Pharmacodynamics of Posaconazole in Experimental Invasive Pulmonary Aspergillosis: Utility of Serum Galactomannan as a Dynamic Endpoint of Antifungal Efficacy

Aspergillus galactomannan antigenemia is an accepted tool for the diagnosis of invasive pulmonary aspergillosis (IPA) in neutropenic patients. Little is known, however, about the utility of this biomarker to assess the efficacy of antifungal therapies. The pharmacokinetics (PK) and pharmacodynamics (PD) of posaconazole in treatment and prophylaxis were investigated in the persistently neutropenic rabbit model of Aspergillus fumigatus IPA at doses between 2 and 20 mg/kg per day.

ABSTRACT

Aspergillus galactomannan antigenemia is an accepted tool for the diagnosis of invasive pulmonary aspergillosis (IPA) in neutropenic patients. Little is known, however, about the utility of this biomarker to assess the efficacy of antifungal therapies. The pharmacokinetics (PK) and pharmacodynamics (PD) of posaconazole in treatment and prophylaxis were investigated in the persistently neutropenic rabbit model of Aspergillus fumigatus IPA at doses between 2 and 20 mg/kg per day. Sparse plasma sampling was used to obtain PK data at steady state, and the serum galactomannan index (GMI), as a dynamic endpoint of antifungal response, was obtained every other day, in addition to conventional outcome parameters including survival and fungal tissue burden. Nonparametric PK/PD model building was performed using the Pmetrics package in R. A one-compartment model with linear elimination best described the PK of posaconazole. The PD effect of posaconazole exposure in plasma on the GMI in serum was best described by dynamic Hill functions reflecting growth and killing of the fungus. Through calculations of the area under the concentration-time curve from 0 to 24 h (AUC_0–24) at steady state, the exposure-response relationship between posaconazole and the GMI for treatment followed a sigmoidal function with an asymptote forming above an AUC_0–24 of 30 mg · h/liter. All prophylactic doses were able to control the fungal burden. A nonparametric population PK/PD model adequately described the effect of posaconazole in prophylaxis and treatment of experimental IPA. An AUC_0–24 greater than 30 mg · h/liter was associated with adequate resolution of the GMI, which well supports previously suggested exposure-response relationships in humans.

INTRODUCTION

Posaconazole is an antifungal triazole that is structurally related to itraconazole and possesses broad-spectrum antifungal activity in vitro, predictable pharmacokinetics (PK), moderate potential for drug-drug interactions, and an overall favorable safety profile (1–3). Based on a set of carefully designed and well-executed clinical trials performed during the past 2 decades (4–6), the compound has evolved into an important option for prophylaxis and treatment of opportunistic invasive fungal diseases in severely immunocompromised patients. Whereas the usefulness of the initially approved oral suspension was limited by high intraindividual and interindividual bioavailability (7, 8), the subsequently developed delayed-release tablet formulation and the parenteral cyclodextrin formulation allow for more controlled administration of the compound (9–12). Leading international guidelines currently recommend posaconazole for primary antifungal prophylaxis in patients with acute myeloid leukemia/myelodysplastic syndrome and prolonged neutropenia and in patients with acute graft-versus-host disease following allogeneic hematopoietic stem cell transplantation, as well as second-line therapy for treatment of invasive aspergillosis (13–15).

Posaconazole is metabolized predominantly via phase II glucuronidation and, while CYP3A4 inhibition affects drug-drug interactions, only minor metabolization can be attributed to the CYP450 family. The compound is highly protein bound, with serum albumin being the predominant binding protein. Previous studies with the oral suspension have shown that posaconazole exhibits linear elimination with a high apparent volume of distribution, a slow rate of absorption (8), and varying bioavailability. No distribution into deeper compartments was detectable, and a one-compartment PK model was used in previous PK analyses (5). More recent studies with the intravenous (i.v.) formulation revealed smaller volumes of distribution and higher peak concentrations, relative to the oral suspension (2, 3, 16).

Exposure-response relationships have been developed for most antifungal compounds (17, 18). For posaconazole, regulatory guidance in the process of dose-finding studies for the tablet and the i.v. formulations (19) and a European Committee for Antimicrobial Susceptibility Testing (EUCAST) rationale (20) propose a dosing target of a minimum trough concentration of 0.7 mg/liter for prophylaxis, corresponding to the required area under the curve (AUC)/MIC ratio of 167; an average concentration of 1.25 mg/liter is suggested for salvage therapy. However, these relationships rely on observed outcomes linked to the drug exposure in clinical trials, rather than a distinct pharmacodynamic (PD) criterion (21, 22).

