Pharmacokinetic and Pharmacodynamic Evaluation of Liposomal ...

ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, Sept. 2009, p. 3847–3854 0066-4804/09/$08.00⫹0 doi:10.1128/AAC.00872-08 Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Vol. 53, No. 9

Pharmacokinetic and Pharmacodynamic Evaluation of Liposomal Amikacin for Inhalation in Cystic Fibrosis Patients with Chronic Pseudomonal Infection䌤 ´ lanrewaju O. Okusanya,1 Sujata M. Bhavnani,1 Jeffrey Hammel,1 Predrag Minic,2 Lieven J. Dupont,3 O Alan Forrest,1 Geert-Jan Mulder,4 Constance Mackinson,5† Paul G. Ambrose,1* and Renu Gupta5 Institute for Clinical Pharmacodynamics, Ordway Research Institute, Inc., Albany, New York1; Institute for Mother and Child Healthcare, Belgrade, Serbia2; Department of Respiratory Medicine, University Hospital Gasthuisberg, Katholieke Universiteit Leuven, Leuven, Belgium3; Forbion Capital Partners, Naarden, The Netherlands4; and Transave, Inc., Monmouth Junction, New Jersey5 Received 1 July 2008/Returned for modification 4 September 2008/Accepted 12 May 2009

The pharmacokinetics and pharmacodynamics of a novel liposomal amikacin for inhalation were evaluated in cystic fibrosis patients with chronic pseudomonas infection. Twenty-four patients from two studies received 500 mg of liposomal amikacin by inhalation once daily for 14 days. Serum, sputum, and 24-h urine samples were collected on days 1 and 14 of therapy; pulmonary function tests (PFT) and sputum for quantitative microbiology were assessed at baseline and serially for 14 days. Relationships between amikacin exposure in serum and sputum and absolute change in PFT endpoints and log10 CFU of Pseudomonas aeruginosa from baseline on days 7 and 14 of therapy were assessed. On days 7 and 14, absolute change from baseline in forced expiratory volume in 1 s (FEV1), percent predicted forced expiratory volume in 1 s (FEV1 % predicted), and forced expiratory flow between 25 and 75% of forced vital capacity (FEF25–75%) increased by 0.24 (P ⴝ 0.002) and 0.13 (P ⴝ 0.10) liters, 7.49 (P < 0.001) and 4.38 (P ⴝ 0.03), and 0.49 (P < 0.001) and 0.42 (P ⴝ 0.02) liters/s, respectively. In addition, relative change from baseline in FEV1 % predicted was 10.8% (P < 0.001) and 5.62% (P ⴝ 0.073) on days 7 and 14, respectively. While significant relationships between absolute change in PFT endpoints and the ratio of serum or sputum area under the concentration-time curve to the MIC (AUC/MIC) were not observed, relationships between change in log10 CFU and serum AUC/MIC ratio and change in log10 CFU and absolute changes in all PFT endpoints were significant. Together, these findings likely represent drug effect and warrant the further development of liposomal amikacin for inhalation. patients (13). Sustained release of such agents via a liposome delivery system may provide the opportunity to maintain prolonged targeted lung exposures and enhance the uptake of drug to the site of infection. Liposomal amikacin for inhalation is a sustained-release formulation of amikacin encapsulated inside nanoscale liposomal carriers designed for administration via inhalation. Given the promise of enhanced delivery and prolonged therapeutic effect at the infection site, CF patients may benefit from this formulation of amikacin for inhalation. In the analysis described herein, we evaluated the pharmacokinetics (PK) of liposomal amikacin for inhalation using data from two phase 1b/2a studies in which CF patients who were chronically infected with P. aeruginosa received 14 daily doses of the study drug. Subsequently, PK-pharmacodynamic (PD) relationships between exposure of liposomal amikacin and the changes relative to baseline in pulmonary function (change in forced expiratory volume in 1 s [FEV1], percent predicted forced expiratory volume in 1 s [FEV1 % predicted], forced expiratory flow between 25 and 75% of forced vital capacity [FEF25–75%], and forced vital capacity [FVC]) and values of P. aeruginosa CFU in sputum were evaluated.

