In Vitro Pharmacokinetic/Pharmacodynamic Modeling of Voriconazole Activity against Aspergillus Species in a New In Vitro Dynamic Model

Abstract

The pharmacodynamics (PD) of voriconazole activity against Aspergillus spp. were studied using a new in vitro dynamic model simulating voriconazole human pharmacokinetics (PK), and the PK-PD data were bridged with human drug exposure to assess the percent target (near-maximum activity) attainment of different voriconazole dosages. Three Aspergillus clinical isolates (1 A. fumigatus, 1 A. flavus, and 1 A. terreus isolate) with CLSI MICs of 0.5 mg/liter were tested in an in vitro model simulating voriconazole PK in human plasma with C_max values of 7, 3.5, and 1.75 mg/liter and a t_1/2 of 6 h. The area under the galactomannan index-time curve (AUC_GI) was used as the PD parameter. In vitro PK-PD data were bridged with population human PK of voriconazole exposure, and the percent target attainment was calculated. The in vitro PK-PD relationship of fAUC_0-24-AUC_GI followed a sigmoid pattern (global R² = 0.97), with near-maximum activities (10% fungal growth) observed at an fAUC_0-24 (95% confidence interval [CI]) of 18.9 (14.4 to 23.1) mg · h/liter against A. fumigatus, 26.6 (21.1 to 32.9) mg · h/liter against A. flavus, and 36.2 (27.8 to 45.7) mg · h/liter against A. terreus (F test; P < 0.0001). Target attainment for 3, 4, and 5 mg/kg-of-body-weight voriconazole dosages was 24% (11 to 45%), 80% (32 to 97%), and 93% (86 to 97%) for A. fumigatus, 12% (5 to 26%), 63% (17 to 93%), and 86% (73 to 94%) for A. flavus, and 4% (2 to 11%), 36% (6 to 83%), and 68% (47 to 83%) for A. terreus. Based on the in vitro exposure-effect relationships, a standard dosage of voriconazole may be adequate for most patients with A. fumigatus but not A. flavus and A. terreus infections, for which a higher drug exposure may be required. This could be achieved using a higher voriconazole dosage, thus highlighting the usefulness of therapeutic drug monitoring in patients receiving a standard dosage.

INTRODUCTION

Invasive aspergillosis is the most serious infection caused by Aspergillus species, particularly in patients with hematological malignancies or bone marrow transplantation (16). These infections are characterized by high morbidity and mortality despite available antifungal therapy (7), among which voriconazole is the drug of choice (24). Several factors may influence the clinical outcome, related to the host (underlying condition and immunosuppression), drug (timing of administration and suboptimal exposure), and pathogen (resistance and virulence). Although voriconazole-resistant isolates have been described, their low prevalence (<10%) cannot explain the high mortality of these infections (10, 21). Testing of large sets of Aspergillus isolates showed that more than half of them exhibited similar in vitro susceptibilities to voriconazole, with a MIC of 0.5 mg/liter (5, 17). Furthermore, the most frequently isolated species, A. fumigatus, A. flavus, and A. terreus, demonstrated similar in vitro susceptibilities, although in vivo experimental and clinical data show that the efficacy of voriconazole differs for these species (22–24, 26). This lack of association between voriconazole MICs and clinical outcome implies that a different approach is required to correlate in vitro-in vivo voriconazole responses against Aspergillus infections in order to optimize therapeutic regiments (2).

The MIC as a marker of the in vitro activities of antifungal agents against Aspergillus spp. is an important pharmacodynamic parameter, but it does not account for other pharmacodynamic properties of antifungal drugs, such as kinetics of growth inhibition, sub-MIC effects, rate and extent of killing, and postantifungal effect. In addition, during MIC testing, drug concentrations remain stable over time when the fungus is exposed in vivo to nonconstant drug concentrations due to absorption, distribution, metabolism, and excretion processes. The effect of decreasing drug concentrations on Aspergillus spp. was unknown until recently, when we developed an in vitro model simulating human pharmacokinetics of antifungal drugs, including voriconazole (11). This novel pharmacokinetic/pharmacodynamic (PKPD) model demonstrated a differential activity of simulated standard dosages of antifungals against Aspergillus isolates with identical MICs. With this model, the time- and concentration-dependent pharmacodynamic properties of antifungal drugs can be studied and PKPD analysis simulating human pharmacokinetics can be performed.

