The Revival of Aztreonam in Combination with Avibactam against Metallo-β-Lactamase-Producing Gram-Negatives: A Systematic Review of In Vitro Studies and Clinical Cases

Infections caused by metallo-β-lactamase (MBL)-producing Enterobacterales and Pseudomonas are increasingly reported worldwide and are usually associated with high mortality rates (>30%). Neither standard therapy nor consensus for the management of these infections exist. Aztreonam, an old β-lactam antibiotic, is not hydrolyzed by MBLs. However, since many MBL-producing strains co-produce enzymes that could hydrolyze aztreonam (e.g., AmpC, ESBL), a robust β-lactamase inhibitor such as avibactam could be given as a partner drug. We performed a systematic review including 35 in vitro and 18 in vivo studies on the combination aztreonam + avibactam for infections sustained by MBL-producing Gram-negatives. In vitro data on 2209 Gram-negatives were available, showing the high antimicrobial activity of aztreonam (MIC ≤ 4 mg/L when combined with avibactam) in 80% of MBL-producing Enterobacterales, 85% of Stenotrophomonas and 6% of MBL-producing Pseudomonas. Clinical data were available for 94 patients: 83% of them had bloodstream infections. Clinical resolution within 30 days was reported in 80% of infected patients. Analyzing only patients with bloodstream infections (64 patients), death occurred in 19% of patients treated with aztreonam + ceftazidime/avibactam. The combination aztreonam + avibactam appears to be a promising option against MBL-producing bacteria (especially Enterobacterales, much less for Pseudomonas) while waiting for new antimicrobials.


Introduction
The global spread of metallo-β-lactamase-producing Gram-negatives (MBL-GN) is a serious cause of concern for public health. In particular, class B1 β-lactamases including Verona integron-encoded MBLs (VIM), imipenemases (IMP), and New Delhi MBLs (NDM), mostly carried by Enterobacterales and Pseudomonas aeruginosa, have now spread worldwide with multitudes of clinical variants [1]. The increase in population exchange at the global level and the prevalent plasmid-mediated nature, as well as the intestinal carriage of MBL-GN, contributed to the uncontrolled MBL-related resistance spread worldwide [2]. Invasive infections by MBL-GN are associated with high mortality rates (>30%), especially in the hospital setting when critically ill patients are involved [3,4]. MBL producers are mostly resistant to all β-lactams, including carbapenems and β-lactam/β-lactamase inhibitor combinations (BLBLICs). The optimal treatment of infections sustained by MBLproducing Enterobacterales or P. aeruginosa is not well defined, being associated with limited tibiotics 2021, 10
Only cumulative MIC values of ATM in combination with AVI or CZA were reported for 527 (23.8%) isolates; hence, the specific antimicrobial activity of ATM cannot
Only cumulative MIC values of ATM in combination with AVI or CZA were reported for 527 (23.8%) isolates; hence, the specific antimicrobial activity of ATM cannot be evaluated against these strains. However, among these isolates, mostly synergistic interactions have been reported for Enterobacterales, while P. aeruginosa isolates mainly did not show synergistic interactions (Table 1). Overall, when reported, MIC values ≤4 mg/L for ATM in combination with AVI or CZA have been described in 79.6% of MBL-producing Enterobacterales, 85.5% of S. maltophilia (few strains, only one report) and 6.2% of MBL-producing P. aeruginosa isolates.
The combination of CZA and ATM was administered as targeted therapy in nearly all cases (99%, 93/94), namely after demonstration of the synergistic activity of the association of the drugs, inactive when considered singularly. The only exception was a case of BSI by a presumptive MBL-producing strain: carbapenemase was not detected but suspicion was raised in light of medical history (previous treatments in India) and of the susceptibility profile of the K. pneumoniae isolate, but synergism was not demonstrated between CZA and ATM [56]. Outcomes' definitions were quite heterogeneous. No adverse event related to CZA plus ATM treatment was reported. With regard to clinical efficacy, clinical resolution within 30 days was achieved in almost four-fifths of patients (80%, 75/94). Early recurrence was described in four cases [56,58,64]: notably, in three patients [58,64], the same treatment was re-administered, obtaining resolution of infection except in a single subject expired owing to chronic lung transplant rejection [58]. In the case series by Shah and colleagues, two late recurrences (within 90 days of follow-up) were described as well [56], and another late recurrence (at 4-month of follow-up) was described in the other case series [61]. Death by all causes occurred in 15 patients in early follow-up (within 30 days) and in another 2 patients when considering longer durations of monitoring after completion of antibiotic courses.
The two cohorts allowed a meta-analytic approach to compare, in the context of BSIs, the combination of CZA and ATM versus other available targeted therapies, based on the administration of at least one active agent, considering 30-day mortality as the endpoint (Figure 3). In total, 64 patients in each group were evaluable: death occurred in 19% of patients receiving CZA plus ATM (12/64) and in 44% of patients not receiving the aforementioned combination (28/64). Therefore, CZA plus ATM was associated with a lower 30-day mortality risk: both under fixed effect and random effect models, the OR was equal to 0.30 (95% CI, 0.13-0.66), and no heterogeneity was detected (I 2 = 0%). Of note, in the comparator group, 30-day mortality crude rate was 58% (21/36) among patients being administered colistin-containing regimen and 25% (7/28) in subjects receiving regimens not including colistin (data not shown but abstracted from the primary studies).