Cystic fibrosis (CF) is a genetic chronic disease that affects approximately 60,000 people worldwide (10). The genetic defect occurs in the CF transmembrane regulator gene, resulting in abnormal production and/or function of its protein. CF transmembrane regulator is an essential protein that is found in the epithelial cells of the body’s internal passageways, such as the lung and gastrointestinal tract. This defective protein results in the production of thick and sticky mucus that impairs normal lung function and leads to chronic life-threatening infections, often involving Pseudomonas aeruginosa (3). Despite the observed benefits of frequent antimicrobial therapy in CF patients with P. aeruginosa infections, eradication of this organism is a difficult endpoint to achieve. Mucus plugs, inactivation of agents by sputum of CF patients, and development of bacterial biofilm are factors which contribute to the poor lung penetration of antimicrobial agents and, ultimately, the emergence of multidrug bacterial resistance. To increase drug exposure at the infection site and minimize systemic exposure (thereby reducing the potential for systemic adverse events), drug delivery via inhalation has been a preferred route for agents such as the aminoglycosides in CF

* Corresponding author. Mailing address: ICPD, 43 British American Blvd., Latham, NY 12110. Phone: (518) 429-2603. Fax: (518) 429-2601. E-mail: [email protected]. † Present address: Ortho-McNeil Pharmaceutical, Inc., Raritan, NJ 08869-0602. 䌤 Published ahead of print on 18 May 2009.

MATERIALS AND METHODS Patient population. The patients included in this analysis were enrolled in one of two phase 1b/2a multiple-dose studies which were conducted to evaluate the safety and tolerability of liposomal amikacin for inhalation. CF patients enrolled in both of these studies were required to be at least 13 years of age or older and

3847

3848

OKUSANYA ET AL.

to have mild to moderate obstructive lung disease (FEV1 % predicted, ⱖ40%). The patients were also required to have chronic infection with P. aeruginosa (defined by ⱖ1 year of positive cultures) and a screening sputum sample positive for the growth of P. aeruginosa, which was susceptible to amikacin with a MIC of ⱕ16 ␮g/ml. Patients were excluded if they had been administered an investigational drug up to 4 weeks prior to screening, had an emergency room visit or were hospitalized due to CF or a respiratory-related illness within 4 weeks prior to screening, or had changes to their physiotherapy technique or schedule during the week prior to screening. Other exclusion criteria included the following: a history of alcohol or drug abuse within 1 year of screening, a history of lung transplantation, the use of antipseudomonal antibiotics within 14 days of study treatment, the initiation of chronic therapy (such as inhaled tobramycin, highdose ibuprofen, recombinant human DNase, and macrolides) within 60 days of screening, a history of sputum or throat swab culture yielding Burkholderia cepacia within 2 years prior to or at screening, a history of mycobacterial infection at screening, a history of daily continuous oxygen supplementation or a requirement for ⬎2 liters/min of oxygen at night, and changes in chest X-ray at screening (or within 6 months of screening) with new onset of infiltrates or changes that would compromise patient safety or the quality of the data. Other interventions performed for enrolled patients were consistent with their normal standard of care. Drug dosage and administration. Patients enrolled in each of the two studies received a once-daily 500-mg dose of a 50-mg/ml formulation of liposomal amikacin for inhalation (as two 20-min sessions with a 5-min rest period between inhalations via the Pari LC Star nebulizer and a DeVilbiss compressor operating at 30 lb/in2) for a period of 14 days. Sample collection and drug assay. Serum samples were collected on days 1 and 14 prior to dosing and at 1, 2, 4, 6, 8, 12, and 24 h postdose. Urine samples were generally collected at 6-h intervals for up to 24 h on days 1 and 14, with spot urine collected on days 3 through 13. Sputum samples were collected on days 1 and 14, just after the dose was administered, between 4 and 6 h after dosing, and prior to dose administration on the following day. Sputum samples were also collected for microbiological testing during screening (between ⫺14 and 0 days prior to study initiation) and prior to drug administration on days 1, 7, and 14. Serum, urine, and sputum samples were assayed for amikacin by using liquid chromatography-tandem mass spectrometry with lower limits of quantification of 0.15 ␮g/ml, 0.5 ␮g/ml, and 0.1 ␮g/ml for the serum, urine, and sputum assays, respectively. The percent coefficient of variation (% CV) of the assay was 4.2% for serum, 11% for urine, and 8.9% for the sputum (data on file at Transave, Inc.). Sputum samples for microbiology were collected, processed, and examined using routine laboratory methods. The sputum samples were processed in order to identify P. aeruginosa by standard biochemical tests. CFU determination and susceptibility testing for the different subpopulations (phenotypic isolates) were performed at the study site labs using the Etest method in accordance with guidelines recommended by the manufacturer (AB Biodisk North America Inc., Culver City, CA). In a given sample, the sum of the CFU for all the phenotypic isolates of P. aeruginosa was obtained to determine the bacterial burden of the patient. The highest MIC associated with the above-described distinct colonies of P. aeruginosa was selected for the subsequent analysis. PFTs. Pulmonary function tests (PFTs), which included FEV1, FEV1 % predicted, FEF25–75%, and FVC, were carried out during screening from day ⫺14 to 0 and at baseline (i.e., prior to dose administration on day 1) and on days 1, 7, and 14. PFTs were also carried out 1.5 and 3 h postdose on days 1 and 14. PK analysis. All patients meeting the eligibility criteria and who also had at least one evaluable amikacin serum concentration were included in the PK analysis. A population PK model was developed to characterize the serum amikacin and urinary excretion data simultaneously by using Monte Carlo parametric expectation maximization, as implemented in the open-source software S-ADAPT 1.53 (2), running on a Windows XP operating system and compiled using Intel Fortran 9.1. Weighting of the serum concentrations and urine amounts was based on the reciprocal of the estimated observation variance (the “error” standard deviation [SD] squared for that observation), which was predicted as a function of the fitted concentrations. Relationships between concentration and variance (using “error” variance models) were estimated based on the performance (as assessed by interday % CV and sensitivity) of the assays for drug in serum and urine. An additive variance model was used for the serum output, while an additive plus proportional variance model was used for the urine output; the additive component in the urine was not estimated but rather fixed at a reasonable value for the population, while the proportional component was estimated but assumed to be constant across the population. The standard error (SE) of the mean param-

