Improved characterization of aminoglycoside penetration into human lung epithelial lining fluid via population pharmacokinetics

ABSTRACT

Aminoglycosides are important treatment options for serious lung infections, but modeling analyses to quantify their human lung epithelial lining fluid (ELF) penetration are lacking. We estimated the extent and rate of penetration for five aminoglycosides via population pharmacokinetics from eight published studies. The area under the curve in ELF vs plasma ranged from 50% to 100% and equilibration half-lives from 0.61 to 5.80 h, indicating extensive system hysteresis. Aminoglycoside ELF peak concentrations were blunted, but overall exposures were moderately high.

INTRODUCTION

Aminoglycosides are an important part of our armamentarium to treat serious lung infections caused by multidrug-resistant Gram-negative pathogens, such as Pseudomonas aeruginosa, Acinetobacter baumannii, and Klebsiella pneumoniae, and often used in combination with β-lactam antibiotics (1–12). Aminoglycosides are polar polycations (Table S1) (13), have small volumes of distribution (11, 14, 15), low (<20%) protein binding (16–19), and diminished activity in acidic pH (20, 21). A few studies report lung epithelial lining fluid (ELF) concentrations to range from 10% to 30% of those in plasma or serum, though these values only considered one or two time points during the first 2 h after the start of a short-term infusion (22–25). In contrast, ELF-to-plasma concentration ratios increased over time and reached 100% in all studies that determined ELF concentrations over up to 6 to 24 h post dose (Fig. S1) (26–31).

Time-course modeling can estimate the rate and extent of penetration, as employed to assess penetration of other antibiotic classes into ELF (32–37), bone (38–41), and cerebrospinal fluid (42, 43). Physiologically based modeling has been applied to predict the pulmonary pharmacokinetics (PK) of fluoroquinolones (44). Population PK modeling is highly beneficial to handle data sets with sparse sampling (e.g., only one ELF concentration per patient) and has been employed to model ELF concentrations of aminoglycosides in mice (45–47). However, we are not aware of published PK modeling analyses of the rate and extent of ELF penetration for aminoglycosides in humans. Instead, all but one prior study reported ELF-to-plasma concentration ratios at single or several time points (22–30) without even applying a non-compartmental analysis (NCA) (48). One study used bronchoscopic micro-sampling and NCA to calculate the area under the curve (AUC) in ELF and plasma and reported a ratio of 67.6% for arbekacin (31). Thus, despite extensive clinical use of aminoglycosides to combat serious lung infections for decades, their extent and rate of human ELF penetration have never been characterized via time-course modeling. Consequently, the impact of system hysteresis with ELF concentrations lagging behind those in plasma has never been quantitatively examined (22). Failure to consider the time course of exposures may lead to underestimating drug exposures in ELF in patients.

We performed model-based meta-analyses by simultaneously modeling all data from each aminoglycoside based on eight published PK studies assessing the ELF penetration in humans after intravenous or intramuscular dosing (23, 24, 26–31). We excluded studies with inhaled dosing (14, 49), one study on plazomicin reported as a poster (25, 50), and one study (51) where tobramycin concentrations in bronchoalveolar lavage fluid were expressed as a function of creatinine. Plasma (or serum) and ELF concentration values were used as reported (23, 24, 26, 28) or digitized (27, 29–31) (see Table S2 for details). For each drug, all plasma (or serum) and ELF concentrations were analyzed by population PK modeling in the S-ADAPT (version 1.57; importance sampling algorithm) and SADAPT-TRAN software packages (52–54) using previously described approaches (41, 55–59). The systemic PK of aminoglycosides was described by linear one or two compartment models (Fig. S2), plus an additional ELF compartment with a small, non-influential volume of distribution (fixed to 0.1 L) (39, 58). We estimated the ratio (F_ELF) for the AUC in ELF vs plasma or serum and the ELF-to-plasma equilibration half-life (t_1/2,eq). The F_ELF characterizes the overall extent and t_1/2,eq the rate of penetration.

Population PK estimated the mean ELF-to-plasma AUC ratios (F_ELF) between 0.502 and 1.00 for all aminoglycosides with good precision (relative standard errors, RSE ≤12%, except for 31% for amikacin, Table 1). The between-patient variability of F_ELF was large for amikacin (84.4% coefficient of variation) and smaller for the other aminoglycosides (≤27.6%; Fig. 1; Fig. S3). The individual subject estimates for the ELF-to-plasma AUC ratio ranged from 0.138 to 1.60 for amikacin and from 0.468 to 1.94 for gentamicin, netilmicin, and tobramycin (Table 1).

TABLE 1.