In Vitro Studies
The quality assessment of the in vitro studies is reported in Supplementary Table S1. The large majority of articles seemed to fulfill most of the qualitative criteria set by the tool adapted for in vitro studies (https://jbi.global/critical-appraisal-tools, last accessed 22 June 2021). The included studies seem to be a reliable source to describe the antimicrobial

In Vitro Studies
The quality assessment of the in vitro studies is reported in Supplementary Table S1. The large majority of articles seemed to fulfill most of the qualitative criteria set by the tool adapted for in vitro studies (https://jbi.global/critical-appraisal-tools, last accessed 22 June 2021). The included studies seem to be a reliable source to describe the antimicrobial activity of the association between ATM and AVI or CZA against MBL-GN. Some studies do not provide a detailed description of isolation background and/or resistance determinants other than MBLs, since these aspects do not represent the central aim of the studies.

In Vivo Studies
The quality assessment of the clinical studies (case report, case series, cohort study) is reported in Supplementary Tables S2-S4. From a qualitative perspective, the large majority of articles appeared to meet most of the criteria established by the different tools, adopted for each type of study design (https://jbi.global/critical-appraisal-tools, last accessed 22 June 2021). Therefore, the included studies seem to be a trustworthy source to describe the real-life use of the association between CZA and ATM to address difficult-to-treat infection by MBL-producing strains.