ANTIMICROB. AGENTS CHEMOTHER. TABLE 1. Baseline demographics of the analysis population Variable Patient variables Age (yr) Weight (kg) Height (cm) Ideal body wt (kg) Body surface area (m2) CLCR (ml/min/1.73 m2) Sex Male Female Baseline pharmacodynamic variables FEV1 (liters) FEV1 % predicted FEF25–75% (liters/s) FVC (liters) Log10 CFU MIC (␮g/ml)

n

24 24 24 24 24 24

Mean (SD)

Median (minimum, maximum or range)

23.7 (6.96) 59.1 (13.0) 168 (8.10) 61.4 (8.99) 1.66 (0.118) 125 (20.9)

22.5 (14.0, 38.0) 58.6 (43.4, 99.6) 168 (155, 194) 60.0 (47.9, 87.7) 1.64 (1.40, 2.17) 126 (76.8, 173)

2.38 (1.07) 65.5 (18.9) 1.71 (1.26) 3.32 (0.92) 7.05 (1.30) 25.3 (39.8)

2.18 (1.15–6.10) 62.5 (40.0–119) 1.49 (0.55–5.50) 3.27 (1.67–5.28) 7.30 (3.51–8.98) 10.0 (2.0–192)

10 (41.7%) 14 (58.3%)

24 24 24 24 21 21

eter estimates was derived from the full second derivative matrix using third and fourth central moments (2). A change in ⫺2 · log likelihood of at least 3.84 (␣ ⫽ 0.05, 1 df) was used to define statistical significance for addition or deletion of a single parameter during the model building process. Model discrimination was accomplished according to the “rule of parsimony” based on Akaike’s information criterion (1, 9). The 24-h serum area under the concentration-time curve (AUC) at steady state was determined for each of the two study periods by dividing the administered dose by the estimated clearance for the associated study period. The area under the sputum concentration-time curve on days 1 and 14 was calculated for each patient by using the linear trapezoidal rule. The average 24-h serum and sputum AUCs were computed by taking the average of the day 1 and 14 AUC values. PK-PD analysis. Statistical analyses were carried out in R 2.4.1 (11). The measure of exposure for this analysis was the average 24-h AUC for serum and sputum, and the measure of drug potency was the baseline MIC of amikacin for P. aeruginosa. The ratio of the average 24-h AUC for serum and sputum to the baseline MIC (AUC/MIC) for P. aeruginosa was computed. Both the absolute and relative changes on days 7 and 14 relative to baseline in PFT values for FEV1, FEV1 % predicted, FEF25–75%, FVC, and log10 CFU of P. aeruginosa were computed, and the mean change from baseline was tested for statistical significance by using repeated-measure analysis of variance. Relative changes in PFT values were calculated by subtracting the baseline PFT value from the PFT value on the study day, dividing by the baseline PFT value, and multiplying by 100%. The relationship between the absolute change from baseline in each of the PFT endpoints or change in log10 CFU and AUC/MIC ratio for serum and sputum was assessed graphically using scatter plots. The direction and strength of the relationship between the absolute change from baseline for each of the PFT endpoints and AUC/MIC ratio for serum and sputum were assessed using Spearman’s rank correlation test (8). Similarly, the direction and strength of the relationship between change from baseline in each of the PFT values and log10 CFU were also assessed using Spearman’s rank correlation test.

