Voriconazole Pharmacokinetics and Pharmacodynamics in Children
- 1 Division of Pediatric Infectious Diseases, University of Southern California, Los Angeles, California
- 2 Laboratory of Applied Pharmacokinetics, Keck School of Medicine, University of Southern California, Los Angeles, California
- 3 Childrens Hospital Los Angeles, Los Angeles, California
- Reprints or correspondence: Dr Michael Neely, Childrens Hospital Los Angeles, 4650 Sunset Blvd, MS 51, Los Angeles, CA 90027 (mneely{at}usc.edu).
-
Presented in part: Meeting of the Population Approach Group Europe (PAGE), 23–26 June 2009, St Petersburg, Russia.
Abstract
Background Voriconazole pharmacokinetic and pharmacodynamic data are lacking in children.
Methods Records at the Childrens Hospital Los Angeles were reviewed for children with ⩾ 1 serum voriconazole concentration measured from 1 May 2006 through 1 June 2007. Information on demographic characteristics, dosing histories, serum concentrations, toxicity and survival, and outcomes was obtained.
Results A total of 207 voriconazole measurements were obtained from 46 patients (age, 0.8–20.5 years). A 2-compartment Michaelis-Menten pharmacokinetic model fit the data best but explained only 80% of the observed variability. The crude mortality rate was 28%, and each trough serum voriconazole concentration < 1000 ng/mL was associated with a 2.6-fold increased odds of death (95% confidence interval, 1.6–4.8; P p.002). Serum voriconazole concentrations were not associated with hepatotoxicity. Simulations predicted an intravenous dose of 7 mg/kg or an oral dose of 200 mg twice daily would achieve a trough >1000 ng/mL in most patients, but with a wide range of possible concentrations.
Conclusions We found a pharmacodynamic association between a voriconazole trough >1000 ng/mL and survival and marked pharmacokinetic variability, particularly after enteral dosing, justifying the measurement of serum concentrations.
Voriconazole is a triazole antifungal drug with activity against yeasts, endemic fungi, and certain molds, notably Aspergillus species, for which it is currently first-line therapy in adults [1]. In children, voriconazole use has increased steadily since 2002, accounting for ∼10% of prescribed systemic antifungal therapy in nearly 63,000 US pediatric inpatients in 2006 [2]. Strikingly, voriconazole has almost completely replaced the use of any amphotericin B formulation for the treatment of aspergillosis. However, to our knowledge there are only 2 peer-reviewed publications that document voriconazole pharmacokinetics obtained from prospective clinical trials in children <12 years of age [3, 4] and no studies to measure a pharmacodynamic link between concentration and outcome in children.
Although a single target trough voriconazole concentration to maximize efficacy has not been defined, several independent studies have demonstrated a relationship between higher voriconazole serum concentrations in adults and improved outcomes [5–8], such that a target range has emerged, and trough serum voriconazole concentrations are recommended to be 500–2000 ng/mL and <6000 ng/mL [9, 10].
Given the paucity of information on voriconazole pharmacokinetics in children, the known interindividual and intraindividual variability, and the associations between serum concentrations and outcomes in adults, voriconazole concentrations are frequently measured at our hospital as part of the clinical care of children receiving this drug. Therefore, we reviewed our therapeutic drug management experience in hospitalized patients with available voriconazole drug concentrations. Our goal was to refine our voriconazole dosing guidelines by achieving 3 aims: (1) to analyze associations between voriconazole concentrations and various patient outcomes, including toxic effects and survival; (2) to create a population model of voriconazole pharmacokinetics across all ages of childhood and adolescence; and (3) to select the optimal dose(s) based on concentration profiles simulated from the model.
Patients and Methods
Data collection
We reviewed the laboratory and pharmacy records at the Childrens Hospital Los Angeles (CHLA), a tertiary care, 324-bed pediatric referral hospital with an electronic medical record system. Eligible patients were identified as those who had voriconazole serum concentrations measured during hospitalization for a 1-year period from 1 May 2006 through 1 June 2007. This project was approved by the institutional review board with a waiver of informed consent for anonymous data collection and analysis. Extracted data included demographic information; voriconazole serum concentrations; timing, amount, route, and formulation of voriconazole doses; underlying and infectious diagnoses; concomitant use of drugs known or strongly suspected to influence voriconazole serum concentrations; alanine transferase (ALT) and alkaline phosphatase (ALKP) concentrations; and mortality. Consistent with definitions used in previous adult and pediatric studies [8], we defined hepatotoxicity as a 1.5-fold to 5-fold increase in ALT or ALKP level over baseline, depending on the age-specific degree of elevation at baseline.