We therefore studied the effect of simulated human pharmacokinetics of increasing voriconazole dosages against the most frequently encountered clinical isolates of A. fumigatus, A. flavus, and A. terreus with voriconazole MICs of 0.5 mg/liter. The magnitude of PKPD parameters associated with maximal activity was determined for each species. Finally, in vitro PKPD data were bridged with previously obtained human PK data in order to assess the efficacy of clinically administered voriconazole dosages against these life-threatening infections.

MATERIALS AND METHODS

Strains.

Three clinical strains of A. fumigatus, A. flavus, and A. terreus, isolated from patients with invasive pulmonary aspergillosis, were used. The CLSI MICs were 0.5 mg/liter for voriconazole and were confirmed by testing in triplicate in our laboratory (4). The strains were maintained at −70°C in 10% glycerol and were subcultured twice on Sabouraud dextrose agar at 30°C for 5 to 7 days before testing. A conidial suspension was prepared in normal saline with 1% Tween 20, and the inoculum size was adjusted to 1 × 10⁵CFU/ml using a Neubauer chamber slide; the concentration was quantified by spread plate cultures on Sabouraud dextrose agar.

Antifungal drug and nutrient medium.

A clinical formulation of voriconazole (Vfend; Pfizer) was reconstituted according to the manufacturer's instructions at 10,000 mg/liter and stored at −70°C. The nutrient medium used throughout contained 10.4 g/liter RPMI 1640 with glutamine without sodium bicarbonate (Sigma-Aldrich, St. Louis, MO) and 0.165 M buffer morpholinepropanesulfonic acid (MOPS) (Invitrogen, Carlsbad, CA), pH 7.0, with 100 mg/liter chloramphenicol (Sigma-Aldrich, St. Louis, MO). CLSI MICs were determined in medium without chloramphenicol.

In vitro pharmacokinetic-pharmacodynamic model.

The in vitro PKPD simulation model is shown in Fig. 1. It consists of the following: (i) a 10 ml-volume dialysis tube (internal compartment [IC]) made of a semipermeable cellulose membrane (Float-A-Lyzer; Spectrum Europe BV, Breda, The Netherlands) allowing free diffusion of small molecules (molecular weight < 20 kDa); this was placed in (ii) a glass beaker containing 700 ml medium (external compartment [EC]), the content of which is continuously diluted by (iii) a peristaltic pump (Minipuls Evolution; Gilson Inc., Villiers le Bel, France) removing drug-containing medium from the EC and adding drug-free medium to the EC at a rate equivalent to that of drug clearance from human plasma. Dialysis tubes, glass beakers, and tubings were sterilized by gamma irradiation (20 G), autoclavation (121°C for 20 min), and 70% ethanol, respectively, while the system was kept free of bacterial contamination using chloramphenicol.

Fig 1.

In vitro pharmacokinetic/pharmacodynamic model that simulates the pharmacokinetics of voriconazole in humans and determines a drug's effect on Aspergillus growth. The model consists of an external compartment (EC), a glass beaker containing 700 ml broth medium, and an internal compartment (IC), a dialysis tube containing 10 ml broth medium; the tube is made of a semipermeable cellulose membrane that allows free passage of small molecules (<20 kDa), such as like antifungals, but not galactomannan. The EC is placed on a heated magnetic stirrer (37°C and 2 rpm). A peristaltic pump introduces drug-free medium in the EC and removes its content concurrently in order to maintain a constant volume. The flow rate is adjusted to achieve drug concentrations corresponding to their clearance from human plasma. At time zero, 10⁵ CFU/ml of Aspergillus conidia are inoculated into the IC, while the drug is introduced into both the EC and the IC for rapid concentration equilibration. Subsequently, the drug concentration declines over time. Galactomannan levels of the IC medium are measured at regular time points (adapted from ref. 11).

The IC was inoculated with 10 ml medium containing 1 × 10⁵ CFU/ml of Aspergillus conidia. The cellulose membrane of the IC allowed free diffusion of nutrients and voriconazole until an equilibrium was reached with the EC, while at the same time it retained the conidia, hyphae, and macromolecular products, such as galactomannan (molecular weight of 25 to 75 kDa). Thus, galactomannan was concentrated within the IC and was used as a biomarker of fungal growth. In order to ensure that galactomannan did not diffuse into the EC, galactomannan levels were determined in the EC.

