Imipenem/Cilastatin/Relebactam Alone and in Combination against Pseudomonas aeruginosa in the In Vitro Pharmacodynamic Model

Combination therapy may enhance imipenem/cilastatin/relebactam’s (I/R) activity against Pseudomonas aeruginosa and suppress resistance development. Human-simulated unbound plasma concentrations of I/R at 1.25 g every 6 h (h), colistin at 360 mg daily, and amikacin at 25 mg/kg daily were reproduced alone and in combination against six imipenem-nonsusceptible P. aeruginosa isolates in an in vitro pharmacodynamic model over 24 h.

ABSTRACT

Combination therapy may enhance imipenem/cilastatin/relebactam’s (I/R) activity against Pseudomonas aeruginosa and suppress resistance development. Human-simulated unbound plasma concentrations of I/R at 1.25 g every 6 h (h), colistin at 360 mg daily, and amikacin at 25 mg/kg daily were reproduced alone and in combination against six imipenem-nonsusceptible P. aeruginosa isolates in an in vitro pharmacodynamic model over 24 h. For I/R alone, the mean reductions in CFU ± the standard errors by 24 h were −2.52 ± 0.49, −1.49 ± 0.49, −1.15 ± 0.67, and −0.61 ± 0.10 log₁₀ CFU/ml against isolates with MICs of 1/4, 2/4, 4/4, and 8/4 μg/ml, respectively. Amikacin alone also resulted in 24 h CFU reductions consistent with its MIC, while colistin CFU reductions did not differ. Resistant subpopulations were observed after 24 h in 1, 4, and 3 I/R-, colistin-, and amikacin-exposed isolates, respectively. The combination of I/R and colistin resulted in synergistic (n = 1) or additive (n = 2) interactions against three isolates with 24-h CFU reductions ranging from −2.62 to −4.67 log₁₀ CFU/ml. The combination of I/R and amikacin exhibited indifferent interactions against all isolates, with combined drugs achieving −0.51- to −3.33-log₁₀ CFU/ml reductions. No resistant subpopulations were observed during I/R and colistin combination studies, and when added to amikacin, I/R prevented the emergence of amikacin resistance. Against these six multidrug-resistant P. aeruginosa, I/R alone achieved significant CFU reductions against I/R-susceptible isolates. Combinations of I/R plus colistin resulted in additivity or synergy against some P. aeruginosa, whereas the addition of amikacin did not provide further antibacterial efficacy against these isolates.

INTRODUCTION

Antibiotic resistance among Pseudomonas aeruginosa, a nonfermenting Gram-negative rod capable of causing severe lower respiratory tract infections, bacteremia, meningitis, intra-abdominal infections, and urinary tract infections, has evolved into a worldwide threat (1–6). Of the 4,175 P. aeruginosa collected by a 2015-2016 surveillance program, 21.1% were multidrug resistant (MDR), and 9.4% were extensively drug resistant (7). Although carbapenems have traditionally served as last-line agents for MDR strains, increasing resistance to this individual class has unfortunately diminished this role. Only 20.9 and 5.3% of MDR and extensively drug-resistant isolates evaluated in the program were susceptible to meropenem (7). This diminishing susceptibility highlights the necessity of new therapies.

Imipenem/cilastatin/relebactam (I/R; Recarbrio; Merck & Co., Inc., Kenilworth, NJ) is a β-lactam/β-lactamase inhibitor combination antibiotic approved to treat complicated urinary tract and intra-abdominal infections, as well as more recently, hospital-acquired and ventilator-associated bacterial pneumonia (8). An additional clinical trial was conducted for the treatment of imipenem-nonsusceptible Gram-negative infections (9). Two large surveillance studies observed that relebactam restored imipenem susceptibility in 74 to 78% of imipenem-nonsusceptible P. aeruginosa (10, 11). These findings were translated clinically in the RESTORE-IMI 1 trial, where 77% of patients (n = 31) were infected with an imipenem-nonsusceptible P. aeruginosa strain, and 71% (n = 21) achieved a favorable I/R response (9).

