We searched relevant literature in PubMed, MEDLINE, CINAHL, Cochrane, Embase, Scopus, and Web of Science in December 2022. The search terms, MeSH terms and publication filters used are presented in the Supplementary material. The references were imported into an EndNote (Clarivate Analytics, Philadelphia, Pennsylvania, United States) database, and duplicated and non-English studies were not included. References within the recruited articles were searched manually to capture additional studies.

The titles and abstracts of all records were screened by 2 reviewers (AA and YE) independently to evaluate full-text eligibility. Any disagreements were resolved by discussion. Articles were included after full-text review if they met the following inclusion criteria: (1) original research, (2) in vitro, in vivo or clinical studies comparing CZA alone vs. CZA-containing combinations against CRE and/or MDR-PA isolates (regardless of their resistance mechanisms) or infections, and (3) published up to December 1, 2022. Otherwise, articles were excluded for the following reasons: (1) review articles; (2) the full text was not available; (3) grey literature; (4) abstracts, comments, letters, and editorials; (5) non-peer reviewed articles; (6) studies published in a language other than English; (7) in vitro synergy studies conducted on only one isolate.

Data were extracted into predesigned spreadsheets in duplicate by two independent reviewers. In in vitro studies, data pertaining with authors, publication year, country, AMR phenotype, predominant AMR mechanisms, CZA susceptibility rate, type of bacterial strain, number of isolates tested, method(s) used for antimicrobial susceptibility tests and criteria used for assessment of their results, type of in vitro synergy test applied, companion antibiotic(s) used, definition of synergism, additive effect, and antagonism, concentrations of CZA and companion antibiotics used in experiments, and number of repeats in experiments were extracted. In in vivo studies, authors, publication year, country, type of bacterial strain, AMR phenotype, main molecular determinants of AMR, type of in vivo infection model, number of isolates tested, method(s) used for antimicrobial susceptibility tests and criteria used for interpretation of their results, doses of CZA and companion antibiotics adjusted in experiments, and killing activity of tested antibiotics were extracted. In clinical studies, we extracted authors, publication year, country, the study design, statistical method, and relevant data, including the number of patients in comparison groups, type of companion antibiotics, in vitro activity of concomitant antibiotics, median time from onset of infection to active treatment, demographic information of patients, the infection type, number of bacteremic patients, carbapenem non-susceptibility rate of the causative pathogens, main AMR mechanism, presence of septic shock, and outcomes parameters of individual studies (i.e., all-cause mortality, clinical cure, microbiological cure, adverse events, acute kidney injury, emergence of CZA resistance and infection relapse). Because of the substantial heterogeneity of study designs, treatment regimens, definitions of study outcomes and combination regimens (i.e., in vitro inactive antibiotics being considered as co-antibiotics in some studies), we did not perform quantitative synthesis of the data using a meta-analysis and a narrative synthesis was undertaken to evaluate the efficacy and safety of CZA monotherapy and CZA combination therapies.

Some antibiotics can be combined with CZA to expand or enhance its antibacterial capacity. For example, avibactam protects aztreonam from the hydrolytic activity of ESBLs and AmpC β-lactamases, which are frequently co-produced by metallo-β-lactamase-producing gram-negative bacteria (22). As aztreonam is not hydrolyzed by metallo-β-lactamases, it can show significant antibacterial activity in the presence of CZA against metallo-β-lactamase-producing gram-negative bacteria (22). Furthermore, CZA plus aztreonam may substantially influence the “divisome” of these bacteria by acting on different types of penicillin-binding proteins (22). Another potential way of enhancing CZA’s in vitro activity is to take advantage of enhanced susceptibility to antibiotics brought about by mutations that compromise CZA’s activity. A unique mutation (D179Y) within or proximal to KPC-3-Ω-loop is responsible for CZA resistance in KPC-producing CRE isolates (19). Intriguingly, the same mutation may restore meropenem susceptibility in some strains. However, the susceptibility of these mutant strains to meropenem is generally unsustainable and, as a result, the clinical relevance of this finding is unclear (23). Nevertheless, the combination of CZA plus meropenem may open a new avenue for the treatment of KPC-producing CRE infections. Other combination regimens containing antibiotics with mechanisms of action different from CZA may increase cumulative antibacterial activity through synergistic interactions by less clearly explained mechanisms.