At time zero, voriconazole was injected into the EC and the IC simultaneously in order to achieve rapid equilibration of drug concentrations between the two compartments simulating iv bolus administration. The drug-containing medium in the EC was then continuously diluted with drug-free medium by the peristaltic pump adjusted to a specific flow rate in order to reproduce average drug half-lives observed in human plasma after intravenous administration of voriconazole (6 h) in accordance with previous clinical studies (18, 19). The EC was then placed on a heated magnetic stirrer adjusted at 37°C and 2 rpm. The temperature and the flow rate were checked regularly throughout the experiment using a thermometer and by measuring the volume of medium pumped out of the EC within 1 min, respectively.

Bioassay of voriconazole.

The drug levels in EC and IC were determined by a microbiological method using the voriconazole-susceptible strain Candida kefyr NCPF 3234 (15). Briefly, 3 × 10⁵ CFU/ml of C. kefyr was inoculated in prewarmed medium (54°C) containing 15 g/liter agar (Bacto Agar Difco; BD Hellas SA, Athens, Greece). The medium was poured into square (10- by 10-cm) plastic petri dishes, and after solidification, 2-mm-diameter wells were made by using a 2-mm-diameter cork borer. One hundred microliters of serial 2-fold drug dilutions (range, 0.25 to 16 mg/liter) and 100 μl of samples obtained from the EC or IC medium were added in the holes. The plates were incubated at 37°C for 24 h, and the diameter of the inhibition zones around the holes was measured with a ruler. A standard curve between the drug concentration and diameter of the inhibition zones was constructed and analyzed by linear regression. Based on this standard curve, the drug concentrations of the EC and the IC were determined at any time point. In order to ensure that drug concentration within the IC was uniform, samples from the center and periphery of the top, middle, and bottom parts of the IC were tested.

Pharmacokinetic analysis.

The in vitro system simulated steady-state voriconazole pharmacokinetics observed in patients with maximum total plasma concentrations of 1.75, 3.5, and 7 mg/liter, respectively, and an average half-life of 6 h (18, 19). The voriconazole concentrations at 0 h, 2 h, 4 h, 6 h, 8 h, 10 h, 12 h, and 24 h after the introduction of the drug in the system were determined both in the IC and EC with the bioassay. The data were analyzed by nonlinear regression (Prism 5.0 software program; GraphPad Inc., La Jolla, CA) based on the compartment pharmacokinetic model described by the equation C_t = C_o × e^−k/t, where C_t (dependent variable) is the concentration of drug at a given time, t (independent variable), C_o is the initial concentration of the drug at time t = 0 h, e is the physical constant 2.18, and k is the rate of drug removal. The half-life was calculated using the equation t_1/2 = k/0.693 for EC and IC separately and compared with the respective values obtained with human plasma. Finally, the calculated area below the curve of drug concentration and time (area under the curve [AUC]) within 24 h (fAUC_0–24) was determined.

Galactomannan levels for determination of fungal growth.

Fungal growth in the IC was monitored by galactomannan production. Galactomannan levels were measured by enzyme-linked immunosorbent assay (ELISA) (Platellia; Bio-Rad Laboratories, Athens, Greece), and results were expressed as a galactomannan index (GI) according to the manufacturer's instructions. In order to correlate GI levels with fungal growth, IC tubes were inoculated with 10³, 10⁴, 10⁵, or 10⁶ CFU/ml of A. fumigatus and incubated without drug for 24 h at 37°C, and GI levels were determined at regular time intervals. The kinetics of galactomannan production were analyzed with nonlinear regression analysis (Prism 5.0; GraphPad Inc., La Jolla, CA) based on the E_max model described by the equation E = E_max · T^γ/(T^γ + T₅₀), where E is the GI (dependent variable), E_max is the maximum GI, T (independent variable) is the time, T₅₀ is the time corresponding to 50% of E_max, and γ is the slope of the curve. In order to capture differences in the extent, rate, or time of galactomannan production reflected by the E_max, γ, or T₅₀ parameter, respectively, the area under the galactomannan index-time curve (AUC_GI) was calculated for each inoculum. The AUC_GI was then correlated with the initial inoculum using linear regression analysis.

Pharmacodynamic analysis.