RESULTS Patient population. There were 11 and 13 patients enrolled in each of the two phase 1b/2a multiple dose studies, respectively. There were no remarkable differences in the patient demographics between studies, and the baseline characteristics for all the patients are provided in Table 1. All 24 patients had data sufficient for inclusion in the PK and PK-PD analysis. One patient, however, did not have a baseline MIC and was excluded from the PK-PD analysis. While the inclusion criteria required that the MICs be ⬍16 ␮g/ml, 6 out of 11 patients in only one study had MICs of ⱖ16 ␮g/ml. These patients were not excluded from the subsequent PK-PD analysis.

VOL. 53, 2009

PK-PD OF INHALED LIPOSOMAL AMIKACIN IN CYSTIC FIBROSIS

TABLE 2. Structural population PK model for liposomal amikacin for inhalation with interoccasional variability—parameter estimates and SEsa Population mean Parameterb Final estimate

CLt/F day 1 (liters/h) Vc/F day 1 (liters) ka (h⫺1) CLR (liters/h) CLt/F day 14 (liters/h) Vc/F day 14 (liters) SDintserum SDslpurine SDinturine

68.4 286 3.34 3.40 45.2 250 0.05 0.70 0.03

% SE

10.3 12.3 32.5 15.4 8.01 8.51 6.02 9.16

Interindividual variability (% CV) Final estimate

% SE

48.7 59.0 99.8 63.9 37.1 27.0

29.9 29.7 50.5 36.7 30.7 30.8

Minimum value of the objective function is ⫺258.6. SDintserum, residual error for serum data; SDslpurine and SDinturine, residual error slope and intercept for urine data, respectively. a b

PK analysis. The concentration-time profiles for all 24 patients on days 1 and 14 were evaluated. Amikacin serum concentrations were below 1.5 ␮g/ml at all times in all patients. The most robust fit to the serum concentration data was obtained using a two-compartment model (one absorption site, the lung, and one central compartment) with zero-order drug input into the lungs, a first-order process from lungs to the central compartment (ka), and first-order elimination with interoccasional variability on the apparent volume of distribution (Vc/F) and apparent total clearance (CLt/F). Amikacin in the urine was modeled as accumulating in a urine compartment after clearance from the central compartment (CLR) by a first-order renal process. The urine compartment was modeled as being emptied at the end of each collection interval. Population PK parameter estimates and SEs, with interoccasional variability, are provided in Table 2. Allowing interoccasional variability on renal clearance did not result in a statistically significant improvement in fit or objective function and thus was not incorporated into the final model. As evidenced by the relatively low percent SE values, the precision of

3849

the mean PK parameters was high. With the exception of ka, the magnitude of the interindividual variability (% CV) was moderate. There was a significant systematic decrease in the mean CLt/F on day 14 compared to that on day 1 (P ⫽ 0.002). However, the decline in mean Vc/F on day 1 compared to that on day 14 was not statistically significant (P ⫽ 0.24). This indicates that while the data could statistically justify different CLt/F and Vc/F by day, there was a significant decrease in CLt/F within study patients but no statistical net change in Vc/F, indicating that some patients had higher Vc/F values on day 1 than on day 14 and vice versa. It is of note that the fitted CLR from the model, which should closely reflect the patients’ renal function, was statistically significantly lower than the predicted glomerular filtration rate (GFR) (P ⱕ 0.001). This discordant result was due to the poor collection of urine in the study and resulted in an inadequate characterization of the renal clearance. The mean half-life of absorption for amikacin from the lung into the body was 0.208 h, while the mean elimination half-life values on days 1 and 14 were 2.9 and 3.8 h, respectively. As evidenced by an overall r2 of 0.98 for observed versus individual fitted serum concentrations, the fits to the serum data were excellent. Also, as seen in Fig. 1A, the line of best fit did not differ from the line of identity. As evidenced by an overall r2 of 0.38 for observed versus individual fitted urine data, shown in Fig. 1B, the fit for urine data was poor. Serum AUC values were obtained by dividing the dose of amikacin administered by the conditional clearance on days 1 and 14. The mean (SD) and median (range) for resultant serum AUC values were 8.27 (4.3) and 6.88 (3.67 to 20.1) ␮g · h/ml on day 1 and 12.0 (5.1) and 10.8 (5.65 to 30.1) ␮g · h/ml on day 14, respectively. Visual inspection of the sputum concentration-time profiles revealed a high degree of intersubject variability at each time point. The mean (SD) and median (range) for sputum AUC values were 3,830 (4,500) and 1,970 (78.7 to 17,200) ␮g · h/g on day 1 and 12,500 (11,400) and 10,578 (10,578 to 50,000) ␮g · h/g on day 14, respectively. PK-PD analysis. (i) PFT endpoints. As the observed results for PFTs obtained predose or 1.5 or 3.0 h postdose on a given