Among in vitro studies, 22 were included in final data synthesis (Figure 1). The main features of all included studies are shown in Table 1. Among 22 studies, 5 reported data on Enterobacterales, 10 K. pneumoniae, 3 P. aeruginosa, 2 both K. pneumoniae and P. aeruginosa, and 1 both Enterobacterales and P. aeruginosa and resistance mechanisms of tested pathogens were quite heterogeneous (Table 2). In vitro synergy was examined by time-kill (n = 13), gradient diffusion (n = 3), E-test MIC:MIC (n = 4), checkerboard (n = 4), and pharmacokinetic/pharmacodynamic (n = 2) studies. In four of these studies, two different methods were used to analyze the in vitro interactions of antibiotic combinations (29, 31, 32, 35). In studies utilizing E-test MIC:MIC, gradient diffusion and checkerboard, the fractional inhibitory concentration index (FICI) was calculated to define synergy. Synergy was defined as FICI ≤ 0.50, additivity >0.50 to ≤1.00, indifference >1.00 to ≤4.00, and antagonism >4.00. In time-kill and pharmacokinetic/pharmacodynamic studies, reduction in bacterial load ≥2log₁₀ colony forming unit (CFU) as compared to most active single antibiotic was defined as synergy, while ≥2-log₁₀ increase in bacterial burden was defined as antagonism. In approximately half of the studies (n = 10), the number of repetitions of the experiments was not provided (26, 29–32, 36, 40, 43–45). The remainder repeated the experiments either 2 (24, 25, 33, 35, 37–39, 41) or ≥ 3 (27, 28, 34, 42) times. All in vitro synergy studies were between CZA and commercially available antibiotics, except for one study in which Idowu et al. analyzed the synergy between CZA and tobramycin-cyclam conjugate, which abrogates the antimicrobial effects of tobramycin but potentiates the full antimicrobial activity of the concomitant antibiotic (32).

Table 2 presents the in vitro synergy and antagonism rates of the combination regimens tested for CRE and MDR-PA. The synergy rate between CZA and aztreonam was >80% in all studies using metallo-β-lactamases-producing CRE and MDR-PA (25, 27, 34, 37, 43). Even though the results were quite heterogeneous as shown in Table 2, CZA plus meropenem or imipenem (30, 38, 40, 44) and CZA plus amikacin can show synergistic activity, particularly against CRE isolates (24, 29, 31, 38, 39). Similarly, synergy rates for the combination of CZA plus fosfomycin were highly variable against CRE isolates (26, 38, 41, 44) but very low against MDR-PA in one study (38). In other combinations, the synergy rates were generally low against CRE and MDR-PA isolates (Table 2). For CZA-resistant MDR-PA isolates, Montero et al. found high rate of synergy between colistin and CZA (39). Conversely, Shields et al. demonstrated significant antagonism between colistin and CZA against KPC-expressing CZA-susceptible Enterobacterales isolates (45).

After assessment of relevant articles, seven were included in narrative data synthesis. All studies except one from Canada were reported from the United States. Five studies included carbapenem-resistant Klebsiella pneumoniae (CRKP) isolates (3 KPC, 1 NDM and 1 NDM plus OXA-48 producers) and two assessed MDR-PA. The in vivo infection models used are as follows: Galleria mellonella survival model (n = 3), neutropenic mouse thigh infection model (n = 2), and neutropenic mouse pneumonia model (n = 2).

In a Galleria mellonella survival model, CZA (1.56/1.56 mg/kg) ensured 70% survival of MDR-PA-challenged larvae after 24 h, while CZA (1.56/1.56 mg/kg) plus tobramycin-cyclam (3.12 mg/kg) provided 100% survival after 24 h (32). In another Galleria mellonella survival model, human simulated regimens of CZA had better antibacterial activity when combined with either meropenem or amikacin against KPC-producing K. pneumoniae isolates (n = 2) with CZA MIC level of 8/4 mg/L (40). In contrast, these combination regimens did not have significant killing effect compared to single CZA regimen against K. pneumoniae isolates (n = 2) with lower (1/4 mg/L) MIC of CZA in virtue of excellent in vivo activity of CZA monotherapy (40). In the last Galleria mellonella survival model, CZA and polymyxin B were tested with concentrations of being four times MIC alone and in combination against KPC-producing K. pneumoniae isolates (n = 3). The CZA plus polymyxin B combination did not significantly improve larval survival in comparison with CZA alone against any strain (28). Likewise, mean survival times were similar in both groups.