The IC was inoculated with 10⁵ CFU/ml A. fumigatus, A. flavus, or A. terreus and incubated for 72 h. Galactomannan levels in the IC were measured at regular time intervals with ELISA (Platellia; Bio-Rad Laboratories, Athens, Greece), and results were expressed as a GI according to the manufacturer's instructions. The kinetics of galactomannan production were studied by nonlinear regression (Prism 5.0; GraphPad Inc., La Jolla, CA) based on the E_max model described above. The E_max, γ, or T₅₀ parameter at each voriconazole dose (E_max,D, γ_D, or T_50,D) was compared with the corresponding value of the drug-free growth control (E_max,GC, γ_GC, or T_50,GC). The AUC_GI for the growth control and each dosing regimen at 24 h, 48 h, and 72 h was estimated and used as a surrogate marker of fungal growth. All the experiments were repeated at least twice.

Pharmacokinetic-pharmacodynamic analysis.

The pharmacodynamic parameter AUC_GI was associated with the pharmacokinetic parameter fAUC_0–24 for each species with nonlinear regression analysis using the E_max model described above. The near-maximum activity fAUC_0–24 was calculated for each species as the fAUC_0–24 corresponding to 10% of AUC_GI of drug-free growth control (AUC_GI,GC) of each species. Differences among the species were assessed with analysis of variance (ANOVA). In order to validate the in vitro PKPD model, the AUC/MIC associated with near-maximum activity was compared with the corresponding value found in an experimental animal model of A. fumigatus infection (9). Finally, in vitro PKPD data were bridged to human PK data combining in vitro AUC_GI/fAUC_0–12 and clinical fAUC_0–12 taking into account the interpatient variation. The % target attainment of the upper and lower 95% confidence interval limits of voriconazole fAUC_0–12s observed previously in patients (18, 19) were calculated based on the in vitro AUC_GI/fAUC_0–12 relationship for each species.

RESULTS

Bioassay of voriconazole.

The standard curve of the diameter of inhibition zone-drug concentration is shown in Fig. 2. The lowest concentration of drug detected by this technique was 0.5 mg/liter, and the concentration range was 0.5 mg/liter to 16 mg/liter. The diameter of the inhibition zone correlated linearly with the drug concentration (r² > 0.86). The coefficient of variation of the intra- and interexperimental variation ranged from 5% to 20% (median, 8%) among all drug concentrations.

Fig 2.

Microbiological method for determining voriconazole levels. Linear regression analysis between the diameter of the inhibition zone (y axis) and the drug concentration (mg/liter) (x axis). The intra- and interexperimental variation was <20% among all drug concentrations. Error bars represent standard deviations.

Surrogate marker of fungal growth.

The production of galactomannan for each clinical Aspergillus isolate followed a sigmoid curve described very well with the E_max model (R²> 0.98) (Fig. 3 depicts results obtained by testing A. fumigatus). The area under the GI-time curve (AUC_GI) was linearly associated with the initial inoculum (r² = 0.96; slope, 46 ± 6) (Fig. 3). Therefore, the AUC_GI was used as a surrogate marker of fungal growth.

Fig 3.

Use of the area under the galactomannan index-time curve (AUC_GI) as a surrogate marker of fungal growth. Kinetics of galactomannan production by increasing inocula of A. fumigatus (left graph) and correlation of the AUC_GI with the initial inoculum (right graph) are shown.

Pharmacokinetic analysis.

The one-compartment pharmacokinetic model described well the drug levels in the IC (R²> 0.97). Intra-and interexperimental variation of the in vitro PK data was <10% (Fig. 4). In vitro voriconazole pharmacokinetics were close to the target values observed in human plasma after administration of 3, 4, and 5 mg/kg-of-body-weight voriconazole (18, 19). The half-life of voriconazole in the in vitro system was 5.7 to 6.5 h, which was similar to the half-life of 4.7 to 7.3 h observed in patients' plasma. The voriconazole fAUC_0–12 values were 11.6, 23.5, and 43.8 mg · h/liter for the three doses; likewise, in human plasma, the respective values for voriconazole dosages of 3, 4, and 5 mg/kg corresponded to 5.8, 12.4, and 18.2 mg · h/liter (calculated based on 58% protein binding and 13.9, 29.5, and 43.4 mg · h/liter total AUC_0–12) (18, 19).

Fig 4.