The polysaccharide galactomannan is a major component of the cell wall of Aspergillus spp. and is released into the systemic circulation during fungal degradation (23). It is detectable in the serum in some patients even before characteristic symptoms of invasive aspergillosis are present. Studies using galactomannan in serum as a diagnostic marker, with a threshold of 0.5 for proven or probable disease status, resulted in a test sensitivity of 82% with a specificity of 81% (24). The role of serum galactomannan as a surrogate marker of success or failure of antifungal interventions is an area of active investigation (25, 26). It has been shown that a galactomannan index (GMI)-based response criterion, as an aspergillosis-specific marker in hematological cancer patients, compares favorably with the European Organization for Research and Treatment of Cancer/Mycoses Study Group (EORTC/MSG) invasive aspergillosis response definition, as well as survival outcomes, and can be beneficial regarding earlier assessments of treatment response (27–29). Therefore, we investigated the PK and PD of posaconazole in experimental invasive pulmonary aspergillosis (IPA), using the GMI as a dynamic endpoint of antifungal response (Fig. 1 and 2).

FIG 1.

Overview of the study cohorts and their disposition in the analysis. For development of the population PK model, data from the 53 posaconazole-treated animals included in cohorts 1 to 4 were used. For investigation of the PD in experimental IPA, data from 25 posaconazole-treated animals in cohorts 2 and 3 and 17 untreated controls for which serial QOD sampling of serum galactomannan (GMI) was available were used. Cohort 4 was not included in the PK/PD model, since only the last available serum galactomannan values were determined. EOT, last available sample before the end of treatment.

FIG 2.

Overview of the PK/PD study setup. Observed data and interventions are depicted against time. Dotted lines, start of posaconazole therapy in the treatment and prophylaxis groups (time the first dose was applied); dashed lines, time of inoculation with 1 × 10⁸ conidia of A. fumigatus; black connected dots, GMI values; gray connected dots, posaconazole plasma concentrations.

RESULTS

Antifungal therapy.

There was a significant improvement in survival postinoculation for rabbits treated with posaconazole, compared to untreated controls. Through the entire study, 29 (76%) of 38 rabbits treated with posaconazole survived in cohorts 3 and 4, and none of the 8 untreated controls survived (P < 0.001 by Fisher’s exact test). There also was a significant quantitative reduction in the growth of Aspergillus fumigatus in lung tissues from rabbits treated with posaconazole, in comparison with that of untreated controls, as measured by the mean log CFU per gram ± standard error of the mean (SEM) at the end of the experiment (0.28 ± 0.07 versus 1.49 ± 0.17 log CFU/g; P < 0.001 by Mann-Whitney U test). Consistent with the improvements in survival and organismal clearance from lung tissue, rabbits treated with posaconazole had a significantly lower mean ± SEM GMI, relative to that of untreated controls, in the last of serial (every other day [QOD]) measurements obtained during the experiments (1.78 ± 0.37 versus 4.66 ± 0.54; P = 0.002 by Mann-Whitney U test) (Fig. 3).

FIG 3.

Effects of treatment and prophylaxis with posaconazole on IPA in persistently neutropenic rabbits, as measured by the residual fungal burden in lung tissue (log CFU per gram) at the end of the experiment and the last available GMI. Posaconazole dosage groups (2, 6, and 20 mg/kg/day) were combined (light columns) and compared to untreated controls (dark columns). Survival rates for animals receiving posaconazole were 76% (29/38 animals) with treatment and 89% (8/9 animals) with prophylaxis. For comparison, none of the 8 and 9 animals in the control cohorts survived through the end of the experiment (P < 0.001 by Fisher’s exact test). (A) Residual fungal burden in lung tissue. Treatment and prophylaxis with posaconazole resulted in highly significant reductions in the mean ± SEM residual fungal tissue burden versus untreated controls (0.28 ± 0.07 versus 1.49 ± 0.1 log CFU/g and 0.13 ± 0.08 versus 1.40 ± 0.72 log CFU/g, respectively; P < 0.001 by Mann-Whitney U test). (B) Last available serum GMI. Concordant with the residual fungal burden, there was a significant reduction in the GMI in animals receiving posaconazole for treatment and prophylaxis (0.78 ± 0.37 versus 4.66 ± 0.54 and 0.34 ± 0.08 versus 4.28 ± 0.83, respectively; P = 0.002 and P < 0.001 by Mann-Whitney U test).