Despite the potency of I/R, treatment failure and resistance development remain concerns, causing clinicians to conservatively consider combination therapy for severe P. aeruginosa infections (12). Polymyxin and aminoglycoside antibiotics remain popular candidates for combination given their differing mechanisms of action and high susceptibility rates (12–14). Our lab previously assessed the in vitro activity of I/R alone and with colistin or amikacin using time-kill methodology (15). Synergistic interactions were observed against most of the isolates, encouraging further investigation. The objective of this study was to compare the antibacterial activity of I/R alone and in combination with colistin or amikacin against six imipenem-nonsusceptible P. aeruginosa isolates in the in vitro pharmacodynamic (IVPD) model.

RESULTS

Study isolates.

Six P. aeruginosa isolates were selected for this study. MIC data are reported in Table 1. Against I/R, three of the six isolates were susceptible, two were intermediate, and one was resistant (16). All isolates were intermediate to colistin and susceptible or intermediate to amikacin. Finally, all six isolates met the definition for MDR strains (15, 17).

TABLE 1.

MICs against the six multidrug-resistant P. aeruginosa study isolates

Isolate	Modal MIC (μg/ml)
	Imipenem-relebactama	Imipenemb	Colistinc	Amikacind	Ceftolozane-tazobactame	Ceftazidime-avibactamf
CAIRD M8-29	1/4	4	1	8	1/4	8/4
CDC0270g	2/4	16	0.5	16	NAi	NAi
CDC0526h	2/4	16	1	8	2/4	2/4
CDC0527h	4/4	32	0.5	0.5	1/4	8/4
CAIRD M23-3	4/4	32	1	32	2/4	4/4
CAIRD M1-4	8/4	32	0.5	32	4/4	8/4

^aImipenem-relebactam MIC breakpoints (16): ≤2/4 μg/ml (susceptible), 4/4 μg/ml (intermediate), and ≥8/4 μg/ml (resistant).

^bImipenem MIC breakpoints (34): ≤2 μg/ml (susceptible), 4 μg/ml (intermediate), and ≥8 μg/ml (resistant).

^cColistin MIC breakpoints (34): ≤2 μg/ml (intermediate) and ≥4 μg/ml (resistant).

^dAmikacin MIC breakpoints (34): ≤16 μg/ml (susceptible), 32 μg/ml (intermediate), ≥64 μg/ml (resistant).

^eCeftolozane-tazobactam MIC breakpoints (34): ≤4/4 μg/ml (susceptible), 8/4 μg/ml (intermediate), and ≥16/4 μg/ml (resistant).

^fCeftazidime-avibactam MIC breakpoints (34): ≤8/4 μg/ml (susceptible) and ≥16/4 μg/ml (resistant).

^gSource: CDC and FDA Antibiotic Resistance Isolate Bank Source panel: Pseudomonas aeruginosa.

^hSource: CDC and FDA Antibiotic Resistance Isolate Bank Source panel: imipenem/relebactam.

ⁱNA, not available.

Human-simulated exposures.

The I/R, colistin, and amikacin concentrations achievable in critically ill patients were targeted to simulate unbound human plasma exposures in the model. Table 2 displays observed versus targeted free maximum concentration (fC_max), half-life (t_1/2), and free area under the curve for 0 to 24 h (fAUC_0–24) for applicable regimens. Mean exposures were within 20% of target for all regimens except relebactam, which resulted in a mean fAUC_0–24 that was 28.6% higher than targeted. Free time above the MIC (fT>MIC) or fAUC_0–24/MIC exposures by isolate are provided in supplemental materials Table S1.

TABLE 2.