Pharmacokinetic analysis of simulated doses with C_max values of 7, 3.5, and 1.75 mg/liter of voriconazole in the in vitro pharmacokinetic/pharmacodynamic system. The horizontal dotted line represents the lower limit of detection of the bioassay. Error bars represent standard deviations.

Pharmacodynamic analysis.

The pharmacodynamic data for each simulated voriconazole dose against the three strains are shown in Fig. 5. The E_max model described the data well (R² > 0.86). The three simulated voriconazole doses resulted in different GI-time curves in terms of the extent and rate of galactomannan production as reflected by the different E_max, T₅₀, and γ parameters of the E_max model. In particular, all three voriconazole doses delayed galactomannan production by A. fumigatus (T_50,GC = 15 h, versus a T_50,D value of 19 to 29 h), whereas only the two highest doses decreased maximum galactomannan production compared to that for the growth control (E_max,GC = 3.8, versus an E_max,1.75 value of 3.8, E_max,3 value of 2, and E_max,7 value of 1.8). In the case of A. flavus, there were no significant differences in T₅₀ parameters (T_50,GC = 5.8 h, versus a T_50,GC value of 4.2 to 5.3), whereas maximum galactomannan production was reduced as the voriconazole dose increased (E_max,GC = 3.9, versus an E_max,1.75 value of 3.6, E_max,3.5 value of 3.2, and E_max,7 value of 2.6). Regarding A. terreus, voriconazole delayed galactomannan production (T_50,GC = 10.5 h, versus a T_50,1.75 value of 14 h, a T_50,3 value of 18 h, and a T_50,7 value of 21 h), whereas maximum galactomannan production was reduced as the voriconazole dose increased (E_max,GC = 4, versus an E_max,1.75 value of 3.7, E_max,3.5 value of 3.2, and E_max,7 value of 3).

Fig 5.

Galactomannan index-time curves in the internal compartment of the in vitro PKPD model for each voriconazole simulated dose against the three Aspergillus species.

Pharmacokinetic-pharmacodynamic analysis.

In order to capture the above-described changes in E_max model parameters of GI-time curves, the AUC_GI, which was used as a surrogate marker of fungal growth, was correlated to the area under the concentration-time curve for each voriconazole dose (fAUC_0–24). The AUC_GI at 24 h decreased from 34 GI · h in the growth control to 10.5 GI · h at the highest voriconazole dose for A. fumigatus, following a sigmoid pattern (R² > 0.97) (Fig. 6, left graph). The same pattern of AUC_GI reduction was observed for A. flavus (60.3 GI · h in the growth control to 42.3 GI · h at the highest voriconazole dose) and for A. terreus (50.1 GI · h to 20.5 GI · h, respectively). Because of the different dynamic AUC_GI ranges among the three Aspergillus spp., the AUC_GI was normalized from 0% to 100% based on the minimum and maximum AUC_GI values, respectively (Fig. 6, right graph). Based on the normalized PKPD relationship, the fAUC_0-24 associated with near-maximum activity (10% AUC_GI,GC) was 18.9 (14.4 to 23.1) mg · h/liter against A. fumigatus, 26.6 (21.1 to 32.9) mg · h/liter against A. flavus, and 36.2 (27.8 to 45.7) mg · h/liter against A. terreus (F_2,19 = 17.22; P < 0.0001 where 2 and 19 are the degrees of freedom for the numerator and denominator of F ratio, respectively). These fAUC_0-24s were similar when the AUC_GIs at 48 h and 72 h were used as markers of fungal growth. Of note, the fAUC_0-24 of 18.9 mg · h/liter for the tested A. fumigatus isolate (MIC = 0.5 mg/liter) corresponds to an fAUC/MIC of 37.8, which is close to the fAUC/MIC of 36.4 associated with 90% survival in an animal model of experimental aspergillosis (9), thus providing a validation of the in vitro PKPD model.

Fig 6.

In vitro PKPD relationship of voriconazole. The relationship between the area under the galactomannan index curve (AUC_GI) (left graph) or normalized AUC_GI (right graph) and the area under the concentration-time curve (fAUC_0–24) for each Aspergillus species. The AUC_GI for the first 24 h was used as the surrogate marker of fungal growth.