Antifungal prophylaxis.

Similar to antifungal therapy, rabbits receiving antifungal prophylaxis with posaconazole had significant improvements in the survival rate, in comparison to untreated controls (8/9 versus 0/9 rabbits surviving; P < 0.001 by Fisher’s exact test), a significant reduction of the residual pulmonary fungal burden at the end of the experiments (0.13 ± 0.08 versus 1.40 ± 0.72 log CFU/g; P < 0.001 by Mann-Whitney U test), and a significantly lower mean ± SEM GMI in the last of serial (QOD) measurements obtained during the experiments (0.34 ± 0.08 versus 4.28 ± 0.83; P < 0.001 by Mann-Whitney U test) (Fig. 3).

Population PK model.

For development of the population PK model, 270 concentration-time points from the 53 posaconazole-treated animals included in cohorts 1 to 4 (Fig. 1) were used. A one-compartment PK model with first-order oral absorption and linear elimination best described the PK of posaconazole in the densely sampled healthy rabbits studied after administration of a single dose of 20 mg/kg (cohort 1). Using additional compartments or implementing a nonlinear elimination did not improve the model in terms of statistical or graphical criteria.

The final model developed in the healthy rabbits was transferred to the infected rabbits receiving posaconazole as treatment or prophylaxis, in which sparse sampling was performed at presumed steady state after multiple daily doses ranging from 2 to 20 mg/kg. In this step, the absorption rate constant was not estimated sufficiently and thus was fixed at 0.35 h⁻¹, the mean estimate derived from the data obtained in the group 4 preanalysis.

For the final PK model component, a linear regression of the individual predictions through utilization of the Bayesian posterior versus the observed values resulted in a mean intercept of −0.018 (95% confidence interval, −0.11 to 0.075). A mean slope of 1.04 (95% confidence interval, 0.99 to 1.09) and a correlation coefficient of 0.934 were determined.

PD model.

For development of the PD model, data from 25 posaconazole-treated animals of cohorts 2 and 3 and 15 untreated controls for which serial QOD sampling of serum galactomannan levels was available were used (Fig. 2). The combined group 2 and 3 data set included a total of 125 posaconazole concentration-time points and a total of 211 individual GMI measurements.

The effect of posaconazole on the GMI was implemented via integration of an effect compartment. In a first step, only data from the treatment cohort were used. Data from the prophylaxis cohort were added consecutively.

The evolution of the GMI was best described via sigmoidal Hill functions,

$$\begin{gathered} \frac{\mathit{dGMI}}{\mathit{dt}}={\mathit{Kg}}_{\mathit{max}}\cdot [1-\frac{{\frac{POC}{Vc}}^{Hg}}{{C50g}^{Hg}+{\frac{POC}{Vc}}^{Hg}}]\cdot \left[\frac{1-CEFF}{{POP}_{\mathit{max}}}\right]\cdot \mathit{CEFF} \\ -{\mathit{Kk}}_{\mathit{max}}\cdot \frac{{\frac{POC}{Vc}}^{Hk}}{{C50k}^{Hk}+{\frac{POC}{Vc}}^{Hk}}\cdot \hspace{0.17em}\mathit{CEFF} \end{gathered}$$

where POC is the amount of posaconazole in the PK compartment at time t, CEFF is the GMI in the effect compartment at time t, POP_max is the population maximal growth reflected by the GMI, Hg is the Hill coefficient for growth, Hk is the Hill coefficient for killing, Kg_max is the maximum rate of growth, Kk_max is the maximum rate of killing; C_50g is the concentration for half-maximal growth, C_50k is the concentration for half-maximal killing, and Vc is the central volume of distribution.

The implemented Hill functions were able to describe the GMI decline in treatment and prophylaxis together with the GMI increase in the control group. Due to the nonparametric modeling approach, only one set of parameters was necessary for the entire population, with the respective support point distribution being estimated. A summary of the estimated PK/PD model parameters is shown in Table 1.

TABLE 1.

