Indian Society of Critical Care Medicine Guidelines on Evidence-based Recommendations for the Treatment of Carbapenem-resistant Enterobacterales, Carbapenem-resistant Acinetobacter baumannii, and Carbapenem-resistant Pseudomonas aeruginosa in Critically Ill Patients: A GRADE-guided Approach

Abstract

Background

Carbapenem-resistant Enterobacterales (CRE), Acinetobacter baumannii (CRAB), and Pseudomonas aeruginosa (CRPA) are major antimicrobial resistance (AMR) threats in Indian intensive care units, with mortality rates often exceeding 40–70%. Despite the clinical burden, there is a lack of standardized treatment guidelines. This study aimed to develop Grading of Recommendations Assessment, Development and Evaluation (GRADE)-based recommendations for managing carbapenem-resistant gram-negative (CRGN) infections in critically ill patients.

Methodology

A systematic review was conducted following the GRADE framework. PubMed was searched through October 2024 for studies comparing polymyxin B/colistin with alternative regimens for CRE, CRAB, and CRPA, and sulbactam-based therapy vs comparators for CRAB. Meta-analyses were performed using RevMan. Evidence of certainty was graded, and recommendations were formulated through structured expert consensus. Outcomes included clinical cure, microbiological eradication, mortality, and adverse events (AEs).

Results

Fifty-two studies were included: 36 evaluating polymyxin/colistin and 16 evaluating sulbactam. For CRE, β-lactam/β-lactamase inhibitor (BL/BLI) combinations achieved higher cure rates (58.7 vs 48.9%), fewer AEs (10.3 vs 13.3%), and trends toward reduced mortality compared with polymyxin-based regimens, leading to a strong recommendation for non-polymyxin therapy (very low certainty). For CRAB, sulbactam-based regimens, particularly sulbactam-durlobactam, showed superior cure rates (61.9 vs 39.7%) and reduced nephrotoxicity (13 vs 38%), supporting a strong recommendation (very low certainty). For CRPA, efficacy appeared comparable, but polymyxin therapy caused higher toxicity (42.1 vs 17.9%), resulting in a weak recommendation for non-polymyxin options (very low certainty).

Conclusion

These findings provide GRADE-based treatment recommendations for carbapenem-resistant infections in Indian ICUs. Non-polymyxin regimens are preferred for CRE and CRPA, while sulbactam-based therapy is favored for CRAB. This shift away from polymyxin use may improve outcomes and reduce toxicity, though robust randomized trials are urgently needed.

Introduction

India stands at the forefront of the global antimicrobial resistance (AMR) crisis, with multidrug-resistant (MDR) gram-negative organisms posing an escalating threat, particularly among critically ill patients. Among these, carbapenem-resistant Enterobacterales (CRE), carbapenem-resistant Acinetobacter baumannii (CRAB), and carbapenem-resistant Pseudomonas aeruginosa (CRPA) represent the most challenging pathogens encountered in clinical practice.¹ These organisms have been designated as “critical priority pathogens” by the World Health Organization (WHO) due to their resistance to carbapenems and other last-resort antibiotics, and their association with severe, invasive infections and high mortality rates in intensive care settings.² India reported over 1 million cases of carbapenem-resistant gram-negative (CRGN) infections in 2019—the highest reported burden globally. Despite the availability of effective antimicrobials, only 7.8% of patients received appropriate therapy, reflecting a critical gap in antimicrobial stewardship and timely clinical decision-making.³ Moreover, MDR-gram-negative bacilli (GNB) are no longer confined to hospital settings; they are now increasingly implicated in community-acquired infections. Predominant pathogens include Escherichia coli, Klebsiella pneumoniae, Acinetobacter spp., and Pseudomonas aeruginosa.^4,5 Surveillance from Indian tertiary care centers indicates that up to 35% of GNB isolated from blood cultures are MDR, with increasing prevalence over time. In India, NDM carbapenemases account for 32.5–56% of carbapenem-resistant isolates, while OXA-48-like enzymes comprise 24–34%. In intensive care units (ICUs), CRE and CRAB together constitute over 90% of isolates, largely driven by resistance mechanisms such as NDM and OXA-48-like carbapenemases.^5–9 These infections are prevalent among high-risk populations—including critically ill ICU patients, neonates, the elderly, and immunocompromised individuals—where they are associated with rapid clinical deterioration and high mortality.^5,10 A recent study from South India reported 41% mortality in CRGNB infections, rising to 72% with septic shock.¹¹ In ICU patients, CRE colonization was linked to higher in-hospital mortality (27%), while CRAB-related bloodstream and device-associated infections showed mortality rates exceeding 20%.¹² These clinical consequences of MDR gram-negative infections have a significant clinical impact, leading to higher rates of persistent bacteremia, septic shock, multiorgan failure, and in-hospital mortality compared to susceptible infections. An epidemiological study from Indian ICUs reported mortality rates exceeding 69% in patients with MDR infections, highlighting their severity in critical care. More broadly, India bears one of the highest infectious disease mortality burdens globally, with a crude mortality rate of nearly 417 per 1,00,000 population—driven in large part by the rise of MDR organisms.¹³ The high prevalence of healthcare-associated MDR infections also contributes to prolonged hospitalization and significantly increased healthcare costs.

Recent post-COVID modeling studies highlight a rising global and regional burden of AMR. In 2021, AMR was directly responsible for 1.27 million deaths and associated with ~4.95 million deaths worldwide, with South Asia bearing a substantial share. In India, AMR-attributable deaths are estimated to be up to 267,000 annually, particularly affecting critically ill populations. Post-pandemic data further indicate increased ICU mortality from CRGN infections, likely driven by healthcare disruptions, delayed diagnosis, and widespread empirical antibiotic use during COVID-19.^14–16

India lacks a national strategy for managing MDR gram-negative infections. Conditions such as CRE, CRAB, and CRPA continue to be managed through institution-specific protocols or adaptations of international guidelines, often without the methodological rigor of frameworks like the Grading of Recommendations Assessment, Development and Evaluation (GRADE). This absence of standardized national guidance is further complicated by high rates of antimicrobial misuse, inconsistent infection control practices, limited access to advanced antimicrobials, and a scarcity of real-time, region-specific susceptibility data.^17,18

These MDR organisms are currently being treated with a colistin/polymyxin B-based regimen. However, the continued reliance on polymyxins is driven largely by their historical availability, widespread familiarity, and cost considerations, particularly in resource-limited settings, rather than clear evidence of superior clinical outcomes. With the introduction of novel antibiotics with efficacy against carbapenem-resistant gram-negative pathogens, there is an increasing need to generate robust comparative evidence evaluating these agents against polymyxin-based regimens, particularly with respect to patient-centered outcomes such as mortality, clinical cure, and safety.

Considering the rising incidence, high mortality, and complex treatment challenges associated with MDR gram-negative infections in India, there is a critical need for rigorous, evidence-based therapeutic guidance tailored to the local clinical landscape. This manuscript aims to bridge that gap by systematically evaluating the current evidence on polymyxin B/colistin-based regimens for the management of CRE, CRAB, and CRPA infections, as well as sulbactam-based therapy for CRAB, using the GRADE framework. Through a structured synthesis of available data, this review seeks to support informed, guideline-driven decision-making in the care of critically ill patients and to lay the groundwork for the development of standardized, context-relevant treatment protocols in India. Importantly, this guideline prioritizes patient-centered outcomes such as mortality, clinical cure, and safety over microbiological endpoints alone, thereby ensuring a clinically meaningful and practice-oriented approach to therapeutic decision-making.

Methodology

Formulation of Population, Intervention, Comparator, Outcomes (PICO) Question

Indian Society of Critical Care Medicine (ISCCM) convened a task force of domain experts to systematically analyze the current evidence on treatment of carbapenem-resistant gram-negative infections, utilizing the GRADE framework for guideline recommendations.

Following the GRADE approach, we formulated clearly defined clinical questions using the PICO framework, with a focus on patient-centered outcomes. These questions were designed to address key areas of therapeutic uncertainty in the management of CRGN infections in critically ill patients.

The specific clinical questions formulated were as follows:

• In critically ill patients with CRE infections, how does polymyxin B/colistin-based therapy compare to non-polymyxin B/non-colistin regimens?

• In critically ill patients with CRAB infections,

  • – How does polymyxin B/colistin-based therapy compare to non-polymyxin B/non-colistin regimens?

  • – How does sulbactam-based therapy compare to non-sulbactam-based therapy?

• In critically ill patients with CRPA infections, how does polymyxin B/colistin-based therapy compare to non-polymyxin B/non-colistin regimens?

Each question was addressed by evaluating the impact of the intervention on patient-centered outcomes, including clinical cure, microbiological eradication, mortality, and adverse events (AEs). Specifically, indirectness was assessed when outcomes were derived from mixed infection populations or when the distribution of infection sites differed across studies. The certainty of evidence supporting each comparison was assessed using the GRADE methodology to ensure methodological rigor and transparency in recommendation development. These guidelines are intended to inform clinical decision-making for infectious disease specialists, intensivists, microbiologists, and other stakeholders managing MDR-GNB infections in Indian ICUs.

Search Strategy and Study Selection

A systematic literature search was conducted across PubMed (up to October 30, 2024) to identify studies evaluating polymyxin B/colistin-based therapy vs non-polymyxin regimens for CRE, CRAB, and CRPA infections and sulbactam-based therapy vs alternatives for CRAB infections. Search strategies combined MeSH terms and keywords (“Polymyxin B,” “Colistin,” “Sulbactam,” “CRE/CRAB/CRPA,” “clinical cure,” “mortality,” microbiological cure, bacterial eradication, and AEs) using Boolean operators. Filters restricted results to published studies between January 2000 and October 2024, prioritizing randomized trials, cohort studies, and case-control designs. Systematic reviews and meta-analyses (SRMAs) from PubMed search were screened to identify relevant randomized controlled trials (RCTs) or observational studies matching the predefined PICO criteria. Eligible primary studies cited within these SRMAs were retrieved and assessed independently. Initial searches yielded 4,175 records (3,992 records for polymyxin/colistin and 183 records for sulbactam-based therapy. After removing duplicates, 2,754 unique records underwent title and abstract screening. Of these, 49 full-text articles related to polymyxin/colistin-based therapy and 44 full-text articles related to sulbactam-based therapy were assessed for eligibility. Following the full-text review, 52 studies were included in the final analysis. This comprised 36 studies evaluating polymyxin/colistin-based therapy for CRE, CRAB, or CRPA infections, including 31 observational studies and 5 RCTs, and 16 studies evaluating sulbactam-based therapy for CRAB infections, including 11 observational studies and 5 RCTs.

Two independent reviewers screened the titles and abstracts, followed by full-text screening to assess eligibility. Disagreements were resolved through discussion or consultation with a third reviewer. Inclusion criteria were RCTs, cohort studies, and case-control studies involving critically ill patients with CRE, CRAB, or CRPA infections. Studies without clinical outcome data or inappropriate comparator arms were excluded.

Data Extraction and Management

Data from included studies were extracted independently by two reviewers using a standardized data sheet developed in Microsoft Excel (version 2021). Extracted data included study design, population characteristics, type of infection, intervention and comparator details, clinical and microbiological outcomes, mortality rates, and AEs. Discrepancies were resolved through consensus.

Risk of Bias (RoB) Assessment

The RoB for RCTs was assessed using the Cochrane RoB 2 tool. This tool evaluates five key domains: (1) bias arising from the randomization process, (2) bias due to deviations from intended interventions, (3) bias due to missing outcome data, (4) bias in measurement of the outcome, and (5) bias in selection of the reported result. Each domain was rated as “low risk,” “some concerns,” or “high risk” of bias. Two reviewers assessed the studies independently, with disagreements resolved through discussion or a third-party adjudicator.¹⁹

For observational studies, we employed the Newcastle–Ottawa Scale (NOS), which evaluates studies across three domains: Selection of study groups, comparability of cohorts, and outcome assessment. Each study was rated on a scale of 0–9 stars, with higher scores indicating a lower RoB.²⁰ As with RCTs, assessments were conducted independently by two reviewers, and any discrepancies were resolved by consensus.

Evidence Synthesis and Statistical Analysis

Meta-analyses were performed where applicable using RevMan software (Version 5.4.1). Effect estimates were calculated based on the outcome type: Odds ratio (OR) for clinical cure rate, microbiological cure, mortality rate, and safety outcomes. Each estimate was reported with a 95% confidence interval (CI). A random-effects model was applied due to anticipated clinical and methodological heterogeneity. Statistical heterogeneity was assessed using the I² statistic.

Forest Plot Generation

Forest plots were generated to visually represent effect sizes and CIs for patient-centered outcomes. These were created using RevMan (version 5.4.1).²¹

Funnel Plot and Publication Bias

Funnel plots were constructed using RevMan (Version 5.4.1) to assess potential publication bias. Asymmetry in the plots was interpreted cautiously due to the small number of studies in some analyses.