FIG. 1. Scatter plot of observed versus individual fitted amikacin serum concentrations (A) and urine amounts (B). Note that the solid line represents the line of identity.

3850

OKUSANYA ET AL.

ANTIMICROB. AGENTS CHEMOTHER.

TABLE 3. Summary statistics and results of one-sample t test for absolute and relative change in PD endpoints from baseline PD endpoint

FEV1 (liters)

Change from baselinea

Absolute Relative (%)

FEV1 % predicted


FEF25–75% (liters/s)


FVC (liters)


Log10 CFU

Absolute

Day

n

Mean

SD

Minimum

Median

Maximum

One-sample t test P value

7 14 7 14 7 14 7 14 7 14 7 14 7 14 7 14 7 14

23 24 23 24 23 24 23 24 23 24 23 24 23 24 23 24 19 20

0.24 0.13 10.3 5.05 7.49 4.38 10.8 5.62 0.49 0.42 25.2 18.3 0.21 0.11 5.12 2.19 ⫺0.15 ⫺0.32

0.34 0.36 0.12 0.15 8.13 9.18 0.12 0.15 0.62 0.80 0.28 0.42 0.41 0.39 0.11 0.11 1.13 1.39

⫺0.32 ⫺0.59 ⫺0.09 ⫺0.17 ⫺5.5 ⫺8.3 ⫺0.09 ⫺0.18 ⫺0.56 ⫺0.25 ⫺0.36 ⫺0.36 ⫺0.4 ⫺0.41 ⫺0.20 ⫺0.25 ⫺3.26 ⫺3.03

0.2 0.045 0.10 0.016 6.9 2.5 0.104 0.029 0.36 0.085 0.27 0.10 0.14 0.055 0.043 0.014 0.04 ⫺0.072

1.24 0.94 0.33 0.43 25 28.2 0.329 0.434 2.02 2.88 0.83 1.5 1.38 1.1 0.26 0.22 1.66 2.00

0.002 0.10 ⬍0.001 0.114 ⬍0.001 0.029 ⬍0.001 0.073 0.001 0.016 ⬍0.001 0.044 0.023 0.18 0.035 0.355 0.56 0.32

a Relative change from baseline for each PFT was expressed as percent change from baseline and was calculated as follows: (PFT value on study day ⫺ baseline PFT value) ⫻ 100%/baseline PFT value.

study day were not considered to be remarkably different, the analysis was based on the predose PFT values. Summary statistics for absolute and relative changes on days 7 and 14 relative to baseline in PFT values are shown in Table 3. Mean absolute and relative changes in PFT values on day 7 compared to baseline values were statistically significant for all PFT endpoints (P ⱕ 0.035). Mean absolute changes on day 14 relative to baseline in FEV1 % predicted and FEF25–75% were also statistically significant (P ⫽ 0.029 and P ⫽ 0.016, respectively).

Scatter plots, with a linear regression line provided as a visual guide, showing the relationship between absolute change on day 14 relative to baseline in each PFT endpoint and AUC/ MIC ratios for serum and sputum are provided in Fig. 2 and 3, respectively. Visual inspection of these plots, supported by correlations between change from baseline in PFT values and either serum or sputum AUC/MIC ratio, did not reveal a statistically significant relationship between the evaluated variables.

FIG. 2. Scatter plots of absolute change on day 14 relative to baseline in PFT values versus AUC/MIC ratio for serum. The squares and circles represent data from the two studies included in this analysis. A linear regression line has been provided as a visual guide. The squared Spearman rank correlation, rs2, and the P value for a corresponding test of association are reported.

