearance and volumes of distribution included in the power functions on weight were estimated. Additionally, body surface area-based scaling through linear functions, centered to the population median, was tested. Neither scaling option improved the model in comparison to allometric scaling. For the remaining covariates, covariate testing was performed using stepwise covariate modeling (SCM) with the PsN module with a 5% forward inclusion criterion and a 1% backward elimination criterion (29). However, no additional significant covariate relation was found through the use of these methods. Parameter-versus-covariate plots were created as well but did not show additional relations. Parameter estimates for the final model are listed in Table 2.

TABLE 2.

Parameter estimates of the final pharmacokinetic model of voriconazole disposition in children 2 to ≤12 years of age^a

Parameter	Parameter estimate	IIV (%)	Relative SE (%)
Ka (h−1)b	1.19
F (%)	59.4	Logit 1.34	17.8
V1 (liter/70 kg)	228	45.4	13.5
Q (liter/h/70 kg)	21.9	67	19.7
V2 (liter/70 kg)	1,430		22.6
Km (mg/liter)b	1.15
Vmax (mg/h/70 kg)	51.5	63.6	15
Proportional error (%)	37.8		7.2
Additive error (mg/liter)	0.0049		2.1

^aIIV, interindividual variability; K_a, absorption rate constant; F, bioavailability; V₁, central volume of distribution; Q, intercompartmental clearance, V₂; peripheral volume of distribution; K_m, Michaelis-Menten rate constant; V_max, maximal kinetic velocity for nonlinear Michaelis-Menten clearance.

^bFixed to values estimated by Friberg et al. (28).

Internal model validation performed by the use of goodness-of-fit plots showed an even distribution of residuals over time after dose (TAD) and population predictions (PRED). When individual predictions (IPRED) and PRED were compared to the observed values, the model revealed the underprediction of PRED, which also became visible for IPRED at higher concentrations. A prediction-corrected visual predictive check (pcVPC) was created to investigate the appropriateness of the model and is shown in Fig. 2. In this visual predictive check (VPC), the observed percentiles and medians matched the corresponding simulated confidence intervals.

FIG 2.

Prediction-corrected (Pred Corr) visual predictive check. Grey areas, 90% confidence intervals of the predictions; lines, the median and 5th and 95th percentiles of the observations; circles, observed values.

Simulations.

The final population pharmacokinetic model was used to simulate the currently recommended two-times-daily (BID) intravenous dosing regimen for pediatric patients aged 2 to 12 years (9, 10), as well as alternative dosing strategies for this age group, using three-times-daily (TID) dosing and regimens of TID loading for 1 to 3 days with a step-down to BID dosing. All simulations were performed for 4 weeks (672 h), since hematopoietic stem cell transplantation (HSCT) patients with invasive fungal infections and, in particular, invasive aspergillosis usually require prolonged antifungal treatment.

As shown in Fig. 3, in comparison to the currently recommended BID regimen, TID dosing achieved higher voriconazole trough concentrations early during treatment, at 24 h. Target trough concentrations of between 1 and 6 mg/liter were reached in 30.9% of simulated subjects following the 9-mg/kg TID regimen versus 3.9% of simulated subjects following the 9-mg/kg BID regimen. However, continuation of TID dosing for prolonged periods of time resulted in faster accumulation, and the difference from adequately dosed patients became more and more negligible. For example, the regimen with an initial loading of 9 mg/kg TID on day 1 and maintenance dosing of 6 mg/kg TID resulted in 59.1% of simulated patients reaching the target attainment at 168 h, whereas 46.5% of the simulated patients reached the target attainment at 168 h with the current BID dosing regimen, and the proportion of patients with concentrations above the putative toxicity cutoff became higher with continued TID dosing (Fig. 3 and 4). The starting point for accumulation is depicted in Fig. 4a to c as the change in the slope of the proportion of patients with concentrations above the toxicity cutoff over time, where the accumulation after TID dosing starts at approximately 150 h, whereas it starts at approximately 250 h with the recommended BID dosing regimen. Prolonging the initial loading phase of the TID regimens to 48 or 72 h resulted in faster accumulation relative to that with the 24-h loading regimens. Among the seven TID regimens tested, TID dosing with 9 mg/kg on day 1 followed by 6-mg/kg TID maintenance therapy had the best performance in providing improved early exposure and avoiding unwarranted accumulation (Fig. 3 and 4).