Observed versus targeted concentrations and pharmacokinetic parameters attained for imipenem, relebactam, colistin, and amikacin in the in vitro pharmacodynamic model^a

Antibiotic	fCmax (μg/ml)		t1/2 (h)		fAUC0–24 (mg⋅h/liter)
	Target	Observed	Target	Observed	Target	Observed
Imipenem	25.0	24.7 (22.7–28.6)	1.2	1.1 (1.1–1.2)	162	162 (150–190)
Relebactam	13.0	15.1 (13.9–18.0)	1.2	1.2 (1.1–1.2)	84	108 (97–125)
Colistin	0.73b	0.85 (0.79–0.91)b	NA	NA	18	21 (19–22)
Amikacin	NA	NA	NA	NA	348	377 (350–410)

^aData are reported as medians (interquartile ranges). fC_max, free maximum concentration; t_1/2, half-life; fAUC_0–24, area under the curve over the 24-h experiment. NA, not applicable.

^bDue to continuous infusion in the model, colistin fC_max is the mean concentration observed over the 24-h experiment.

Antibacterial efficacy.

Time-kill curves are provided in Fig. 1. The starting inoculum across all experiments achieved a mean bacterial density of 6.24 ± 0.15 log₁₀ CFU/ml. Control reactors verified bacterial growth throughout the study and grew to 7.25 ± 0.29 log₁₀ CFU/ml over 24 h. Figure 2 depicts the observed 24-h differences in bacterial density. For I/R alone, CFU reductions by 24 h were −2.52 ± 0.49, −1.49 ± 0.49, −1.15 ± 0.67, and −0.61 ± 0.10 log₁₀ CFU/ml against isolates with MICs of 1/4, 2/4, 4/4, and 8/4 μg/ml, respectively. For amikacin alone, CFU reductions were –4.13 ± 0.44, –1.65 ± 1.31, 0.21 ± 1.05, and 1.34 ± 0.31 log₁₀ CFU/ml against isolates with amikacin MICs of 0.5, 8, 16, and 32 μg/ml. Finally, for colistin alone, the CFU reductions were −2.49 ± 0.78 and –2.03 ± 0.91 log₁₀ CFU/ml against isolates with colistin MICs of 0.5 and 1 μg/ml. Table 3 lists the interaction category of each combination regimen. The results of the log ratio (LR) of the area under the curve (AUCFU) analyses are provided in Table 4.

FIG 1.

Mean number of CFU over 24 h by isolate. The data are presented as the mean CFU from experiment replicates of each regimen. (A) CAIRD M8-29; (B) CDC0270; (C) CDC0526; (D) CDC0527; (E) CAIRD M23-3; (F) CAIRD M1-4. Sold line/no symbol, control; dashed line/no symbol, I/R alone; solid line/square, colistin alone; solid line/triangle, amikacin alone; dashed line/square, I/R plus colistin combination therapy; dashed line/triangle, I/R plus amikacin combination therapy; dotted line, lowest limit of detection (LLOD).

FIG 2.

Mean change at 24 h in bacterial density from 0 h. (A) CAIRD M8-29; (B) CDC0270; (C) CDC0526; (D) CDC0527; (E) CAIRD M23-3; (F) CAIRD M1-4. All treatments were significantly different from their controls (P < 0.05) except for I/R against CAIRD M23-3 and amikacin against CDC0270, CAIRD M23-3, and CAIRD M1-4. Statistical results between combination therapies and their constituent agents alone are reported in the upper left corner of each graph.

TABLE 3.

Antimicrobial interactions observed for imipenem-relebactam combinations at 24 h against P. aeruginosa

Isolate	Antimicrobial interactiona
	I/R+CST	I/R+AMK
CAIRD M8-29	Additive	Indifferent
CDC0270	Indifferent	Indifferent
CDC0526	Synergy	Indifferent
CDC0527	Indifferent	Indifferent
CAIRD M23-3	Additive	Indifferent
CAIRD M1-4	Indifferent	Indifferent

^aI/R, imipenem-relebactam; CST, colistin; AMK, amikacin.

TABLE 4.