Population PK parameter estimates for one- or two-compartment models describing plasma and lung ELF concentrations of five aminoglycosides after intravenous (or intramuscular) dosing^a

Drug	Population	AUCELF/AUCPlasma			t1/2,eq (h)	CL (L/h)	CLd (L/h)	V1 (L)	V2 (L)
		Mean (RSEc)	BSVb (RSEc)	Median [range]	Mean	Mean	Mean	Mean	Mean
Amikacin	ICU patients (with VAP)	0.502 (31%)	0.844 (39%)	0.461 [0.138–1.60]	5.80 (58%)	4.88		24.6
Amikacin	Ventilated neonates					0.138		1.61
Arbekacin	Healthy volunteers	0.535 (12%)	0.10 (fixed)	–f	0.613 (24%)	3.56	6.77	3.00	4.77
Gentamicin	VABP critically ill patients	0.600 (7%)	0.165 (160%)	0.600 [0.507–0.782]	0.857 (18%)	3.63	14.9	13.5	13.9
Netilmicin	ICU patients (ventilated, with pneumonia)e	1.00 (11%)	0.276 (67%)	0.989 [0.627–1.94]	1.68 (13%)	5.61	6.90	9.36	7.87
Tobramycind	VABP critically Ill patients and healthy volunteers	0.642 (9%)	0.248 (84%)	0.635 [0.468–0.901]	3.38 (17%)	5.60		20.1

^a The volume of distribution of the ELF compartment was set to a small, non-influential value (0.1 L). The estimated ELF-to-plasma equilibration half-life characterized the extent of system hysteresis. AUC_ELF, area under the ELF concentration time curve from time zero to infinity (for a single aminoglycoside dose); AUC_Plasma, area under the plasma concentration time curve from time zero to infinity (for a single aminoglycoside dose); ICU, intensive care unit; t_1/2,eq, equilibration half-life between the ELF and the plasma compartment. This half-life characterizes the extent of system hysteresis and represents the slower rise of ELF concentrations compared to the rapid rise of the plasma or serum concentrations. If an aminoglycoside was dosed as a continuous infusion, the t_1/2,eq would be the half-life of approaching a constant steady-state concentration in ELF; VAP / VABP, ventilator-associated (bacterial) pneumonia.

^b BSV, between subject variability reported as coefficient of variation.

^c Relative standard errors.

^d The estimated absorption half-life (t_1/2,abs) after intramuscular dosing of tobramycin was 13.4 min.

^e Patients were intubated and ventilated for a variety of reasons and received antibiotics due to the development of pneumonia.

^f Not applicable.

Fig 1.

Fitted (lines) and observed (markers) plasma (or serum) and ELF concentrations based on published data of five aminoglycosides from eight human PK studies (23, 24, 26–31). The lines represent different subjects if individual subject data were reported (see Table S2 for details on data sets).

The estimated t_1/2,eq differed between aminoglycosides with a range of 0.857 to 5.80 h in patients and 0.613 h in healthy volunteers (Table 1). It is unknown, whether these differences arose from the aminoglycoside structures (Table S1) (13), clinical factors (e.g., type of infection and inflammation) (60, 61), or both. The longest half-life was observed in neonates receiving amikacin (Fig. 1; Fig. S1). Neonates might have had an altered expression of pulmonary transporters compared to adults (62–65). Future research on such transporters and their impact on lung penetration is warranted (66–71). Owing to the sparse nature of the data sets, the differences in t_1/2,eq between aminoglycosides should be interpreted cautiously.

The system hysteresis (Fig. S1) yielded blunted peak concentrations (C_max) in ELF, which were on average 2.3- to 4-fold lower than those in plasma. Despite this, the AUC in ELF was 50% to 100% of the plasma AUC. Plazomicin modeling results are shown in Fig. S4 (50). Thus, estimating the system hysteresis via time-course modeling is important when determining the ELF penetration of aminoglycosides. To support future studies, we provided Monte Carlo simulation code to simulate ELF and plasma concentrations, as well as D-optimal sampling designs [in the PopED Lite software (72)] for amikacin, gentamicin, and tobramycin in the supplementary materials.

In mice, PK analyses estimated average ELF-to-plasma AUC ratios of 0.60 to 0.88 for amikacin, tobramycin, plazomicin, and apramycin (4, 45–47, 73), consistent with our results (Table 1). However, the ELF-to-plasma equilibration half-lives were substantially faster in mice [3 to 5 min for tobramycin and plazomicin (45, 46), and 22 to 36 min for amikacin (73)] compared to those in patients (0.857 to 5.80 h; Table 1).

This study represents the first time-course modeling to characterize the rate and extent of ELF penetration for five aminoglycosides in humans based on eight published studies. Our model-based meta-analyses revealed the average AUC in ELF to be 50% to 100% compared to those in plasma for humans. The individual subject ELF penetration ratios displayed considerable variability. Due to extensive system hysteresis, the C_max in ELF were blunted and lower than those in plasma. With both C_max and AUC being correlated to bacterial killing and clinical efficacy of aminoglycosides (2–11), future studies are warranted to assess whether or not blunted C_max in ELF are clinically important. Moreover, future research is warranted to assess the impact of pH and different oxygen tensions on aminoglycoside efficacy for lung infections (74, 75). This study supports translational research to simulate the time course of ELF concentrations in in vitro infection models (76) and future clinical ELF penetration studies in animals and humans.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aac.01393-23.

Tables S1 to S4, Fig. S1 to S4, optimal design analyses, and simulation model code. aac.01393-23-s0001.docx.

Supplementary figures, tables. and simulation model code.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.