FIG 3.

Percentage of simulated subjects (indicated on the y axis) receiving various TID dosing regimens of voriconazole with trough concentrations of <1 mg/liter, between 1 and 6 mg/liter, and >6 mg/liter at 24 h (a), 168 h (b), and 672 h (c) after first dose. The bars represent the following regimens from left to right, respectively: BID, 9 mg/kg BID on day 1 followed by 8 mg/kg BID; TID1, 6 mg/kg TID on day 1 followed by 5.34 mg/kg BID (the total daily dose was equal to that achieved with the BID regimen); TID2, 7 mg/kg TID on day 1 followed by 6 mg/kg BID; TID3, 8 mg/kg TID on day 1 followed by 6 mg/kg BID; TID4, 8 mg/kg TID on day 1 followed by 7 mg/kg BID; TID5, 9 mg/kg TID on day 1 followed by 7 mg/kg BID; TID6, 9 mg/kg TID on day 1 followed by 6 mg/kg BID; and TID7, 9 mg/kg TID on day 1 followed by 5 mg/kg BID (Table 4).

FIG 4.

Probability of target attainment (PTA) for various simulated intravenous TID regimens of voriconazole in children 2 to ≤12 years of age. (a to c) PTA for toxic trough concentrations of >6 mg/liter; (d to f) PTA for target range of trough concentrations of 1 to 6 mg/liter; (g to i) PTA for subtherapeutic trough concentrations of <1 mg/liter. BID, 9 mg/kg BID on day 1 followed by 8 mg/kg BID; TID1, 6 mg/kg TID on day 1 followed by 5.34 mg/kg BID (the total daily dose was equal to that achieved with the BID regimen); TID2, 7 mg/kg TID on day 1 followed by 6 mg/kg BID; TID3, 8 mg/kg TID on day 1 followed by 6 mg/kg BID; TID4, 8 mg/kg TID on day 1 followed by 7 mg/kg BID; TID5, 9 mg/kg TID on day 1 followed by 7 mg/kg BID; TID6, 9 mg/kg TID on day 1 followed by 6 mg/kg BID; and TID7, 9 mg/kg TID on day 1 followed by 5 mg/kg BID (Table 4).

Since TID regimens provided improved early exposure but had no benefit relative to the recommended BID regimen during continued dosing, select regimens with TID loading for 1 to 3 days and a step-down to BID dosing were also simulated (Fig. 5 and 6). Loading with 9 mg/kg TID for 24, 48, and 72 h resulted in higher voriconazole trough concentrations at 24, 48, and 72 h (Fig. 5 and 6) without relevant accumulation, with the percent probability of target attainment (PTA) for the toxic range being similar to that for the recommended BID dosing regimen.

FIG 5.

Simulated concentration-versus-time profiles of voriconazole after treatment with 9 mg/kg TID for 24, 48 and 72 h, followed by 8 mg/kg BID thereafter, in children 2 to ≤12 years of age. Black dots, 5th and 95th percentiles; gray dots, median concentrations.

FIG 6.

Probability of target attainment (PTA) for simulated combined TID/BID dosing regimens of voriconazole in children 2 to ≤12 years of age. (a) PTA for trough concentrations of >6 mg/liter; (b) PTA for target trough concentrations of between 1 and 6 mg/liter; (c) PTA for subtherapeutic trough concentrations of <1 mg/liter. Comb 1, 9 mg/kg TID for 24 h followed by 8 mg/kg BID as maintenance therapy; Comb 2, 9 mg/kg TID for 48 h followed by 8 mg/kg BID as maintenance therapy; Comb 3, 9 mg/kg TID for 72 h followed by 8 mg/kg BID as maintenance therapy (Table 4).

Simulated area under the curve (AUC) values are shown in Table 3. When the AUC/MIC ratio of the approved regimen was compared to that of the combined TID/BID regimen, the target ratio of 32.1 was reached at the median only for the combined regimen. Due to the switch to BID treatment after day 3, the AUC/MIC ratio became more similar to that with the recommended BID regimen (Table 3).