Certainty of Evidence and Summary of Findings (SoF) Table

The certainty of evidence for each outcome was assessed using the GRADE approach, facilitated by GRADEproGDT software.²² Domains evaluated included RoB, inconsistency, indirectness, imprecision, and publication bias. Summary of Findings tables were generated to present the certainty of evidence and magnitude of effect for key outcomes.

Formulation of Recommendations

Recommendations were developed using the GRADE Evidence-to-Decision (EtD) framework. A panel of intensivists independently reviewed the evidence and formulated recommendations based on the balance of benefits and harms, quality of evidence, feasibility, and clinical applicability. Where consensus was needed, it was achieved through a panel discussion. Recommendations were classified as “Strong” or “Weak” in accordance with GRADE guidance. In selected situations, strong recommendations were issued despite low or very low certainty of evidence, particularly where there was a clear imperative for harm avoidance, strong biological plausibility, and consistent direction of effect across available studies. This approach is consistent with GRADE guidance, which permits strong recommendations under such conditions when the anticipated benefits outweigh potential harms, and the consequences of inaction may be substantial.²³

Results

Guideline Question

In critically ill patients with CRE infections, should polymyxin B/colistin-based regimens be used instead of non-polymyxin B/non-colistin therapies [e.g., novel BL/BLI combinations such as ceftazidime-avibactam (CZA)] to improve clinical outcomes such as clinical cure rates, microbiological eradication, mortality, and safety?

Clinical Cure Rate

Non-polymyxin agents, particularly CZA and other novel BL/BLIs, consistently demonstrate higher clinical cure rates compared to polymyxin-based therapies with 59.67% (148/248) vs 28.5% (57/200) with OR 0.34 (95% CI: 0.22–0.53). This corresponds to approximately 66% lower odds of clinical cure failure, favoring non-polymyxin-based therapy (particularly CZA-AVI) over colistin/polymyxin regimens. Four observational studies comparing CZA to colistin in critically ill patients with CRE infections, CZA-based regimens reported a higher clinical cure rate [53.5% (91/170) vs 23.2% (45/194)],^24–27 and one observational study comparing CZA to polymyxin B shows [73.1% (57/78) vs 46.2% (12/26)] (Fig. 1).²⁸

Fig. 1.

Forest plot showing the clinical cure rate in critically ill patients with carbapenem-resistant enterobacterales infections treated with polymyxin B/colistin-based regimens compared with non-polymyxin B/non-colistin regimens from observational studies

CI, confidence interval; M-H, Mantel-Hanszael

Microbiological Cure Rate

Non-colistin/polymyxin-based therapies achieved a higher microbiological cure rate (66.9%; 113/169) compared to colistin/polymyxin-based regimens (51.0%; 50/98). The pooled OR was 0.43 (95% CI: 0.19–0.94; p = 0.03), indicating a significant advantage for non-polymyxin therapies. This reflects approximately a 57% reduction in the odds of microbiological cure failure, favoring non-polymyxin-based regimens. In a pooled cohort of 216 patients, microbiological cure was achieved in 58.7% (74/126) of patients treated with CZA, compared to 48.9% (44/90) with colistin (OR: 0.43; 95% CI: 0.16–1.16).^24,25,27 Similarly, in a single-center study (n = 51), CZA demonstrated a cure rate of 90.7% (39/43) vs 75.0% (6/8) with polymyxin (OR: 0.31; 95% CI: 0.05–2.06).²⁸ While statistical significance was not reached due to small sample sizes, the magnitude and direction of effect consistently favor CZA. These agents benefit from more predictable pharmacokinetics and better tissue penetration, which likely contribute to the superior microbial clearance seen in these patients (Fig. 2).

Fig. 2.

Forest plot showing the microbiological cure rate in critically ill patients with CRE infections receiving polymyxin B/colistin-based vs non-polymyxin B/non-colistin therapies from observational studies

CI, confidence interval; CRE, carbapenem-resistant enterobacterales; M-H, Mantel-Hanszael

Mortality Rate

Mortality data further strengthen the clinical rationale favoring CZA over polymyxin-based regimens. Mortality was lower with non-colistin/polymyxin–based regimens (29.5%; 61/207) compared to colistin/polymyxin-based therapies (37.5%; 69/184).

The pooled OR was 1.82 (95% CI: 0.88–3.73; p = 0.10), corresponding to approximately 82% higher odds of death with colistin/polymyxin-based therapy. In pooled 3 non-randomized studies (n = 287), mortality was lower with CZA (30.2%) compared to colistin (36.1%), though not statistically significant (OR 1.75 (95% CI: 0.62–4.92).^24,26,27 A separate observational study (n = 104) reported mortality of 28.2% with CZA vs 46.2% with polymyxin (OR 2.18 (95% CI: 0.87–5.45).²⁸ Although statistical significance was not reached, the consistent trend toward lower mortality across studies, together with the well-established toxicity profile of polymyxins, supports a clinical preference for CZA-based regimens where appropriate. Despite low certainty of evidence, the consistent direction of effect across studies aligns with CZA's superior microbiological efficacy and favorable safety profile, supporting its preferential use in the treatment of MDR gram-negative infections where susceptible (Fig. 3).

Fig. 3.

Forest plot showing mortality rates in critically ill patients with CRE infections treated with polymyxin B/colistin-based regimens vs non-polymyxin B/non-colistin regimens in observational studies

CI, confidence interval; CRE, carbapenem-resistant enterobacterales; M-H, Mantel-Hanszael

Adverse Events Rate

Adverse events (AE) were less frequent with non-colistin/polymyxin-based regimens (8.7%; 17/195) compared to colistin/polymyxin-based therapies (15.3%; 20/131). The pooled OR was 2.11 (95% CI: 1.01–4.40; p = 0.05), indicating more than twofold higher odds of AEs with colistin/polymyxin-based therapy. In a pooled analysis of three non-randomized studies (n = 222), AEs were reported in 13.3% of colistin-treated patients and 10.3% of those receiving CZA (OR: 1.48; 95% CI: 0.60–3.62), with no statistically significant difference.^24,26,27 However, the certainty of evidence was rated very low due to RoB, imprecision, and suspected publication bias. Conversely, a single study (n = 104) reported a significantly higher AE rate with CZA (23.1%) compared to polymyxin (6.4%), with an OR of 4.38 (95% CI: 1.21–15.84) (Fig. 4).²⁸

Fig. 4.

Forest plot showing adverse event rates among critically ill patients with CRE infections treated with polymyxin B/colistin vs non-polymyxin B/non-colistin therapy in observational studies

CI, confidence interval; CRE, carbapenem-resistant enterobacterales; M-H, Mantel-Hanszael

Strength of Recommendation

Strong recommendation to use non-polymyxin-based therapy (Ceftazidime Avibactam or other novel BL/BLI combination) instead of polymyxin B/colistin-based therapy in critically ill patients with CRE.

This recommendation is primarily driven by consistently higher clinical cure rates and a more favorable safety profile with non-polymyxin regimens. While mortality differences were not statistically significant, the direction of effect generally favored non-polymyxin therapy. Importantly, divergence between RCTs (favoring non-colistin therapy) and observational studies (often showing neutral effects) likely reflects methodological limitations inherent to observational data, including confounding by indication and survivor bias, where more severely ill patients are preferentially treated with polymyxins.

Certainty of evidence: Very low (Tables 5 to 4)

Rationale for Strong Recommendation for Use of Novel BL/BLI

Significant results towards increasing clinical cure and microbiological cure, and a trend towards decreasing mortality and increasing safety in comparison to polymyxin-based therapy.

Guideline Question

In critically ill patients with CRAB infections should polymyxin B/colistin-based regimens be preferred over non-polymyxin B/non-colistin therapies to improve clinical outcomes such as clinical cure, microbiological eradication, mortality, and safety?

Clinical Cure Rate

Clinical cure outcomes were evaluated across four RCTs (n = 255) involving patients with Acinetobacter infections. In the pooled analysis, clinical cure was achieved in 39.7% of patients treated with colistin-containing regimens and 65.3% of those receiving non-colistin-based therapies (OR: 0.35; 95% CI: 0.20–0.61; p < 0.001). This indicates that patients receiving non-colistin-based therapies had approximately a 65% greater likelihood of achieving clinical cure, favoring non-colistin-based regimens over colistin-containing therapies. However, the certainty of evidence was rated very low, downgraded due to methodological limitations, small sample sizes, and suspected publication bias.

Among these trials, two RCTs (n = 204) compared colistin + imipenem–cilastatin vs sulbactam–durlobactam + imipenem–cilastatin, reporting clinical cure in 38.1% (40/105) and 62.6% (62/99) of patients, respectively (OR: 0.37; 95% CI: 0.21–0.65).^29,30 One small RCT (n = 28) comparing colistin vs ampicillin–sulbactam showed cure in 60.0% (9/15) vs 69.2% (9/13) (OR: 0.67; 95% CI: 0.14–3.19).³¹ Another trial (n = 23) comparing colistin + levofloxacin vs levofloxacin + high-dose sulbactam infusion reported cure in 27.3 (3/11) vs 83.3% (10/12) (OR: 0.67; 95% CI: 0.14–3.19).³² Although some individual comparisons numerically favored colistin, the pooled data showed better outcomes with non-colistin regimens. Limitations included open-label design, small sample sizes, and lack of adjustment for confounding, contributing to very low certainty of evidence (Fig. 5).

Fig. 5.

Forest plot showing the clinical cure rate in critically ill patients with carbapenem-resistant Acinetobacter baumanii infections treated with polymyxin B/colistin-based regimens compared with non-polymyxin B/non-colistin regimens from RCTs

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; M-H, Mantel-Hanszae; RCTs, randomized controlled trials

Observational data were more extensive but yielded inconsistent results. In a pooled analysis of 12 non-randomized observational studies (n = 951), clinical cure was achieved in 51.8% of patients receiving non-colistin/non-polymyxin-based regimens and 48.7% of those treated with colistin or polymyxins (OR: 0.86; 95% CI: 0.44–1.68; p = 0.62), with no statistically significant difference. This indicates that treatment with colistin/polymyxin-based regimens was associated with a 14% decrease in the odds of achieving clinical cure compared to non-colistin/non-polymyxin-based therapies, favoring non-colistin/non-polymyxin-based regimens. The overall certainty of evidence was again rated very low, owing to RoB, imprecision, and suspected publication bias.

Two observational studies (n = 155) compared colistin vs imipenem–cilastatin, with cure rates of 45.7% (37/81) and 68.9% (51/74), respectively (OR: 0.47; 95% CI: 0.00–1,199.37).³³ A single study (n = 107) comparing colistin monotherapy vs non-colistin-based combinations showed cure in 30.6% (11/36) vs 42.3% (30/71) of patients (OR: 0.60; 95% CI: 0.26–1.41).³⁴ One study (n = 33) compared polymyxins vs doxycycline, reporting higher cure with doxycycline 75.9% (22/29) vs 50.0% (2/4); (OR: 3.14; 95% CI: 0.37–26.62).³⁵ Another (n = 102) compared polymyxins + tigecycline vs tigecycline alone, with cure rates of 21.6% (11/51) vs 27.5% (14/51), respectively (OR: 0.73; 95% CI: 0.29–1.80).³⁶

Three observational studies (n = 180) comparing colistin and tigecycline reported similar cure rates 48.3% (57/118) vs 45.2% (28/62); OR: 1.14; 95% CI: 0.88–1.46, with no significant difference.^37–40 Two studies (n = 215) comparing colistin vs cefiderocol found numerically higher cure with cefiderocol 70.0% (77/110) vs colistin 54.3% (57/105) (OR: 0.52; 95% CI: 0.02–11.07), although results were not statistically significant.^44,45 In a single study (n = 98) comparing colistin vs ampicillin–sulbactam, cure was achieved in Colistin 47.0% (31/66) vs ampicillin–sulbactam 56.3% (18/32) of patients (OR: 0.69; 95% CI: 0.29–1.61).⁴⁶

A single observational study (n = 61) comparing polymyxins with other antibiotics reported a significantly higher cure rate in the polymyxins 95.7% (22/23) vs 23.7% (9/38); with (OR: 70.89; 95% CI: 8.35–601.92).⁴⁴ Despite statistical significance, the study was at high RoB and did not specify comparator details, limiting confidence in the results (Fig. 6).

Fig. 6.

