Development of Modernized Acinetobacter baumannii Susceptibility Test Interpretive Criteria for Recommended Antimicrobial Agents Using Pharmacometric Approaches

ABSTRACT

Acinetobacter baumannii-Acinetobacter calcoaceticus complex (referred to herein as A. baumannii) treatment guidelines contain numerous older antimicrobial agents with susceptibility test interpretive criteria (STIC, also known as susceptibility breakpoints) set using only epidemiological data. We utilized a combination of in vitro surveillance data, preclinical murine thigh and lung infection models, population pharmacokinetics, simulation, and pharmacokinetic/pharmacodynamic (PK/PD) target attainment analyses to evaluate A. baumannii STIC for four commonly recommended antimicrobials from different classes (amikacin, ceftazidime, ciprofloxacin, and minocycline). Antimicrobial in vitro surveillance data were based on 1,647 clinical A. baumannii isolates obtained from 109 centers in the United States and Europe. Among these isolates, 5 were selected for evaluation in murine infection models based on fitness and MIC variability. PK and dose-ranging studies were conducted using neutropenic murine thigh and lung infection models The MIC ranges for the 5 isolates evaluated were as follows: amikacin, 2 to 32 μg/mL; ceftazidime, 4 to 16 μg/mL; ciprofloxacin, 0.12 to 2 μg/mL; minocycline, 0.25 to 4 μg/mL. All organisms grew ≥1.5 log₁₀ CFU in both models in untreated controls. Plasma and epithelial lining fluid (ELF) pharmacokinetics for all drugs were determined in mice using liquid chromatography-tandem mass spectrometry (LC-MS/MS) methods. For each isolate, 5 dose levels of each drug were tested individually in the thigh and lung infection model. The inoculum ranged from 7.9 to 8.4 and 6.8 to 7.7 log₁₀ CFU/mL for the lung and thigh models, respectively. PK/PD targets associated with net bacterial stasis and 1- and 2-log₁₀ CFU reductions from baseline were identified for each organism/infection model using Hill-type models. Population pharmacokinetic models for each agent were identified from the literature. Using demographic variables for simulated patients with hospital-acquired or ventilator-associated bacterial pneumonia or urinary tract infections (including acute pyelonephritis) who were administered maximal dosing regimens of each agent, estimates of protein binding, and ELF penetration ratios based on data from the literature, free-drug plasma and total-drug concentration-time profiles were generated, and PK/PD indices by MIC were calculated. Percent probabilities of attaining median and randomly assigned PK/PD targets associated with the above-described endpoints were determined. Recommended susceptible breakpoints for each agent were those representing the highest MIC at which the percent probabilities of achieving PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline approached or were ≥90%. The following susceptible breakpoints for A. baumannii were identified: amikacin, ≤8 μg/mL for pneumonia; ceftazidime, ≤32 and ≤8 μg/mL for pneumonia; ciprofloxacin, ≤1 μg/mL; and minocycline, ≤0.5/≤1 μg/mL which correspond to the standard and high minocycline dosing regimens of 200 mg per day and 200 mg every 12 h, respectively. Implementation of appropriate STIC will help clinicians optimally use the above-described agents and improve the likelihood of successful patient outcomes.

INTRODUCTION

Acinetobacter baumannii-Acinetobacter calcoaceticus complex (referred to herein as A. baumannii) represents a common pathogen, and isolates manifesting resistance to carbapenem (i.e., carbapenem-resistant A. baumannii [CRAB]) were added to the CDC urgent threat list in 2019 (1). Recently, there has been an alarming rate of increase in CRAB. In all patient settings and among hospital-acquired infections, the percentage of CRAB increased by 35 and 78%, respectively, from 2019 to 2020 (2). At present, there are limited clinical trial data to guide therapy, and current treatment decisions are based upon in vitro susceptibility test interpretive criteria (STIC), colloquially known as susceptibility breakpoints.

While four older antimicrobial agents approved by the United States Food and Drug Administration (FDA), amikacin (1976), ceftazidime (1984), ciprofloxacin (1987), and minocycline (1971) (3–7), representing four drug classes, aminoglycoside, β-lactam, quinolone, and tetracycline, have been used to treat A. baumannii, these agents have not undergone rigorous preclinical and clinical testing to determine their appropriateness for treatment of A. baumannii infections. Additionally, the in vitro antimicrobial STIC for each of these “older” antimicrobial agents were supported primarily by epidemiologic data alone rather than in combination with clinical and pharmacokinetic/pharmacodynamic (PK/PD) data, as is the norm for contemporary data packages. Unfortunately, much of the STIC established prior to 2001 are not working properly (8). Malfunctioning STIC have significant negative consequences for patients and society. Specifically, inappropriate STIC guidance promotes the misuse of old and new antimicrobial agents, encourages the use of antimicrobial agents with questionable efficacy and increased toxicity (e.g., colistin), and fosters resistance selection (9).

The goal of the present set of studies and analyses was to utilize pharmacometric approaches to evaluate four antimicrobials, amikacin, ceftazidime, ciprofloxacin, and minocycline, against A. baumannii to discern if current STIC need to be revisited. Specifically, we utilized contemporary in vitro surveillance data to characterize MIC distributions for these four agents against A. baumannii. We then carried out PK/PD studies with each drug and multiple A. baumannii isolates with widely varying MICs in in vivo infection models that are used for translation for the most common infection scenarios, the neutropenic murine thigh and murine lung infection models. We utilized this MIC distribution and nonclinical PK/PD targets for efficacy derived from these studies, together with population pharmacokinetic (PK) models with each antibiotic obtained from the literature and simulation, to estimate the likelihood that dosing regimens for these four agents would attain the nonclinical PK/PD targets by MIC. We defined the pharmacometrics-based susceptible breakpoint as the highest MIC for each at which the probability of PK/PD target attainment approached or was ≥90%.

RESULTS

In vitro surveillance data and selection of A. baumannii isolates for in vivo study.

The MIC distributions for the four antimicrobial agents tested against the collection of 755 A. baumannii isolates and the subset of 239 carbapenem-resistant A. baumannii (CRAB) isolates from 71 medical centers in the United States and 892 A. baumannii isolates and the subset of 617 CRAB isolates from 38 European medical centers are shown in Table S1 in the supplemental material. A. baumannii isolates were collected from medical centers in both regions from 2018 to 2020. CRAB was defined as A. baumannii resistant to imipenem and meropenem per CLSI criteria (MIC ≥ 8 μg/mL) (10). The MIC distribution was bimodal for each antibiotic. Notably, MICs for isolates collected in Europe compared to those from the United States were higher for all four antimicrobial agents (Table S1). The tentative epidemiological cutoff (ECOFF) values for isolates collected in the United States and Europe and ECOFF values for each agent obtained from the EUCAST website (https://mic.eucast.org/search/) are shown in Table S2 and Fig. S1.