TABLE 3.

Simulated AUCs for approved and combined regimens

Simulated regimena	AUC/MIC	Value on:
		Day 1	Day 14
BID	5th percentile	8.8	9.6
	Median	21.7	48.1
	95th percentile	50.4	172.5
Comb 1	5th percentile	13.2	9.6
	Median	32.7	50.6
	95th percentile	75.7	179.4
Comb 2	5th percentile	13.2	9.6
	Median	32.7	48.9
	95th percentile	75.7	189.1
Comb 3	5th percentile	13.2	9.6
	Median	32.7	52.8
	95th percentile	75.7	199.3

^aBID, 9 mg/kg BID on day 1 followed by 8 mg/kg BID; Comb 1, 9 mg/kg TID for 24 h followed by 8 mg/kg BID as maintenance therapy; Comb 2, 9 mg/kg TID for 48 h followed by 8 mg/kg BID as maintenance therapy; Comb 3, 9 mg/kg TID for 72 h followed by 8 mg/kg BID as maintenance therapy (Table 4).

DISCUSSION

The pharmacokinetics of voriconazole are complex and involve oxidative metabolization through several CYP450 isoenzymes with genetic polymorphisms in its major metabolic pathway, resulting in nonlinear pharmacokinetics and unpredictable dose-exposure relationships that are further influenced by drug-drug interactions, inflammation, and variable oral bioavailability, particularly in children (17, 30–32). On the basis of the significant correlations between exposure and effect and the significant effects of TDM on treatment responses and adverse effects, TDM is advocated for children and adults in current management guidelines (16–20, 33). Given the large fraction of children with an inadequate exposure to voriconazole (21–24), the delays in correcting inadequate exposures through TDM, and the necessity of early and adequate treatment in children with life-threatening IFDs (25, 27, 34), we used population pharmacokinetic modeling and Monte Carlo simulations to explore alternative approaches to the dosing of voriconazole in children 2 to 12 years of age to optimize voriconazole exposure, particularly in the first days of treatment. The results of this study indicate that TID loading at 9 mg/kg per dose for up to 3 days may result in a substantially improved percentage of children with adequate exposure to voriconazole during early treatment. Whereas continued TID dosing led to relevant accumulation after 7 days, continuation with BID dosing at the approved 8 mg/kg per dose may maintain plasma exposures without undue accumulation to potentially toxic concentrations. However, validation of exposure, safety, and tolerability in a carefully designed clinical trial would be needed before implementation of this regimen in clinical practice.

The concentration data observed in the 23 children and adolescents receiving voriconazole for prophylaxis after allogeneic HSCT were best described by a two-compartment model with first-order absorption, nonlinear Michaelis-Menten elimination, and allometric scaling. The overall structure of the model is similar to that of the models that were used in the process of pediatric dose finding (28, 35). The central and the peripheral volumes of distribution, however, were slightly higher, since the observed average plasma concentrations in our study population were overall lower. Apart from the fact that sampling after intravenous administration did not include the true peak concentration, this difference in exposure may also be related to the fact that the cohort investigated in this study was more homogeneous than the previously investigated study populations in terms of interventions and concomitant treatments and thus differs from the previously investigated study populations. Since reasonable estimates of K_m and K_a could not be obtained due to the small sample size, these parameters were fixed in the final model to values reported by Friberg et al. (28), as their model, like ours, included patients with an age range wider than that of the previous models. Clearance terms, variabilities, as well as parameter precision, represented by relative standard errors, had orders of magnitude comparable to those of the previous models, and the estimated residual error of the model of 37.8%, which reflects the remaining unexplained variability in the disposition of voriconazole, was also comparable to that of the previous models (28, 35). Of note, while the population model presented here may describe the pharmacokinetics of voriconazole in the investigated study population with sufficient accuracy, it may not be easily transferable to a different population. Recently, a methodology for the development of a hybrid model for voriconazole that incorporated information from prior models in a biologically plausible manner in order to better predict the kinetics of voriconazole in a new population has been presented (36). Further validation of the hybrid model is required to better delineate its value for individualization of the voriconazole dose in patients.