Forest plot showing the clinical cure rate in critically ill patients with carbapenem-resistant Acinetobacter baumanii infections treated with polymyxin B/colistin-based regimens compared with non-polymyxin B/non-colistin regimens from observational studies

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; IV, independent variable; a, Hartung-Knapp-Sidik-Jonkman method

Microbiological Cure

Microbiological cure data were limited, with only one RCT (n = 28) reporting this outcome in patients with Acinetobacter infections. This RCT compared colistin vs ampicillin–sulbactam and reported microbiological cure in 66.7% (10/15) of patients receiving colistin and 61.5% (8/13) in those on ampicillin–sulbactam (OR: 1.25; 95% CI: 0.27–5.89; p = 0.78).³¹ This indicates that patients treated with colistin had 25% higher odds of achieving microbiological cure compared to those receiving ampicillin–sulbactam. Although the numerical difference was small and not statistically significant, the certainty of evidence was rated high, as the study had no major concerns regarding bias, indirectness, or imprecision, despite its limited sample size (Fig. 7).

Fig. 7.

Forest plot showing the microbiological cure rate in carbapenem-resistant Acinetobacter baumanii infections treated with polymyxin B/colistin-based vs non-polymyxin B/non-colistin regimens from RCTs

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; M-H, Mantel-Hanszael; RCTs, randomized controlled trials

Observational data were more extensive, though less consistent. In a pooled analysis of 12 non-randomized studies (n = 792), microbiological cure was achieved in 210/408 (51.5%) of patients treated with colistin or polymyxins, compared to 173/384 (45.1%) of those receiving non-colistin-based therapies (OR: 1.49; 95% CI: 1.10–2.01; p = 0.01). This indicates that patients receiving colistin/polymyxin-based regimens had 49% higher odds of achieving microbiological cure compared to those on non-colistin-based therapies, with a statistically significant difference, thereby favoring colistin/polymyxin-based therapy. However, despite statistical significance, the overall certainty of evidence remains low due to the non-randomized study designs and potential residual confounding.

Across individual comparisons, results varied. In a single-center study (n = 35), colistin showed a numerically higher cure rate than imipenem–cilastatin [66.7% (14/21) vs 50.0% (7/14); OR: 2.0; 95% CI: 0.5–8.0].³⁸ Similarly, in the colistin vs tigecycline comparison across two studies (n = 125), cure was achieved in 38% (30/79) patients receiving colistin and only 4% (2/46) receiving tigecycline (OR: 10.27; 95% CI: 2.61–40.42), suggesting a significant benefit with colistin, though evidence certainty remained very low.^39,40 Similarly, a study (n = 50) comparing colistin vs ampicillin–sulbactam reported better outcomes with colistin [48.5% (16/33) vs 17.6% (3/17); OR: 4.39; 95% CI: 1.06–18.20].⁴⁶

Comparisons of polymyxin vs doxycycline (1 observational study; n = 24) showed microbiological cure in 81.0% (14/21) of patients receiving polymyxin and 66.7% (2/3) on doxycycline (OR: 2.13; 95% CI: 0.15–29.66) and a observational study (n = 78), colistin monotherapy vs non-colistin-based combinations showed a cure of 69% (20/29) vs 83% (49/59); (OR: 0.45; 95% CI: 0.16–1.28), respectively.^34,35 In colistin vs tobramycin (1 observational study; n = 64), cure was reported in 50.0% (16/32) and 56.3% (18/32) of patients, respectively (OR: 0.78; 95% CI: 0.29–2.08).⁴⁸ In contrast, in two observational studies comparing colistin vs cefiderocol (n = 215), microbiological cure was achieved in 56.4% (62/110) of patients receiving cefiderocol and 53.3% (56/105) of those on colistin (OR: 0.82; 95% CI: 0.47–1.43).^44,45 None of these comparisons showed statistically significant differences. The certainty of evidence was rated low to very low, primarily due to imprecision and non-randomized study design. In another study (n = 102), combination therapy with polymyxin + tigecycline was not superior to high-dose tigecycline monotherapy (high-dose tigecycline resulted in 29.4% (15/51) vs 25.5% (13/51) in polymyxin + tigecycline; OR: 0.82; 95% CI: 0.34–1.96).³⁷ A observational study (n = 28) comparing colistin + tigecycline vs tigecycline alone demonstrated low cure for Colistin+ Tigecycline, 50% (7/14) vs 57.1 (8/14).⁴⁹ One small study (n = 61) comparing polymyxins with other antibiotics showed significantly higher cure in the polymyxin group [91.3% (21/23) vs 18.4% (7/38); OR: 46.5; 95% CI: 8.79–246.09], though it was rated very low certainty due to extreme imprecision (Fig. 8).⁴⁷

Fig. 8.

Forest plot showing the microbiological cure rate in carbapenem-resistant Acinetobacter baumanii infections treated with polymyxin B/colistin-based vs non-polymyxin B/non-colistin regimens from observational studies

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; M-H, Mantel-Hanszael

Mortality Rate

Mortality outcomes were evaluated in two RCTs (n = 153) comparing colistin-based regimens with non-colistin therapies. In the pooled analysis, mortality was 32.5% (25/77) in the colistin arm vs 21.1% (16/76) in the comparator arm, with an OR: 1.79; 95% CI: 0.86–3.72, p = 0.12. This indicates that patients treated with colistin had 79% higher odds of death compared to those receiving non-colistin/non-polymyxin-based therapy. In the larger ATTACK trial (n = 125), which compared colistin + imipenem–cilastatin (20/62; 32.3%) with sulbactam–durlobactam + imipenem–cilastatin (12/63; 19.0%), with OR: 2.02; 95% CI: 0.89–4.61.³⁰ A smaller RCT by Betrosian et al. (n = 28) comparing colistin (5/15; 33.3%) vs ampicillin–sulbactam (4/13; 30.8%) reported an OR: 1.13; 95% CI: 0.23–5.54.³¹ While neither trial reached statistical significance, both demonstrated a numerical trend toward higher mortality with colistin. The certainty of evidence from RCTs was rated high (Fig. 9).

Fig. 9.

Forest plot comparing mortality rates in critically ill patients with carbapenem-resistant Acinetobacter baumanii infections treated with polymyxin B/colistin vs non-polymyxin B/non-colistin therapy in RCTs

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; IV, independent variable; RCTs, randomized controlled trials

In 18 non-randomized observational studies (n = 1,608), pooled mortality was 41.8% (367/878) in the colistin/polymyxin group and 40.5% (296/730) in the non-colistin group (OR: 1.14; 95% CI: 0.81–1.59; p = 0.45). This indicates a 14% higher odds of death with colistin/polymyxin-based therapy. The certainty of evidence was rated very low due to imprecision, inconsistency, and suspected publication bias.

Subgroup analyses revealed mixed findings. In three studies (n = 341), colistin was associated with higher mortality than cefiderocol [47.8% (88/184) vs 38.2% (60/157); OR: 1.44; 95% CI: 0.82–2.54].^44,50 In six studies (n = 560), colistin vs tigecycline showed similar mortality [40.1% (132/329) vs 45.0% (104/231); OR: 0.84; 95% CI: 0.49–1.43].^{36,37,40,44,51,52} One study (n = 64) comparing colistin to tobramycin reported mortality of 34.4% (11/32) and 21.9% (7/32), respectively (OR: 1.87; 95% CI: 0.62–5.68).⁴⁸ Another study (n = 98) comparing colistin monotherapy with non-colistin combinations found mortality rates of 74.3% (26/35) vs 60.3% (38/63) (OR: 1.90; 95% CI: 0.76–4.72).³⁴ A study comparing colistin or ampicillin–sulbactam vs TMP-SMX (n = 136) reported mortality of 38.6% (32/83) and 22.6% (12/53), respectively (OR: 2.14; 95% CI: 0.98–4.68).⁵³ In a separate study (n = 104), colistin was associated with significantly higher mortality than ampicillin–sulbactam [25.8% (17/66) vs 7.9% (3/38); OR: 4.05; 95% CI: 1.10–14.88].⁴⁶ Two studies (n = 155) comparing colistin vs imipenem–cilastatin reported similar mortality [46.9% (38/81) vs 43.2% (32/74); OR: 1.09; 95% CI: 0.57–2.09].³³ Another study comparing colistin vs carbapenems showed mortality of 51.6% (16/31) vs 43.3% (13/30) (OR: 1.39; 95% CI: 0.51–3.83).⁵⁴ One observational study (n = 61) comparing polymyxins to other antibiotics reported significantly lower mortality in the polymyxin group: 8.7% (2/23) vs 18.4% (7/38) (OR: 0.08; 95% CI: 0.02–0.38).⁴⁷ Finally, a study comparing colistin + tigecycline vs tigecycline monotherapy (n = 28) found 35.7% (5/14) vs 42.9% (6/14) respectively (OR: 0.74; 95% CI: 0.16–3.39) (Fig. 10).⁴⁶

Fig. 10.

Forest plot comparing mortality rates in critically ill patients with CRAB infections treated with polymyxin B/colistin-based vs non-polymyxin B/non-colistin regimens from observational studies

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; IV, independent variable; TMP/SX, trimethoprim-sulfamethaxazole

AE Rate

Safety data from three RCTs (n = 221) demonstrated a numerically higher incidence of AEs with colistin/polymyxin-based regimens compared to non-colistin comparators, though statistical significance varied across studies. The pooled AE rate was 12.6% (14/111) in the colistin/polymyxin arm vs 6.4% (7/110) in the non-colistin arm, corresponding to a pooled OR of 2.22 (95% CI: 0.10–49.00; p = 0.38), with modest heterogeneity (I² = 25%). This indicates a more than twofold higher odds of AEs with colistin/polymyxin-based therapy.

In the ATTACK trial (n = 170), which compared colistin + imipenem–cilastatin with sulbactam–durlobactam + imipenem–cilastatin, both arms reported 2/85 (2.4%) AEs (OR: 1.00; 95% CI: 0.14–7.27; p = 1.00).³⁰ In the trial by Betrosian et al. (n = 28), comparing colistin (6/15; 40.0%) vs ampicillin–sulbactam (4/13; 30.8%), the AE rate was numerically higher in the colistin arm (OR: 1.67; 95% CI: 0.31–8.84; p = 0.71).³¹ The third trial by Mosaed et al. (n = 23) evaluated colistin + levofloxacin (6/11; 54.5%) vs levofloxacin + high-dose ampicillin–sulbactam (1/12; 8.3%), and found a statistically significant increase in AEs in the colistin-containing arm (OR: 13.00; 95% CI: 1.24–140.68; p = 0.03) (Fig. 11).³²

Fig. 11.

Forest plot showing adverse event rates among critically ill carbapenem-resistant Acinetobacter baumanii-infected patients treated with polymyxin B/colistin-based vs non-polymyxin B/non-colistin regimens from RCTs

AMP/SUL, ampicillin-sulbactam; CI, confidence interval; IV, independent variable; RCTs, randomized controlled trials

Twelve observational studies (n = 1,041) assessing AEs in Acinetobacter infections showed that colistin or polymyxin-based regimens were associated with higher AE rates compared to non-colistin therapies. The pooled AE rate was 29.5% (175/593) in the colistin/polymyxin group vs 17.2% (77/448) in the comparator arm, corresponding to a pooled OR of 2.01 (95% CI: 1.45–2.78). This indicates that patients treated with colistin/polymyxin-based regimens had approximately twice the odds of experiencing AEs compared to those receiving non-colistin-based therapies, favoring the non-colistin group in terms of safety. The certainty of evidence was rated as very low, primarily due to serious risks of bias, inconsistency, and imprecision.

In subgroup analyses, colistin vs cefiderocol (3 studies; n = 338) showed significantly higher AE rates with colistin [27.6% (50/181) vs 7.6% (12/157); OR: 5.18; 95% CI: 2.59–10.38].^44,45,50 Similarly, colistin vs tigecycline (4 studies; n = 348) showed 24.8% (50/202) vs 8.9% (13/146) (OR: 2.53; 95% CI: 1.26–5.07).^37–40,51 The comparison of colistin vs imipenem–cilastatin (n = 35) showed AE rates of 23.8% (5/21) vs 42.9% (6/14) (OR: 0.42; 95% CI: 0.10–1.79), favoring colistin.³⁸ Similarly, colistin vs tobramycin (n = 44) showed 19.0% (4/21) vs 8.7% (2/23) (OR: 2.47; 95% CI: 0.40–15.15).⁴⁸

In other comparisons, polymyxins vs tigecycline (n = 37) showed 21.2% (7/33) in the polymyxin group vs 0% (0/4) in the tigecycline group (OR: 2.55; 95% CI: 0.12–52.83).³⁵ A study comparing TMP–SMX vs colistin or ampicillin–sulbactam (n = 136) showed lower AE rates in the TMP–SMX arm: 34.0% (18/53) vs 42.2% (35/83) in colistin or ampicillin–sulbactam (OR: 1.42; 95% CI: 0.69–2.90).⁵⁰ Lastly, polymyxins + tigecycline vs high-dose tigecycline (n = 103) showed similar AE rates: 46.2% (24/52) vs 51.0% (26/51) (OR: 0.82; 95% CI: 0.38–1.79) (Fig. 12).³⁶

Fig. 12.