Voriconazole assay
Serum concentrations of voriconazole were determined by Focus Diagnostics, Inc., using a validated high-performance liquid chromatography assay. Voriconazole was extracted from serum and concentrations determined using a C18 column, autosampler, column oven, degasser, UV-VIS diode array detector and Chemstation software. The serum standard curve for voriconazole was linear through a concentration range of ∼200–30,000 ng/mL. The mean recovery from serum was 94%. The validation interassay coefficient of variation was < 5%. No interference was detected with 14 other antimicrobial agents, including 4 additional antifungal agents. Turnaround time ranged from 7 to 10 days.
Data analysis
Descriptive and analytic statistics and plots were generated using the open-source software R, version 2.9.2 (R Foundation; http://cran.r-project.org). For survival and hepatotoxicity analyses, Cox proportional hazards models were constructed to examine associations with measures of voriconazole exposure, all adjusted for time receiving therapy.
Because of the retrospective nature of the study, we performed a population analysis to estimate group and individual voriconazole pharmacokinetic parameters (eg, clearance and volume of distribution) and the variability in those parameters. Population analysis was essential given the diversity in number and timing of samples from each patient, as well the dose amounts, routes, and other clinical characteristics of the patients. For the population analysis, we used the nonparametric adaptive grid modeling component of the MM-USCPACK software (available by license from the University of Southern California; http://www.lapk.org) to test the fit of the observed data to various candidate models, including the one used by Karlsson et al [4].
All models included weight as a covariate on volume and clearance terms. In addition, we explored the influence on model variables of age, sex, creatinine clearance, ALT level, and ALKP level. We did not include the CYP2C19 genotype in the model, which is known to be associated with fast, intermediate, and slow phenotypes in voriconazole clearance [11], because this information was not available. However, nonparametric modeling excels at capturing subpopulations such as fast and slow metabolizers because no parameter distribution (eg, normal) is assumed; therefore, the absence of the CYP2C19 data was unlikely to affect the model fit except for increasing the variability in parameter estimates. Models were evaluated on the basis of the calculated log-likelihood and inspection of observed versus predicted plots.
MM-USCPACK was also used to simulate the effect of differing voriconazole dosing strategies on the predicted range of concentration profiles: 10,000 sets of parameter values were randomly selected, with replacement, from the weighted, multivariate distribution of parameter values in the final model, including the full covariance matrix, and predicted concentrations were calculated for each parameter set.
Results
Study population and sample characteristics
From 1 May 2006 until 1 June 2007, a total of 207 voriconazole concentrations were measured from 46 CHLA patients (Table 1). As indicated in Table 1, there was a wide range in doses and serum voriconazole concentrations, both between and within individual patients. Overall, 90% of the doses were administered enterally versus intravenously.
View this table:
Table 1
Characteristics of 46 Patients with 207 Voriconazole Measurements
Of the 207 samples, 99 (48%) were obtained after outpatient dosing with no verified preceding dose time; these were excluded from population modeling because time after dose was a necessary factor. Twenty-six (13%) of the 207 samples were reported as below the limit of quantification of 200 ng/mL, all of which were obtained after outpatient dosing. Because the mean outpatient dose of 6.4 mg/kg was not significantly different from the mean inpatient enteral dose of 5.8 mg/kg (P p.14 ), it is possible that suboptimal adherence contributed to these low concentrations.
For the remaining 108 inpatient samples, the median sampling time was 9.0 h (range, 1.3–36.0 h) after the preceding dose, reflecting the clinical practice of monitoring trough serum voriconazole. These samples, from 40 patients, were used for modeling.
Of the medications listed in the voriconazole package insert known to alter voriconazole serum concentrations, only 2 patients also received phenytoin, which decreases serum voriconazole levels; none received other medications known to alter serum voriconazole levels.
Toxic effects
Overall, 16 (42%) of 28 patients with baseline and on-therapy ALT or ALKP measurements met the definition for significant hepatoxic effects by ALT level (10 patients), ALKP level (2 patients), or both (4 patients). The median increase in ALT level over baseline was 8.3-fold (range, 3.7–30.5-fold). The ALT level peaked 63 days (range, 5–238 days) days after starting voriconazole therapy and resolved to baseline 49 days (range, 11–461 days) later in 71%, at least two-thirds of whom were still receiving voriconazole. The median increase in ALKP level over baseline was 6.6-fold (range, 5.2-fold to 12.7-fold). The ALKP level peaked 69 days (range, 33–116 days) after starting voriconazole therapy and was not documented to resolve to baseline in any patients by the end of the data extraction period. By Cox proportional hazards analysis, the risk of hepatotoxicity was not significantly associated with voriconazole dose, area under the curve, average concentration, or maximum or minimum concentration.