On the basis of the high rate of inadequate exposure to voriconazole after standard doses (21–24) and the delays in correcting them through TDM, the primary goal of the pharmacokinetic/pharmacodynamic simulations of this study was to explore the utility of alternative TID dosing regimens to optimize voriconazole exposures in children 2 to 12 years of age, particularly early during treatment. Using our model parameters, Monte Carlo simulations of the currently recommended dosing regimen of voriconazole described the plasma concentration-time course that is usually observed in the clinical setting: whereas some individuals start accumulating the drug after a short time, some do not seem to build up adequate levels. Thus, the simulations depict how the concentration-time course evolves in the whole population under a certain treatment regimen, rather than being able to give a precise prediction for a single individual. For these reasons, we used the percent PTA, which reflects the proportion of the simulated population reaching a certain target with trough concentrations of between 1 and 6 mg/liter, as our primary endpoint for adequate exposure.

Relative to the currently recommended BID regimen, TID dosing achieved higher voriconazole trough concentrations in children 2 to 12 years of age early during treatment, at 24 to 72 h. However, continuation of TID dosing resulted in an apparent saturation of clearance mechanisms and relevant accumulation after 7 days, with increasing proportions of simulated patients having concentrations above the putative toxicity cutoff over time. In contrast, exploration of regimens with TID loading for 1 to 3 days and a step-down to BID dosing at the approved 8 mg/kg per dose revealed higher voriconazole trough concentrations at 24, 48, and 72 h without relevant accumulation to the toxic range. A simulated TID dosing regimen of 9 mg/kg per dose for up to 3 days resulted in a substantially higher percentage of children with adequate exposure to voriconazole during early treatment; continuation with BID dosing at the approved 8 mg/kg per dose maintained plasma exposures without undue accumulation to potentially toxic concentration ranges.

Of note, while the simulation of TID loading suggests substantial improvements in the initial percent PTA, the improvements were only gradual and did not resolve the issue of high interindividual pharmacokinetic variability. Thus, while TID loading with a dose of 9 mg/kg for a maximum of 72 h may improve exposure to voriconazole in children 2 to 12 years of age early during therapy, it would not replace the need for TDM. Indeed, reflecting the delays encountered in correcting inadequate exposures through TDM, the addition of a second agent during the initial treatment phase in critically ill patients with a new diagnosis of proven or probable invasive aspergillosis may be indicated in clinical practice, in order to provide optimal antifungal coverage until achievement of target plasma concentrations of voriconazole. Independent of these considerations, however, a regimen consisting of 9 mg/kg intravenously TID for the first 3 days is clearly investigational, as it has been assessed by simulations only. For validation, a carefully designed clinical trial would be needed to validate exposure and safety, particularly in patients younger than 5 years of age, in whom accurate dosing is even more challenging (37).

TDM and individualized dose adaptation will be part of future dosing recommendations for voriconazole in pediatric patients. In the absence of the MIC of the infecting organism, the pharmacodynamic target of trough concentrations of 1 to 6 mg/liter, although widely accepted, is derived from relatively low-quality evidence (20, 33, 38). The delivery of TDM as a future standard of care will require real-time measurement of drug concentrations at the bedside and software algorithms for dosage adjustment (34, 39). Finally, better measures of the pharmacodynamic effect, including lesion volumes or circulating galactomannan levels, are required to deliver therapy that is truly individualized (38, 40).

MATERIALS AND METHODS

Design of pharmacokinetic study.

The study was an open-label, single-center observational phase II study of voriconazole prophylaxis in pediatric patients undergoing allogeneic hematopoietic stem cell transplantation (HSCT) in 2008 and 2009. Voriconazole was administered starting from day +10 after transplantation onward at the doses recommended at the time of the study. Patients aged up to 12 years received voriconazole at 7 mg/kg two times daily (BID) as an intravenous infusion given over 1 h. For patients older than 12 years of age, intravenous dosing included a loading dose of 6 mg/kg BID for the first 24 h, followed by 4 mg/kg BID thereafter. Whenever feasible, dosing was switched to oral administration with a fixed dose of 200 mg BID for all age groups (median time to switch from intravenous to oral dosing, 72 h; range, 23 to 311.5 h). The study was conducted in accordance with the Declaration of Helsinki 2013. Written informed consent for antifungal therapy as part of the medically indicated measures of supportive care and for sample and data collection was obtained within the consent procedure for haploidentical hematopoietic stem cell transplantation (HSCT) approved by the Institutional Review Board of the Faculty of Medicine of the Goethe University of Frankfurt am Main (protocol code no. ZKI-SCT-HAPLO-0106; Eudra-CT no. 2006-000 393-76).