Forest plot showing adverse event rates among critically ill carbapenem-resistant Acinetobacter baumanii-infected patients treated with polymyxin B/colistin-based vs non-polymyxin B/non-colistin regimens from observational studies

AMP + SL, ampicillin-sulbactam; CI, confidence interval; M-H, Mantel-Hanszael; TMP/SX, trimethoprim-sulfamethaxazole

Strength of Recommendation

Weak recommendation to use non-polymyxin-based therapy instead of polymyxin B/colistin-based therapy in critically ill patients with CRAB infection

Certainty of evidence: Very low (Tables 5 to 4)

Rationale for Weak Recommendation

A non-significant trend towards increasing clinical cure and microbiological cure, and a decrease in adverse effects with non-polymyxin-based therapy, and no difference in mortality outcome in critically ill patients with CRAB infection.

Guideline Question

Should sulbactam-based therapies or sulbactam-based therapies be used instead of non-sulbactam-based regimens (e.g., colistin-based) to improve clinical outcomes such as clinical cure, microbiological eradication, mortality, and safety? (refer to Table 4).

Table 4.

Summary of findings table presenting the certainty of evidence and effect estimates for adverse events in critically ill patients with CRE, CRAB, and CRPA infections treated with polymyxin B/colistin-based regimens vs non-polymyxin B/non-colistin regimens

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non-colistin/non-polymyxins	With colistin/polymyxins		Risk with non-colistin/non-polymyxins	Risk difference with colistin/polymyxins
Acinetobacter (observational study)-AER—colistin vs imipenem + cilastatin29
0 cases 0 controls 5/21 exposed 6/14 unexposed (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	None	⨁◯◯◯ Very low	0 cases 0 controls 5/21 exposed 6/14 unexposed		OR: 0.42 (0.10–1.79)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational study)-AER—colistin vs tobramycin44
44 (1 non-randomized study)	Serious	Not serious	Not serious	Very serious	None	⨁◯◯◯ Very low	2/23 (8.7%)	4/21 (19.0%)	OR: 2.47 (0.40–15.15)	2/23 (8.7%)	103 more per 1,000 (from 50 fewer to 504 more)
Acinetobacter (observational study)-AER—colistin vs cefiderocol36,37,46
0 cases 0 controls 50/181 exposed 12/157 unexposed (3 non-randomized studies)	Serious	Not serious	Not serious	Serious	None	⨁◯◯◯ Very low	0 cases 0 controls 50/181 exposed 12/157 unexposed		OR: 5.18 (2.59–10.38)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational study)-AER—colistin vs tigecycline33–35,47
0 cases 0 controls 50/202 exposed 13/146 unexposed (4 non-randomized studies)	Serious	Serious	Not serious	Very serious	None	⨁◯◯◯ Very low	0 cases 0 controls 50/202 exposed 13/146 unexposed		OR: 2.53 (1.26–5.07)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational study)-AER—polymyxins vs tigecycline32
37 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	None	⨁◯◯◯ Very low	0/4 (0.0%)	7/33 (21.2%)	OR: 2.55 (0.12–52.83)	0/4 (0.0%)	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational study)-AER—colistin/AMP+SL Vs TMP/SMX49
136 (1 non-randomized study)	Serious	Not serious	Not serious	Serious	None	⨁◯◯◯ Very low	18/53 (34.0%)	35/83 (42.2%)	OR: 1.42 (0.69–2.90)	18/53 (34.0%)	82 more per 1,000 (from 78 fewer to 259 more)
Acinetobacter (observational study)-AER—polymyxin + tigecyclin vs high dose—tigecycline38
103 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	None	⨁◯◯◯ Very low	26/51 (51.0%)	24/52 (46.2%)	OR: 0.82 (0.38–1.79)	26/51 (51.0%)	50 fewer per 1,000 (from 227 fewer to 141 more)
Acinetobacter (observational study)-AER—total
0 cases 0 controls 175/593 exposed 77/448 unexposed (12 non-randomized studies)	Serious	Serious	Not serious	Very serious	None	⨁◯◯◯ Very low	0 cases 0 controls 175/593 exposed 77/448 unexposed		OR: 2.01 (1.45–2.78)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Pseudomonas aeruginosa (Observational Study)-AER - Colistin vs Ceftolozane-tazobactam41–43
0 cases 0 controls 69/164 exposed 21/117 unexposed (3 non-randomized studies)	Serious	Serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 69/164 exposed 21/117 unexposed		OR: 2.89 (0.93–9.04)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Pseudomonas aeruginosa (observational study)-AER—total
0 cases 0 controls 69/164 exposed 21/117 unexposed (3 non-randomized studies)	Serious	Serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 69/164 exposed 21/117 unexposed		OR: 2.89 (0.93–9.04)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Enterobacteriaceae (observational study)-AER—colistin vs CZA20,22,23
0 cases 0 controls 14/105 exposed 12/117 unexposed (3 non-randomized studies)	Serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 14/105 exposed 12/117 unexposed		OR: 1.48 (0.60–3.62)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Enterobacteriaceae (observational study)-AER—polymyxin vs CZA24
104 (1 non-randomized study)	Serious	Not serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	5/78 (6.4%)	6/26 (23.1%)	OR: 4.38 (1.21–15.84)	5/78 (6.4%)	167 more per 1,000 (from 12 more to 456 more)
Enterobacteriaceae (observational study)-AER—total
0 cases 0 controls 20/131 exposed 17/195 unexposed (4 non-randomized studies)	Serious	Not serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 20/131 exposed 17/195 unexposed		OR: 2.11 (1.01–4.40)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (RCTs)-AER—colistin + imipenem–cilastatin vs sulbactam–durlobactam + imipenem–cilastatin26
170 (1 RCT)	Very serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	2/85 (2.4%)	2/85 (2.4%)	OR: 1.00 (0.14–7.27)	2/85 (2.4%)	0 fewer per 1,000 (from 20 fewer to 126 more)
Acinetobacter (RCTs)-AER—colistin vs AMP/SUL27
28 (1 RCT)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	4/13 (30.8%)	6/15 (40.0%)	OR: 1.50 (0.31–7.19)	4/13 (30.8%)	92 more per 1,000 (from 187 fewer to 454 more)
Acinetobacter (RCTs)-AER—colistin + levofloxacin vs levofloxacin + high dose ampicillin/sulbactam infusion28
23 (1 RCT)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	1/12 (8.3%)	6/11 (54.5%)	OR: 13.20 (1.24–140.68)	1/12 (8.3%)	462 more per 1,000 (from 18 more to 844 more)
Acinetobacter (RCTs)-AER—total
221 (3 RCTs)	Very serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	7/110 (6.4%)	14/111 (12.6%)	OR: 2.33 (0.84–6.43)	7/110 (6.4%)	73 more per 1,000 (from 10 fewer to 240 more)

Outcome: AE rate. AER, adverse event rate; AMP/SUL, ampicillin-sulbactam; CI, confidence interval; CRAB, carbapenem-resistant Acinetobacter baumanii; CRE, carbapenem-resistant Enterobacterales; CRPA, carbapenem-resistant Pseudomonas aeruginosa; CZA, ceftazidime-avibactam; RCT, randomized controlled trial; RoB, risk of bias; TMP/SX, trimethoprim-sulfamethaxazole

Table 1.

Summary of findings table presenting the certainty of evidence and effect estimates for clinical cure in critically ill patients with CRE, CRAB, and CRPA infections treated with polymyxin B/colistin-based regimens vs non-polymyxin B/non-colistin regimens

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non-colistin/non-polymyxins	With colistin/polymyxins		Risk with non-colistin/non-polymyxins	Risk difference with colistin/polymyxins
Acinetobacter (observational study)-CCR—colistin vs imipenem + cilastatin29,30
0 cases 0 controls 37/81 exposed 51/74 unexposed (2 non-randomized studies)	Serious	Serious	Very serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 37/81 exposed 51/74 unexposed		Outcome measured −0.19 (−2.04 to 1.66) 0 per 1,000	Low
										-- per 1,000 (from -- to --)
Acinetobacter (observational Study)-CCR—colistin monotherapy vs non-colistin-based combinations31
107 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	30/71 (42.3%)	11/36 (30.6%)	Outcome.measure.rd −0.12 (−0.31 to 0.07)	30/71 (42.3%)	-- per 1,000 (from -- to --)
Acinetobacter (observational Study)-CCR—polymyxins vs doxycycline32
33 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	2/4 (50.0%)	22/29 (75.9%)	Outcome.measure.rd 0.26 (−0.26 to 0.77)	2/4 (50.0%)	-- per 1,000 (from -- to --)
Acinetobacter (observational study)-CCR—colistin vs tigecycline33–35
180 (3 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	28/62 (45.2%)	57/118 (48.3%)	Outcome.measure.rd 0.03 (−0.03 to 0.09)	28/62 (45.2%)	-- per 1,000 (from -- to --)
Acinetobacter (observational study)-CCR—colistin vs cefiderocol36,37
215 (2 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	77/110 (70.0%)	57/105 (54.3%)	Outcome.measure.rd −0.15 (−0.96 to 0.67)	77/110 (70.0%)	-- per 1,000 (from -- to --)
Acinetobacter (observational study)-CCR—polymyxins + tigecycline vs tigecycline38
102 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	14/51 (27.5%)	11/51 (21.6%)	Outcome.measure.rd −0.06 (−0.23 to 0.11)	14/51 (27.5%)	-- per 1,000 (from -- to --)
Acinetobacter (observational study)-CCR—colistin vs AMP/SUL39
98 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	18/32 (56.3%)	31/66 (47.0%)	Outcome.measure.rd −0.09 (−0.30 to 0.12)	18/32 (56.3%)	-- per 1,000 (from -- to --)
Acinetobacter (Observational Study)-CCR - Polymyxins vs Other Antibiotics40
61 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	9/38 (23.7%)	22/23 (95.7%)	Outcome.measure.rd 0.72 (0.56 to 0.88)	9/38 (23.7%)	-- per 1,000 (from -- to --)
Acinetobacter (observational study)-CCR—total
951 (12 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	229/442 (51.8%)	248/509 (48.7%)	Outcome.measure.rd 0.01 (−0.16 to 0.18)	229/442 (51.8%)	-- per 1,000 (from -- to --)
Pseudomonas aeruginosa (observational study)-CCR—colistin vs ceftolozane-tazobactam41–43
281 (3 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	79/117 (67.5%)	80/164 (48.8%)	Outcome.measure.rd −0.13 (−0.45 to 0.19)	79/117 (67.5%)	-- per 1,000 (from -- to --)
Pseudomonas aeruginosa (observational study)-CCR—total
281 (3 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	79/117 (67.5%)	80/164 (48.8%)	Outcome.measure.rd −0.13 (−0.45 to 0.19)	79/117 (67.5%)	-- per 1,000 (from -- to --)
Enterobacteriaceae (observational study)-CCR—colistin vs CZA20–23
364 (4 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	91/170 (53.5%)	45/194 (23.2%)	Outcome.measure.rd −0.19 (−0.28 to −0.10)	91/170 (53.5%)	-- per 1,000 (from -- to --)
Enterobacteriaceae (observational study)-CCR—polymyxin vs CZA24
104 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	57/78 (73.1%)	12/26 (46.2%)	Outcome.measure.rd −0.27 (−0.48 to −0.05)	57/78 (73.1%)	-- per 1,000 (from -- to --)
Enterobacteriaceae (Observational Study)-CCR - total
468 (5 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	148/248 (59.7%)	57/220 (25.9%)	outcome.measure.rd −0.20 (−0.29 to −0.12)	148/248 (59.7%)	-- per 1,000 (from -- to --)
Acinetobacter (RCTs)-CCR—colistin + imipenem–cilastatin vs sulbactam–durlobactam + imipenem–cilastatin25,26
204 (2 RCTs)	Very serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	62/99 (62.6%)	40/105 (38.1%)	Outcome.measure.rd −0.25 (−0.38 to −0.11)	62/99 (62.6%)	-- per 1,000 (from -- to --)
Acinetobacter (RCTs)-CCR—colistin vs AMP/SUL27
28 (1 RCT)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁⨁◯◯ Low	9/13 (69.2%)	9/15 (60.0%)	Outcome.measure.rd −0.09 (−0.45 to 0.26)	9/13 (69.2%)	-- per 1,000 (from -- to --)
Acinetobacter (RCTs)-CCR—colistin + levofloxacin vs levofloxacin + high dose ampicillin/sulbactam infusion28
23 (1 RCT)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁⨁◯◯ Low	10/12 (83.3%)	3/11 (27.3%)	Outcome.measure.rd −0.56 (−0.90 to −0.22)	10/12 (83.3%)	-- per 1,000 (from -- to --)
Acinetobacter (RCTs)-CCR—total
255 (4 RCTs)	Very serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	81/124 (65.3%)	52/131 (39.7%)	Outcome.measure.rd −0.27 (−0.42 to −0.13)	81/124 (65.3%)	-- per 1,000 (from -- to --)

Outcome: Clinical cure rate

Table 2.