In two neutropenic mouse thigh infection models, the authors demonstrated significant synergy between CZA plus aztreonam and CZA plus fosfomycin, respectively. Marshall et al. showed that as compared to CZA (32 mg/kg q8h) and aztreonam (32 mg/kg q8h) alone, CZA (32/8 mg/kg q8h) plus aztreonam (32 mg/kg q8h) combination reduces bacterial titers of an NDM-producing K. pneumoniae by 3.79 log₁₀ CFU and 2.08 log₁₀ CFU per thigh, respectively (22). Papp-Wallace et al. revealed that the human simulated doses of CZA and fosfomycin combination provides significant reduction in CFUs of CZA-susceptible MDR-PA isolates (n = 2) by approximately 5 logs, compared with the vehicle-treated control (46). Separate administration of CZA and fosfomycin reduced bacterial loads by approximately 1 log and 2 logs, respectively, compared with those of the vehicle-treated control (46).

Huang et al. performed a neutropenic mouse pneumonia model with inoculation of a KPC-producing K. pneumoniae isolate (31). At 24 h, mean bacterial lung concentrations (± the standard deviation) for the untreated control group, gentamicin, CZA, and gentamicin plus CZA combination were 8.97 ± 0.09, 4.64 ± 0.34, 5.95 ± 0.25, and 2.95 ± 1.07 log₁₀ CFU/lung, respectively. In a neutropenic murine lung infection model utilizing NDM, OXA-48 and CTX-M expressing K. pneumoniae strain (n = 1), the humanized dose of CZA monotherapy (2.5 g q8h) led to substantial in vivo killing activity (2.43 log₁₀ CFU reduction at 24 h), and consequently, no significant benefit was found compared to CZA monotherapy with CZA plus aztreonam (2 g every 6 h) or CZA plus tigecycline (100 mg q12h) (2.54 log₁₀ CFU reduction with CZA plus aztreonam and 2.15 log₁₀ CFU reduction with CZA plus tigecycline at 24 h) (47). Even with CZA plus aztreonam plus tigecycline combination, bacterial load can be reduced by 2.18 log₁₀ CFU at 24 h, which did not provide significant advantage over CZA monotherapy (47).

In summary, CZA plus fosfomycin and CZA plus aztreonam were shown to have more robust antibacterial activity than CZA alone in neutropenic mouse thigh infection models utilizing MDR-PA and NDM-producing K. pneumoniae, respectively. Intriguingly, in a neutropenic murine lung infection model, CZA alone showed excellent killing activity against NDM, OXA-48 and CTX-M-expressing K. pneumoniae, raising the clinical impact of NDM production on these infections into question (47). In another study, more effective antibacterial activity was shown with CZA plus meropenem or amikacin against KPC-producing K. pneumoniae strains if CZA MIC level is near to susceptibility breakpoint. For other combinations, no benefit of the combination regimens was demonstrated over CZA alone.