LR of AUCFU for antibiotics alone and in combination against P. aeruginosa

Isolate	Log difference in AUCFUa
	I/R/control	CST/control	AMK/control	I/R+CST/I/R alone	I/R+CST/CST alone	I/R+AMK/I/R alone	I/R+AMK/AMK alone
CAIRD M8-29	−2.80	−2.67	–2.06	0.10	0.11	0.14	–0.55
CDC0270	–2.76	–2.72	–1.35	–0.08	–0.14	0.42	–1.11
CDC0526	–2.54	–2.65	–2.38	–0.30	–0.19	–0.11	–0.09
CDC0527	–2.92	–2.87	–2.89	–0.09	0.04	–0.20	–0.02
CAIRD M23-3	–2.32	–2.62	–0.49	–0.68	–0.33	0.15	–1.48
CAIRD M1-4	–1.96	–2.82	–0.54	–0.71	0.10	–0.15	−1.75

^aLog difference is reported as log ratio (LR), or the number of log₁₀CFU(regimen 1/regimen 2). AUCFU, area under the curve for CFU; I/R, imipenem-relebactam; CST, colistin; AMK, amikacin.

Resistant subpopulations.

After 24 h of growth within the untreated control reactors, neither I/R- nor amikacin-resistant subpopulations emerged on antibiotic containing agar plates at 3× the baseline MIC in any of the isolates. However, colistin-resistant subpopulations were identified in four untreated isolates (CAIRD M8-29, CDC0270, CDC0526, and CAIRD M23-3), suggesting a baseline presence of heteroresistance. After 24 h of exposure to each antibiotic alone, resistant subpopulations were observed in one isolate to I/R (CAIRD M23-3), in all four heteroresistant isolates to colistin (CAIRD M8-29, CDC0270, CDC0526, and CAIRD M23-3), and in three isolates to amikacin (CAIRD M8-29, CDC0270, and CAIRD M1-4). No I/R- or colistin-resistant subpopulations emerged when isolates were exposed to I/R plus colistin. The combination of I/R and amikacin prevented the emergence of amikacin-resistant subpopulations in all isolates, but I/R-resistant subpopulations for M23-3 remained in one of two chemostat models.

DISCUSSION

Although carbapenems have historically been used to treat MDR P. aeruginosa infections, recent surveillance studies have observed resistance rates of >20% in North America (18). As the class’s susceptibility declines, optimal utilization of newer therapies becomes more important, especially in critically ill patients who are most likely to be afflicted by MDR pathogens. Although relebactam itself restores imipenem susceptibility, combination with other antibiotics offers the potential of enhanced antibacterial effects and suppression of resistance development. In this study, the IVPD model was used to determine the antibacterial effects adding colistin or amikacin to I/R. Although our lab previously performed a time-kill study investigating the same drug combinations (15), our observations here differed slightly. Colistin plus I/R achieved synergistic (n = 1) or additive (n = 2) antibacterial activity in half of the isolates and suppressed the emergence of resistant subpopulations. In contrast, benefits from the combination of I/R plus amikacin were restricted to amikacin resistance suppression alone.

When interpreting these results, it is critical to consider the pharmacodynamics of each component of I/R. Because relebactam serves to restore imipenem’s activity by preventing enzymatic degradation, the main pharmacodynamic driver of imipenem remains the fT>MIC as a percentage of the dosing interval (19). Imipenem requires ∼40% fT>MIC against most Gram-negative bacteria to achieve at least 1-log₁₀ CFU/ml kill. An I/R regimen simulating the mean human exposure obtained with the approved adult regimen of 1.25 g q6h (i.e., 500 mg of imipenem, 500 mg of cilastatin, and 250 mg relebactam) was utilized and provided at least 40% fT>MIC for imipenem against isolates with I/R MICs up to 4/4 μg/ml (see Table S1 in the supplemental material). As a result, the three I/R-susceptible isolates (MICs ≤ 2/4 μg/ml) and one of the intermediate isolates (MIC = 4/4 μg/ml) yielded >1-log₁₀ CFU/ml 24-h reductions.