Pharmacokinetic sampling and recording of covariates.

Blood samples were taken prior to the first intravenous and oral dose, at 2 h, 6 h, and 8 h after the first administration, and prior to the third and fifth doses, at 24 h and 48 h after the start of BID treatment, respectively. Blood samples were obtained in heparinized vials, immediately centrifuged for 10 min at 2,000 × g, and stored at −70°C until analysis. At the baseline, the following demographic parameters were documented: underlying condition, body weight, height, body surface area (BSA), age, and sex. Additionally, the C-reactive protein (CRP) concentration was documented for each day of measurement. For assessment of covariates for hepatic and renal function, the serum concentration or activity of bilirubin, aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyltransferase (γ-GT), alkaline phosphatase (AP), and creatinine was recorded on the day of transplantation (day 0) and at the start and end of treatment with voriconazole. The maximum values during the intravenous and the oral treatment phases were also recorded.

Analytical method.

The concentrations of voriconazole in plasma were determined using a previously described and in-house-validated high-performance liquid chromatography (HPLC) method (41). The method was validated according to the U.S. Department of Health and Human Services, Food and Drug Administration, Guidance for Industry, Bioanalytical Method Validation (42), as well as International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) guideline Q2(R1), Validation of Analytical Procedures (43).

Calculated limits of detection (LOD) and lower limits of quantification (LLOQ) were 0.002 mg/liter and 0.01 mg/liter, respectively. Accuracies and precision values for assays of the intraday and interday variations for quality control samples with low, middle, and high concentrations were ±10.5% and ±5.6%, respectively. For model building, concentrations below the LLOQ were set as LLOQ/2 (44). This was relevant for 7.6% of the collected samples.

Population pharmacokinetic modeling.

Model building was performed using the NONMEM program (version 7.2; ICON Development Solutions, Ellicott City, MD) with the ADVAN 6 subroutine and the FOCE+I estimation method. Prediction-corrected visual predictive checks (pcVPC) and SCM building were performed with the PsN module (version 4.2.0) (45). Graphical output was created in R (version 3.3.2; R Foundation for Statistical Computing, Vienna, Austria), with the additional use of the Xpose package (version 4.5.3) (46).

During the model-building process, concentration data for samples obtained after intravenous administration were used to define the basic structural model: one to three compartments with linear clearance, nonlinear (Michaelis-Menten) clearance, and a combination of both were tested. Subsequently, the parameters were fixed and data from oral administration were added to characterize bioavailability and oral absorption. Bioavailability was modeled using a logit transformation. Finally, the previously fixed parameters were unfixed again and the structural model was reestimated.

For nested models, the difference in the objective function value (OFV) was considered the best parameter to quantify model improvement. A drop in the OFV of 6.63, corresponding to a 1% level of significance, was considered an adequate model improvement, when a single parameter was added. Nonhierarchical models were compared using the Akaike information criterion (AIC) in combination with goodness-of-fit (GOF) plots (47). As the pediatric patient population included individuals with a wide age range, allometric scaling was included in the structural model-building process (48). For a better comparison with former models, the allometric inclusion of weight was standardized to 70 kg (49).

Potential continuous covariates included the CRP concentration, age, and baseline values for the recorded parameters of renal and hepatic function. Sex was tested as a categorical covariate. The influence of the covariate was tested on the central and peripheral volumes of distribution, oral bioavailability, and V_max.

Model performance was evaluated by creating GOF plots, including individual predictions (IPRED) and population predictions (PRED) versus the observed values and plots for conditional weighted residuals (CWRES) versus time, time after dose (TAD), and PRED. Visual predictive checks (VPCs) with prediction correction were created to test the predictive performance of the model.