Summary of estimated PK and PD parameters of the final model describing the relationship between posaconazole exposure and the GMI

Parametera	Mean	SD
CL (liter/h)	0.60	0.56
V (liter)	117	98.6
Kgmax (GMI/h)	0.03	0.02
Hg	196	98.4
Hk	55.8	88.2
POPmax (GMI)	6.58	1.73
Kkmax (GMI/h)	1.97	1.69
C50g (mg/liter)	0.19	0.13
C50k (mg/liter)	3.99	1.95

^aCL, clearance; V, volume of distribution; Kg_max, maximum rate of growth; Hg, Hill coefficient for growth; Hk, Hill coefficient for killing; POP_max, population maximal growth reflected by GMI; Kk_max, maximum rate of killing; C_50g, concentration for half-maximal growth; C_50k, concentration for half-maximal killing.

The linear regression of the individual GMI predictions through utilization of the Bayesian posterior versus the observed values resulted in a mean intercept of 0.11 (95% confidence interval, −0.002 to 0.224). A mean slope of 0.933 (95% confidence interval, 0.882 to 0.983) and a correlation coefficient of 0.864 were determined. The goodness of fit for individual posaconazole and GMI predictions is shown in Fig. 4 and 5.

FIG 4.

PK model goodness-of-fit plot for individual posaconazole predictions. Black dots, observed and individual predicted values; solid line, line of identity; dashed line, locally estimated scatterplot smoothing (LOESS) fit across predictions.

FIG 5.

PD model goodness-of-fit plot for individual GMI predictions. Black dots, observed and individual predicted values; solid line, line of identity; dashed line, LOESS fit across predictions.

Exploration of PK/PD relationships.

The final PK/PD model was used to explore the relationship between posaconazole exposure and the GMI. For this purpose, the AUC_0–24 was calculated for posaconazole and for the GMI in rabbits included in the treatment group (cohort 3).

To quantify the relationship and to derive a PD threshold, a function was fit to the derived AUCs (Fig. 6). A sigmoidal function best described the relationships of both AUCs,

$$\text{AUC}\hspace{0.17em}\mathit{GMI}=\alpha -\frac{\mathit{Emax}\hspace{0.17em}\cdot \hspace{0.17em}\mathit{AUC}\hspace{0.17em}\mathit{PO}{C}^{\gamma }}{E{C}_{50}{}^{\gamma }+\mathit{AUC}\hspace{0.17em}\mathit{PO}{C}^{\gamma }}$$

where POC is the posaconazole concentration, GMI is the galactomannan antigen index, γ is the Hill coefficient, E_max is the maximal GMI depicting fungal effect, and EC₅₀ is the posaconazole concentration AUC for half-maximal effect (α = 118, E_max = 93.7 GMI, EC₅₀ = 11.6 mg/liter/h, and γ = 2.3).

FIG 6.

PK/PD relationship between posaconazole exposure and the GMI in treatment of experimental IPA, as assessed by the AUC_0–24 of posaconazole levels and the AUC_0–24 of the GMI. Both AUCs were calculated at day 5 after inoculation. Gray line, fitted curve from regression analysis, resulting in displayed equation.

In the prophylaxis group (cohort 2), the posaconazole trough concentration was determined at the time of inoculation and compared to the AUC_0–24 for the GMI on day 5 postinoculation. When posaconazole prophylaxis was administered, the calculated GMI AUC with day 5 postinoculation did not exceed 24, suggesting the GMI did not cross the threshold of 1 with this 24-h time window, whereas the control group (visible as AUCs at a posaconazole concentration of 0 mg/liter) showed high GMI AUC values.

In Fig. 6 and 7, the GMI AUCs for treatment and prophylaxis are compared to posaconazole exposure, showing the effect of posaconazole treatment and prophylaxis on A. fumigatus infections throughout the tested posaconazole dosing range. The quantification of these relationships was facilitated through the PK/PD modeling approach, which enables extrapolation of posaconazole concentration and GMI to the necessary time points and thus is able to display more dynamically the PD effect in prophylaxis and treatment.

FIG 7.

PK/PD relationship between posaconazole exposure and the GMI in prophylaxis of experimental IPA, as assessed by the posaconazole trough and the AUC_0–24 of the GMI, calculated at the day of inoculation.