For relebactam, the driver when combined with a fixed-dose imipenem was fAUC/MIC, and thresholds of 8.2, 12, and 18 were required for imipenem to achieve stasis, 1 log, and 2 logs of bacterial kill in the in vitro hollow-fiber model, respectively (20). Furthermore, a translational semimechanistic model identified a relebactam fAUC/MIC of 7.5 to be predictive of 2-log kill (21). During our experiments, the attained relebactam fAUC_0–24 exposures were elevated above the conservative target from healthy adult volunteers by 28.6%. Despite the higher relebactam exposures, the observed CFU reductions were concordant with imipenem fT>MIC and the reported I/R MIC. Importantly, the attained relebactam exposures were well within range of fAUC_0–24 reported in humans during clinical trials. In a phase 2 dose-ranging study, Lucasti et al. observed a median total relebactam AUC_0–24 of ∼350,000 nmol⋅h/liter (22). When corrected for molecular weight and protein binding, the free drug steady-state average concentration (C_ssavg) in the population was 4.1 μg/ml. Our free relebactam C_ssavg was 4.5 μg/ml, a difference unlikely to further potentiate the imipenem MIC against P. aeruginosa (20, 23, 24). Collectively, the CFU reductions for I/R alone in this experiment support the current U.S. Food and Drug Administration (FDA) susceptibility breakpoint of ≤2/4 μg/ml against P. aeruginosa for the approved dosing regimen (16).

Colistin and amikacin are infrequently administered alone to treat serious P. aeruginosa infections due to a high likelihood of heteroresistant subpopulations and subsequent development of resistance. Indeed, this was observed here as well. Colistin had a rapid and profound initial in vitro effect when tested alone, producing >4 logs of kill in all strains over the first hour, followed by rapid regrowth over the last 12 to 18 h. In addition, colistin-resistant subpopulations were identified in four of the six isolates despite all isolates having MICs in the intermediate range of 0.5 to 1 μg/ml. These were the same isolates that demonstrated heteroresistance at baseline, signaling a predisposition toward resistance development. These observations are consistent with other IVPD studies against P. aeruginosa (25, 26). The observed free colistin steady-state concentrations during these experiments were >0.5 μg/ml but generally <1 μg/ml. This is consistent with the concentrations achieved in patients, even after an aggressive regimen. However, variability in achievable colistin concentrations can be substantial in patients, and we cannot extrapolate our observations to free drug concentrations above 1 μg/ml, which are attainable with the European Medicines Agency (EMA)-approved dose (27). Amikacin MICs for the included organisms ranged between 0.5 and 32 μg/ml (susceptible to intermediate range). Despite the simulation of an aggressive, high-dose, once-daily amikacin regimen, all isolates with MICs of ≥8 μg/ml displayed substantial killing over the first 4 h, followed by regrowth nearly to control. These observations were not surprising. Aminoglycosides require fAUC/MIC of ∼80 to achieve 1-log₁₀ CFU/ml reductions (28), and only CDC0527 with an amikacin MIC of 0.5 μg/ml (fAUC/MIC 847; see Table S1) succumbed to maximal CFU reductions over 24 h. Our CFU observations with amikacin alone are also consistent with past in vitro chemostat experiments using P. aeruginosa with MICs of 2 to 8 μg/ml (25).

The combination studies used two accepted endpoints to assess the value of adding a second antibiotic for P. aeruginosa. First, antibacterial effects were assessed based on the change in 24-h CFU relative to the antibiotic that achieved the greatest kill when studied alone (Fig. 2 and Table 3). In addition, the LR AUCFU was assessed (Table 4). The combination of I/R plus colistin resulted in synergistic or additive effects against three of the six isolates and indifferent effects against the remaining isolates. Synergy was only observed against a single isolate that was I/R-susceptible with an MIC of 2/4 μg/ml (CDC0526, Fig. 2C). Additive effects were observed for CAIRD M8-29 (I/R MIC 1/4 μg/ml, Fig. 2A). Finally, additive effects were observed for CAIRD M23-3 (I/R MIC 4/4 μg/ml, Fig. 2E), where I/R and the combination with colistin resulted in −0.69- and −3.87-log₁₀ CFU/ml reductions (P = 0.009), respectively. However, the combination regimen did not achieve a statistically significant CFU reduction from colistin alone.

