Serial Plasma Voriconazole Concentrations after Allogeneic Hematopoietic Stem Cell Transplantation

Plasma voriconazole concentrations vary considerably betweenpatients receiving standard dosing, and trough voriconazoleconcentrations are known to affect efficacy and toxicity. Temporalvariations in serial plasma voriconazole concentrations throughthe course of therapy in hematopoietic stem cell transplantationpatients has not been carefully described. Paired voriconazoleconcentrations in 64 patients were studied to determine thepredictability of the second concentration based on the first.The difference between the two values was $≤$ 5% in six patients.In 25 patients, the second concentration was higher by a medianof 40%. In 33 patients, the subsequent concentration was lowerby a median of 59%. For patients with an initial concentrationof <2 µg/ml, the correlation between the two valueswas poor (r = 0.24; P < 0.17). For those with an initialconcentration of $≥$ 2 µg/ml, the correlation was good (r= 0.72; P < 0.0001). There was no relationship between themagnitude of the change and the time elapsing between the twomeasurements. Among the 43 patients who had an initial concentrationof $≥$ 1 µg/ml, the two voriconazole measurements were stronglycorrelated (r = 0.66, P < 0.0001), but only 67% had a voriconazoleserum concentration of $≥$ 1 µg/ml on the second measurement.No studied variables were reliable predictors in identifyingconcentrations above or below 1 or 2 µg/ml. Our data suggestthat variations in voriconazole concentrations are unpredictabledespite standard dosing, and the acceptability of a concentrationon one occasion cannot be extrapolated to future concentrationsin the same patient. This suggests that ongoing therapeuticdrug monitoring and dose adjustment may be beneficial in patientsrequiring prolonged voriconazole therapy.

INTRODUCTION

Top

INTRODUCTION

Invasive fungal infections are a significant source of morbidityand mortality following allogeneic hematopoietic stem cell transplantation(HSCT). Voriconazole, a triazole antifungal agent, is frequentlyused in the treatment of many yeast and mold infections followingHSCT (2, 7). Use of voriconazole for prevention of opportunisticfungal infections has also shown merit (16). Voriconazole ismetabolized by the cytochrome P450 isoenzyme system, primarilyCYP 3A4, CYP 2C9, and CYP 2C19 (3). Genetic polymorphism ofthe CYP 2C19 isoenzyme, interactions with several drugs usedduring the posttransplantation period, self-induced voriconazolemetabolism, and direct hepatotoxicity from the conditioningregimen can alter the metabolism and disposition of voriconazolesignificantly (1, 8). This may increase or decrease voriconazoleexposure in an unpredictable fashion, and large inter- and intrapatientvariations in voriconazole plasma concentrations have been observedin several reports (11, 12, 14). Recent studies have suggestedthat there is an association between voriconazole plasma concentrationsand successful treatment outcomes, indicating a possible needfor therapeutic drug monitoring (TDM) (5, 9, 13).

Although the optimum duration of voriconazole prophylaxis andtherapy for most fungal infections has not been defined clearly,most allogeneic HSCT recipients require prolonged antifungaltherapy—often for several months. Since physiologic (e.g.,absorption, metabolism, and protein binding) and pharmacologic(e.g., drug interactions and nonlinear kinetics) conditionschange over time following HSCT, plasma voriconazole concentrationswould be expected to vary.

The temporal variation of plasma voriconazole concentrationsthrough the course of therapy in HSCT patients when measuredserially has not been carefully described. We report the resultsof paired plasma voriconazole measurements in 64 patients followingallogeneic HSCT.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Adult patients with hematologic malignancies who had undergoneallogeneic HSCT at Northwestern Memorial Hospital between June2004 and January 2007 and received voriconazole for prophylaxisof invasive fungal infection were identified. Paired voriconazoletrough plasma concentrations were systematically drawn 30 minprior to the next dose from HSCT recipients. Concentrationswere quantified by standardized high-pressure liquid chromatographyassay at the University of Texas at San Antonio Health ScienceCenter (6).