We selected isolates from this collection with various MICs for each of the four test antibiotics, amikacin, ceftazidime, ciprofloxacin, and minocycline, for use in the PK/PD studies. The 15 isolates selected for study were collected in 2020 (4 isolates; 26.7%), 2019 (5 isolates; 33.3%), and 2018 (6 isolates; 40.0%) from patients with pneumonia (7 isolates; 46.7%), skin and skin structure infections (4 isolates; 26.7%), bloodstream infections (2 isolates; 13.3%), urinary tract infections (1 isolate; 6.7%), and intra-abdominal infections (1 isolate; 6.7%). The MIC ranges were as follows: amikacin, 1 to >128 μg/mL; ceftazidime, 2 to >128 μg/mL; ciprofloxacin, 0.12 to >32 μg/mL; minocycline, 0.5 to 16 μg/mL.

Each of the 15 isolates was assessed for in vivo fitness in both the neutropenic thigh and lung models by measuring control growth between 2 and 24 h after infection. Five isolates were chosen based upon control growth and relative MIC variation among the isolates and four antibiotics (Table 1). Control growth in pilot studies ranged from 1.50 to 3.10 log₁₀/organ for these strains. The isolates included those in the current CLSI/FDA susceptible and resistant categories. MICs varied 16-fold for amikacin, ciprofloxacin, and minocycline, but for ceftazidime, isolate MICs varied 4-fold.

TABLE 1.

A. baumannii isolates utilized for in vivo studies, MICs, and relative fitness in the murine infection models

Isolate	MIC (μg/mL)				Control growth (log10 CFU/organ)
	Amikacin	Ceftazidime	Ciprofloxacin	Minocycline	Thigh	Lung
1064791	4	8	1	1	2.79 ± 0.29	1.94 ± 0.77
1170594	8	16	2	4	1.50 ± 0.36	2.89 ± 0.30
1082759	2	4	0.12	0.25	1.83 ± 0.05	2.05 ± 0.65
1109046	2	8	0.25	0.5	2.16 ± 0.41	2.57 ± 0.13
1108210	32	16	1	0.5	1.74 ± 0.28	3.10 ± 0.18

PK/PD characterization of four guideline-recommended A. baumannii antibiotics using murine thigh and lung infection models.

The neutropenic mouse thigh and lung infection models have been used extensively for PK/PD target determination and, thus, to support the determination of STIC for multiple bacterial species and nearly every antimicrobial drug class that has been approved for systemic use by the FDA (11). We utilized this established animal infection model platform to define the PK/PD targets of antimicrobial agents from four different classes that are recommended for treatment of A. baumannii.

Murine pharmacokinetics.

Plasma and epithelial lining fluid (ELF) pharmacokinetics following subcutaneous single doses of amikacin, ceftazidime, ciprofloxacin, and minocycline were calculated using a noncompartmental model. The plasma and ELF pharmacokinetics with each antibiotic over the 64-fold dose range are shown graphically in Fig. 1 and Fig. 2, respectively. The plasma and ELF pharmacokinetics were relatively linear over the dose range for amikacin, ceftazidime, and minocycline (R², 0.95 to 0.99). The ciprofloxacin kinetics over the range were more dose dependent for both plasma and ELF (R², 0.70 to 0.92). The difference in PK characteristics of ciprofloxacin were the result of using a solubilized powder formulation instead of the intravenous (i.v.) formulation, which was necessary to be able to provide the concentration range in the studies, and the former has slower systemic absorption. The degree of penetration from plasma into the ELF based on the area under the concentration-time curve (AUC) from 0 to 24 h values ranged from 0.5 to 1.4 (with mean ± standard deviation [SD] values as follow: amikacin, 0.8 ± 0.14; ceftazidime, 0.5 ± 0; ciprofloxacin, 1.4 ± 0.34; and minocycline, 0.9 ± 0.1) and was similar over the dose range studied. Protein binding values in mice were similar over the concentration range for each antibiotic (amikacin, 24%; ceftazidime, 10%; minocycline, 85%; and ciprofloxacin, 24%). In general, the results obtained were similar to those previously reported based on murine PK studies.

FIG 1.

Single-dose, murine plasma pharmacokinetics following subcutaneous administration of amikacin (A), ceftazidime (B), ciprofloxacin (C), and minocycline (D). Samples were obtained at six or seven time points over 12 h. Each symbol and error bar represents the mean and standard deviation from three mice following four escalating doses. C_max, AUC, and t_½ are reported in the legend for each dose.

FIG 2.

Single-dose murine epithelial lining fluid pharmacokinetics following subcutaneous administration of amikacin (A), ceftazidime (B), ciprofloxacin (C), and minocycline (D). Samples were obtained at six or seven time points over 12 h. Each symbol and error bar represents the mean and standard deviation from three mice following four escalating doses. C_max, AUC, and t_½ are reported in the legend for each dose.

In vivo efficacy assessment and determination of PK/PD targets.

At the start of therapy, the burdens of organisms in mouse thighs and lungs were 7.37 ± 0.34 and 7.25 ± 0.46 log₁₀ CFU/organ, respectively. Control growth in untreated mice was relatively similar in the two models, 2.04 ± 0.54 and 2.46 ± 0.74 log₁₀ CFU/organ in the thigh and lung infection models, respectively. We assessed the bacterial burden after 24 h of therapy using 5 dosing regimens for each of the four antibiotics over a 256-fold dose range (Fig. 3 and 4). The highest-dose regimen resulted in more than a 2 log₁₀ CFU/organ reduction compared to the burden at the start of therapy for most antibiotic-organism combinations. The average decreases in burden were −2.39 ± 1.40 and −2.81 ± 1.59 log₁₀ CFU/organ in the thigh and lung infection models, respectively. The dose-response curves were shifted to the right in the treatment study against isolates with higher MICs for each antibiotic in both infection models, indicating the impact of MIC on outcome.

FIG 3.