DISCUSSION

In this well-established persistently neutropenic rabbit model, posaconazole was highly effective in prophylaxis and treatment of experimental IPA, as documented by the endpoints of survival, residual fungal burden in lung tissue, and suppression of the GMI in serum at the end of treatment. The PK/PD relationship between posaconazole exposure in plasma and the evolution of the GMI during prophylaxis and treatment was best described by dynamic Hill functions reflecting growth and killing of the fungus. The exposure-response relationship for treatment followed a sigmoidal function, with an asymptote forming above an AUC_0–24 of 30 mg · h/liter.

Whereas a link between drug exposure and observed outcomes has been documented in animal models (30, 31) and in clinical studies (6, 32, 33), to the best of our knowledge this is the first published PK/PD modeling study investigating the effects of posaconazole against IPA taking the effects of treatment and prophylaxis into account. In a murine kidney target model of invasive candidiasis using noncompartmental PK, the AUC_0–24/MIC ratio was most predictive of treatment success (34), which is in accordance with previous PK/PD assessments of antifungal triazoles in invasive candidiasis. In neutropenic murine models of disseminated aspergillosis and mucormycosis, AUC_0–24/MIC ratios of >100 were predictive of successful treatment with posaconazole (31). Similarly, in experimental murine IPA, other investigators found an AUC_0–24/MIC ratio of at least 94 to be strongly associated with success in antifungal prophylaxis (35).

Using a PK/PD modeling approach, Howard et al. investigated the exposure-response relationship of posaconazole in an inhalational murine model. There, an AUC/MIC ratio of 167 was associated with half-maximal antifungal effect (22).

Apart from a robust assessment of the efficacy of posaconazole against experimental IPA, the persistently neutropenic rabbit model is well suited to establish the link between posaconazole exposure and surrogate markers for PD effects (36, 37). In the present study, we used the GMI as a biomarker to monitor the decline of the fungal burden in lung tissue and linked it to the plasma concentration-time profile of posaconazole in each infected rabbit. Given the notorious problems in assessing the effects of antifungal interventions in immunocompromised patients by clinical and radiographic methods, the existence of a validated and readily available biomarker would be a major advance to steward treatment in clinical practice and to guide treatment decisions in clinical trials (25, 26).

The one-compartment model with linear elimination found to best describe the PK of posaconazole in the rabbits is well in accordance with previously published PK models of the compound in human subjects (32, 38) and emphasizes the usefulness of this species for PK/PD bridging studies with antifungal agents (39). The final PK/PD individual Bayesian posteriors accounted for 93% of the observed variability in plasma concentrations and for 86% of the observed variability in the GMI.

Linking the GMI to the posaconazole exposure in the prophylaxis cohort showed the strong ability of posaconazole to control tissue invasion after inoculation and to prevent invasive disease in the model. In the treatment cohort, a clear exposure-response relationship was detectable. Depending on the intensity of the posaconazole treatment started 24 h after inoculation, the GMI was able to evolve in the rabbits as a marker of uncontrolled or controlled disease. A sigmoidal function was found to best describe this relationship. The detected function turned asymptotic at AUC_0–24 values greater than 30 to 35 mg · h/liter, indicating that this AUC value is the threshold for fungal suppression and treatment success. This value corresponds well to the previously reported AUC_0–24 that was associated with a 75% response rate in invasive aspergillosis salvage therapy (6). Assuming a common epidemiological cutoff MIC value of 0.25 mg/liter for Aspergillus spp., effective antifungal treatment in the model corresponded to a posaconazole AUC_0–24/MIC ratio of 120 to 140. Of note, the magnitude of this PK/PD target value is identical to current recommendations for therapeutic drug monitoring, which suggests a target AUC_0–24/MIC ratio of 100 to 200 for treatment and of at least 94 for sufficient antifungal prophylaxis (19, 20).

Whereas previously published studies linking PK to antifungal efficacy used noncompartmental PK to estimate key PK parameters, including AUC, clearance rate, half-life, and volume of distribution (31, 34, 35, 40), nonlinear mixed-effects modeling was used in this study to estimate individual and population-based PK parameters, allowing assessment also of intersubject variability and time dependence. The established model was then able to predict plasma concentration-time profiles for each rabbit at the exact time points of individual GMI measurements. This approach allowed for more flexibility in the study setup and the ability to connect the population PK model with a second PD biomarker model. While previous PK/PD investigations linked the fungal burden in target sites such as the kidney or the lungs at the end of the intervention or observation period to the observed PK profiles, the model presented here was able to connect the population PK model with a time-varying marker measurement to actually observe the suppression of fungal growth during the intervention.