Patients were included in the analysis if the paired measurementsfulfilled the following criteria: (i) patients were hospitalizedfor the entire study period, (ii) there were identical voriconazoledoses for both measurements, (iii) both plasma concentrationswere measured a minimum of 6 days after voriconazole therapywas initiated (to assess plasma steady-state concentrations),(iv) only oral voriconazole was administered during the studyperiod (i.e., no intravenous voriconazole), and (v) no patientreceived voriconazole during the study period for treatmentof invasive fungal infection (i.e., only patients receivingprophylactic voriconazole were included).

Although all patients experienced some degree of mucositis anddiarrhea, documentation was not consistent enough to allow classificationaccording to World Health Organization or National Cancer Institutegrading criteria, and thus these parameters could not be analyzed.A clinical pharmacist verified drug compliance with the prescribedregimen via medication administration records. Our cohort ofallogeneic HSCT recipients did not receive any other drugs knownto interact with voriconazole disposition or metabolism. Thisretrospective study was reviewed and approved by the NorthwesternUniversity Institutional Review Board.

Statistical methods. Correlation analysis was used to determine the test-retest reliabilityof the two voriconazole measurements and to evaluate the associationbetween voriconazole measurements, inter-recording interval,and weight. A 95% confidence interval for proportions was constructedfor the proportion of samples that fell below 1 or 2 µg/mlon the second test. Optimal data analysis was used to statisticallydiscriminate patients with voriconazole plasma concentrationsthat did or did not fall below 1 or 2 µg/ml on the secondtest (17). The reason for choosing 1 to 2 µg/ml as a thresholdwas evidence that this plasma concentration may influence efficacy(5, 9, 13).

RESULTS

Top

RESULTS

All patients (n = 64) had hematologic malignancies (40 leukemia,16 lymphoma, 6 myeloma, and 2 myelodysplastic syndrome). Voriconazolewas given orally to all patients in the present study at a doseof 200 mg twice daily in 81% of the patients, 300 mg twice dailyin 13% of the patients, and other doses in 6% of the patients.The drug dose based on the actual body weight was 2.7 to 12.0mg/kg (median, 5.8 mg/kg). Patients received voriconazole forprimary (80%) or secondary prophylaxis (20%). Thirty-six hadreceived a conventional-intensity conditioning regimen, andtwenty-eight had undergone a reduced-intensity allograft. Thetime interval between the two plasma concentrations was 1 to252 days (median, 15 days).

Table 1 shows the paired plasma concentrations. The differencebetween the two plasma concentrations was $≤$ 5% in six patients.In 25 patients, the plasma concentration increased by 6 to 1,865%(median, 40%). In 33 patients, the plasma concentration declinedby 11 to 96% (median, 59%). This is graphically illustratedin Fig. 1 (where two increases of 1,865 and 650% have not beencharted due to scale constraints). When all 128 readings arepooled, the correlation between the drug dose based on bodyweight and drug plasma concentration was modest (r = 0.25) (Fig.2).

View this table:
[in this window]
[in a new window]

TABLE 1. Voriconazole plasma concentrations

View larger version (10K):
[in this window]
[in a new window]

FIG. 1. Percent change in voriconazole plasma concentration in each of the 64 patients (the two highest changes—1,865 and 650%—have not been charted due to constraints of scale). Positive numbers represent an increase from the first to the second level and negative numbers represent a decline.

View larger version (14K):
[in this window]
[in a new window]

FIG. 2. Correlation between voriconazole plasma concentration and drug dose in mg/kg. All levels have been pooled (n = 128).

For patients with an initial plasma concentration of <2 µg/ml,the correlation between the initial and the subsequent plasmaconcentration was poor (r = 0.24; P = 0.17). For those withan initial plasma concentration of $≥$ 2 µg/ml, the correlationwas good (r = 0.72; P < 0.0001) (Fig. 3) and approximateda 1:1 relationship. However, even in the latter population,the subsequent voriconazole plasma concentration was <2 µg/mlin almost half the patients.