Summary of findings table presenting the certainty of evidence and effect estimates for microbiological cure in critically ill patients with CRE, CRAB, and CRPA infections treated with polymyxin B/colistin-based regimens vs non-polymyxin B/non-colistin regimens

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non- colistin/non- polymyxins	With colistin/polymyxins		Risk with non-colistin/non-polymyxins	Risk difference with colistin/polymyxins
Acinetobacter (observational studies)-MCR—colistin vs imipenem + cilastatin29
35 (1 non-randomized study)	Not serious	Not serious	Not serious	Very serious	None	⨁◯◯◯ Very low	7/14 (50.0%)	14/21 (66.7%)	OR: 2.0 (0.5–8.0)	7/14 (50.0%)	167 more per 1,000 (from 167 fewer to 389 more)
Acinetobacter (observational studies)-MCR—colistin vs tobramycin44
64 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	18/32 (56.3%)	16/32 (50.0%)	OR: 0.78 (0.29–2.08)	18/32 (56.3%)	62 fewer per 1,000 (from 291 fewer to 165 more)
Acinetobacter (observational studies)-MCR—colistin monotherapy vs non-colistin-based combination31
88 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	49/59 (83.1%)	20/29 (69.0%)	OR: 0.45 (0.16–1.28)	49/59 (83.1%)	143 fewer per 1,000 (from 391 fewer to 32 more)
Acinetobacter (observational studies)-MCR—polymyxins vs doxycycline32
24 (1 non-randomized study)	Not serious	Not serious	Not serious	Very serious	None	⨁◯◯◯ Very low	2/3 (66.7%)	17/21 (81.0%)	OR: 2.13 (0.15–29.66)	2/3 (66.7%)	143 more per 1,000 (from 436 fewer to 317 more)
Acinetobacter (observational studies)-MCR—colistin vs cefiderocol36,37
0 cases 0 controls 56/105 exposed 62/110 unexposed (2 non-randomized studies)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	0 cases 0 controls 56/105 exposed 62/110 unexposed		OR: 0.82 (0.47–1.43)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational studies)-MCR—colistin vs tigecycline34,35
0 cases 0 controls 30/79 exposed 2/46 unexposed (2 non-randomized studies)	Not serious	Not serious	Not serious	Very serious	None	⨁◯◯◯ Very low	0 cases 0 controls 30/79 exposed 2/46 unexposed		OR: 10.27 (2.61–40.42)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational studies)-MCR—polymyxin + tigecycline vs high dose—tigecycline38
102 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	15/51 (29.4%)	13/51 (25.5%)	OR: 0.82 (0.34–1.96)	15/51 (29.4%)	39 fewer per 1,000 (from 170 fewer to 155 more)
Acinetobacter (Observational Studies)-MCR - Colistin vs AMP/SUL39
50 (1 non-randomized study)	Not serious	Not serious	Not serious	Very serious	None	⨁◯◯◯ Very low	3/17 (17.6%)	16/33 (48.5%)	OR: 4.39 (1.06–18.20)	3/17 (17.6%)	308 more per 1,000 (from 9 more to 619 more)
Acinetobacter (observational studies)-MCR—polymyxin vs other antibiotics40
61 (1 non-randomized study)	Not Serious	Not serious	Not serious	Extremely Serious	None	⨁◯◯◯ Very low	7/38 (18.4%)	21/23 (91.3%)	OR: 46.50 (8.79–246.09)	7/38 (18.4%)	729 more per 1,000 (from 481 more to 798 more)
Acinetobacter (observational studies)-MCR—colistin + tigecycline vs tigecycline45
28 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	8/14 (57.1%)	7/14 (50.0%)	OR: 0.75 (0.17–3.33)	8/14 (57.1%)	71 fewer per 1,000 (from 387 fewer to 245 more)
Acinetobacter (observational studies)-MCR—total
0 cases 0 controls 210/408 exposed 173/384 unexposed (12 non-randomized studies)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁◯◯ Low	0 cases 0 controls 210/408 exposed 173/384 unexposed		OR: 1.49 (1.10–2.01)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Pseudomonas aeruginosa (observational study)-MCR—colistin vs ceftolozane-tazobactum42
51 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	8/18 (44.4%)	5/33 (15.2%)	OR: 0.22 (0.06–0.84)	8/18 (44.4%)	295 fewer per 1,000 (from 399 fewer to 43 fewer)
Pseudomonas aeruginosa (observational study)-MCR—total
51 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	8/18 (44.4%)	5/33 (15.2%)	OR: 0.22 (0.06–0.84)	8/18 (44.4%)	295 fewer per 1,000 (from 399 fewer to 43 fewer)
Enterobacteriaceae (observational study)-MCR—colistin vs CZA20,22,23
0 cases 0 controls 44/90 exposed 74/126 unexposed (3 non-randomized studies)	Not serious	Serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 44/90 exposed 74/126 unexposed		OR: 0.43 (0.16–1.16)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Enterobacteriaceae (observational study)-MCR—polymyxin vs CZA24
51 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	39/43 (90.7%)	6/8 (75.0%)	OR: 0.31 (0.05–2.06)	39/43 (90.7%)	156 fewer per 1,000 (from 579 fewer to 46 more)
Enterobacteriaceae (Observational Study)-MCR - total
0 cases 0 controls 50/98 exposed 113/169 unexposed (4 non-randomized studies)	Not serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 50/98 exposed 113/169 unexposed		OR: 0.43 (0.19–0.94)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (RCTs)-MCR—Colistin vs AMP/SUL27
28 (1 RCT)	Not Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	8/13 (61.5%)	10/15 (66.7%)	OR: 1.25 (0.27–5.89)	8/13 (61.5%)	51 more per 1,000 (from 314 fewer to 289 more)
Acinetobacter (RCTs)-MCR—total
28 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	8/13 (61.5%)	10/15 (66.7%)	OR: 1.25 (0.27–5.89)	8/13 (61.5%)	51 more per 1,000 (from 314 fewer to 289 more)

AMP/SUL, ampicillin-sulbactam; CCR, clinical cure rate; CI, confidence interval; CZA, ceftazidime-avibactam; MCR, microbiological cure rate; RCT, randomized controlled trial; RoB, risk of bias

Table 3.

Summary of Findings table presenting the certainty of evidence and effect estimates for mortality in critically ill patients with CRE, CRAB, and CRPA infections treated with polymyxin B/colistin–based regimens vs non-polymyxin B/non-colistin regimens

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non- colistin/non- polymyxins	With colistin/polymyxins		Risk with non- colistin/non- polymyxins	Risk difference with colistin/polymyxins
Acinetobacter (observational study)-MR—colistin vs tobramycin44
64 (1 non-randomized study)	Not serious	Serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	7/32 (21.9%)	11/32 (34.4%)	RR: 1.57 (0.70–3.54)	7/32 (21.9%)	125 more per 1,000 (from 66 fewer to 556 more)
Acinetobacter (observational study)-MR—colistin vs cefiderocol36,46
341 (3 non-randomized studies)	Not serious	Serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	60/157 (38.2%)	88/184 (47.8%)	RR: 1.23 (0.89–1.68)	60/157 (38.2%)	88 more per 1,000 (from 42 fewer to 260 more)
Acinetobacter (observational study)-MR—colistin vs tigecycline32,33,35,38,47,48
0 cases 0 controls 132/329 exposed 104/231 unexposed (6 non-randomized studies)	Not serious	Serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 132/329 exposed 104/231 unexposed		RR: 0.88 (0.67–1.16)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (Observational Study)-MR - Colistin Monotherapy vs Non-Colistin based Combination31
98 (1 non-randomized study)	Not serious	Not serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	38/63 (60.3%)	26/35 (74.3%)	RR: 1.23 (0.93–1.63)	38/63 (60.3%)	139 more per 1,000 (from 42 fewer to 380 more)
Acinetobacter (observational study)-MR—colistin or AMP/SUL vs TMP/SMX49
136 (1 non-randomized study)	Not serious	Not serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	12/53 (22.6%)	32/83 (38.6%)	RR: 1.70 (0.97–3.00)	12/53 (22.6%)	158 more per 1,000 (from 7 fewer to 453 more)
Acinetobacter (observational study)-MR—colistin vs AMP/SUL49
104 (1 non-randomized study)	Not serious	Not serious	Not serious	Extremely serious	Publication bias strongly suspected	⨁◯◯◯ Very low	3/38 (7.9%)	17/66 (25.8%)	RR: 3.26 (1.02–10.41)	3/38 (7.9%)	178 more per 1,000 (from 2 more to 743 more)
Acinetobacter (observational study)-MR—colistin vs imipenem + cilastatin29,30
0 cases 0 controls 38/81 exposed 32/74 unexposed (2 non-randomized studies)	Not serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 38/81 exposed 32/74 unexposed		RR: 1.03 (0.74–1.44)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (observational study)-MR—colistin vs carbapenems (imipenem or meropenem)50
61 (1 non-randomized study)	Not serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	13/30 (43.3%)	16/31 (51.6%)	RR: 1.19 (0.70–2.03)	13/30 (43.3%)	82 more per 1,000 (from 130 fewer to 446 more)
Acinetobacter (observational study)-MR—polymyxins vs other antibiotics40
61 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	21/38 (55.3%)	2/23 (8.7%)	RR: 0.16 (0.04–0.61)	21/38 (55.3%)	464 fewer per 1,000 (from 531 fewer to 216 fewer)
Acinetobacter (observational study)-MR—colistin+tigecycline vs tigecycline45
28 (1 non-randomized study)	Not serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	6/14 (42.9%)	5/14 (35.7%)	RR: 0.83 (0.33–2.11)	6/14 (42.9%)	73 fewer per 1,000 (from 287 fewer to 476 more)
Acinetobacter (observational study)-MR—total
1608 (18 non-randomized studies)	Not serious	Serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	296/730 (40.5%)	367/878 (41.8%)	RR: 1.08 (0.90–1.28)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Pseudomonas aeruginosa (Observational Study)-MR - Colistin vs Ceftolozane-tazobactam44–46
0 cases 0 controls 73/164 exposed 50/117 unexposed (3 non-randomized studies)	Not serious	Serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 73/164 exposed 50/117 unexposed		RR: 0.92 (0.50–1.70)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Pseudomonas aeruginosa (observational study)-MR—total
0 cases 0 controls 73/164 exposed 50/117 unexposed (3 non-randomized studies)	Not serious	Serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 73/164 exposed 50/117 unexposed		RR: 0.92 (0.50–1.70)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Enterobacteriaceae (observational study)-MR—colistin vs CZA20,22,23
0 cases 0 controls 57/158 exposed 39/129 unexposed (3 non-randomized studies)	Not serious	Serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 57/158 exposed 39/129 unexposed		RR: 1.43 (0.70–2.25)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Enterobacteriaceae (observational study)-MR—polymyxin vs CZA24
104 (1 non-randomized study)	Not serious	Not serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	22/78 (28.2%)	12/26 (46.2%)	RR: 1.64 (0.90–2.42)	22/78 (28.2%)	181 more per 1,000 (from 28 fewer to 401 more)
Enterobacteriaceae (observational study)-MR—total
0 cases 0 controls 69/184 exposed 61/207 unexposed (4 non-randomized studies)	Not serious	Serious	Not serious	Very serious	Publication bias strongly suspected	⨁◯◯◯ Very low	0 cases 0 controls 69/184 exposed 61/207 unexposed		RR: 1.47 (0.91–2.07)	Low
										0 per 1,000	0 fewer per 1,000 (from 0 fewer to 0 fewer)
Acinetobacter (RCTs)-MR—Colistin + imipenem–cilastatin vs sulbactam–durlobactam + imipenem–cilastatin26
125 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	12/63 (19.0%)	20/62 (32.3%)	RR: 1.69 (0.91–3.16)	12/63 (19.0%)	131 more per 1,000 (from 17 fewer to 411 more)
Acinetobacter (RCTs)-MR—Colistin vs AMP/SUL27
28 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	4/13 (30.8%)	5/15 (33.3%)	RR: 1.08 (0.37–3.21)	4/13 (30.8%)	25 more per 1,000 (from 194 fewer to 680 more)
Acinetobacter (RCTs)-MR—total
153 (2 RCTs)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	16/76 (21.1%)	25/77 (32.5%)	RR: 1.52 (0.88–2.60)	16/76 (21.1%)	109 more per 1,000 (from 25 fewer to 337 more)

Outcome: Mortality rate. AMP/SUL, ampicillin-sulbactam; CI, confidence interval; MR, mortality rate; CRE, carbapenem-resistant Enterobacterales; CRAB, carbapenem-resistant Acinetobacter baumanii; CRPA, carbapenem-resistant Pseudomonas aeruginosa; CZA, ceftazidime-avibactam; MCR, microbiological cure rate; RCT, randomized controlled trial; RoB, risk of bias

Clinical Cure Rate

Five RCTs (n = 372) evaluated clinical cure outcomes comparing sulbactam-based regimens with non-sulbactam comparators in patients with Acinetobacter infections. Across all studies, clinical cure was achieved in 66.7% (97/195) of patients receiving sulbactam-based therapy, compared to 49.7% (118/177) in those receiving non-sulbactam regimens. The pooled OR was 2.05; 95% CI: 0.94–4.49; p = 0.06, indicating that sulbactam-based therapy was associated with approximately a 105% higher likelihood of clinical cure compared to non-sulbactam regimens, favoring sulbactam-based therapy, though the difference was not statistically significant. The overall certainty of evidence was rated moderate due to concerns regarding RoB in some included studies (Fig. 13).