Discussion
The treatment of MBL infection has been a difficult challenge for at least two decades [66]. MBLs are capable of hydrolyzing all β-lactams, with the exception of ATM. At any rate, due to the frequent co-production of other β-lactamases within MBL-producing Enterobacterales, ATM is active against just about 30% of these isolates [23]. The therapeutic choice usually relied on individual susceptibility profile, resorting to agents such as aminoglycosides, tetracyclines, fosfomycin, and polymyxins, without an established therapeutic consensus and facing many safety issues [1]. New commercially available BLBLICs are inactive against MBL producers, while cefiderocol and the repurposing of old agents (e.g., intravenous fosfomycin) have expanded the armamentarium of potentially effective drugs against highly resistant Gramnegative bacteria [67], but the therapeutic options remain limited. A tentative algorithm for the empirical treatment of severe infections likely due to MBL-producing strains has been suggested by Bassetti and colleagues: a pre-eminent role, in light of their good safety profile coherent with the one of the β-lactam class, is given to cefiderocol and to the association of CZA and ATM [68]. The latter may exploit the activity of ATM against MBL and the activity of CZA against other β-lactamases, often co-existing and able to hydrolyze ATM. The combination of CZA and ATM derives from the current unavailability in clinical practice of AVI and ATM in a single product formulation, currently in Phase III.
To sum up, this is the first review systematically describing the antimicrobial activity and the clinical use of ATM in combination with AVI or CZA against MBL-GN. In vitro studies have been conducted worldwide, with more than 2000 MBL-producing isolates tested. Overall, ATM in combination with AVI or CZA showed high antimicrobial activity against about 80% of MBL-producing Enterobacterales, mostly NDM producers, providing a >128-fold reduction in the MICs of ATM alone. The combination also showed high antimicrobial activity against isolates that co-produced other acquired resistant determinants, such as KPC and ESBLs. In line with these data, previously reported kinetic assays using purified protein extracts have demonstrated that AVI exerts potent activity against KPC, OXA-48, CTX-M-like and E. cloacae AmpC [69,70], protecting ATM by enzymatic hydrolysis related to these determinants. The combination was also highly active against 85% of S. maltophilia isolates and 6% of P. aeruginosa isolates. It is of note that almost all of P. aeruginosa isolates (>90%) showed MIC values ≥16 mg/L. These data provide impactful information to support clinical decisions about the use of ATM in combination with CZA when facing infections sustained by MBL-producing Enterobacterales and S. maltophilia, but not for MBL-producing P. aeruginosa. Notably, no MBL variants have been directly associated with increased MIC values for ATM in combination with AVI or CZA, nei-ther in Enterobacterales nor P. aeruginosa, but low antimicrobial activity seems to be more likely related to other parameters. Overall, hyperexpression of efflux systems and/or the presence of derepressed bla PDC variants, as well as the mutation in oprD, likely impact resistance to β-lactams in P. aeruginosa. In our search, in the case of P. aeruginosa isolates, an important role could be played by impermeability (porin loss), production of PDC variants, OXA enzymes (other than OXA-48) or hyperexpression of efflux systems. In particular, it is well-known that AVI may show a potent inhibitory activity against the basal AmpC produced by P. aeruginosa, even though extended-spectrum OXA enzymes can escape its wide spectrum of activity [71]. Selection of extended-spectrum OXA-2 or OXA-10 variants such as OXA-539, OXA-681 and OXA-14 has previously been associated with in vivo acquisition of high-level CZA resistance [72][73][74], while OXA-10 and OXA-18 enzymes have been associated with ATM's high MIC values [75]. Moreover, previously reported development of resistance to CZA during treatment of P. aeruginosa infections has mainly been associated with selection of variants of PDC-enzymes [38,73]. Notably, co-production of PDC and OXA determinants conferring resistance to ATM in P. aeruginosa isolates has also previously been reported [76].
ATM in combination with AVI or CZA counts for a low antimicrobial activity or not synergistic interactions against only 3% of Enterobacterales. About 50% of them presented specific amino acid insertion (12 bp duplications) in PBP3 determinants after residue 333 (YRIN or YRIK), providing the reduction in molecular affinity for ATM [17,40]. Structural analysis suggests that this insertion will impact the accessibility of ATM (and other β-lactam drugs) to the transpeptidase pocket of PBP3 [17]. This particular polymorphism of PBP3 was associated with high MIC values for ATM in combination with AVI only in E. coli isolates. For the other remaining isolates, co-production and/or hyperexpression of ESBL and AmpC-type determinants could have contributed to the high MIC values of ATM in combination with AVI. Notably, the association of CMY-42 and the YRIK insertion in PBP3 has been demonstrated to confer resistance to ATM-AVI in E. coli by mutagenesis experiments [77]. ATM-AVI showed low antimicrobial activity against 15% of S. maltophilia isolates, even if potential resistance mechanisms were not investigated. Importantly, very low antimicrobial activity has also been reported against E. anophelis, with all isolates showing MIC values >256 mg/L for ATM-AVI. Since only one report exists, very few data are known about the efficacy of ATM-AVI against this species, as well as the possible contribution of GOB and BlaB to the resistance profile. This species, although sporadically reported, represents a very difficult-to-treat opportunistic pathogen, being resistant to all new commercially available BLBLICs (in addition to ATM-AVI) and allowing very limited therapeutic options for the treatment of related infections.