Survival
Of the 46 patients, 12 (26%) had proven invasive fungal disease [12], with a positive fungal culture from a normally sterile site and radiographic and/or clinical evidence of disease at that site. An additional 7 patients (15%) had probable disease [12], with at least 1 positive fungal culture from respiratory secretions (eg, deep endotracheal suction or bronchoalveolar lavage) and/or galactomannan test, and clinical or radiographic evidence of localized disease. Recovered organisms included Aspergillus species (Aspergillus fumigatus and Aspergillus versicolor), Alternaria species, Scedosporium apiospermum, and Candida species (Candida albicans and Candida glabrata).
As indicated in Table 1, the crude mortality rate in this population was 28%. Seventy-five percent of those who died had at least 1 low voriconazole trough versus only 20% of those who survived. Adjusted for annualized observation time and peak fold-increase in ALT over the age-specific upper limit of normal, each voriconazole trough concentration < 1000 ng/ mL increased the relative risk of death by 6.3-fold (95% confidence interval, 1.6–24.0-fold; P p.008, by Cox proportional hazards).
Modeling: nonparametric population analysis
The final population model was a 2-compartment model with absorption after a delay and Michaelis-Menten elimination, which implies a constant amount of drug cleared per unit of time. As shown in Figure 1, overall, the model explained 80% of the observed variability in individual voriconazole concentrations, with no difference in accuracy according to dosing route, despite the small number of samples obtained after intravenous dosing. The model underpredicted observed concentrations by an average of 36% at the lower limit of the assay (200 ng/mL) and only 3% at the highest concentrations. The 5 most underpredicted concentrations occurred in 3 children <12 years of age, with 3 of them from a single 9-year-old child with unexpectedly high trough concentrations.
Figure 1
Observed versus individual model-predicted concentrations. Solid line, linear regression through the points; dashed line, line of unity; triangles, patients aged ≥ 12 years; squares, patients aged <12 years; filled symbols, concentrations obtained after enteral dosing; open symbols, concentrations obtained after intravenous dosing.
Estimates for the population model variables are detailed in Table 2. Because nonparametric analysis does not assume any distribution in the values of model variables (such as a normal or gaussian distribution), means and standard deviations can obscure information. Therefore, Figure 2 shows the distributions of values in the patients for each variable in the model. Volumes and clearances were indexed to weight in kilograms, and there was a modestly significant additional effect of age (Table 2 and Figure 2). Sex, race, ethnicity, and ALT and ALKP concentrations were not significantly associated with mean values of any of the variables in the model. Inclusion of all of these covariates accounted for only 10%–40% of the variability, suggesting that unmeasured covariates, such as CYP219C genotype, disease state, gut motility, or others, were important.
Figure 2
Distribution of individual values for nonparametric model variables in the study population. Vertical dashed line, geometric mean for patients aged ⩾ 12 years; dashed-dotted line, geometric mean for patients aged < 12 years. P values are for the difference between the geometric means according to age. CL, clearance; F, bioavailability; Ka, rate constant of absorption; Km, Michaelis-Menten constant; Q, intercompartmental clearance; Tlag, lag (delay) before absorption; Vc, volume of the central compartment; Vp, volume of the peripheral compartment.
View this table:
Table 2
Geometric Mean (GM) and Geometric Relative Standard Error (GRSE%) of Voriconazole Pharmacokinetic Parameter Estimates in Patients Stratified by Age
Simulation scenarios
Figure 3 shows the distribution of 10,000 model-simulated steady-state concentration-time curves after dosing 5.8 mg/kg every 12 h, with 10% administered intravenously (the median dose and percentage of intravenous doses in the CHLA population). The median simulated 12-h trough concentration was 1315 ng/mL, and the 5th and 95th percentiles were 242 and 6411 ng/mL, respectively. Included in Figure 3 are the observed voriconazole concentrations in the CHLA patients, stratified symbolically by age 112 years and dosing route. These observed concentrations have been normalized to the median dose of 5.8 mg/kg and condensed to a single dosing interval of 12 h to enable comparison with the simulated concentrations. On the basis of the observed versus predicted voriconazole concentrations in Figure 1 and Figure 3, the model was judged a good representation of the data and appropriate for further simulations.