The addition of amikacin to I/R achieved only indifferent effects against all isolates, and none of the observed CFU reductions were significantly different from I/R alone. Based on attainable fAUC/MIC ratios, it is plausible that any potential benefits for combination therapy disappear when amikacin MICs increase to ≥8 μg/ml. However, synergistic or additive effects were also absent for the three susceptible isolates when combined with I/R, which we believe is due to the bactericidal effects of at least one of the backbone agents. For example, only against CDC0527 (amikacin MIC 0.5 μg/ml, Fig. 1D) was amikacin alone able to achieve maximal CFU reductions at 24 h, thus preventing any opportunity for observations of synergy.

CAIRD M8-29, CDC0526, CAIRD M23-3, and CAIRD M1-4 were also included in a recent time-kill synergy study and were selected for this study to bridge the static and dynamic in vitro experiments (15). In contrast to observations in the chemostat IVPD, the addition of colistin to I/R resulted in synergy against all of these isolates in the time-kill studies except for CAIRD M8-29, against which indifference was observed. Amikacin plus I/R resulted in synergy against all isolates. The primary difference between models was the use of static average concentrations compared to human-simulated profiles in the IVPD model. Notably, colistin exposures would have been nearly identical because it was simulated as a continuous C_ssavg in the chemostat. However, despite using average steady-state concentrations for I/R and amikacin in the time-kill experiments that were based on the AUC exposures used in the chemostat experiments, the shape of the concentration time curves are very different and likely responsible for the observed differences. For example, imipenem concentrations of 5.9 μg/ml in the time-kill study would provide 100% fT>MIC against all isolates except for CAIRD M1-4 with an MIC of 8/4 μg/ml. However, in the IVPD, 100% fT>MIC exposure was not achieved for any of the organisms except CAIRD M8-29 (see Table S1).

The LR for AUCFU provides an alternative method to assess antibacterial activity (29). Unlike the above categorizations, which focused on the bacterial density at a single time point, it assesses overall activity over 24 h. A value of –1 log indicates a 90% decrease in bacteria population and −2 logs would be equivalent to a 99% decrease. Although substantial reductions were observed compared with amikacin alone (Table 4), the most meaningful comparisons would be between the combination regimens and I/R alone, since amikacin and colistin monotherapy are discouraged clinically due to the high risk of resistance emerging. Using I/R alone as the reference regimen, the highest LR values occurred when colistin was included against I/R-nonsusceptible isolates, notably CAIRD M23-3 (−0.68 log or 79% reduction) and CAIRD M1-4 (−0.71 log or 81%). Importantly, colistin alone achieved substantial overall antibacterial effects against these isolates, and we propose the combination allows colistin to achieve this kill without emergence of resistant subpopulations over the 24-h experiment. While I/R nonsusceptibility in P. aeruginosa is still rare, such isolates have varied likelihoods of activity to other β-lactam/β-lactamase inhibitor combination antibiotics. Should I/R treatment be required for P. aeruginosa with MICs of 4/4 to 8/4 μg/ml and no other available β-lactam based therapies exist, a combination regimen with colistin appears to be a reasonable consideration. In contrast, I/R-plus-amikacin combinations resulted in little overall bacterial load reduction and appear to be less beneficial unless the amikacin MIC was very low.

One perceived benefit of combination therapy is its potential to suppress resistance development (30, 31). Colistin and amikacin have both previously demonstrated their ability to prevent the emergence of resistance when combined with β-lactam-based therapies (32, 33). In our study, only one isolate generated I/R-resistant subpopulations over 24 h, whereas four colistin- and amikacin-resistant subpopulations were identified in four and three isolates, respectively, after exposure to the respective antibiotics alone. Of note, the isolate producing I/R-resistant subpopulations was already defined as nonsusceptible before exposure (CAIRD M23-3, I/R MIC 4/4 μg/ml). Upon the addition of colistin to I/R, no resistant subpopulations developed to either drug, whereas the addition of amikacin prevented the emergence of amikacin-resistant subpopulations. Although we view these data positively, additional studies longer than 24 h and those conducted at higher inocula would be of value to determine the durability of these observations.