In conclusion, posaconazole was highly effective in the treatment and prophylaxis of experimental invasive aspergillosis in persistently neutropenic rabbits at exposures comparable to those achieved in human subjects. All prophylactic dosing regimens were able to suppress the surrogate marker GMI below 1.0 throughout the experiment. In the treatment experiments, a sigmoidal exposure-response relationship was detected, leading to an asymptote at an AUC_0–24 of greater than 30 mg · h/liter, which was associated with significant resolution of the GMI and maximum fungal eradication of A. fumigatus in lung tissue.

MATERIALS AND METHODS

Study overview.

In order to develop a PK/PD model of posaconazole in prophylaxis and treatment of IPA using galactomannan as a dynamic endpoint of efficacy, raw data from experiments of previously published studies investigating the PK and antifungal activity in normal and persistently neutropenic rabbits were used. (30, 40). The study included data from a total of 70 animals studied in four different experimental cohorts, i.e., (i) 6 healthy, noninfected rabbits that received a single dose of 20 mg/kg posaconazole followed by serial plasma sampling to explore the plasma PK of the compound in rabbits; (ii) 9 neutropenic rabbits with experimental IPA that received posaconazole at 2, 6, and 20 mg/kg every day (QD) as prophylaxis starting 4 days prior to inoculation; (iii) 16 neutropenic rabbits with experimental IPA that received posaconazole at 2, 6, and 20 mg/kg QD as treatment starting 24 h after inoculation; and (iv) 22 neutropenic rabbits with experimental IPA that received posaconazole at 1, 2, and 3 mg/kg twice a day (BID) as treatment starting 24 h after inoculation, as well as 17 rabbits with experimental IPA that served as untreated controls in cohorts 2 and 3 (30, 40).

For development of the population PK model, data from the 53 posaconazole-treated animals included in cohorts 1 to 4 were used. For investigation of the PD in experimental IPA, data from 25 posaconazole-treated animals in cohorts 2 and 3 and 17 untreated controls for which serial QOD sampling of serum galactomannan levels was available were used. Cohort 4 was not included in the PK/PD model, since only the last available serum galactomannan values were determined. An overview of the study cohorts is provided in Fig. 1.

Animals.

Healthy female New Zealand White rabbits weighing 2.6 to 3.7 kg (Hazleton, Deutschland, PA) were used in all experiments. Rabbits were individually housed and maintained with water and standard rabbit feed ad libitum according to National Institutes of Health (NIH) guidelines and in fulfillment of the criteria of the American Association for Accreditation of Laboratory Animal Care (NRC 1996). Vascular access was established by placement of a silastic tunneled central venous catheter (41).

Organism and inoculation.

Aspergillus fumigatus (NIH isolate 4215 [ATCC MYA1163]) obtained from a fatal case of pulmonary aspergillosis was used in all experiments. The MIC was determined using NCCLS methods (42, 43), and the minimum fungicidal concentration for posaconazole was 0.125 μg/ml.

Pulmonary aspergillosis was established as described previously (30, 44). For each experiment, the A. fumigatus inoculum was prepared from a frozen isolate that was subcultured on Sabouraud dextrose slants (BBL, Cockeysville, MD). The slants were incubated for 24 h at 37°C and then kept at room temperature for 5 days before use. Conidia were harvested under a laminar airflow hood with a solution of 10 ml of 0.025% Tween 20 (Thermo Fisher Scientific, Fair Lawn, NJ) in 0.9% NaCl (Quality Biological, Inc., Gaithersburg, MD), transferred to a 50-ml conical tube, washed, and counted with a hemocytometer. The concentration was adjusted to a predetermined inoculum of 1 × 10⁸ conidia of A. fumigatus in a volume of 250 to 350 μl and was confirmed by serial dilutions cultured on Sabouraud glucose agar (SGA).