In vivo dose effect of amikacin (A), ceftazidime (B), ciprofloxacin (C), and minocycline (D) against 5 selected A. baumannii isolates using a neutropenic murine thigh infection model, expressed as total dose. Each symbol represents the mean and standard deviation for four thighs from two mice. Five dose levels were administered every 6 h. The burden of organisms was measured at the start and end of therapy. The study period was 24 h. The horizontal dashed line at 0 represents the burden of organisms at the start of therapy. Data points above the line represent net growth, and those below the line represent a reduction in burden.

FIG 4.

In vivo dose effect of amikacin (A), ceftazidime (B), ciprofloxacin (C), and minocycline (D) against 5 selected A. baumannii isolates using a neutropenic murine lung infection model, expressed as total dose. Each symbol represents the mean and standard deviation for lungs from three mice. Five dose levels were administered every 6 h. The burden of organisms was measured at the start and end of therapy. The study period was 24 h. The horizontal dashed line at 0 represents the burden of organisms at the start of therapy. Data points above the line represent net growth, and those below the line represent a reduction in burden.

PK/PD relationships describing the relationships between the change in log₁₀ CFU from baseline and PK/PD indices for each agent based on the Hill-type models constructed using data from the thigh and lung infection models are shown graphically in Fig. S2 and S3, respectively.

For amikacin, ciprofloxacin, and minocycline, the ratios of the area under the total- and free-drug plasma and total-drug ELF concentration-time curves from to 0 to 24 h to the MIC (AUC/MIC ratios) were assessed. For ceftazidime, the percentage of time from 0 to 24 h that the total- and free-drug plasma and total-drug ELF concentrations were above the MIC (%T>MIC) were assessed. In general, the data were well fitted by the Hill-type models based on visual inspection and the moderately high R² values for amikacin, ciprofloxacin, and minocycline (R² range, 0.60 to 0.75). For ceftazidime, there was observed activity at very low %T>MIC, including several values at or near zero, which led to the relatively lower R² and the uncharacteristic appearance of the Hill-type models. Summary statistics for the magnitude of the PK/PD indices associated with net bacterial stasis and 1- and 2-log₁₀ CFU reductions from baseline are presented in Tables 2 and 3 for the neutropenic murine thigh and lung infection models, respectively (isolate level calculations are shown in Tables S3 and S4, respectively). The magnitudes of the PK/PD indices associated with a given endpoint for amikacin, ceftazidime, and ciprofloxacin against the A. baumannii isolates were considerably lower than we and others have observed in treatment of other Gram-negative bacilli (12–14). For example, the free-drug plasma AUC/MIC ratio target associated with a 1-log₁₀ CFU reduction from baseline was 67.4 for quinolones against Enterobacterales based on data from a neutropenic murine thigh infection model (13) but was 17.0 for ciprofloxacin against A. baumannii for the same endpoint and infection model, as described in Table 2. The magnitude of the median free-drug plasma PK/PD targets based on the data from the two infection models were reasonably similar for each agent. We also noted that the magnitudes of the free-drug plasma and ELF PK/PD targets based on data from the neutropenic murine lung infection model for each agent were congruent.

TABLE 2.

Summary free-drug plasma PK/PD targets for efficacy for amikacin, ceftazidime, minocycline, and ciprofloxacin against 5 A. baumannii in the neutropenic murine thigh infection model^a

Drug	Measure	Bacterial reduction endpoint
		Net bacterial stasis			1 log10 reduction from baseline			2-log10 CFU reduction from baseline
		Total dose (mg/kg/24 h)	PK/PD index		Total dose (mg/kg/24 h)	PK/PD index		Total dose (mg/kg/24 h)	PK/PD index
			Total-drug plasma	Free-drug plasma		Total-drug plasma	Free-drug plasma		Total-drug plasma	Free-drug plasma
Amikacin	Mean	88.8	10.4	7.9	179.6	24.3	18.5	431.8	75.8	57.6
	Median	75.6	8.4	6.4	220.8	16.1	12.2	397.6	25.3	19.2
	SD	38.5	8.2	6.2	118.8	28.9	22.0	371.9	110.7	84.1
Ceftazidime	Mean	131.6	8.8	8.0	466.8	26.3	24.2	2284.9	53.3	51.3
	Median	60.2	0.0	0.0	235.4	21.9	19.7	1,754.7	47.5	45.5
	SD	162.1	12.2	11.1	421.9	19.4	19.4	1,831.6	21.9	22.0
Ciprofloxacin	Mean	159.4	9.9	7.6	213.2	18.0	13.7	NA
	Median	75.1	7.7	5.9	55.6	22.4	17.0	NA
	SD	223.6	7.6	5.8	314.3	7.7	5.9	NA
Minocycline	Mean	18.3	24.9	3.7	44.7	60.6	9.1	114.5	149.4	22.4
	Median	6.9	7.8	1.2	11.6	13.1	2.0	23.3	26.0	3.9
	SD	26.6	36.9	5.5	71.8	96.9	14.5	193.2	249.9	37.5

^aThe PK/PD indices were AUC/MIC ratio for amikacin, ciprofloxacin, and minocycline and %T>MIC for ceftazidime. NA, not available.

TABLE 3.

Summary plasma and ELF PK/PD targets for amikacin, ceftazidime, minocycline, and ciprofloxacin against 5 A. baumannii isolates the neutropenic murine lung infection model^a

Drug	Measure	Bacterial reduction endpoints
		Net bacterial stasis			1 log10 reduction from baseline			2-log10 CFU reduction from baseline
		Total dose (mg/kg/24 h)	PK/PD index		Total dose (mg/kg/24 h)	PK/PD index		Total dose (mg/kg/24 h)	PK/PD index
			Total-drug plasma	Free-drug plasma		Total-drug plasma	Free-drug plasma		Total-drug plasma	Free-drug plasma
Amikacin	Mean	23.8	2.9	4.7	31.0	3.9	6.1	54.2	6.1	8.7
	Median	24.0	3.3	5.2	25.9	4.0	6.6	30.9	5.5	8.9
	SD	13.6	1.6	2.6	28.1	2.1	3.0	66.5	4.5	5.3
Ceftazidime	Mean	141.6	2.3	8.1	345.4	11.0	17.7	1,068.1	31.4	34.2
	Median	132.4	0.0	0.0	330.7	0.0	13.0	957.6	23.3	29.1
	SD	106.7	5.2	12.4	292.1	17.7	19.4	1,112.1	38.7	30.5
Ciprofloxacin	Mean	129.5	11.2	8.8	366.8	17.6	13.8	NA
	Median	26.9	11.8	9.3	52.1	21.1	16.4	NA
	SD	196.9	6.1	4.7	585.2	10.3	7.9	NA
Minocycline	Mean	25.5	4.2	4.0	33.8	5.4	5.1	44.1	7.0	6.6
	Median	19.4	4.7	4.5	25.6	6.0	5.7	35.3	8.1	7.6
	SD	19.1	1.4	1.2	25.8	1.9	1.7	32.0	2.4	2.2

^aThe PK/PD indices were AUC/MIC ratio for amikacin, ciprofloxacin, and minocycline and %T>MIC for ceftazidime. NA, not available.