The use of ATM in combination with CZA is currently considered a reasonable option for clinical use in the management of infections sustained by MBL producers. The paramount issue is that the optimum pharmacokinetic/pharmacodynamic (Pk/Pd) target for ATM in combination with CZA is still to be determined. According to in vitro models, optimal eradication and resistance suppression may be achieved by administering 8 g instead of 6 g per day of ATM (as continuous or 2-h infusion each 6 h of 2 g) plus 2.5 g each 8 h of CZA [78]. Monte Carlo simulations run from a PK analysis of clinical samples of MBL-producing isolates from highly comorbid patients demonstrate that standard dosage (CZA 2.5 g and ATM 2 g every 8 h, the most frequent according to our systematic review) fulfilled the time-dependent Pd targets for these agents [79]. However, pediatric patients seemed to well tolerate very high dosages: for instance, 150 mg/kg/day for CZA (calculated on the basis of ceftazidime component) [55,65] and 100 mg/kg/day [55] or 150 mg/Kg/day for ATM [65], which would translate to more than 10 g daily for each drug in normal weight adult subjects. Notably, for the phase 3 study of ATM plus AVI, Pk data supported the selection of a maintenance dose equal to 6 g of ATM and equal to 2 g of AVI, both divided into four administrations every 6 h, after a loading dose equal to 500/167 mg (ATM and AVI, respectively), in patients with creatinine clearance >50 mL/min, for the Phase III development program [10].
Beyond the not negligible Pk/Pd issues, the question is whether current data permit the suggestion of CZA plus ATM against MBL infections preferentially over other available options. In the prospective international cohort described by Falcone and colleagues, a propensity score-based matched analysis, allowing a minimal loss of initial information (only two not-matched patients from the inception cohort of 102 subjects), showed that CZA plus ATM was associated with a lower 30-day mortality rate compared with other active agents for the treatment of MBL-producing Enterobacterales causing BSI: hazard ratio 0.31 (95% CI 0.15-0.66) [47]. This value overlaps with the OR for mortality emerging from the meta-analytic results in the present study, favoring the combination of CZA and ATM over alternative targeted therapies. Of note, in the described casuistry of nearly 100 patients affected by MBL infection in a wide array of clinical scenarios, this combination demonstrated a very good safety profile (neither adverse events nor treatment discontinuations were registered) along with a high success rate as targeted treatment, both as first-line and salvage option.
The strength of the present work is rooted in the strict inclusion criteria, extensive literature search, granular description of clinical use of CZA plus ATM on a case-by-case basis and the large amounts of data regarding in vitro studies. However, this study presents some limitations. From a clinical standpoint, it was based only on observational studies, preeminently case series or case reports. Large prospective studies are urgently needed to better understand the in vivo efficacy of the ATM-AVI combination. Inevitably, the present study inherits their intrinsic limitations, specifically selection bias, impossibility to appropriately account for potential lurking variables, huge heterogeneity and subjectivity of outcomes' definitions as well as vast variability pertaining to CZA plus ATM use (indication, dose, duration, first-line or salvage treatment). The meta-analytic results comparing CZA plus ATM versus other targeted therapy against MBL-producing isolates (only Enterobacterales) responsible for BSI draw solely on two non-randomized studies, having small sample sizes and not originally conceived to weigh up different treatment options; therefore, no rock-solid evidence can be inferred on the superiority of one antibiotic regimen over another. Moreover, some in vitro studies provided cumulative data only, making it difficult (or impossible) to extrapolate data referring to specific MBL-producing isolates.
Notwithstanding, this study provides 'real-world data' that, if properly interpreted in the wider framework of the healthcare evidence ecosystem, may contribute to recommendations and guidelines when a higher source of information in the hierarchy of evidence is lacking. At any rate, further studies are needed to better define the clinical efficacy of CZA plus ATM, and only randomized clinical trials will provide high-quality evidence on the best therapeutic option for infections by MBL-producing strains.