Of the 20 studies evaluated, 9 were retrospective multi-center cohort (48–56), 8 retrospective single-center cohort (15, 57–63), 1 prospective multi-center cohort (64), 1 prospective single-center cohort (65) and 1 multi-center case–control study (66). There were no RCTs which address combination therapy and monotherapy with CZA. The number of patients included ranged from 13 to 577. Only four studies assessed bloodstream infections alone and the rest included patients with different types of infections. These studies were conducted in various countries as follows: Spain (6), United States (4), Italy (4), Greece (1), Saudi Arabia (1), Brazil (1), India (1), China (1), Europe, Israel and Australia (1). As shown in Table 3, in vitro susceptibility of carbapenems was not tested in two studies (55, 60) and carbapenem non-susceptibility rate was <100% among isolates collected in four studies (49, 57, 64, 65). Similarly, CZA susceptibility rate was not documented in three studies (49, 50, 55), as well as it was found as 89.5% and 79% in two studies, respectively (57, 61). More than half of the studies had CRE infections, while six included CRKP infections (15, 52–54, 56, 60). MDR/XDR-PA infections were assessed only in one study (57) and two studies included mixed types of infections caused by CRE or carbapenem-resistant P. aeruginosa (CRPA) and CRE or MDR-PA, respectively (49, 51). The causative pathogens were susceptible to concomitantly administered antibiotic(s) in the vast majority of the studies, five studies did not document in vitro antimicrobial activity of partner antibiotics (48, 56, 58, 61, 66) and 70% of concomitant antibiotics had in vitro antimicrobial activity against the causative pathogens in one study (62). Intriguingly, none of the co-administered antibiotics were in vitro active against the identified pathogens in one study (53). Although median time between the onset of infection and administration at least one in vitro active antibiotic was not reported in 12 studies, it was generally quite long in other studies as shown in Table 3. Except four studies included only bloodstream infections, secondary bacteremia was frequent in other studies (Table 3). The definitions of investigated outcomes were highly heterogeneous and are portrayed in Table 4. In three studies, 30-day mortality rate was explored along with mortality rates at other time points (56, 63, 65). Since the majority of the studies included in this scoping review analyzed 30-day mortality, we preferred to report the 30-day mortality rates in these studies. All-cause mortality was investigated in all but one and ranged from 6.3% to 47.6% in patients treated with CZA monotherapy and from 0% to 44.0% in those receiving CZA combination regimens (Table 5). Only four studies performed a multivariable analysis while comparing monotherapy and combination therapies. In the study conducted by Zheng et al., CZA as a component of combination regimens was found to be associated with lower risk of mortality in comparison with CZA monotherapy (24.4% vs. 47.6%; HR, 0.167; 95 CI, 0.06–0.46) (53). In contrast, CZA combination regimens were associated with higher mortality in another study (27.8% vs. 11.8%; HR, 2.49; 95% CI, 1.10–5.63) (64). Alqahtani et al. reported no significant difference in mortality between CZA combination therapies and CZA monotherapy (27% vs. 16%; OR, 1.5; 95% CI, 0.61–3.64) (63). Lastly, in an Italian retrospective matched cohort study, there was no difference between CZA plus fosfomycin and CZA ± other antibiotic groups in terms of 30-day mortality (14.8% vs. 18.0%; HR, 0.72; 95% CI, 0.28–1.85) (56).

Among all studies, 14 assessed clinical cure or improvement of recruited patients by applying definitions shown in Table 4. The clinical cure rate ranged from 55.9% to 86.6% in the CZA monotherapy group and 44.8% to 100% in the CZA combination therapy group. Only the study showing lowest rate of clinical cure with CZA combination regimens executed multivariable analysis and reported that CZA combination therapy is significantly associated with lower clinical cure rate as compared to CZA monotherapy (44.8% vs. 62.5%; OR, 0.02; 95% CI, 0.01–0.38) (57). Only eight studies documented microbiological eradication rates in the CZA monotherapy and combination therapy groups (Table 5). While they were between 35.2 and 67.9% in the monotherapy groups, they ranged from 38.2% to 66.6% in the CZA combination therapy groups. In the study conducted by Oliva et al., blood culture negativity 72 h after treatment initiation for KPC-producing CRKP was higher in CZA plus fosfomycin arm than in CZA ± another antibiotic arm. However, there was no difference between the microbiological eradication rates of the two groups on the 7th and 14th days (56). Adverse event rates of monotherapy and combination therapies were reported in four studies and they seem to be quite similar between the two treatment groups in all of these studies (Table 5). Additionally, six studies explored the acute kidney injury (AKI) rates of CZA monotherapy and CZA combination regimens (Table 5). The AKI rates were numerically higher in CZA combination regimens in all but one study in which no AKI episode was detected in both groups (50).

One of the biggest issues of CZA treatment is the development of resistance during or following treatment. Except one study conducted by Ackley et al. (48), none of nine studies demonstrated a substantial difference between monotherapy and combination therapies in terms of development of CZA resistance during or following therapy (Table 5). In the study of Ackley et al., resistance development was not observed in any patients treated with CZA combination regimens, while 7.3% of monotherapy group developed resistance against CZA (48). Similarly, except the studies conducted by Ackley et al. and Oliva et al. (48, 56), five studies indicated numerically higher infection relapse rates in CZA combination therapy groups and no infection relapse was observed in one study (Table 5).