Human population pharmacokinetic models, simulated patients, and drug exposures.

The population PK models describing the disposition of amikacin, ceftazidime, ciprofloxacin, and minocycline in hospitalized or critically ill patients that were selected from the literature (15–18) are summarized in Table S5 in the supplemental material. The ELF penetration ratios for each of these agents based on data selected from the literature (19–23; data on file, Institute of Clinical Pharmacodynamics, Inc.) are summarized in Table S6.

Summary statistics for continuous and categorial demographic variables for patient populations enrolled in clinical trials that consisted of 1,748 patients with complicated urinary tract infections (cUTI), including acute pyelonephritis (AP) (cUTI/AP) and 592 patients with hospital-acquired or ventilator-associated bacterial pneumonia (HABP/VABP) with creatinine clearance (CL_CR) ≥30 mL/min are presented in Table S7. These data are also shown graphically in Fig. S4 and S5, respectively. Given that distributions of most of the demographic variables evaluated, including CL_CR and weight, were similar between patients with cUTI/AP and HABP/VABP, data for the 1,748 patients with cUTI/AP and 592 patients with HABP/VABP were pooled (total population n = 2,340) and replicated twice to generate one patient population containing 4,680 simulated patients.

PK/PD target attainment results.

Using free-drug plasma and total-drug ELF PK/PD indices by MIC generated for simulated patients after administration of the amikacin, ceftazidime, ciprofloxacin, and minocycline dosing regimens described in Table 4 (4–6, 14, 24), the percent probability of PK/PD target attainment by MIC was assessed. Percent probabilities of PK/PD target attainment by MIC based on median and randomly assigned free-drug plasma PK/PD targets for A. baumannii derived from the neutropenic murine thigh infection model are shown in Tables S8 and S9, respectively. Such tabular data based on median and randomly assigned free-drug plasma and total-drug ELF PK/PD targets from the neutropenic murine lung infection model are shown in Tables S10 to S13. Percent probabilities of PK/PD target attainment by MIC based on the assessment of median and randomly assigned amikacin, ceftazidime, ciprofloxacin, and minocycline free-drug plasma and total-drug ELF AUC/MIC ratio targets associated with a 1-log₁₀ CFU reduction from baseline for A. baumannii isolates evaluated in neutropenic murine thigh or lung infection models are shown in Table 5. Percent probabilities of PK/PD target attainment by MIC based on PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline across all assessments are shown graphically, overlaid on the MIC distributions for the collection of 755 A. baumannii isolates and the subset of 239 CRAB isolates collected in the United States, in Fig. 5. Figure S6 shows the percent probabilities of PK/PD target attainment by MIC based on PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline across all assessments graphically, overlaid on the MIC distributions for the collection of 892 A. baumannii isolates and 617 CRAB isolates collected in Europe.

TABLE 4.

Summary of amikacin, ceftazidime, ciprofloxacin, and minocycline dosing regimens administered to simulated patients

Antimicrobial agent	Dosing regimen for simulated patients
	Normal renal function	Renal impairmenta
Amikacin	20 mg/kg i.v. q24hb	CLCR >60 to ≤90 mL/min: 20 mg/kg q36hb CLCR > 45 to ≤60 mL/min: 20 mg/kg q48hb CLCR > 30 to ≤45 mL/min: 20 mg/kg q48hb
Ceftazidime	2 g i.v. q8hc	CLCR, 31–50 mL/min: 2 g q12hc
Ciprofloxacin	400 mg i.v. q8hd	No dose adjustmentd
Minocycline	200 mg i.v. q24h,e 200 mg i.v. q12hoursf	No dose adjustmente

^aSimulated patients had a CL_CR of ≥30 mL/min, and dosing regimens were based on adjustment for this minimal CL_CR or higher.

^bBased on dosing regimens used in clinical practice as summarized in the USCAST report describing the aminoglycoside in vitro susceptibility test interpretive criterion (STIC) evaluations (14).

^cBased on the ceftazidime package insert (4).

^dBased on ciprofloxacin package insert (5).

^eBased on minocycline package insert (6).

^fBased on the Infectious Diseases Society of America Guidance on the treatment of carbapenem-resistant Acinetobacter baumannii infections (24).

TABLE 5.

Percent probabilities of PK/PD target attainment by MIC^a

Drug	PK/PD index	MIC (μg/mL)	Percent probability of PK/PD target attainment in modelb
			Thigh		Lung
			Free-drug plasma targets		Free-drug plasma targets		Total-drug ELF targets
			Medianc	Randomly assignedd	Medianc	Randomly assignedd	Medianc	Randomly assignedd
Amikacin	AUC/MIC ratio	0.25	100	100	100	100	99.1	99.1
		0.5	100	100	100	100	99.0	99.0
		1	100	100	100	100	99.0	98.9
		2	100	99.6	100	100	98.7	98.7
		4	100	96.2	100	100	98.3	98.1
		8	98.1	84.5	100	99.0	96.3	95.0
		16	75.4	63.4	97.2	91.8	88.1	84.1
		32	23.2	38.0	70.2	71.9	59.6	63.6
		64	0.7	16.3	18.2	42.5	20.1	36.4
Ceftazidime	%T>MIC	0.25	100	100	100		98.5
		0.5	100	100	100		98.4
		1	100	100	100		98.4
		2	100	100	100		98.1
		4	100	100	100		97.0
		8	100	100	100		93.3
		16	100	99.0	100		75.6
		32	99.1	92.1	99.9		27.1
		64	45.3	49.7	66.4		0.8
Ciprofloxacin	AUC/MIC ratio	0.25	100	100	100	100	99.7	99.3
		0.5	99.8	99.6	99.9	99.2	98.0	96.9
		1	94.9	95.4	95.8	92.1	88.2	87.1
		2	58.7	71.2	62.1	70.0	51.0	63.5
		4	7.5	32.0	9.4	39.2	8.2	34.9
		8	0	5.5	0.1	12.5	0.2	11.4
		16	0	0.4	0	2.0	0	2.1
		32	0	0	0	0.1	0	0.1
		64	0	0	0	0	0	0
Minocyclinee	AUC/MIC ratio	0.25	100	96.5	100	100	99.7	99.7
		0.5	100	88.6	99.9	99.9	98.3	98.2
		1	100	74.9	95.1	93.4	83.5	84.7
		2	99.5	57.0	30.1	47.8	27.5	42.8
		4	75.2	37.4	0.1	5.4	0.6	6.5
		8	4.3	21.1	0	0	0	0.1
		16	0	8.6	0	0	0	0
		32	0	2.2	0	0	0	0
		64	0	0.1	0	0	0	0