Protocol and Registration
This systematic review was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta Analyses) guidelines [80]. The review protocol was registered at the Prospero international prospective register of systematic reviews (registration no. CRD42020220888).

Literature Search Strategy
By using the PubMed and the Embase databases, searches for relevant articles were performed with the following items: "(aztreonam) AND (avibactam)". Searches were limited to peer-reviewed articles published in English up to 31 October 2020. The search was updated to include further articles published until 31 May 2021. Moreover, the reference lists of reports identified by this search strategy were also hand-searched to select further relevant articles.

Study Selection
Two reviewers (LP and CM) independently screened the titles and abstracts to determine eligibility for full-text review. Inclusion criteria were the following: (i) studies published in full; (ii) as mentioned above, studies published in English language; (iii) in vivo and in vitro studies investigating the association between ATM and AVI (with or without other agents) against MBL producers under a clinical and microbiological standpoint, respectively. Studies were excluded if they were commentaries, editorials or review papers.

Data Extraction
After the initial screening, all potential eligible articles were independently reviewed using the full text by the same researchers for final inclusion. Two reviewers (CM, AEM) extracted relevant information from each included study by resorting to a standardized data extraction sheet and then proceeding to cross-check the results. The extracted data included: publication year, study period, geographical setting, relevant clinical and microbiological information. In detail, the former were: study design, sample size, age and gender of study participant(s), type(s) of infection, causative agent(s), resistance mechanism(s), antimicrobial susceptibility testing (AST) profile, therapeutic regimen(s), efficacy and safety outcomes as reported by each study. The latter were, beyond general features of the included studies: number of isolates, MBL determinants, minimum inhibitory concentration (MIC) range, antimicrobial agents tested (ATM plus CZA or AVI), methods to evaluate interactions, other resistant determinants. Since no susceptibility breakpoint for ATM-AVI exists, a current EUCAST Pk/Pd non-species-related susceptibility breakpoint for ATM (4 mg/L) has been arbitrarily taken as reference to assess low (≤4 mg/L) and high (>4 mg/L) antimicrobial activity (https://www.eucast.org/fileadmin/src/media/ PDFs/EUCAST_files/Breakpoint_tables/v_11.0_Breakpoint_Tables.pdf; last accessed on 4 July 2021).

Strategy for Data Synthesis
A qualitative assessment and synthesis of the main characteristics of included studies was carried out. Key findings were tabulated. Meta-analyses were performed on comparable outcomes measured by at least two studies. For dichotomous data, a pooled odds ratio (OR) as a summary estimate was calculated with its 95% confidence interval (CI). A 2-sided p-value less than 0.05 and a 95% CI that did not cross 1 (OR) were considered statistically significant. Meta-analyses were carried out by using both the fixed effect and random effects DerSimonian and Laird methods, compared graphically with forest plots. Heterogeneity between studies was gauged by resorting to I 2 statistics. Statistical analyses were conducted by using the statistical software R, version 1.3.1093 (RStudio Team) and package 'metafor'.

Quality Assessment
Included studies were critically appraised through the tools of the Joanna Briggs Institute, according to the study design (https://jbi.global/critical-appraisal-tools, last accessed on 22 June 2021). The tool named Checklist for Analytical Cross Sectional Studies was adapted for the quality assessment of in vitro studies. Definitions of case report(s), case series and cohort studies were predicated on the work of Dekkers and colleagues [81]. Since summary quality scores may yield misleading results, a global judgement on the methodological quality of included studies was considered more appropriate [82,83].

Ethics
Ethics committee approval was not a prerequisite in this case because the project used anonymized data as pertaining to clinical information and original studies that had already received proper institutional review board approval.

Conclusions
In this review, we resumed at large the in vivo and in vitro activity of the combination ATM-AVI against MBL-GN. Taken together, these data suggest that ATM in combination with AVI or CZA is an important therapeutic option against MBL-producing Enterobacterales, with very few isolates showing high MIC values and mostly providing a favorable outcome in treated patients. Importantly, the presence of ceftazidime has been demonstrated to not affect the in vitro antimicrobial activity of the combination ATM-AVI in Enterobacterales [45]. However, MBL-producing P. aeruginosa remains an important unsolved issue, and treatment alternatives for related infections mainly rely on colistin, (±fosfomycin) or cefiderocol. Accordingly, the use of the ATM-CZA combination in the treatment of infections sustained by MBL-producing P. aeruginosa remains to be elucidated, since very few data are available. This point represents a serious cause of concern, and new antimicrobial options against MBL-producing P. aeruginosa are urgently needed.
Waiting for the approval of the fixed combination between AVI and ATM as a single product formulation, the intriguing synergy involving the two drugs may be exploited in the context of clinical use of CZA and ATM association as a valid therapeutic strategy. Nevertheless, the optimal dosing strategy remains to be elucidated, and another challenge to take into account is the current unavailability of automated susceptibility testing for this combination.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/antibiotics10081012/s1, Table S1: Quality assessment of in vitro studies; Table S2: Quality assessment of clinical studies: case(s) report; Table S3: Quality assessment of clinical studies: case series; Table S4: Quality assessment of clinical studies: cohort study; Figure S1: Prisma 2020 checklist.