View larger version (12K):
[in this window]
[in a new window]

FIG. 3. Correlation between the initial and subsequent voriconazole plasma concentration. There is a lack of correlation if the initial level is <2 µg/ml and significant correlation if the initial level is $≥$ 2 µg/ml.

A lack of obvious relationship between the magnitude of thechange and the time elapsing between measurement of the twoplasma concentrations was observed (Fig. 4A). In the entiregroup, the median change from the first plasma concentrationwas a 22% decline. Figure 4B shows the lack of any obvious relationshipbetween the magnitude of the change and the drug dose basedon patient weight.

View larger version (12K):
[in this window]
[in a new window]

FIG. 4. (A) Percent change in voriconazole plasma concentration by the time interval between the two levels, showing no obvious relationship. Time between levels has been shown on a log scale to depict individual data points better. (B) Percent change in voriconazole plasma concentration by the drug dose based on body weight, showing no obvious relationship.

Among the 29 patients who had a voriconazole plasma concentrationof $≥$ 2 µg/ml on the first measurement, 18 (62%; 95% confidenceinterval, 34 to 80%) had a voriconazole plasma concentrationof $≥$ 2 µg/ml on the second measurement. In these 29 patients,correlations did not exist between plasma concentrations andthe time interval between the measured concentrations (P = 0.60)or patient weight (P = 0.49). Optimal data analysis revealedthat all 10 patients with an initial voriconazole plasma concentrationof $≥$ 4.6 µg/ml had a second voriconazole plasma concentrationof $≥$ 2 µg/ml (model positive predictive value of 100%).This finding was consistent when jackknife validity analysiswas conducted, suggesting that this result is likely to cross-generalizeto an independent random sample.

The optimal data analysis model predicted that when a patient'sinitial plasma concentration is <4.6 µg/ml, the secondplasma concentration will be <2 µg/ml. All 11 patientswho in reality had a second plasma concentration of <2 µg/mlwere correctly predicted by the optimal data analysis modelto have a second plasma concentration <2 µg/ml (modelspecificity of 100%). The model was, however, not perfect, andeight patients with an initial plasma concentration of <4.6µg/ml had a second plasma concentration of $≥$ 2 µg/ml,yielding model negative predictive value of 58% and a sensitivityof 56%.

For the sample of 43 patients who had a voriconazole plasmaconcentration of $≥$ 1 µg/ml on the first measurement, thetwo voriconazole measurements were strongly correlated (r =0.66, P < 0.0001). Plasma concentration was not significantlycorrelated with the inter-recording time intervals (P = 0.21)or weight (P = 0.62). Of the total of 43 patients, 33 (67%)had a voriconazole plasma concentration of $≥$ 1 µg/ml onthe second measurement. The 95% confidence interval for theproportion of patients with a second voriconazole recordingbelow 1 µg/ml, regardless of the inter-recording timeinterval, was 10.5 to 36.0%. The initial voriconazole plasmaconcentration (P = 0.40), the intermeasurement interval (P =0.90), or the weight (P = 0.77) were not statistically reliablediscriminators of patients whose second voriconazole plasmaconcentration fell below 1 µg/ml.

DISCUSSION

Top

DISCUSSION

This report differs from previous efforts in that we assessedpaired voriconazole plasma concentrations. This study confirmsour previous reports of significant inter- and intrapatientvariability in voriconazole disposition at standard doses usedin practice (11, 12), Of the 128 obtained plasma concentrations,30 were included in our first report of voriconazole plasmaconcentrations after HSCT, and 84 were included in our secondreport of variability in voriconazole plasma concentrationsafter HSCT (11, 12). Statistical techniques were applied todetermine whether, despite the inter- and intrapatient variations,it was possible to predict subsequent changes in a patient'svoriconazole plasma concentration based upon the first plasmaconcentration. Our data suggest that with current assessed variables,it is not possible to predict subsequent voriconazole concentrationswith any degree of clinical certainty.