Fig. 13.

Forest plot showing the clinical cure rate in critically ill patients with CRAB infections treated with sulbactam-based therapy vs non-sulbactam-based regimens from RCTs

CI, confidence interval; IV, independent variable; RCTs, randomized controlled trials

The highest-quality and most contemporary evidence in this section is derived from the (ATTACK; n = 126), clinical cure was significantly higher with sulbactam–durlobactam + imipenem–cilastatin (61.9%; 39/63) compared to colistin + imipenem–cilastatin (39.7%; 25/63), with an OR: 2.47; 95% CI: 1.21–5.06, supporting the benefit of the sulbactam-based combination.⁵⁵ An RCT (n = 148) comparing cefoperazone–sulbactam vs cefepime reported cure rates of 73.1% (49/67) and 56.8% (46/81), respectively, with an OR: 2.07; 95% CI: 1.03–4.16, numerically favoring cefoperazone–sulbactam.⁵⁶ In contrast, a smaller trial comparing meropenem + ampicillin–sulbactam vs meropenem + colistin (n = 47) showed similar outcomes [75.0% (18/24) vs 69.6% (16/23)] with (OR: 0.76; 95% CI: 0.21–2.74).⁵⁷ Another RCT (n = 23) comparing levofloxacin + high-dose ampicillin/sulbactam to levofloxacin + colistin demonstrated a substantially higher clinical cure rate with sulbactam [83.3% (10/12) vs 27.3% (3/11)], with a statistically significant (OR: 13.33; 95% CI: 1.78–100.14).³² Lastly, one trial (n = 28) comparing intravenous ampicillin–sulbactam + nebulized colistin with comparator reported comparable cure rates [33.3% (4/12) vs 31.3% (5/16)]; OR: 1.10; 95% CI: 0.22–5.45).⁵⁸

In the analysis of 10 observational studies (n = 1,783), clinical cure was achieved in 66.0% (690/1,045) of patients treated with sulbactam-based regimens compared to 62.9% (464/738) of those receiving non-sulbactam therapies (pooled OR: 0.89; 95% CI: 0.65–1.24, p = 0.45), indicating that Non-sulbactam-based regimens were associated with 11% lower odds of clinical cure compared to non-sulbactam therapies, though the difference was not statistically significant, suggesting similar efficacy between the two groups.

Subgroup comparisons demonstrated mixed findings. In a large study (n = 386), imipenem/cilastatin plus sulbactam vs tigecycline-based therapy showed significantly lower cure rates in the sulbactam group [69.2% (184/266) vs 50.0% (60/120); OR: 0.45; 95% CI: 0.29–0.69].⁵⁹ In contrast, ampicillin–sulbactam vs imipenem–cilastatin across two studies (n = 135) showed marginally lower cure with sulbactam [53.8% (50/93) vs 42.9% (18/42); OR: 0.68; 95% CI: 0.17–2.78].^60,61 Another study comparing sulbactam-based vs tigecycline-based regimens (n = 168) showed no difference (66.7% in both arms; OR: 1.00; 95% CI: 0.53–1.90).⁶²

Among colistin-based combinations, colistin–sulbactam vs colistin–carbapenem demonstrated comparable efficacy [49.0% (50/102) vs 46.4% (32/69); OR: 0.90; 95% CI: 0.49–1.66], while colistin–sulbactam vs colistin plus other agents showed numerically higher cure with sulbactam [39.5% (17/43) vs 46.4% (32/69); OR: 1.32; 95% CI: 0.61–2.87].⁶³ In a small study, cefoperazone–sulbactam vs imipenem–cilastatin had similar results [75.0% (9/12) vs 77.1% (27/35); OR: 1.13; 95% CI: 0.24–5.18].⁶⁴ Likewise, colistin + sulbactam combination vs colistin monotherapy yielded slightly better outcomes [40.0% (14/35) vs 29.8% (14/47); OR: 1.57; 95% CI: 0.63–3.95].⁶⁰ In a large cohort (n = 624), cefoperazone–sulbactam vs piperacillin–tazobactam showed comparable cure rates [80.5% (298/370) vs 82.7% (210/254); OR: 1.15; 95% CI: 0.76–1.75] (Fig. 14).⁴¹

Fig. 14.

Forest plot showing the clinical cure rate in critically ill patients with carbapenem-resistant Acinetobacter baumanii infections treated with sulbactam-based therapy vs non-sulbactam-based regimens from observational studies

CI, confidence interval; IV, independent variable

Microbiological Cure Rate

Microbiological cure outcomes were assessed in both RCTs and observational studies comparing sulbactam-based regimens with non-sulbactam therapies in patients with Acinetobacter infections. In four RCTs (n = 262), microbiological cure was achieved in 65.4% (83/127) of patients receiving sulbactam-based regimens and 45.2% (61/135) in those receiving non-sulbactam therapies, with a pooled OR of 2.48 (95% CI: 1.20–5.13; p = 0.03), indicating that patients receiving sulbactam-based therapy were about 2.5 times more likely to achieve microbiological cure, favoring sulbactam-based regimens with a statistically significant difference. However, the certainty of evidence was rated low due to concerns related to RoB and imprecision.

Subgroup analysis from RCTs showed that in the ATTACK trial (sulbactam–durlobactam + imipenem–cilastatin vs colistin + imipenem–cilastatin; n = 125), microbiological cure was higher with sulbactam–durlobactam (68.3%; 43/63) compared to colistin (41.9%; 26/62) (OR: 2.98; 95% CI: 1.43–6.19; moderate certainty).⁵² Another RCT comparing cefoperazone–sulbactam vs cefepime (n = 62) reported cure rates of 51.7% (15/29) and 39.4% (13/33), respectively (OR: 1.65; 95% CI: 0.60–4.52; low certainty), again numerically favoring sulbactam.⁵³ In a smaller RCT comparing meropenem + ampicillin–sulbactam vs meropenem + colistin (n = 47), both groups showed high cure rates (91.3 vs 87.5%) with a non-significant difference (OR: 1.50; 95% CI: 0.23–9.92; low certainty).⁵⁴ A further small trial (n = 28) assessing intravenous ampicillin–sulbactam + nebulized colistin vs Intravenous colistin + nebulized colistin showed numerically higher cure with the sulbactam arm (33.3 vs 6.3%; OR: 7.50; 95% CI: 0.71–78.90; low certainty) (Fig. 15).⁵⁵

Fig. 15.

Forest plot showing the microbiological cure rate in CRAB infections treated with sulbactam-based vs non-sulbactam-based regimens in RCTs

CI, confidence interval; IV, independent variable; RCTs, randomized controlled trials

Observational data from seven non-randomized studies (n = 944) showed microbiological cure in 49.6% (181/365) of patients on sulbactam-based regimens vs 35.2% (204/579) on non-sulbactam therapies, with a pooled OR of 1.25 (95% CI: 0.88–1.78; p = 0.21), indicating a 25% higher likelihood of microbiological cure with sulbactam-based therapy, although the trend favored sulbactam, the difference was not statistically significant, and the certainty of evidence was rated very low due to serious concerns regarding RoB, publication bias, and imprecision.

Within these observational studies, individual comparisons yielded varied results. In a large study (n = 386), imipenem/cilastatin + sulbactam vs tigecycline demonstrated significantly higher cure in the sulbactam arm (11.7 vs 1.1%; OR: 11.58; 95% CI: 3.26–41.11; very low certainty).⁵⁶ In two smaller studies (n = 135), ampicillin–sulbactam vs imipenem–cilastatin showed numerically higher cure with sulbactam (50.0 vs 37.6%; OR: 1.20; 95% CI: 0.56–2.59; very low certainty).^57,58 Conversely, colistin–sulbactam vs colistin–carbapenem and colistin–sulbactam vs colistin + other agents both showed similar cure rates of 81.4% in the sulbactam arms compared to 71.0% in the comparator arms (ORs: 0.56; very low certainty).⁶⁰

A study comparing colistin + sulbactam vs colistin alone (n = 82) showed lower cure in the sulbactam group (72.3 vs 85.7%; OR: 2.29; 95% CI: 0.73–7.19; very low certainty).⁴²

Sulbactam-based regimens demonstrated a higher microbiological cure in most RCTs and a favorable trend in observational studies. However, variability in comparator agents, study designs, and substantial limitations in methodological quality contributed to low to very low certainty in observational findings (Fig. 16).

Fig. 16.

Forest plot showing the microbiological cure rate in carbapenem-resistant Acinetobacter baumanii infections treated with sulbactam-based vs non-sulbactam-based regimens in observational studies

CI, confidence interval; M-H, Mantel-Hanszael

Mortality Rate

Mortality outcomes were assessed in four RCTs (n = 366) and nine observational studies (n = 1,532) comparing sulbactam-based therapies to non-sulbactam regimens in Acinetobacter infections. Across all RCTs, mortality occurred in 32 of 177 patients (18.1%) receiving sulbactam-based therapy and 38 of 189 patients (20.1%) in the comparator arms, with a pooled OR: 0.94; 95% CI: 0.24–3.67, p = 0.89. Although this difference was not statistically significant, the direction of effect consistently favored sulbactam-based therapy, with numerically lower mortality observed across studies. This indicates that patients receiving sulbactam-based therapy had slightly lower odds of mortality (about 6% less) compared to those on non-sulbactam regimens. The overall certainty of evidence was low, downgraded due to RoB and inconsistency (Fig. 16).

Among individual trials, the ATTACK trial (n = 125) compared sulbactam–durlobactam plus imipenem–cilastatin (12/63; 19.0%) to colistin plus imipenem–cilastatin (20/62; 32.3%), yielding an OR: 0.49; 95% CI: 0.22–1.13 with high-certainty evidence favoring the sulbactam-based regimen.⁵² A separate RCT (n = 166) comparing cefoperazone–sulbactam (7/79; 8.9%) to cefepime (2/87; 2.3%) reported an OR: 4.13; 95% CI: 0.83–20.52, although certainty was rated very low due to imprecision.⁵³ Another trial (n = 47) comparing meropenem plus ampicillin–sulbactam (9/23; 39.1%) with meropenem plus colistin (10/24; 41.7%) showed similar (OR: 0.90; 95% CI: 0.28–2.89).⁵⁴ In a small study (n = 28), IV ampicillin–sulbactam vs nebulized colistin, in which IV ampicillin–sulbactam + nebulized colistin reported (4/12; 33.3%) showed comparable mortality to the IV colistin+ nebulized colistin (6/16; 37.5%), with an OR: 0.83; 95% CI: 0.17–4.01 (Fig. 17).⁵⁵

Fig. 17.

Forest plot comparing mortality rates in critically ill patients with CRAB infections treated with sulbactam-based vs non-sulbactam-based therapies in RCTs

CI, confidence interval; IV, independent variable; RCTs, randomized controlled trials

In the observational dataset (n = 1,532), mortality was observed in 189 of 672 patients (28.1%) treated with sulbactam-based regimens vs 301 of 860 patients (35.0%) in the non-sulbactam group, resulting in a pooled (OR: 0.61; 95% CI: 0.34–1.12, p = 0.10), indicating a 39% lower odds of death with sulbactam-based therapy; however, the difference did not reach statistical significance, showing a trend favoring sulbactam-based regimens. The certainty of this evidence was rated very low due to RoB, inconsistency, and suspected publication bias.