Inoculation was performed on day 2 of the experiments under general anesthesia. Each rabbit was anesthetized with 0.8 to 1.0 ml of a 2:1 mixture (vol/vol) of i.v. ketamine (100 mg/ml) obtained as Ketaset (Phoenix Scientific, Inc., St. Joseph, MO) and xylazine (20 mg/ml) (Bayer Corp., Agriculture Division, Animal Health, Shawnee Mission, KS) obtained as Rompun. A Flagg O straight-blade laryngoscope (Welch Allyn Inc., Skaneateles Falls, NY) was inserted in the oral cavity until the vocal cords were clearly visualized, and the inoculum was administered intratracheally with a tuberculin syringe attached to a 5.25-inch 16-gauge Teflon catheter (Becton, Dickinson Infusion Therapy Systems Inc., Sandy, UT).

Immunosuppression and maintenance of neutropenia.

Profound and persistent neutropenia (neutrophil count of <100 neutrophils/μl) was achieved with an initial course of 525 mg of cytarabine (Cytosar-U; The Upjohn Company, Kalamazoo, MI) per m² for 5 consecutive days starting 1 day before endotracheal inoculation. A maintenance dose of 484 mg of cytarabine per m² was administered for four additional doses on days 8, 9, 13, and 14 of the experiment. Methylprednisolone (Abbott Laboratories, North Chicago, IL) at 5 mg/kg of body weight was administered on days 1 and 2 of the experiment to facilitate establishment of infection.

Ceftazidime (Glaxo, Inc., Research Triangle Park, NC) (75 mg/kg given i.v. BID), gentamicin (Elkins-Sinn, Inc., Cherry Hill, NJ) (5 mg/kg given i.v. QOD), and vancomycin (Abbott Laboratories) (15 mg/kg given i.v. QD) were administered from day 4 of immunosuppression until study completion to prevent opportunistic bacterial infections during neutropenia. To prevent antibiotic-associated diarrhea due to Clostridium spiroforme, rabbits continuously received 50 mg of vancomycin per liter of drinking water.

Antifungal compound.

Posaconazole was provided by the Schering-Plough Research Institute. Drug stock solution (30 mg/ml) was prepared by dissolving the antifungal powder in a solution of distilled water and Tween 80 (Thermo Fisher Scientific) according to the manufacturer’s instructions.

Treatment regimen.

Study groups consisted of either untreated controls or animals treated with posaconazole administered orally at dosages of 2, 6, and 20 mg/kg QD (cohort 3) or 1, 2, and 3 mg/kg BID (cohort 4). Antifungal therapy was started 24 h after endotracheal inoculation and continued throughout the course of the experiments for a maximum of 12 days in surviving rabbits.

Prophylaxis regimen.

The prophylaxis experiments used the same methods as described above with the following exceptions. Rabbits received the same dosages of posaconazole (2, 6, or 20 mg/kg/day) administered for 4 days before endotracheal inoculation. On the day of inoculation, posaconazole was administered in the morning and the endotracheal inoculum was administered approximately 4 h later. Posaconazole was then continued for a maximum of 12 more days after inoculation. To simulate the setting of antifungal prophylaxis, the administered inoculum was 5 × 10⁷ conidia.

Outcome variables.

Surviving rabbits were euthanized by i.v. administration of sodium pentobarbital (The Butler Company, Columbus, OH) (65 mg (1 ml)/kg of body weight) 24 h after administration of the last dose of antifungal drug or vehicle (controls). In the primary experiments, a panel of outcome variables was used to assess antifungal efficacy. These variables included survival in days postinoculation, lung weight and pulmonary infarction score as measures of organism-mediated pulmonary injury, and microbiological clearance from lung tissue in log CFU per gram. Blood was collected QOD from each rabbit, and the serum GMI was determined with the exception of cohort 4, for which only the last available specimen was analyzed.

PK sampling.

A sparse sampling strategy was employed to obtain key PK parameters for each individual infected animal and to correlate PD parameters with endpoints of antifungal efficacy. The time points for sparse plasma sampling were determined using optimal sampling theory implemented by the ADAPT II computer program (45) and full concentration-time profiles from 6 healthy rabbits following oral administration of a single dose of 20 mg/kg, with dense sampling prior to dosing and 0.5, 1, 2, 4, 6, 8, 12, 24, 48, 72, and 96 h after dosing (cohort 1) (46). The plasma profiles for these rabbits were fitted to a one-compartment PK model with first-order input, no lag time, and first-order elimination. The model fitted the data well, with a mean r² value of 0.964. The selected time points for sparse sampling were immediately before dosing and 1, 4, 8, and 24 h after dosing (40). Plasma sampling in infected rabbits of cohorts 2, 3, and 4 was performed 6 days after inoculation. Blood samples were immediately centrifuged, and plasma was stored at −80°C until assayed.