^aBased on the assessment of median and randomly assigned amikacin, ceftazidime, ciprofloxacin, and minocycline free-drug plasma and total-drug ELF PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline for A. baumannii isolates evaluated in neutropenic murine thigh and murine lung infection models.

^bPercent probability of PK/PD target attainment by MIC in murine infection models with the given approach for selecting PK/PD targets. Shading indicates probabilities of ≥90%.

^cThe median free-drug plasma AUC/MIC ratio targets associated with a 1-log₁₀ CFU reduction from baseline for A. baumannii against amikacin, ciprofloxacin, and minocycline based on the neutropenic murine thigh infection model were 12.2, 17.0, and 1.96, respectively. The median free-drug plasma %T>MIC target associated with a 1-log₁₀ CFU reduction from baseline for A. baumannii against ceftazidime was 19.7. The median free-drug plasma and total-drug ELF AUC/MIC ratio targets associated with a 1-log₁₀ CFU reduction from baseline for A. baumannii against amikacin, ciprofloxacin and minocycline based on the neutropenic murine lung infection model were 6.6 and 4.0 for amikacin, 16.4 and 21.1 for ciprofloxacin, and 5.7 and 6.0 for minocycline, respectively. The median free-drug plasma and total-drug ELF %T>MIC targets associated with a 1-log₁₀ CFU reduction from baseline for A. baumannii against ceftazidime were 13.0 and 0, respectively. The percent probability of PK/PD target attainment for a simulated patient based on the median value of 0 was defined by a free-drug plasma %T>MIC of >0.

^dFree-drug plasma or total-drug ELF AUC/MIC ratio targets were randomly assigned based on an estimated log normal distribution of the AUC/MIC ratio targets associated with a 1-log₁₀ CFU reduction from baseline. The distribution of targets for each agent and infection model was truncated on a log scale such that randomly selected targets were all within 2 standard deviations of the mean value. Since at least one A. baumannii isolate evaluated in the neutropenic murine lung infection model had a ceftazidime free-drug plasma and total-drug ELF %T>MIC target associated with a 1-log₁₀ CFU reduction of 0, random assignment for these two sets of %T>MIC targets was not performed due to the inability to estimate a log-normal distribution.

^eBased on the assessment of minocycline 200 mg i.v. q12h.

FIG 5.

Probability of percent PK/PD target attainment by MIC simulation for amikacin (A), ceftazidime (B), ciprofloxacin (C), and minocycline (D) based upon a 1-log₁₀ CFU reduction from baseline from murine thigh and lung models against A. baumannii overlaid on the MIC distribution for A. baumannii isolates and carbapenem-resistant A. baumannii isolates from the United States. Note that assessments for minocycline were based on a dosing regimen of 200 mg intravenously q12h.

Recommended pharmacometric-based STIC.

Recommended susceptible breakpoints for each agent against A. baumannii were based upon the highest MIC at which the probabilities of target attainment approached or were ≥90% based upon both median and randomly assigned PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline and consideration of ECOFF values. Recommended STIC for each agent against A. baumannii are shown in comparison to current FDA, CLSI, U.S. Committee on Antimicrobial Susceptibility Testing (USCAST), and EUCAST STIC (10, 25–27) in Table 6.

TABLE 6.

USCAST-recommended STIC for each agent and comparison by organization^a

Drug	USCAST-recommended STIC (μg/mL)b			Current STIC (μg/mL)
				FDAc			CLSId			EUCASTe
	S	I	R	S	I	R	S	I	R	S	I	R
Amikacin	≤8		≥16	CLSI M100	CLSI M100	CLSI M100	≤16	32	≥64	≤8		≥16
Ceftazidime	≤32		≥64	CLSI M100	CLSI M100	CLSI M100	≤8	16	≥32
Ceftazidime-pneumonia	≤8		≥16
Ciprofloxacin	≤1		≥2				≤1	2	≥4	≤0.001		≥2
Minocycline	≤0.5/≤1f		≥1/≥2	CLSI M100	CLSI M100	CLSI M100	≤4	8	≥16

^aS, susceptible; I, intermediate; R, resistant.

^bBased on analysis results. USCAST version 7 (2021) breakpoint tables (26).

^cFDA 2023 interpretive criteria (24). A notation of “CLSI M100” indicates that the FDA recognizes the STIC described in the CLSI M100 document.

^dCLSI M100-33rd edition (2023) interpretive criteria (10).

^eEUCAST 2023 clinical breakpoint tables (27).

^fSusceptible breakpoints of ≤0.5 and ≤1 μg/mL correspond to the standard and high minocycline dosing regimens of 200 mg/day and 200 mg q12h, respectively. While minocycline i.v. dosing regimens were evaluated among simulated patients, STIC apply to i.v. or oral dosing regimens, since published data suggest that oral doses are 95% to 100% bioavailable (44).

DISCUSSION

As part of the objective to develop recommendations for amikacin, ceftazidime, ciprofloxacin, and minocycline STIC against A. baumannii, a comprehensive set of in vitro and in vivo studies and pharmacometric analyses were completed. One goal of the in vitro studies conducted was to characterize the activity of these agents against a robust collection of A. baumannii isolates using contemporary in vitro surveillance data collected from the United States and Europe as part of the SENTRY Antimicrobial Surveillance Program. The in vivo data generated involved the use of neutropenic murine thigh and lung infection models to conduct dose-ranging studies for each drug using multiple A. baumannii isolates with various MICs. Using MIC distributions from the previous studies and nonclinical PK/PD targets for efficacy based on pharmacometric analyses of the latter study data, together with population PK models from the literature, simulations were conducted. Simulated patients resembling the target populations of interest received dosing regimens with maximal dosing for each agent based on current clinical practice. Percent probabilities of PK/PD target attainment were assessed by MIC. As described below, results of these analyses allowed potential STIC to be identified.