Bridging the in vitro PKPD data with human PK data, a low median percent target attainment (<45%) was found for 3-mg/kg voriconazole dosing for all species (Table 1). The standard dosage of 4 mg/kg was associated with a high median percent target attainment (80%) for A. fumigatus but not for A. flavus and A. terreus, for which as low as 17% and 6% target attainment, respectively, were found for fAUC_0-12 at the lower 95% confidence interval limit observed in patients. Even for A. fumigatus, a wide range of percent target attainment (32 to 97%) was detected, reflecting the wide variation of voriconazole fAUC_0-12 values among patients receiving 4 mg/kg. In agreement with these findings is the range of 50 to 80% survival reported in clinical trials among patients with aspergillosis treated with voriconazole (2, 6). Importantly, for patients infected by A. fumigatus isolates with voriconazole MICs of 0.5 mg/liter, the 6-week survival rate was ∼75% (2), which is very close to the 80% median percent target attainment recorded in the present study (2), thus providing clinical validation of the in vitro model. With a dosage of 5 mg/kg, more patients will attain the target for A. fumigatus infections (>86%), whereas >73% and >47% will achieve the target for A. flavus and A. terreus infections. Of note, the percentage of target attainment for the upper 95% confidence interval for 4-mg/kg voriconazole dosage was >83% for all Aspergillus species (Table 1).

Table 1.

Percent target attainment of in vitro PKPD parameter associated with near-maximum activity for each voriconazole dosage^a

Voriconazole dose (mg/kg)	Mean (95% CI) fAUC0–12 (mg · h/liter)b	Mean (95% CI) % target attainment
		A. fumigatus	A. flavus	A. terreus
3	5.85 (4.4–7.75)	24 (11–45)	12 (5–26)	4 (2–11)
4	12.4 (6.55–23.35)	80 (32–97)	63 (17–93)	36 (6–83)
5	18.2 (14.15–23.5)	93 (86–97)	86 (73–94)	68 (47–83)

^aCI, confidence interval.

^bFrom reference 19.

DISCUSSION

The in vitro PKPD modeling of voriconazole activity against A. fumigatus, A. flavus and A. terreus isolates with identical MICs showed important pharmacodynamic differences in a new dynamic model simulating human plasma voriconazole pharmacokinetics. The voriconazole fAUC_0–24 associated with near maximum activity differed among the three Aspergillus spp. with fAUC_0–24 of 18.9 (14.4 to 23.1) mg · h/liter for A. fumigatus, 26.6 (21.1 to 32.9) mg · h/liter for A. flavus and 36.2 (27.8 to 45.7) mg · h/liter for A. terreus. Bridging these data with human PK showed that the standard dosage of 4 mg/kg was associated with a high median percent target attainment for A. fumigatus (80%) but not for A. flavus (63%) and A. terreus (36%), although a wide range of percent target attainment was observed, reflecting the large interpatient variation of voriconazole fAUC_0–12.

In vitro PKPD analysis may reveal differences in antifungal activity that cannot be predicted by an MIC value. Studying the effect of decreasing concentrations of voriconazole provides information about drug pharmacodynamic properties related to sub-MIC effect, postantifungal effect, and time- and concentration-dependent activities. These effects can be quantified by a surrogate marker of fungal growth based on galactomannan production kinetics which captures any difference on the above antifungal effects. Differential antifungal activity against three Aspergillus species with identical MICs of voriconazole, amphotericin B, and caspofungin was recently found, emphasizing the importance of studying nonconstant drug concentrations (11). The growth rate may be an important determinant of antifungal activity. Voriconazole activity was found to be correlated with growth rates of Candida isolates as determined in a pharmacokinetic model (8) and of Aspergillus species using a microdilution assay measuring metabolic activity (1). The three Aspergillus species are characterized by different growth rates, with A. terreus growing slower and A. flavus faster than A. fumigatus (12, 25). Indeed, in the present model, a larger voriconazole fAUC_0–24 was required to inhibit the slow-growing A. terreus than for the other two species. However, the fAUC_0–24 against the slower-growing A. fumigatus was smaller than the fAUC_0–24 against the faster-growing A. flavus, indicating that other factors influence voriconazole activity in the present model, such as time-dependent effects.

Time-dependent activity of voriconazole inhibition may differ for each Aspergillus species. Exposure of Aspergillus conidia to concentrations near the voriconazole MIC for 6 h resulted in a significant amount of fungal growth for A. terreus (64%), A. flavus (24%), and A. fumigatus (15%) isolates (1). After 8 h of exposure, voriconazole median EC₅₀ against A. fumigatus isolates was lower than after 48 h of incubation (0.69 times), whereas against A. flavus isolates they were similar (1.06 times), indicating that voriconazole acts faster against A. fumigatus than against A. flavus. This may explain the smaller fAUC_0–24 observed in the present model for A. fumigatus than for A. flavus.