Subgroup comparisons showed varied trends. In a large study (n = 210), cefoperazone–sulbactam was associated with lower mortality than tigecycline [29.3% (22/75) vs 51.9% (70/135); OR: 0.39; 95% CI: 0.21–0.70].⁴³ Another study comparing sulbactam-based vs tigecycline-based therapy (n = 168) showed comparable mortality [33.3% (28/84) vs 29.8% (25/84); OR: 1.18; 95% CI: 0.62–2.26].⁵⁹ Sulbactam + imipenem/cilastatin significantly outperformed tigecycline + imipenem/cilastatin [14.3% (4/28) vs 64.3% (36/56); OR: 0.09; 95% CI: 0.03–0.30].⁶⁵

In two comparisons evaluating colistin–sulbactam vs colistin–carbapenem and colistin–sulbactam vs colistin + other agents, mortality was numerically lower with sulbactam in both comparisons [46.4% (32/69) vs 54.9% (56/102); OR: 0.71; 95% CI: 0.38–1.31 and 46.4% (32/69) vs 55.8% (24/43); OR: 0.68; 95% CI: 0.32–1.47, respectively].⁶⁰ The comparison of cefoperazone–sulbactam vs imipenem–cilastatin (n = 47) favored sulbactam with lower mortality [20.0% (7/35) vs 50.0% (6/12); OR: 0.25; 95% CI: 0.06–1.02].⁶¹ In a large study comparing cefoperazone–sulbactam vs piperacillin–tazobactam (n = 624), mortality was similar [16.5% (42/254) vs 16.8% (62/370); OR: 0.98; 95% CI: 0.64–1.51].⁴¹ Two additional comparisons—ampicillin–sulbactam vs imipenem–cilastatin [42.9% (12/28) vs 33.3% (10/30); OR: 1.50; 95% CI: 0.52–4.36] did not demonstrate significant differences (Fig. 18).^57,58

Fig. 18.

Forest plot comparing mortality rates in critically ill patients with carbapenem-resistant Acinetobacter baumanii infections treated with sulbactam-based vs non-sulbactam-based therapies in observational studies

CI, confidence interval; IV, independent variable

AE Rate

Adverse event rates were assessed in five RCTs (n = 441) comparing sulbactam-based therapies with non-sulbactam comparators in patients with CRAB infections. Across all studies, AEs were reported in 46.1% (100/217) of patients receiving sulbactam-based regimens and 52.2% (117/224) of those receiving non-sulbactam-based treatments (OR: 0.66; 95% CI: 0.23–1.93; p = 0.35) indicating a 34% lower odds of AEs with sulbactam-based therapy, though the difference was not statistically significant, suggesting comparable safety between the two groups and the overall certainty of evidence was rated low due to serious concerns regarding RoB and inconsistency.

In a large trial (cefoperazone–sulbactam vs cefepime; n = 166), AEs occurred in 73.4% (58/79) of sulbactam-treated patients and 66.7% (58/87) in those receiving cefepime (OR: 1.38; 95% CI: 0.71–2.70; low-certainty evidence), suggesting a numerically higher rate with sulbactam, though the CI was wide and crossed unity.⁵³ The ATTACK trial (sulbactam–durlobactam + imipenem–cilastatin vs colistin + imipenem–cilastatin; n = 177) reported AEs in 39.6% (36/91) of patients in the sulbactam arm compared to 48.8% (42/86) in the colistin arm (OR: 0.69; 95% CI: 0.38–1.24), indicating a numerically lower risk with sulbactam; this was rated as high-certainty evidence.⁵²

A smaller RCT comparing meropenem + ampicillin–sulbactam vs meropenem + colistin (n = 47) reported AEs in 4.3% (1/23) vs 12.5% (3/24), respectively (OR: 0.32; 95% CI: 0.03–3.31; moderate-certainty).⁵⁴ Another trial comparing levofloxacin + high-dose ampicillin–sulbactam vs levofloxacin + colistin (n = 23) showed significantly fewer AEs with sulbactam [8.3% (1/12) vs 54.5% (6/11); OR: 0.08; 95% CI: 0.01–0.81; high-certainty evidence].³² Finally, in a trial evaluating IV ampicillin–sulbactam + nebulized colistin vs IV colistin + nebulized colistin (n = 28), the sulbactam combination had lower AE rates [33.3% (4/12) vs 50.0% (8/16); OR: 0.50; 95% CI: 0.11–2.35; high-certainty] (Fig. 19).⁵⁵

Fig. 19.

Forest plot showing adverse event rates among critically ill carbapenem-resistant Acinetobacter baumanii-infected patients treated with sulbactam-based vs non-sulbactam-based regimens in RCTs

CI, confidence interval; IV, independent variable; RCTs, randomized controlled trials

Strength of Recommendation

Strong recommendation to use sulbactam-based regimen in comparison to a non-Sulbactam regimen in critically ill patients with CRAB infection.

Certainty of evidence: Very low (Tables 5 to 8)

Table 5.

Summary of findings table presenting the certainty of evidence and effect estimates for clinical cure in critically ill patients with CRE, CRAB, and CRPA infections treated with sulbactam-based vs non-sulbactam–based regimens (for CRAB)

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non-sulbactam/non EDTA-based therapy	With sulbactam/EDTA-based therapy		Risk with non-sulbactam/non EDTA-based therapy	Risk difference with sulbactam/EDTA-based therapy
Acinetobacter (RCTs)-CCR—cefoperazone-sulbactam vs Cefepime49
148 (1 RCT)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	46/81 (56.8%)	49/67 (73.1%)	Not estimable	46/81 (56.8%)	160 fewer per 1,000 (from 310 fewer to 10 fewer)
Acinetobacter (RCTs)-CCR—sulbactam-durlobactam + imipenem–cilastatin vs Colistin combined + imipenem–cilastatin48
126 (1 RCT)	Not Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	25/63 (39.7%)	39/63 (61.9%)	Not Estimable	25/63 (39.7%)	220 fewer per 1,000 (from 390 fewer to 50 fewer)
Acinetobacter (RCTs)-CCR—meropenem + ampicillin-sulbactam vs Meropenem + colistin50
47 (1 RCT)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	18/24 (75.0%)	16/23 (69.6%)	Not estimable	18/24 (75.0%)	50 more per 1,000 (from 200 fewer to 310 more)
Acinetobacter (RCTs)-CCR—levofloxacin + high-dose ampicillin/sulbactam vs Levofloxacin +colistin28
23 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	3/11 (27.3%)	10/12 (83.3%)	Not estimable	3/11 (27.3%)	560 fewer per 1,000 (from 900 fewer to 220 fewer)
Acinetobacter (RCTs)-CCR—intravenous Ampicillin–sulbactam + nebulized colistin51
28 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	5/16 (31.3%)	4/12 (33.3%)	Not estimable	5/16 (31.3%)	20 fewer per 1,000 (from 370 fewer to 330 more)
Acinetobacter (RCTs)-CCR—total
372 (5 RCTs)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	97/195 (49.7%)	118/177 (66.7%)	Not estimable	97/195 (49.7%)	170 fewer per 1,000 (from 430 fewer to 90 more)
Acinetobacter (OBS)-CCR—imipenem/cilastatin + sulbactam vs Tigecycline alone or in combination52
386 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	184/266 (69.2%)	60/120 (50.0%)	Not estimable	184/266 (69.2%)	190 more per 1,000 (from 90 more to 300 more)
Acinetobacter (OBS)-CCR - Ampicillin-Sulbactam vs Imipenem-Cilastatin53,54
135 (2 non-randomized studies)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	50/93 (53.8%)	18/42 (42.9%)	Not estimable	50/93 (53.8%)	100 more per 1,000 (from 250 fewer to 450 more)
Acinetobacter (OBS)-CCR—sulbactam-based therapy vs tigecycline-based therapy55
168 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	56/84 (66.7%)	56/84 (66.7%)	Not estimable	56/84 (66.7%)	0 fewer per 1,000 (from 140 fewer to 140 more)
Acinetobacter (OBS)-CCR—colistin-sulbactam group vs colistin-carbapenem group56
171 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	50/102 (49.0%)	32/69 (46.4%)	Not estimable	50/102 (49.0%)	30 more per 1,000 (from 130 fewer to 180 more)
Acinetobacter (OBS)-CCR—colistin-sulbactam group vs colistin plus other agent group56
112 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	17/43 (39.5%)	32/69 (46.4%)	Not estimable	17/43 (39.5%)	70 fewer per 1,000 (from 260 fewer to 120 more)
Acinetobacter (OBS)-CCR—cefoperazone/sulbactam vs imipenem/cilastatin57
47 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	9/12 (75.0%)	27/35 (77.1%)	Not estimable	9/12 (75.0%)	20 fewer per 1,000 (from 300 fewer to 260 more)
Acinetobacter (OBS)-CCR—colistin + sulbactam combination therapy vs colistin alone56
82 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	14/47 (29.8%)	14/35 (40.0%)	Not estimable	14/47 (29.8%)	100 fewer per 1,000 (from 310 fewer to 110 more)
Acinetobacter (OBS)-CCR—cefoperazone-sulbactam vs piperacillin-tazobactam58
624 (1 non-randomized study)	Serious	Not serious	Not serious	Not Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	298/370 (80.5%)	210/254 (82.7%)	Not estimable	298/370 (80.5%)	20 fewer per 1,000 (from 80 fewer to 40 more)
Acinetobacter (OBS)-CCR—total
1783 (10 non-randomized studies)	Serious	Serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	690/1045 (66.0%)	464/738 (62.9%)	Not estimable	690/1045 (66.0%)	20 more per 1,000 (from 50 fewer to 90 more)

Outcome: Clinical cure rate. CCR, clinical cure rate; CI, confidence interval; CRAB, carbapenem-resistant Acinetobacter baumanii; CRE, carbapenem-resistant Enterobacterales; CRPA, carbapenem-resistant Pseudomonas aeruginosa; EDTA, ethylene diamine tetra-acetic acid; OBS, observational; RoB, risk of bias; RCT, randomized controlled trial

Table 8.

Summary of findings table presenting the certainty of evidence and effect estimates for adverse events in critically ill patients with CRE, CRAB, and CRPA infections treated with sulbactam-based vs non-sulbactam-based regimens (for CRAB)

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non sulbactam/non EDTA-based therapy	With sulbactam/EDTA-based therapy		Risk with non sulbactam/non EDTA-based therapy	Risk difference with sulbactam/EDTA-based therapy
Acinetobacter (RCTs)-AER—cefoperazone-sulbactam vs cefepime49
166 (1 RCT)	Serious	Not serious	Not serious	Serious	None	⨁⨁◯◯ Low	58/87 (66.7%)	58/79 (73.4%)	OR: 1.38 (0.71–2.70)	58/87 (66.7%)	67 more per 1,000 (from 80 fewer to 177 more)
Acinetobacter (RCTs)-AER—sulbactam-durlobactam + imipenem–cilastatin vs colistin combined + imipenem–cilastatin48
177 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	42/86 (48.8%)	36/91 (39.6%)	OR: 0.69 (0.38–1.24)	42/86 (48.8%)	91 fewer per 1,000 (from 222 fewer to 54 more)
Acinetobacter (RCTs)-AER—meropenem + ampicillin-sulbactam vs meropenem + colistin50
47 (1 RCT)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	3/24 (12.5%)	1/23 (4.3%)	OR: 0.32 (0.03–3.31)	3/24 (12.5%)	81 fewer per 1,000 (from 121 fewer to 196 more)
Acinetobacter (RCTs)-AER—levofloxacin + high dose ampicillin/sulbactam vs levofloxacin + colistin28
23 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	6/11 (54.5%)	1/12 (8.3%)	OR: 0.08 (0.01–0.81)	6/11 (54.5%)	458 fewer per 1,000 (from 534 fewer to 53 fewer)
Acinetobacter (RCTs)-AER—intravenous ampicillin–sulbactam + nebulized colistin51
28 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	8/16 (50.0%)	4/12 (33.3%)	OR: 0.50 (0.11–2.35)	8/16 (50.0%)	167 fewer per 1,000 (from 401 fewer to 201 more)
Acinetobacter (RCTs)-AER—total
441 (5 RCTs)	Serious	Serious	Not serious	Not serious	None	⨁⨁◯◯ Low	117/224 (52.2%)	100/217 (46.1%)	OR: 0.66 (0.23–1.93)	117/224 (52.2%)	103 fewer per 1,000 (from 321 fewer to 156 more)

Outcome: AE rate. AER, adverse event rate; CI, confidence interval; CRAB, carbapenem-resistant Acinetobacter baumanii; CRE, carbapenem-resistant Enterobacterales; CRPA, carbapenem-resistant Pseudomonas aeruginosa; EDTA, ethylene diamine tetra-acetic acid; RCT, randomized controlled trial; RoB, risk of bias

Table 6.