Recommended susceptible breakpoints for each agent against A. baumannii were those representing the highest MIC at which the percent probabilities of achieving median or randomly assigned PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline approached or were ≥90%. This endpoint was chosen given the expectation that patients with infections arising from A. baumannii would frequently have pneumonia or other systemic infections with potentially poor source control and/or comorbidities (28). Additionally, attaining PK/PD indices associated with this magnitude of bacterial reduction has been shown to be consistent with meeting the primary efficacy endpoint in studies of antimicrobial agents for pneumonia (29–32). When weighing the different sets of results across assessments, priority was given to the PK/PD target attainment results based on randomly assigned PK/PD targets. For STIC for pneumonia, priority was also given to assessments based on total-drug ELF exposures. For amikacin, susceptible and resistant breakpoints of ≤8 and ≥16 μg/mL, respectively, were identified. Using amikacin MICs of ≤8 for the susceptible breakpoint, 80.9 and 47.5% of A. baumannii isolates and the CRAB subset in the United States, respectively, would be considered susceptible. This susceptible breakpoint recommendation is consistent with an ECOFF determination of 8 μg/mL. For ceftazidime, susceptible and resistant breakpoints of ≤32 and ≥64 μg/mL, respectively, were identified. Ceftazidime breakpoints for pneumonia shifted downward by 2 doubling MIC dilutions to ≤8 and ≥16 μg/mL, respectively. Using ceftazidime MICs of ≤32 and ≤8 μg/mL for susceptible breakpoints, 77.1 and 66.0% of A. baumannii isolates in the United States, respectively, would be considered susceptible. Of the CRAB subset, 38.1 and 20.9%, respectively, would be considered susceptible. The latter susceptible breakpoint recommendation for pneumonia is below the ECOFF determination of 16 μg/mL. The recommended ciprofloxacin susceptible and resistant breakpoints were ≤1 and ≥2 μg/mL, respectively. Using this definition, 61.1 and 2.1% of A. baumannii isolates and the CRAB subset would be considered susceptible. As described for amikacin, this susceptible breakpoint recommendation for ciprofloxacin is consistent with ECOFF determination of 1 μg/mL. Lastly, minocycline susceptible and resistant breakpoints of ≤0.5 and ≥1 μg/mL, respectively, for the 200-mg/day dosing regimen and ≤1 and ≥2 μg/mL, respectively, for the regimen of 200 mg every 12 h (q12h), were identified. Using these minocycline MICs for the susceptible breakpoints, 66.5 and 73.9%, respectively, of A. baumannii isolates and 16.3 and 31.4%, respectively, of the CRAB subset would be considered susceptible. Also, while a EUCAST-published ECOFF value was not available for minocycline, the minocycline susceptible breakpoints of ≤0.5 and ≤1 μg/mL were consistent with tentative ECOFF values for the assessment of 95.0 to 99.5% endpoints based on the SENTRY Program. Using the above-described susceptible breakpoints, a lower percentage of A. baumannii and CRAB isolates from Europe would be considered susceptible for all four agents.

Based on the above-described MIC distributions and ECOFF values, bisection of wild-type MIC distribution by the susceptible breakpoints identified in the present investigation was evident only for ceftazidime using a susceptible breakpoint of 8 μg/mL. Thus, implementation of these recommended STIC may be easier with the minimization of such concerns. However, one challenge is that susceptibility testing will be difficult because the above-described susceptible breakpoints for each agent are largely in the middle of the MIC distributions for carbapenem-resistant isolates and other resistant subsets. Thus, there will be greater uncertainty of the test results, as error rates will be higher (33). These errors notwithstanding, the low percentage of CRAB isolates that would be considered susceptible by the recommended STIC suggests a need for new agents for the treatment of patients with infections caused by this group of pathogens. Appropriate STIC for the agents studied will enable optimal use of these agents and, thus, a higher likelihood of improved patient outcome. While clinical outcomes by MIC data are ideally needed to provide further support for these STIC recommendations, such data are often difficult to find in the literature, as is the case for infections arising from A. baumannii. Thus, greater reliance in such circumstances is placed on the results of robust in vitro and in vivo studies and pharmacometric analyses, as described here. Given the importance of these data for guiding STIC recommendations, it is important to consider the data inputs further and compare these to the existing literature.

Compared to PK/PD targets for efficacy for other Gram-negative pathogens from neutropenic murine infection models, the magnitude of the PK/PD targets associated with a given bacterial reduction endpoint was lower for A. baumannii (12–14). While pathogenicity and growth control in the animal model do not explain these differences, changes in the magnitude of PK/PD targets associated with a given bacterial reduction endpoint among pathogen groups have been previously reported, including for A. baumannii (34). These sets of in vivo PK/PD targets for amikacin, ceftazidime, ciprofloxacin, and minocycline efficacy against A. baumannii efficacy were also compared to those reported in the literature using neutropenic murine infection models. Despite the common use of these four agents for the treatment of infections arising from A. baumannii, in vivo efficacy data are limited and were only found for minocycline. Zhou and colleagues evaluated the PK/PD of minocycline against 5 A. baumannii isolates using a neutropenic murine lung infection model (35). They reported total-drug ELF AUC/MIC ratio targets associated with net bacterial stasis and a 1-log₁₀ CFU reduction from baseline of 140 and 410, respectively. In contrast, Tarazi and colleagues used a neutropenic rat lung infection model to evaluate PK/PD targets of minocycline against 6 A. baumannii isolates and reported free-drug plasma AUC/MIC ratio targets associated with net bacterial stasis and a 1-log₁₀ CFU reduction from baseline that ranged from 10.6 to 20.1 and 13.1 to 33.9, respectively (36). Herein, we reported median (minimum, maximum) free-drug plasma AUC/MIC ratio targets associated with these endpoints based on data from a neutropenic murine lung infection model of 4.5 (2.6, 4.9) and 5.7 (3.2, 6.5), respectively. Median (minimum, maximum) total-drug ELF AUC/MIC ratio targets were 4.7 (2.6, 5.3) and 6.0 (3.3, 6.9), respectively. The reason for the differences among these three evaluations is unknown but may be a function of differences in host species, the isolates evaluated, drug preparation, administration route, dose selected for PK studies, and methods undertaken to mount the infection model. The implication of the lower minocycline AUC/MIC ratio targets is that a wider range of MICs among those in the MIC distribution would be categorized as susceptible. Additional in vivo data using the study design described here and more A. baumannii isolates will provide more support for the PK/PD targets for efficacy assessed, including the minocycline AUC/MIC ratio targets.