Analytical method.

After solid-phase extraction, concentrations of posaconazole were determined by liquid chromatography-tandem mass spectrometry at the Schering-Plough Research Institute (Kenilworth, NJ, USA). The analytical procedure involved dilution of the samples in control plasma prior to extraction. The quantifiable range of the assay was 4 to 1,000 ng/ml. Accuracies (bias) and intraday and interday variabilities (precision) were within ±15% and ±20%, respectively, at the lower limit of quantitation (30, 40).

PD sampling.

To describe the PD of posaconazole in prophylaxis and treatment of experimental IPA, the GMI was used as a dynamic endpoint of antifungal efficacy. Blood was collected from each infected rabbit prior to inoculation and QOD thereafter, and the serum GMI was determined. In cohort 4, the GMI was determined only for the last available specimen from each rabbit (30, 40).

Galactomannan assay.

Serum galactomannan levels were determined by the Platelia Aspergillus enzyme immunoassay (EIA) sandwich microplate assay (Genetic Systems/Sanofi Diagnostic Pasteur, Redmond, WA) as described previously (30). EIA data were expressed as a serum GMI. The GMI for each test serum sample is equal to the optical density (OD) (determined by microplate spectrophotometer at 450 nm) for the sample divided by the OD for threshold serum.

Population PK/PD modeling.

PK/PD model building was performed using the nonparametric adaptive grid (NPAG) approach with the Pmetrics software package in R (version 1.5.1; Laboratory of Applied Pharmacokinetics, Los Angeles, CA, USA) (47). The additive lambda approach was chosen to describe the residual error. Model building consisted of a two-step process. First, the population PK model was created. Second, the model was extended to a full PK/PD model. An overview of the PK/PD analysis study setup is given in Fig. 2.

To explore the PK of posaconazole in rabbits, an initial model was built on the basis of dense concentration data for 6 healthy rabbits after a single oral dose of 20 mg/kg posaconazole (cohort 1). In a next step, the steady-state concentration data obtained from infected rabbits receiving posaconazole as prophylaxis or treatment were added (cohorts 2, 3, and 4).

During the model-building process, different structural options were tested. Models consisting of one or two compartments with either linear or nonlinear Michaelis-Menten-type elimination were considered. To compare the different models, the log-likelihood profile was used on nested models. Nonhierarchical models were compared using the Akaike information criterion (AIC). In addition to statistical criteria, graphical output was used for model evaluation, including goodness-of-fit plots comparing individual and population predictions with observed plasma concentrations, as well as graphical evaluations of the residuals.

The PD effect of posaconazole in prophylaxis and treatment of experimental IPA, reflected by the serial assessments of the GMI (cohorts 2 and 3), was added to the final PK model. Since the GMI for each rabbit was determined only for the last available specimen, data from rabbits of cohort 4 were not included in this analysis. The previous PK modeling thus informed PK support point distributions for the full PK/PD model, for which PK samples were available only at steady state.

In a first step, adequate functions to represent the evolution of the GMI with and without antifungal treatment were explored. For this purpose, a subset including only the treatment data was formed. The PD effect was modeled linearly, with power functions as well as sigmoidal Hill functions. The type of function that best depicted the PD of posaconazole was then expanded to reflect the prophylaxis arm. During the model-building process, the selection of the most appropriate model was guided by the inspection of the AIC and goodness-of-fit plots, as well as residual plots.

The final PK/PD model was used to calculate the individual Bayesian posterior for each parameter and was subsequently used to determine the AUC for posaconazole levels and for the GMI. To determine the effectiveness of the treatment regimens, the AUC_0–24 values for the two variables were compared. For posaconazole plasma concentration AUC calculations, day 5 of the study was chosen, because posaconazole was thought to be at steady state and the GMI to be reasonably evolved at that time point. The prophylactic regimen was evaluated by comparing the posaconazole trough level on the day of the inoculation with the GMI on day 5 after inoculation, to allow for a reasonable time frame for IPA evolution under prophylaxis.

This manuscript is dedicated to the memory of Diana Mickiene, a dear colleague and friend, who has made lasting contributions to the preclinical development of antifungal agents.