Figure 3
Simulated voriconazole concentration-time plot. Shaded bands are the indicated percentiles of 10,000 concentration-time profiles simulated from the final nonparametric model. Triangles, measured concentrations obtained in patients aged ⩾ 12 years; squares, concentrations from patients aged < 12 years; filled symbols, concentrations obtained after enteral dosing; open symbols, concentrations obtained after intravenous dosing; horizontal dashed line, target minimum concentration of 1000 ng/mL. Concentrations have been normalized to the median population dose, which was the same dose used for the simulations; hence, 2 measured concentrations appear to be at or below the assay limit of quantification (200 ng/mL).
With a regimen of 7 mg/kg intravenously every 12 h, as approved for use in Europe [13], the median simulated steady-state trough concentration was 1287 ng/mL, with the 5th and 95th percentiles of 176 and 3406 ng/mL; 66% of patients with this dose would be expected to be above a trough concentration of 1000 ng/mL. Figure 4 shows the simulated range of steady-state voriconazole trough concentrations after the recommended European enteral dose of 200 mg twice daily at all ages from 2 to 12 years. The simulated median trough concentration ranges from 3600 ng/mL at the age of 2 years to 880 ng/mL at the age of 12 years, although there is significant variability around these median values. As a fixed dose, 200 mg can result in doses in excess of 15 mg/kg in the youngest children. The highest enteral dose in our population was 12.9 mg/kg, so the model may not be generalizable to higher doses.
Figure 4
Predicted voriconazole trough concentrations based on 1000 simulated patients at each age at the licensed European dose of 200 mg enterally every 12 h. Lines are (top to bottom) the 95th, median, and 5th percentiles of simulated concentrations. The target trough concentration is the horizontal reference line at 1000 ng/mL.
Finally, the relationship between the median simulated voriconazole trough concentration and enteral dose is shown in Figure 5. There is a constant increase of ∼520 ng/mL for every increase in dose of 1 mg/kg, which is consistent with the Michaelis-Menten kinetics of the drug. For example, a 50% increase in dose from 4 to 6 mg/kg does not increase the median trough by 50% but from 750 to 1800 ng/mL, an increase of 240%.
Figure 5
Median simulated trough voriconazole concentration for a given enteral dose, showing the regression line with associated P value.
Discussion
The data from the present report indicate that monitoring serum concentrations is important in children. A statistically significant association was found between crude mortality and voriconazole 12-h trough concentrations <1000 ng/mL, even in this relatively small population. This breakpoint is consistent with several other investigations [5–7, 14, 15].
There was also a great deal of between-patient and inter-occasion pharmacokinetic variability, as noted by Karlsson et al [4]. Failure to account for interoccasion variability increases model error and introduces biases in some parameter estimates [16]. Because 98% of the samples used for our model were obtained during different dosing intervals, in some cases during a period of months, and often after different dose amounts or routes, we were unable to model the changes in parameter values within a single individual over time. This likely contributed to the large variability in estimates of pharmacokinetic parameters (Table 2). Despite this, simulation of 10,000 concentration profiles closely matched the observed distribution of concentrations in the CHLA patients (Figure 3).
Significant features of our model that differ from that of Karlsson et al [4], as indicated in Table 2, are the delay in voriconazole absorption after an oral dose and the higher bioavailability. These differences may be due to our relative lack of samples early in the dosing interval and within-individual, interoccasion variability. Our oral bioavailability was ∼80%, compared with 96% reported for adults in the package insert and 45% reported by Karlsson et al [4]. Although not known to be pH dependent, absorption of voriconazole is reduced by food [17], which may explain some of the differences. Because the calculation of bioavailability depends on ratios of achievable concentrations after enteral and intravenous dosing, the small number of samples obtained after intravenous dosing in our population may have caused bias. However, our estimate was closer to that in the package insert, and we are unaware of any drug that has as large a discrepancy in bioavailability as that reported by Karlsson et al [4] that could be attributed to age alone.
On the basis of our simulations, an intravenous dose of 7 mg/kg every 12 h in children up to the age of 12 years, as approved in Europe, is likely to achieve a voriconazole trough concentration of >1000 ng/mL in about two-thirds of children and appears to be a good starting dose. The recommended European enteral dose of 200 mg every 12 h, regardless of age or weight, was chosen to best match the voriconazole exposure (area under the curve) in children to that in adults who also receive 200 mg every 12 h. The somewhat surprising dose equivalence, which can result in doses in excess of 15 mg/kg in the youngest children, is due to the single report of voriconazole's markedly decreased bioavailability and increased clearance in children relative to adults [4].