VOL. 53, 2009


3851

FIG. 3. Scatter plots of absolute change on day 14 relative to baseline in PFT values versus AUC/MIC ratio for sputum. The squares and circles represent data from the two studies included in this analysis. A linear regression line has been provided as a visual guide. The squared Spearman rank correlation, rs2, and the P value for a corresponding test of association are reported.

(ii) Microbiologic endpoints. Summary statistics for change relative to baseline on days 7 and 14 in log10 CFU were evaluated. Regardless of the study day considered, mean change from baseline in log10 CFU was not statistically significant (P ⫽ 0.56 on day 7 and P ⫽ 0.32 on day 14). When the relationship between change in log10 CFU and baseline MIC was assessed, statistically significant relationships were found for evaluations on days 7 (rs ⫽ 0.57, P ⫽ 0.011) and 14 (rs ⫽ 0.64, P ⫽ 0.0026). Scatter plots, with a linear regression line provided as a visual guide, showing the relationship between absolute change on day 14 relative to baseline in log10 CFU and AUC/MIC ratios for serum and sputum are provided in Fig. 4. Visual inspection of these plots demonstrated a relationship between

increasing serum AUC/MIC ratios and decreases in log10 CFU. Of note, approximately half of the patients evaluated demonstrated a decrease from baseline in log10 CFU across the range of serum AUC/MIC ratios. The correlation between change from baseline in log10 CFU and serum AUC/MIC ratio was statistically significant for both days 7 and 14. Increasing serum AUC/MIC ratios were associated with larger decreases relative to baseline in log10 CFU on days 7 (rs ⫽ ⫺0.46, P ⫽ 0.048) and 14 (rs ⫽ ⫺0.45, P ⫽ 0.048). Despite the statistical significance of this relationship, only a modest amount of variability in change from baseline in log10 CFU was explained by serum AUC/MIC ratio (as evidenced by the low values of rs2, ⱕ0.21). The correlation between change from baseline in log10 CFU and sputum AUC/

FIG. 4. Scatter plots of absolute change on day 14 relative to baseline in log10 CFU versus AUC/MIC ratio for serum and sputum. The squares and circles represent data from the two studies included in this analysis. A linear regression line has been provided as a visual guide. The squared Spearman rank correlation, rs2, and the P value for a corresponding test of association are reported.

3852

OKUSANYA ET AL.

ANTIMICROB. AGENTS CHEMOTHER.

FIG. 5. Scatter plots of absolute change on day 14 relative to baseline in PFT values versus change relative to baseline in log10 CFU. The squares and circles represent data from the two studies included in this analysis. A linear regression line has been provided as a visual guide. The squared Spearman rank correlation, rs2, and the P value for a corresponding test of association are reported.

MIC ratio was not statistically significant for either day 7 or day 14 (P ⱖ 0.6). (iii) Relationship between change in pulmonary function and log10 CFU. Scatter plots, with a linear regression line provided as a visual guide, showing the relationship between the absolute change on day 14 relative to baseline in each of the PFT endpoints and log10 CFU are shown in Fig. 5. Visual inspection of these plots demonstrated an apparent relationship between decreases in the change in PFT values and increases in the change in log10 CFU relative to baseline. However, as evidenced by the scatter of the data, some patients who had stable values or increases in log10 CFU had PFT values that did not change from baseline or that worsened. Correlations between change relative to baseline on days 7 and 14 in PFT values and log10 CFU were statistically significant for FEV1, FEV1 % predicted, and FVC (P ⬍ 0.05). Increases in the change in log10 CFU relative to baseline were associated with decreases in change in these three PFT endpoints relative to baseline on days 7 (rs ⱕ ⫺0.56, P ⱕ 0.026) and 14 (rs ⱕ ⫺0.52, P ⱕ 0.043). Despite the statistical significance of these relationships, only a modest amount of variability in change in these three PFT endpoints was explained by change in log10 CFU on either day 7 or day 14 (as evidenced by the low values of rs2, ⱕ0.31). DISCUSSION The first two objectives of these analyses were to use population PK modeling to characterize amikacin systemic exposure, including the approximate systemic bioavailability, and to characterize the liposomal amikacin exposure in sputum.