In conclusion, previous STIC guidance for A. baumannii was developed decades ago and was largely based on epidemiological MIC distributions and limited clinical data, the latter of which remains limited. We used robust in vitro and in vivo data together with a pharmacometrics-based approach to determine potential STIC for amikacin, ceftazidime, ciprofloxacin, and minocycline against A. baumannii. We propose the use of these STIC in clinical practice to help better differentiate candidate treatment options and ultimately improve patient outcome. The evaluation of in vitro surveillance data based on revised STIC for these agents will also serve to better inform the landscape of empirical treatment options so that the need for new agents for the treatment of patients with infections arising from A. baumannii can be better assessed.

MATERIALS AND METHODS

In vitro surveillance and antimicrobial susceptibility testing.

A total of 1,647 A. baumannii isolates, 755 from 71 medical centers in the United States and 892 from 38 European medical centers participating in the SENTRY Program from 2018 to 2020, were evaluated. Susceptibility testing was performed by the reference broth microdilution method according to CLSI standards (37). Frozen-form 96-well MIC panels were prepared with cation-adjusted Mueller-Hinton broth at JMI Laboratories. Although a high proportion of isolates had off-scale high MICs, tentative ECOFF values were calculated for isolates collected from the United States and Europe as part of the SENTRY Program (38, 39), ECOFF values for each agent were also obtained from the EUCAST website (https://mic.eucast.org/search/).

The isolates selected for the in vivo study from among the above-described collection were tested in triplicate, and the most common MIC was selected as the final MIC. If three different MICs were obtained, the median value was selected as the final MIC. MICs were validated by concurrently testing the following CLSI-recommended (10) ATCC strains and/or EUCAST (27)-recommended quality control (QC) reference strains: Escherichia coli ATCC 25922, Pseudomonas aeruginosa ATCC 27853, A. baumannii NCTC 13304.

Murine thigh and lung models.

Animals were maintained in accordance with the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care International. All animal studies were approved by the Animal Research Committees of the William S. Middleton Memorial VA Hospital and the University of Wisconsin. Six-week-old, specific-pathogen-free, female ICR/Swiss mice weighing 24 to 27 g were used for all studies (Envigo, Indianapolis, IN). Mice were rendered neutropenic (<100 neutrophils/mm³) by cyclophosphamide (Mead Johnson Pharmaceuticals, Evansville, IN) injection subcutaneously 4 days (150 mg/kg) and 1 day (100 mg/kg) before thigh infection. Broth cultures of freshly plated bacteria were grown to logarithmic phase overnight to an absorbance at 580 nm of 0.3 (Spectronic 88; Bausch and Lomb, Rochester, NY). After a 1:10 dilution into fresh Mueller-Hinton broth, bacterial counts of the inoculum ranged from 10^6.8 to 10^7.7 CFU/mL for the thigh model. Thigh infections with each of the isolates were produced by the injection of 0.1 mL of the inoculum into the thighs of isoflurane-anesthetized mice. Therapy with either amikacin, ceftazidime, ciprofloxacin, or minocycline was initiated 2 h after the induction of infection, and therapy continued for 24 h, at which point the treatment groups and controls were sacrificed for CFU enumeration. The organism burden was quantified by CFU counts from serial dilutions of thigh homogenates. The lung model of infection was similar in that 6-week-old, specific-pathogen-free, female ICR/Swiss mice weighing 24 to 27 g were used for all studies. Mice were rendered neutropenic by cyclophosphamide injection, and logarithmic-phase cultures were used to infect the mice as described above. After a 1:10 dilution, bacterial counts of the inoculum ranged from 10^7.9 to 10^8.4 CFU/mL. Lung infection was produced by the intranasal administration of 50 μL of the inoculum to isoflurane-anesthetized mice held upright to produce aspiration into the lungs. Therapy with one of the four antibiotics was initiated 2 h after the induction of infection, and therapy continued for 24 h, at which point the treatment groups and controls were sacrificed for CFU enumeration. The organism burden was quantified by CFU counts from serial dilutions of lung homogenates.

Murine pharmacokinetics.

Single-dose plasma and ELF pharmacokinetics of amikacin (6.25, 25, 100, and 400 mg/kg), ceftazidime (6.25, 25, 100, and 400 mg/kg), ciprofloxacin (5, 20, 80, and 320 mg/kg), and minocycline (2.5, 10, 40, and 160 mg/kg) were performed in infected mice following subcutaneous administration. Plasma was collected from groups of three mice for drug concentration determinations at six or seven time points over a 12- to 24-h period using three mice per time point. Plasma was obtained from each animal by centrifugation of anticoagulated blood obtained by cardiac puncture into Na-EDTA tubes, except for minocycline samples, which were collected in lithium-heparin tubes. Plasma was stored at −70°C. Drug concentrations were determined using liquid chromatography-tandem mass spectrometry (LC-MS/MS).

Single-dose pharmacokinetics were completed in infected mice (ICR/Swiss mice; weight, 24 to 27 g). Drug was administered by the subcutaneous route. Both plasma and bronchoalveolar lavage (BAL) fluid were collected from three infected mice per time point over 7 time points. EDTA was utilized as the anticoagulant for plasma collection. BAL fluid samples were collected from mice following lavage with 1 mL of sterile saline. The recovered BAL fluid was then centrifuged to separate the alveolar macrophages (cell pellet) and the ELF (supernatant). Plasma and ELF urea concentrations were measured to allow calculation of ELF drug concentrations by the method described by Rennard et al. (40). Urea concentrations were determined in triplicate by use of a commercial colorimetric assay kit (Pointe Scientific, Inc.). Corrected ELF concentrations were calculated by urea correction methodology using the following formula: [drug]_ELF = [drug]_BAL × ([urea]_plasma/[urea]_BAL). Total-drug concentrations from ELF were utilized for all analyses. The penetration of each drug into the ELF space was calculated by comparing ELF/plasma AUC ratios. Drug assay was performed by LC-MS/MS. Protein binding was assessed using equilibrium dialysis over a 100-fold concentration to allow comparison of mouse and human free-drug concentrations.