Despite the higher bioavailability in our data, simulation predicts that a fixed dose of 200 mg will result in reasonable trough concentrations close to 1000 ng/mL in many children. However, physicians should be aware that concentrations may be markedly above or below this target because of the variability in absorption. Furthermore, we are not aware of any actual reported experience of children receiving an enteral dose as great as 18 mg/kg. The largest enteral dose in the CHLA patients was 12.9 mg/kg twice daily (with associated trough concentrations of 13,200 ng/mL and 4200 ng/mL on different occasions), and there is 1 report of a 4-year-old child who tolerated 13.3 mg/kg twice daily based on measured serum concentrations [18]. The tolerability and absorption of higher doses are not reported in the peer-reviewed literature, although a prospective trial to study this dose is under way.
Given the uncertainty associated with the pharmacokinetics of enteral voriconazole in children, it may be appropriate to prolong intravenous therapy until there is clear evidence of sustained improvement. These doses should be considered as initial, with measurement of serum trough concentrations at a minimum, followed by rational dosage adjustment as necessary. Our data suggest that for every increase in dose of 1 mg/kg, serum concentrations will increase ∼520 ng/mL. However, measured concentrations outside the desired range should prompt a careful examination of other factors, such as fasting status, concomitant medication use, or organ function.
We did not find any significant association between voriconazole concentrations and ALT or ALKP levels. One-half of the CHLA children started voriconazole therapy with elevated ALT and ALKP levels, and further elevations after starting voriconazole therapy were common, regardless of baseline values. However, most patients returned to baseline ALT levels or below by 2–4 weeks after peak values, without interrupting voriconazole therapy.
The obvious limitation of this study is the retrospective design. We did not control for which children had concentrations measured or any other factor associated with voriconazole therapy. However, population modeling is the analytic technique of choice under these circumstances because the individual doses, sample times, sample numbers, serum concentrations, and patient characteristics are all included in the modeling process. It is reassuring that our findings do not differ from the published prospective study [4] in terms of recommendations for initial dosing.
In conclusion, we wish to emphasize 2 major points for the physician who uses voriconazole therapy in children. First, we found a significant survival benefit when voriconazole trough concentrations were >1000 ng/mL. To our knowledge, this is the first such demonstration in children, and it is consistent with evidence in adults. Second, this study confirms that voriconazole pharmacokinetics in children are highly variable, particularly with the enteral formulations. Although 7 mg/kg intravenously and 200 mg enterally every 12 h may be reasonable starting doses, we suggest, as has been done for adults, that voriconazole concentrations be measured and the dose adjusted to maintain a trough of at least 1000 ng/mL. This strategy is especially important before the prospective validation of these doses that is currently under way.
Acknowledgments
We sincerely thank Rosa Cruz for data abstraction.
Financial support. The National Institute of Allergy and Infectious Diseases (1 K23 AI076106–01 to M.N.) and The National Institute of Biomedical Imaging and Bioengineering (R01 EB005803–01A1 to M.N. and R.J.).
Potential conflicts of interest. M.N. is on the speaker's bureau for Virco and Merck and has attended an advisory meeting for Virco. All other authors: no conflicts.
- Received March 25, 2009.
- Accepted August 17, 2009.
- Final version accepted December 1, 2009.
- © 2010 by the Infectious Diseases Society of America
References
This Article
- AbstractFree
- » Full Text (HTML)Free
- Full Text (PDF)Free
Classifications
-
- Articles and Commentaries
Services
Citing Articles
Google Scholar
PubMed
Related Content
Share
- Email this article
-
CiteULike
Connotea
Delicious
Facebook
Google+
Mendeley
Twitter
Published on behalf of
IDSA Members: For your free access to this journal, log in via the IDSA members area.
The Journal
Impact Factor: 8.186
5-Yr impact factor: 7.898
For The Media
For Authors
- Services for authors
- Self-archiving Policy
- IDSA Conflict of Interest policy
- Instructions for Authors
- ICMJE Form
- Submit Now!
Open access options for authors visit Oxford Open
This journal enables compliance with the NIH Public Access Policy
Corporate Services
Alerting Services
Related IDSA Publications