In the present analysis, which was based on multiple-dose serum and urine PK data, the population PK of liposomal amikacin for inhalation was best described by a two-compartment model, in which bioavailable drug leaves the absorptive compartment (postulated as the lung) according to a firstorder process into the central compartment and drug is eliminated (mainly unchanged) via a first-order process. Since the systemic bioavailability is not known, the volume of distribution and clearance are both conditioned on the fraction of the dose which is systemically absorbed (F). The best model to describe the serum and urine amikacin concentration was one that allowed for interoccasional variability in clearance and volume on days 1 and 14. This finding was not unexpected, as it highlights the importance of accounting for the variability in the drug delivery associated with different administrations. Of note, the apparent fraction absorbed was higher on day 14 than on day 1 (as evidenced by lower estimates of CLt/F and Vc/F on day 14 than on day 1). The fitted CLR, which should be closely related to renal function, was significantly lower than the predicted GFR or creatinine clearance (CLCR). Amikacin is eliminated primarily by glomerular filtration, as evidenced by previous reports of total CLR and CLCR of 100 ml/min and 94 ml/min, respectively, in healthy subjects (12). Therefore, close agreement between CLR and CLCR was expected. Possible explanations for the differences between CLR and CLCR include falsely high plasma concentrations, falsely small urine amounts (either of which may be due to limitations of the assay), incomplete urine collection and reporting, or reabsorption in the renal tubules. Incomplete collection and reporting of urine PK data were

VOL. 53, 2009


noted in the two studies described herein and thus represent an important limitation of the present analysis. As a result of the incomplete collection and reporting of urine PK data, the exposure of the kidneys per unit time to the drug and the actual systemic bioavailability of liposomal amikacin for inhalation could not be reliably estimated. The amount of drug that passes through the kidneys can be inferred to be relatively small due to the low systemic exposure seen in serum. In addition, means of 5.26% and 6.84% of the administered nebulized dose were recovered in the urine on days 1 and 14, respectively, indicating low systemic bioavailability. This is also supported by the finding that the serum AUC values obtained on days 1 and 14 were 8.7% and 12.6%, respectively, of the expected mean AUC of 95.1 mg · h/liter that would be obtained using the PK parameters reported by Byl et al. (4) after intravenous administration of amikacin to CF patients, the estimated total clearance for which was 5.26 liters/h. Median sputum AUC values based on the noncompartmental analysis were 286- and 978-fold greater than serum AUC values based on the population PK model on days 1 and 14, respectively. However, given the sparseness of the sputum concentration-time data, the inherent variability of sputum specimens from CF patients, the high degree of variability in noncompartmental sputum AUC estimates (% CV of 112 on day 1 and 91.2 on day 14) compared to fitted serum AUC values, and the technical challenges in obtaining relevant samples and measuring concentrations, limited inferences about pulmonary exposure can be made. The high degree of variability observed with sputum exposures is consistent with what has been reported for inhalation tobramycin (5, 7). Whether assessing after the first dose or after 20 weeks of tobramycin therapy, sputum concentrations ranged from 35 to 8,085 ␮g/g, with average concentrations of 1,237 and 1,154 ␮g/g after day 1 and 20 weeks, respectively (5). In addition to the challenges of obtaining and measuring sputum concentrations, it is likely that this variability is also due to the complex relationship between the drug delivery mode, the device, and patient-specific factors such as lung pathology, disease state, and sputum quantity and distribution within the lung (5). The objectives of the PK-PD analyses were to evaluate the relationships between amikacin exposure in serum and sputum after 7 and 14 days of therapy with liposomal amikacin for inhalation delivered by a Pari LC Star nebulizer and change relative to baseline in PFT values and log10 CFU of P. aeruginosa. Although PK-PD relationships for the absolute change from baseline in PFT values and AUC/MIC ratios for serum and sputum were not detected, statistically significant and clinical relevant absolute changes from baseline were evident for each of the four PFT endpoints on day 7 and for two of the four endpoints on day 14. The most consistent change over time was associated with FEF25-75%, for which the observed absolute change from baseline was 0.49 liters/s on day 7 (P ⫽ 0.001) and 0.42 liters/s on day 14 (P ⫽ 0.016), both of which are considered to be clinically relevant changes. This is not surprising given that FEF25–75% is a measure that generally reflects small airway function, which is usually affected by CF. There was also a statistically significant and clinically relevant absolute change on day 7 compared to baseline for FEV1