In agreement with a differential voriconazole time-dependent inhibition among species is the fact that complete inhibition of galactomannan production was not observed for any of the strains tested in the present dynamic model. This effect may be related to the drug mechanism of action, since azole-induced complete inhibition of ergosterol synthesis requires at least 1 h and complete exchange of ergosterol by its methylated precursors occurs after about 6 h of azole exposure (20). The kinetics of voriconazole inhibition may differ among Aspergillus species, as reflected in the present dynamic model by the different maxima of galactomannan production for each species. These differences were taken into account during the analysis by normalization of data.

In order to validate the present in vitro PKPD model, the obtained results were compared with those found in a murine model of experimental aspergillosis (9). In the latter, voriconazole efficacy was tested against four A. fumigatus isolates demonstrating increasing MIC values. The AUC/MIC associated with 90% survival in mice was 36.4, which was very close to the AUC/MIC of 37.8 associated with near-maximum activity (i.e., 10% fungal growth) in the present study. Voriconazole dosages of 10 and 40 mg/kg, which corresponded to fAUC_0–24s of 1 and 20 mg · h/ml, against a wild-type A. fumigatus isolate resulted in 90% and 0% mortalities, respectively (9). Similarly, in our in vitro dynamic model, fAUC_0–24s of 1 and 20 mg · h/ml (0 and 1.3 log₁₀ fAUC_0–24, respectively) corresponded to 100% and 10% of fungal growth (right graph of Fig. 6). Finally, voriconazole at a dosage of 10 mg/kg was active against A. fumigatus (13, 23) but not A. flavus (27) infections in murine models of invasive aspergillosis. Likewise, the present study demonstrated that voriconazole was more active against A. fumigatus than A. flavus. Thus, the results of the in vitro dynamic model can predict the outcome of voriconazole treatment in vivo.

The results of the PKPD model are in agreement with clinical data, since the percent target attainment (32 to 97%) for the 4-mg/kg dosage of voriconazole was similar to the survival rates among patients with A. fumigatus infections who were treated with standard dosing (50 to 80%) (2, 6). In addition, the median percent target attainment for this species (80%) was very close to the median 6-week survival rate (∼75%) previously observed for hematological patients infected by A. fumigatus isolates with MICs corresponding to 0.5 mg/liter (2). The AUC/MIC index of 37.8, which was found to be associated with near-maximum activity for A. fumigatus in the present study, can be reached with standard dosing of 4 mg/kg for isolates with MICs of 0.5 mg/liter, supporting voriconazole epidemiological cutoff value of 1 mg/liter (5). Regarding the other species, the survival rate of patients with invasive aspergillosis due to A. terreus was generally lower than that for aspergillosis due to A. fumigatus after treatment with voriconazole (6, 22). This was also reflected in the present study by a higher percent target attainment for A. fumigatus than for A. terreus. These correlations provide further clinical validation of the in vitro model.

In summary, higher voriconazole exposure was required to inhibit A. terreus than for A. flavus and A. fumigatus. Based on the in vitro exposure-effect relationships found in the novel PKPD model, a standard dosage of voriconazole may be adequate for most patients with A. fumigatus infections; a higher level of drug exposure appears to be required for A. flavus and A. terreus infections. This could be achieved using a higher voriconazole dosage or by therapeutic drug monitoring in patients receiving the standard dosage given the wide distribution of voriconazole fAUC_0-12s among them. Optimizing voriconazole exposure in order to obtain fAUC_0-12 at the upper 95% confidence interval limit would result in percent target attainment of >83% for all three Aspergillus species. Therapeutic drug monitoring for voriconazole is becoming an important tool to improve the efficacy and safety of this agent (3). This is corroborated by clinical studies where adequate voriconazole levels have been associated with efficacy, while drug underexposure was associated with a worse outcome (14, 15). There are a multitude of factors that can influence patient outcomes. In that respect, in vitro PKPD modeling assists in providing information on critical parameters, such as drug kinetics and dynamics, thus offering a powerful tool for estimating the impact of drug underexposure and improving efficacy through drug exposure optimization.