The limitations present in these experiments are similar to those in other IVPD studies. Due to the nature of the chemostat model, isolate growth was not impeded by an immune system, as might be available in an animal or human model. Only a single exposure, representing the mean of the population, for each drug regimen was simulated; therefore, the effects of lower or higher exposures of each antibiotic, even those within the naturally occurring variability found in patients, could influence the antibacterial effects of the combinations. As previously mentioned, future studies may benefit from experiments longer than 24 h to observe if CFU reductions and the suppression of resistant subpopulations are sustained. Of the selected isolates, none were resistant to amikacin or colistin; however, such resistance phenotypes are still rare. Additional studies would be worthwhile to determine whether a combination of colistin plus I/R would prove beneficial for colistin-resistant strains.

In summary, against these six MDR P. aeruginosa, I/R exposures consistent with the approved 1.25-g q6h regimen achieved significant CFU reductions in the in vitro chemostat model against I/R-susceptible isolates when tested alone. The addition of colistin to I/R resulted in synergistic or additive interactions at 24 h against three of six isolates and suppressed resistance emergence, whereas a combination regimen containing amikacin demonstrated indifference against all isolates. Additional reductions in overall bacterial burden compared to I/R alone were observed for I/R plus colistin against two I/R-nonsusceptible isolates. These data warrant further studies evaluating the combination of I/R and colistin against MDR P. aeruginosa.

MATERIALS AND METHODS

Study isolates.

Six imipenem-nonsusceptible clinical P. aeruginosa isolates were included in this study. Isolates were selected to span an I/R MIC range between 1/4 and 8/4 μg/ml, which would be within 1 to 2 dilutions of the current imipenem susceptibility breakpoint for P. aeruginosa (2/4 μg/ml). CAIRD M8-29, M23-3, and M1-4 were clinical isolates obtained between July 2017 and June 2018 from U.S. hospitals. Remaining isolates were obtained from the Centers for Disease Control and Prevention (CDC)/FDA Antibiotic Resistance (AR) Bank (www.cdc.gov/drugresistance/resistance-bank/index.html). The modal MICs of I/R, colistin, amikacin, ceftolozane-tazobactam, and ceftazidime-avibactam were obtained in triplicate via broth microdilution according to the Clinical and Laboratory Standards Institute (CLSI) (34).

Antibacterial agents.

Imipenem (lot number 0000685746) and relebactam (lot number 002D039) were supplied by Merck & Co., Inc. (Kenilworth, NJ). Colistin sulfate analytical powder (lot number SLCB7174) was purchased from Sigma-Aldrich (St. Louis, MO). Amikacin sulfate analytical powder (lot number 115837/B) was purchased from Medisca (Plattsburgh, NY). Stock solutions were prepared per CLSI recommendations (34) and frozen at –80°C until needed.

Human-simulated drug exposures.

The I/R regimen was designed to simulate fC_max and t_1/2 of imipenem and relebactam observed in healthy human volunteers receiving I/R at 1.25 g (500 mg of imipenem/250 mg of relebactam; cilastatin was not added to the models) every 6 h as 0.5-h infusions (35, 36). Assuming 20% protein binding for both imipenem and relebactam, target free C_max values were 25.0 and 13.0 μg/ml, respectively; the simulated t_1/2 was 1.2 h for both drugs. Colistin was infused continuously to achieve the free steady-state average concentration per the recommended EMA total daily dose of colistin 360 mg for critically ill patients (27); assuming 50% protein binding, this was 0.73 μg/ml. Amikacin was administered to simulate the free C_max and fAUC_0–24 produced by a regimen of 25 mg/kg every 24 h (37). Assuming 11% protein binding, free C_max and fAUC_0–24 were 63.6 and 348.0 mg⋅h/liter, respectively. The elimination rate was matched to I/R, so supplemental amikacin bolus doses were administered to retain the fAUC_0–24 target.

In vitro pharmacodynamic chemostat model.