3853

(0.240 liter, P ⫽ 0.002) and FEV1 % predicted (7.49, P ⬍ 0.001). By day 14, absolute change from baseline either remained statistically significant (change in FEV1 % predicted, P ⫽ 0.029) or was of borderline significance (FEV1, P ⫽ 0.10). Together, these changes may represent a signal of drug effect more specific than FEF25–75%. While statistically significant changes from baseline in log10 CFU were not identified on day 7 or 14, relationships between change from baseline in log10 CFU on day 7 and 14 and serum AUC/MIC ratio were observed. Although these relationships were statistically significant (P ⫽ 0.048), increases in serum AUC/MIC ratio explained only a small proportion in the variability in change in log10 CFU (rs2 ⱕ 0.21). While PK-PD relationships for change in log10 CFU were apparent for serum AUC/MIC ratio, such relationships were not observed for sputum AUC/MIC ratio. The lack of such findings may have been due in part to the aforementioned high degree of variability associated with sputum AUC values. Statistically significant relationships (P ⬍ 0.05) were also observed between the absolute change in FEV1, FEV1 % predicted, and FVC and change in log10 CFU, and as described above, only a small amount of variability in change in these three PFT endpoints was explained by the change in log10 CFU on either day 7 or day 14 (rs2 ⱕ 0.31). While the transitive logic of the aforementioned relationships between (i) drug exposure in serum and the reduction in log10 CFU and (ii) the reduction in log10 CFU and improvement in PFT endpoints is appealing, it is important to note that, across the combined population, there was no overall change from baseline in log10 CFU over time. At the current dosing regimen of 500 mg/day, some patients demonstrated improvement in PFT values or log10 CFU over time, while others either did not or remained stable. This suggests that larger doses of liposomal amikacin for inhalation may be required to be more reliably effective in large patient populations. In conclusion, through these analyses, serum exposures for liposomal amikacin for inhalation were reliably estimated, thus enabling exploration of PK-PD relationships for efficacy. The evaluation of additional data, as described herein, collected over a wider dose range will enable further characterization of PK-PD relationships using PFT and/or log10 CFU endpoints in patients with CF. Results from such analyses will be of benefit to support dose selection for further clinical development of liposomal amikacin for inhalation. Clinical studies evaluating the safety and efficacy of a 70-mg/ml formulation of liposomal amikacin for inhalation are ongoing. REFERENCES 1. Akaike, H. 1979. A Bayesian extension of the minimum AIC procedure of autoregressive model fitting. Biometrika 66:237–243. 2. Bauer, R. J. 2006. S-ADAPT/MCPEM user’s guide: software for pharmacokinetic, pharmacodynamic and population data analysis. 3. Brennan, A. L., and D. M. Geddes. 2004. Bringing new treatments to the bedside in cystic fibrosis. Pediatr. Pulmonol. 37:87–98. 4. Byl, B., D. Baran, F. Jacobs, A. Herschuelz, and J.-P. Thys. 2001. Serum pharmacokinetics and sputum penetration of amikacin 30 mg/kg once daily and of ceftazidime 200 mg/kg/day as a continuous infusion in cystic fibrosis patients. J. Antimicrob. Chemother. 48:325–327. 5. Chiron Corporation. November 2004. TOBI (tobramycin inhalation solution, USP) prescribing information. Chiron Corporation, Emeryville, CA. 6. Reference deleted. 7. Geller, D. E., W. H. Pitlick, P. A. Nardella, W. G. Tracewell, and B. W. Ramsey. 2002. Pharmacokinetics and bioavailability of aerosolized tobramycin in cystic fibrosis. Chest 122:219–226.

3854

OKUSANYA ET AL.

8. Hollander, M., and D. A. Wolfe. 1973. Nonparametric statistical inference, p. 185–194. John Wiley & Sons, New York, NY. 9. Jusko, W. 1992. Applied pharmacokinetics: principles of therapeutic drug monitoring. Applied Therapeutics, Inc., Vancouver, WA. 10. Moss, R. B. 2002. Long-term benefits of inhaled tobramycin in adolescent patients with cystic fibrosis. Chest 121:55–63. 11. R Development Core Team. 2006. R: a language and environment for

ANTIMICROB. AGENTS CHEMOTHER. statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 12. Sicor Pharmaceuticals. June 2005. Amikacin sulfate injection, USP. Package insert. Sicor Pharmaceuticals, Inc., Irvine, CA. 13. Touw, D. J., R. W. Brimicombe, M. E. Hodson, H. G. Heijerman, and W. Bakker. 1995. Inhalation of antibiotics in cystic fibrosis. Eur. Respir. J. 8:1594–1604.