Simulations with a reference population.

Monte Carlo-based simulations for alternative dosing regimens were performed in NONMEM using the final population pharmacokinetic model. For this purpose, a pediatric reference population consisting of 16 pediatric patients with an even age distribution of between 2 and 12 years was created. In the first step, age and weight were extracted from the model-building data set of patients in the predefined age group. In the resulting cohort, patients younger than 8 years were underrepresented. Thus, to achieve an equally distributed reference population with respect to age, weights for fictive patients derived from WHO growth charts were added (50). The weights for the fictive patients were based on the tabulated data for the median and 5th and 95th percentiles of weight for 2-, 4-, and 6-year-olds.

Simulated dosing regimens.

On the basis of the data for the resulting reference population, simulations of different intravenous administration schedules with dosing for 672 h (4 weeks) were performed. Since the approved dosing regimen for children 2 to 12 years of age was modified to 8 mg/kg BID with a loading dose of 9 mg/kg BID on day 1 after the completion of the study, simulations with this modified dosing regimen served as a reference for simulations of alternative dosing schedules. In the first step, six different dosing regimens with TID loading and TID maintenance dosing were simulated. In the second step, combined TID and BID regimens were simulated, with the initial dosing being administered TID and the maintenance dosing being administered BID (Table 4). The TID/BID regimens were tested first with 24 h of initial TID dosing and then with prolongation of the initial TID dosing to 48 h and 72 h prior to the step-down to BID dosing.

TABLE 4.

Intravenous dosing regimens of voriconazole explored by Monte Carlo simulation for children 2 to ≤12 years of age

Regimena	Loading dose			Maintenance doseb
	Dose (mg/kg)	Interval (h)	Duration (days)	Dose (mg/kg)	Interval (h)
Approved BID regimen	9	12	1	8	12
TID/TID regimen 1	6	8	1	5.34	8
TID/TID regimen 2	7	8	1	6	8
TID/TID regimen 3	8	8	1	6	8
TID/TID regimen 4	8	8	1	7	8
TID/TID regimen 5	9	8	1	7	8
TID/TID regimen 6	9	8	1	6	8
TID/TID regimen 7	9	8	1	5	8
TID/BID regimen 1	9	8	1	8	12
TID/BID regimen 2	9	8	2	8	12
TID/BID regimen 3	9	8	3	8	12

^aBID, two times daily; TID, three times daily. First, seven different regimens with TID loading and maintenance dosing were simulated. In a second step, combined TID loading and BID maintenance regimens were simulated. The currently approved BID dosing regimen served as a reference.

^bDosing simulated for 648 h total.

In addition, voriconazole areas under the curve (AUCs) were calculated within the simulation step by integration of the drug concentration in the central compartment over time. Simulated AUCs included the AUC from time zero to 24 h on day 1 and on day 14 of treatment.

Evaluation of simulated data.

For the first 120 h (5 days) after administration of the first dose, full plasma concentration-time profiles were simulated. In addition, trough concentrations were simulated every 24 h for up to 672 h (4 weeks) following the first dose.

For evaluation of the probability of target attainment (in percent), three categories were created on the basis of the simulated trough concentrations: category 1 was trough concentrations of <1 mg/liter, considered inadequate exposure; category 2 was trough concentrations of between 1 and 6 mg/liter, considered the effective and safe exposure range (19); and category 3 was trough concentrations exceeding 6 mg/liter, representing the putative cutoff for potentially toxic exposures. For evaluation of target attainment at 24, 48, and 72 h after the start of TID loading doses, simulated values of trough concentrations prior to the switch from TID to BID dosing before the 4th dose, before the 7th dose, and before the 10th dose, respectively, were used.

Simulated AUCs were evaluated using the AUC/MIC ratio. In a dynamic in vitro model of invasive aspergillosis, an AUC/MIC ratio of 32.1 was associated with a nearly maximum effect (51). A MIC of 1 mg/liter, representing the epidemiological cutoff (ECOFF) value of voriconazole for Aspergillus fumigatus set forth by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (52), was used.