Pharmacokinetic parameters and exposure measures, including the beta-elimination half-life (t_1/2), 24-h AUC, and maximum concentration of drug in plasma (C_max), were calculated using noncompartmental analyses. The half-life was determined by linear least-squares regression. The AUC was calculated from the mean concentrations using the trapezoidal rule. Pharmacokinetic parameter estimates and exposure measures for dose levels that were not measured were calculated using linear interpolation or extrapolation.

In vivo efficacy assessment and determination of PK/PD targets.

Hill-type models were used to identify the magnitude of the dose and PK/PD index associated with net bacterial stasis (i.e., suppressing bacterial growth compared to the bacterial burden at the start of therapy) and the 1- and 2-log₁₀ CFU reduction from baseline (i.e., compared to start of therapy) for each antimicrobial agent, organism, and infection site. The choice of these PK/PD indices for these agents, AUC/MIC ratio for amikacin, ciprofloxacin, and minocycline and %T>MIC for ceftazidime, was supported by results of dose fractionation studies for these classes of agents that were carried out using murine infection models (41).

Human population pharmacokinetic models.

Population PK models describing the disposition of amikacin, ceftazidime, ciprofloxacin, and minocycline were obtained from the literature and qualified as described below. To identify available population models described in the literature for consideration, searches were performed using PubMed with keywords and terms including, “amikacin,” “ceftazidime,” “ciprofloxacin,” “minocycline,” “population pharmacokinetic,” and “pharmacokinetic model.” The population PK model for each drug was chosen prioritizing the following criteria: models for adult patients with infections, which were given priority over other patient populations; models developed using robust data sets in which patients were administered a broad range of doses; and models capable of predicting exposures in patients with various degrees of renal function.

Additional searches were performed to identify the ELF penetration ratio and protein binding values of each drug. PubMed search terms included “amikacin,” “ceftazidime,” “ciprofloxacin,” “minocycline,” “ELF penetration,” and “protein binding.”

Generation of simulated patients and drug exposures.

Using R version 4.0.4 (42) and a demographic database consisting of patients with cUTI/AP or HABP/VABP (data on file, Institute for Clinical Pharmacodynamics, Inc.) and CL_CR of ≥30 mL/min, simulated patient populations were generated. The observed databases of patients were replicated a sufficient number of times to achieve a population of 3,000 or more simulated patients in each group. In an effort to represent the target patient populations with infections arising from A. baumannii, demographic variables among simulated patients remained the same as for the observed patients.

Typical PK values for each simulated patient in the above-described populations were calculated using demographic values for relevant covariates predictive of PK in conjunction with the fixed-effect parameter estimates for each population PK model selected for each agent. A description of the process to generate individual PK parameters values for each agent is provided in the supplemental material.

Using the selected population PK models for amikacin, ceftazidime, ciprofloxacin, and minocycline and resultant individual PK parameters, total-drug concentration-time profiles for each drug were generated for each simulated patient after administration of the selected dosing regimens of each agent, including those adjusted for renal impairment, described in Table 4 (4–6, 14, 24). Given the severity of disease and relatively higher MICs expected with A. baumannii infections, dosing regimens with maximum total doses for each antibiotic were considered. Individual total-drug concentration-time profiles were generated for each drug from 0 to 48 h after the first dose. Using the protein binding estimates of 0, 10, 30, and 76% for amikacin, ceftazidime, ciprofloxacin, and minocycline, respectively (4, 5, 14, 43), free-drug plasma concentrations were determined by multiplying the individual predicted total-drug plasma concentrations by the free fraction. The process to determine free-drug plasma and total-drug ELF %T>MIC for ceftazidime and free-drug plasma and total-drug ELF AUC/MIC ratios for amikacin, ciprofloxacin, and minocycline based on concentration-time profiles generated from 0 to 48 h is described in the supplemental material. The evaluation of drug exposures over 48 h allowed comparison of agents across different dosing intervals for dosing regimens adjusted for renal impairment that were administered to simulated patients with lower CL_CR.

PK/PD target attainment analyses.

Using the above-described average 24-h free-drug plasma and total-drug ELF PK/PD indices calculated for the simulated patients, percent probabilities of attaining median and randomly assigned PK/PD targets for A. baumannii were evaluated. Randomly assigned %T>MIC targets for ceftazidime and AUC/MIC ratio targets for amikacin, ciprofloxacin, and minocycline were based on estimated log normal distributions derived from the targets associated with each of the endpoints, net bacterial stasis, and 1- and 2-log₁₀ CFU reductions from baseline. As described previously (31), each distribution was truncated at ± 2 standard deviations on the log scale. Percent probabilities of PK/PD target attainment were assessed at fixed MICs and relative to the in vitro surveillance data shown in Table S1 for each agent against A. baumannii.

Amikacin, ceftazidime, ciprofloxacin, and minocycline PK/PD targets associated with net bacterial stasis and a 1- and/or 2-log₁₀ CFU reduction from baseline for A. baumannii were assessed. PK/PD targets associated with net bacterial stasis and a 1-log₁₀ CFU reduction from baseline are considered appropriate when evaluating STIC for patients with cUTI, since such infections are associated with lower bacterial inocula (28). For decisions about STIC for dosing regimens for patients with HABP/VABP, PK/PD targets associated with 1- and 2-log₁₀ CFU reductions from baseline were considered. The former endpoint is supported by the results of an analysis for which a relationship between attaining a PK/PD target associated with a 1-log₁₀ CFU reduction from baseline based on data from neutropenic mice was associated with a high probability of regulatory approval for antimicrobial agents for the treatment of patients with pneumonia. Antimicrobial dosing regimens that met this criterion had the greatest probability of meeting the clinical protocol-defined primary efficacy endpoints (29–32). Among the assessments carried out, priority was given to the PK/PD target attainment results based on randomly assigned PK/PD targets. For STIC for pneumonia, priority was also given to assessments based on total-drug ELF exposures.

Susceptible breakpoints for each agent against A. baumannii that represented the highest MIC at which the percent probabilities of target attainment approached or were ≥90% based on both median and randomly assigned PK/PD targets associated with a 1-log₁₀ CFU reduction from baseline were identified. This bacterial reduction endpoint was chosen given the expectation that patients with infections arising from A. baumannii would frequently have pneumonia or other systemic infections with potentially poor source control and/or comorbidities. STIC recommendations in the context of these results and the range of ECOFF values generated were considered. The latter data were evaluated in order to identify STIC that minimized bisection of the wild-type MIC distributions when possible.