TDM is recommended to improve the safety and efficacy of drugtherapy when drugs exhibit wide inter- or intrapatient variabilityand there is a relationship between measured concentrationsand safety or efficacy. A lack of prospective studies has hamperedour understanding of the role of TDM for oral antifungal therapy.However, growing evidence suggests that there is a relationshipbetween voriconazole plasma concentrations for efficacy as wellas toxicity (5, 9, 13). Hence, these findings implicate a rolefor TDM at least in some patients, and recently formulated evidence-basedguidelines for the management of aspergillosis now recommendvoriconazole TDM, along with clinical assessment for optimaltreatment efficacy and safety (15).

Two retrospective studies have shown a correlation between voriconazoleplasma concentrations of $≥$ 2 µg/ml and successful treatmentoutcomes for patients with aspergillosis (9) or prevention ofbreakthrough Candida glabrata infections (13). Recently, a prospectivestudy of 52 patients with invasive fungal infections which incorporatedTDM to adjust voriconazole dosing to achieve plasma concentrationsbetween 1 and 5.5 µg/ml found significantly more favorableoutcomes among patients whose plasma concentrations were maintainedwithin the target range. All six patients with subtherapeuticvoriconazole plasma concentrations and inadequate clinical responsehad successful outcomes when the voriconazole dose was increasedto attain plasma concentrations in the target range. Increasedtoxicity was also seen as voriconazole plasma concentrationsincreased beyond 5.5 µg/ml (5).

Voriconazole may display nonlinear pharmacokinetics, and ittakes approximately 5 days to reach steady-state plasma concentrations(10). We conservatively measured plasma concentrations 6 to7 days after starting therapy. Interestingly, we found thatthe voriconazole plasma concentrations drawn 6 days after theinitiation of therapy tend to be higher than plasma concentrationsmeasured at a later date in the same patient. This suggeststhat reliance upon a single early plasma concentration couldoverestimate the proportion of patients in whom drug plasmaconcentrations are therapeutic. This undesirable change in voriconazoleplasma concentrations out of the target range could not be predictedby patient weight or the time interval between the two voriconazolemeasurements. This report is consistent with the findings ofprevious studies; that there are large and unpredictable inter-and intrapatient variations in voriconazole plasma concentrationswith doses commonly used in clinical practice (11, 12, 14).

Finally, consistent with prior reports (4), ca. 20% of the voriconazoleplasma concentrations were undetectable, and 35% were belowthe MICs for mold infections. Neither of these outcomes couldbe predicted by dose, weight, or time interval between measurements.The cause of the decline in voriconazole plasma concentrationsover time is probably multifactorial and may include changesin absorption, patient protein status, liver function, and diseasemodifying effects (e.g., graft-versus-host disease).

There are limitations with our study. It is quite possible thatsome of the changes in plasma concentrations were the resultof changes in concomitant therapy or changes in patient variables.However, all concomitant therapy was solely dictated by clinicalneeds. Even with a prospective study, most patients would requireextensive adjunct therapies due to the nature of the underlyingillness and intervention. Thus, the data, despite being retrospective,were acquired under circumstances routinely seen in clinicalpractice. The results are from a single institution in a definedpatient population and may not necessarily be entirely applicableto other patient populations.

Conclusions. Large intrapatient variability in plasma voriconazole concentrations(especially in patients with initially low measured plasma concentrations)suggests the need for repeated TDM in subjects requiring prolongedvoriconazole administration.

ACKNOWLEDGMENTS

S.M.T. attended a Pfizer-sponsored conference in 2007 to discussthe role of monitoring voriconazole plasma concentrations.

FOOTNOTES

* Corresponding author. Mailing address: 676 N. St. Clair St., Ste. 850, Chicago, IL 60611. Phone: (312) 695-4443. Fax: (321) 695-6189. E-mail: j-mehta{at}northwestern.edu

Published ahead of print on 17 February 2009.

REFERENCES

Top