The one-compartment in vitro chemostat model was employed as previously described (25). Each isolate-drug combination experiment was composed of two treatment reactors and one antibiotic-free control reactor. The reactors sat in a 35°C water bath and contained 150 ml of cation-adjusted Mueller-Hinton broth (lot number 9239528; Becton Dickinson, Sparks, MD) and a magnetic stir bar. Thirty minutes before antibiotic administration, the reactors were inoculated with a bacterial suspension of 10⁶ log₁₀ CFU/ml to permit log phase of growth. At 0 h, bolus doses of the antibacterial agents were injected to reach target peak concentration(s), and sterile antibiotic-free broth was infused through a peristaltic pump (Masterflex model 7519-05 and Masterflex L/S model 7519-15; Cole-Palmer Instrument Company, Vernon Hills, IL) to target the I/R t_1/2. For experiments including colistin, the infused broth contained a colistin concentration necessary to achieve the 0.73 μg/ml target. Experiments were performed over 24 h after antibiotic administration.

To follow the bacterial density over time, samples were collected at 0, 1, 4, 8, 16, and 24 h, diluted in normal saline, and plated on Trypticase soy agar with 5% sheep blood (BBL stacker plate; Becton Dickinson). Colony counts were read after incubation for 16 to 24 h at 37°C. The lower limit of detection (LLD) was 1.7 log₁₀ CFU/ml.

Resistant subpopulation determination.

Samples were collected at 0 and 24 h to assess the emergence of resistant subpopulations, which was defined as the presence of colonies (LLD 1.7 log₁₀ CFU/ml) on antibiotic-containing plates at least 3× higher than the baseline MIC. Portions (10 μl) of sample were plated on I/R-, colistin-, or amikacin-containing agar plates (lot number 9002580; Becton Dickinson). For imipenem and amikacin, plate concentrations were isolate dependent, targeting 3× MIC. Relebactam was maintained at 4 μg/ml. Colistin-containing plates were prepared to 3 μg/ml to target a concentration above the P. aeruginosa intermediate breakpoint of 2 μg/ml (34). All antibiotic-containing plates were confirmed for targeted concentration by plating isolates with known MICs above and below the target concentration on the morning of each experiment. Inoculated plates were incubated at 37°C for 18 to 24 h.

Antibiotic concentrations and exposures.

Samples taken to measure antibiotic concentrations were obtained and stored at –80°C until analysis. Prior to freezing, a stabilizing buffer of 0.01 M phosphate buffer (pH 7.2) and acetonitrile was added to I/R-containing samples (75 μl of sample, 75 μl of buffer, and 150 μl of acetonitrile). Imipenem and relebactam were assayed by Merck & Co. using a high-performance liquid chromatography assay. The LLD for both drugs was 0.03 μg/ml. Amikacin concentrations were determined by Quest Diagnostics (Chantilly, PA) via an enzyme multiple-immunoassay technique (LLD, 2.5 μg/ml). Keystone Bioanalytical, Inc. (North Wales, PA), analyzed the colistin samples via a validated liquid chromatography-tandem mass spectrometry assay. Colistin A and colistin B concentrations were obtained and summed for the final colistin concentration result reported (LLD, 0.05 μg/ml).

Statistical analyses.

Antibiotic activity was determined by the difference in log₁₀ CFU/ml observed between 0 and 24 h. These changes were compared through analysis of variance with the Holm-Sidak method in SigmaPlot (version 14.0; Systat Software, Inc., San Jose, CA). An a priori P value of <0.05 was considered statistically significant. Synergy was defined as a >2-log₁₀ CFU/ml difference from the change observed with the most active agent alone (25, 38, 39). Additivity was defined as a 1- to 2-log₁₀ CFU/ml difference, and indifference was defined as a <1-log₁₀ CFU/ml. Antagonism was defined as a statistically significant worsening relative to the efficacy of the least active agent alone. The AUCFU was calculated using the trapezoidal rule. The LR difference in AUCFU was determined to analyze the change in bacterial burden between two regimens (25).