Summary of findings table presenting the certainty of evidence and effect estimates for microbiological cure in critically ill patients with CRE, CRAB, and CRPA infections treated with sulbactam-based vs non-sulbactam–based regimens (for CRAB)

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non-sulbactam/non EDTA-based therapy	With sulbactam/EDTA-based therapy		Risk with non-sulbactam/non EDTA-based therapy	Risk difference with sulbactam/EDTA-based therapy
Acinetobacter (RCTs)-MCR—cefoperazone-sulbactam vs cefepime49
62 (1 RCT)	Serious	Not serious	Not serious	Serious	None	⨁⨁◯◯ Low	13/33 (39.4%)	15/29 (51.7%)	OR: 1.65 (0.60–4.52)	13/33 (39.4%)	124 more per 1,000 (from 113 fewer to 352 more)
Acinetobacter (RCTs)-MCR—sulbactam-durlobactam + imipenem–cilastatin vs colistin combined + imipenem–cilastatin48
125 (1 RCT)	Not serious	Not serious	Not serious	Serious	None	⨁⨁⨁◯ Moderate	26/62 (41.9%)	43/63 (68.3%)	OR: 2.98 (1.43–6.19)	26/62 (41.9%)	263 more per 1,000 (from 89 more to 398 more)
Acinetobacter (RCTs)-MCR—meropenem + ampicillin-sulbactam vs meropenem + colistin50
47 (1 RCT)	Serious	Not serious	Not serious	Serious	None	⨁⨁◯◯ Low	21/24 (87.5%)	21/23 (91.3%)	OR: 1.50 (0.23–9.92)	21/24 (87.5%)	38 more per 1,000 (from 258 fewer to 111 more)
Acinetobacter (RCTs)-MCR—intravenous ampicillin–sulbactam + nebulized colistin51
28 (1 RCT)	Not serious	Not serious	Not serious	Very serious	None	⨁⨁◯◯ Low	1/16 (6.3%)	4/12 (33.3%)	OR: 7.50 (0.71–78.90)	1/16 (6.3%)	271 more per 1,000 (from 17 fewer to 778 more)
Acinetobacter (RCTs)-MCR—total
262 (4 RCTs)	Serious	Not serious	Not serious	Serious	None	⨁⨁◯◯ Low	61/135 (45.2%)	83/127 (65.4%)	OR: 2.48 (1.20–5.13)	61/135 (45.2%)	220 more per 1,000 (from 45 more to 357 more)
Acinetobacter (OBS)-MCR—ampicillin-sulbactam vs imipenem-cilastatin53,54
135 (2 non-randomized studies)	Serious	Serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	35/93 (37.6%)	21/42 (50.0%)	OR: 1.20 (0.56–2.59)	35/93 (37.6%)	44 more per 1,000 (from 124 fewer to 233 more)
Acinetobacter (OBS)-MCR—imipenem/cilastatin + sulbactam vs tigecycline alone or in combination52
386 (1 non-randomized study)	Serious	Not serious	Not serious	Extremely serious	Publication bias strongly suspected	⨁◯◯◯ Very low	3/266 (1.1%)	14/120 (11.7%)	OR: 11.58 (3.26–41.11)	3/266 (1.1%)	105 more per 1,000 (from 25 more to 308 more)
Acinetobacter (OBS)-MCR—colistin-sulbactam group vs colistin-carbapenem group56
171 (1 non-randomized study)	Serious	Not serious	Not serious	Not Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	83/102 (81.4%)	49/69 (71.0%)	OR: 0.56 (0.27–1.15)	83/102 (81.4%)	104 fewer per 1,000 (from 273 fewer to 20 more)
Acinetobacter (OBS)-MCR - Colistin-Sulbactam Group vs Colistin plus Other Agent Group56
112 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	35/43 (81.4%)	49/69 (71.0%)	OR: 0.56 (0.22–1.42)	35/43 (81.4%)	104 fewer per 1,000 (from 324 fewer to 47 more)
Acinetobacter (OBS)-MCR—colistin + sulbactam combination therapy vs colistin alone56
82 (1 non-randomized study)	Serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	34/47 (72.3%)	30/35 (85.7%)	OR: 2.29 (0.73–7.19)	34/47 (72.3%)	134 more per 1,000 (from 67 fewer to 226 more)
Acinetobacter (OBS)-MCR—total
944 (7 non-randomized studies)	Serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	204/579 (35.2%)	181/365 (49.6%)	OR: 1.25 (0.88–1.78)	204/579 (35.2%)	52 more per 1,000 (from 29 fewer to 140 more)

Outcome: Microbiological cure rate. CI, confidence interval; CRAB, carbapenem-resistant Acinetobacter baumanii; CRE, carbapenem-resistant Enterobacterales; CRPA, carbapenem-resistant Pseudomonas aeruginosa; EDTA, ethylene diamine tetra-acetic acid; MCR, microbiological cure rate; OBS, observational; RCT, randomized controlled trial; RoB, risk of bias

Table 7.

Summary of findings table presenting the certainty of evidence and effect estimates for mortality in critically ill patients with CRE, CRAB, and CRPA infections treated with sulbactam-based vs non-sulbactam–based regimens (for CRAB)

Certainty assessment							Summary of findings
Participants (studies) Follow-up	RoB	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
							With non sulbactam/non EDTA-based therapy	With sulbactam/EDTA-based therapy		Risk with non sulbactam/non-EDTA-based therapy	Risk difference with Sulbactam/EDTA-based therapy
Acinetobacter (RCTs)-MR—cefoperazone-sulbactam vs cefepime49
166 (1 RCT)	Serious	Not serious	Not serious	Very serious	None	⨁◯◯◯ Very low	2/87 (2.3%)	7/79 (8.9%)	RR: 3.85 (0.82–18.01)	2/87 (2.3%)	66 more per 1,000 (from 4 fewer to 391 more)
Acinetobacter (RCTs)-MR—sulbactam-durlobactam + imipenem–cilastatin vs colistin combined + imipenem–cilastatin48
125 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	20/62 (32.3%)	12/63 (19.0%)	RR: 0.59 (0.32–1.10)	20/62 (32.3%)	132 fewer per 1,000 (from 219 fewer to 32 more)
Acinetobacter (RCTs)-MR—meropenem + ampicillin-sulbactam vs meropenem + colistin50
47 (1 RCT)	Serious	Not serious	Not serious	Not serious	None	⨁⨁⨁◯ Moderate	10/24 (41.7%)	9/23 (39.1%)	RR: 0.94 (0.47–1.88)	10/24 (41.7%)	25 fewer per 1,000 (from 221 fewer to 367 more)
Acinetobacter (RCTs)-MR - Intravenous Ampicillin–sulbactam + nebulized colistin
28 (1 RCT)	Not serious	Not serious	Not serious	Not serious	None	⨁⨁⨁⨁ High	6/16 (37.5%)	4/12 (33.3%)	RR: 0.89 (0.32–2.47)	6/16 (37.5%)	41 fewer per 1,000 (from 255 fewer to 551 more)
Acinetobacter (RCTs)-MR—total
366 (4 RCTs)	Serious	Serious	Not serious	Not serious	None	⨁⨁◯◯ Low	38/189 (20.1%)	32/177 (18.1%)	RR: 0.88 (0.35–2.26)	38/189 (20.1%)	24 fewer per 1,000 (from 131 fewer to 253 more)
Acinetobacter (OBS)-MR—cefoperazone-sulbactam vs tigecycline60
210 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	70/135 (51.9%)	22/75 (29.3%)	RR: 0.57 (0.38–0.83)	70/135 (51.9%)	223 fewer per 1,000 (from 321 fewer to 88 fewer)
Acinetobacter (OBS)-MR—sulbactam-based therapy vs tigecycline-based therapy55
168 (1 non-randomized study)	Serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	25/84 (29.8%)	28/84 (33.3%)	RR: 1.12 (0.72–1.75)	25/84 (29.8%)	36 more per 1,000 (from 83 fewer to 223 more)
Acinetobacter (OBS)-MR—colistin-sulbactam group vs colistin-carbapenem group56
171 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	56/102 (54.9%)	32/69 (46.4%)	RR: 0.84 (0.62–1.15)	56/102 (54.9%)	88 fewer per 1,000 (from 209 fewer to 82 more)
Acinetobacter (OBS)-MR—colistin-sulbactam group vs colistin plus other agent group56
112 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	24/43 (55.8%)	32/69 (46.4%)	RR: 0.83 (0.58–1.20)	24/43 (55.8%)	95 fewer per 1,000 (from 234 fewer to 112 more)
Acinetobacter (OBS)-MR—cefoperazone/sulbactam vs imipenem/cilastatin57
47 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	6/12 (50.0%)	7/35 (20.0%)	RR: 0.40 (0.17–0.96)	6/12 (50.0%)	300 fewer per 1,000 (from 415 fewer to 20 fewer)
Acinetobacter (OBS)-MR—sulbactam + imipenem/cilastatin vs tigecycline + imipenem/cilastatin61
84 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	36/56 (64.3%)	4/28 (14.3%)	RR: 0.22 (0.09–0.56)	36/56 (64.3%)	501 fewer per 1,000 (from 585 fewer to 283 fewer)
Acinetobacter (RCTs)-MR - Intravenous Ampicillin–sulbactam + nebulized colistin
624 (1 non-randomized study)	Serious	Not serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	62/370 (16.8%)	42/254 (16.5%)	RR: 0.99 (0.69–1.41)	62/370 (16.8%)	2 fewer per 1,000 (from 52 fewer to 69 more)
Acinetobacter (OBS)-MR—ampicillin-sulbactam vs imipenem-cilastatin54
58 (1 non-randomized study)	Serious	Not serious	Not serious	Serious	Publication bias strongly suspected	⨁◯◯◯ Very low	10/30 (33.3%)	12/28 (42.9%)	RR: 1.29 (0.66–2.49)	10/30 (33.3%)	97 more per 1,000 (from 113 fewer to 497 more)
Acinetobacter (OBS)-MR—total
1,532 (9 non-randomized studies)	Serious	Serious	Not serious	Not serious	Publication bias strongly suspected	⨁◯◯◯ Very low	301/860 (35.0%)	189/672 (28.1%)	RR: 0.78 (0.56–1.09)	301/860 (35.0%)	77 fewer per 1,000 (from 154 fewer to 32 more)

Outcome: Mortality rate. CI, confidence interval; CRAB, carbapenem-resistant Acinetobacter baumanii; CRE, carbapenem-resistant Enterobacterales; CRPA, carbapenem-resistant Pseudomonas aeruginosa; EDTA, ethylene diamine tetra-acetic acid; MR, mortality cure rate; OBS, observational; RCT, randomized controlled trial; RoB, risk of bias

Rationale for Strong Recommendation

Significant clinical cure rate, microbiological cure and decreased mortality with fewer adverse effects in sulbactam-based therapy. These findings are supported by at least one high-quality RCT (ATTACK), which demonstrated a significant improvement in clinical cure with sulbactam-based therapy. Although mortality benefit was not statistically significant, the direction of effect consistently favored sulbactam-based regimens, further supporting their clinical use.

Guideline Question

In critically ill patients with CRPA infections, should polymyxin B/colistin-based regimens be used instead of non-polymyxin B/non-colistin therapies (e.g., novel BL/BLI combinations such as CZA) to improve clinical outcomes such as cure rates, microbiological eradication, mortality, and safety?

Clinical Cure Rate

Clinical cure rates in CRPA infections were evaluated across three observational studies (n = 281) comparing colistin-based regimens with ceftolozane–tazobactam. Clinical cure was achieved in 67.5% of patients treated with ceftolozane–tazobactam (79/117) compared to 48.8% with colistin (80/164), yielding (OR: 0.56; 95% CI: 0.14–2.29; p = 0.42), indicating that ceftolozane–tazobactam therapy was associated with a numerically higher clinical cure rate, though the difference was not statistically significant, favoring ceftolozane–tazobactam.^66–68 The overall certainty of evidence was rated very low due to study design limitations, serious RoB, and suspected publication bias (Fig. 20).

Fig. 20.

Forest plot showing the clinical cure rate in critically ill patients with CRPA infections treated with polymyxin B/colistin-based regimens compared with non-polymyxin B/non-colistin regimens in observational studies

CI, confidence interval; CRPA, carbapenem-resistant Pseudomonas aeruginosa; M-H, Mantel-Hanszael

Microbiological Cure Rate

Microbiological cure in CRPA infections was evaluated in a single non-randomized study (n = 51) comparing colistin with ceftolozane–tazobactam. The ceftolozane–tazobactam group demonstrated a significantly higher microbiological cure rate, with 44.4% (8/18) achieving cure vs 15.2% (5/33) in the colistin group (OR: 0.22; 95% CI: 0.06–0.84; p = 0.03) indicating that colistin-treated patients had 78% lower odds of achieving microbiological cure compared to those receiving ceftolozane–tazobactam, favoring ceftolozane–tazobactam therapy.⁶⁷ The certainty of evidence was rated very low due to the inherent limitations of the observational study design and suspected publication bias (Fig. 21).

Fig. 21.

Forest plot showing the microbiological cure rate in CRPA infections treated with polymyxin B/colistin-based vs non-polymyxin B/non-colistin regimens from observational studies

CI, confidence interval; CRPA, carbapenem-resistant Pseudomonas aeruginosa; M-H, Mantel-Hanszael

Mortality Rate

Mortality outcomes in CRPA infections were reported across three non-randomized studies (n = 281) comparing colistin-based regimens with ceftolozane–tazobactam. Mortality was observed in 44.5% (73/164) of patients treated with colistin and 42.7% (50/117) of those receiving ceftolozane–tazobactam, corresponding to OR 0.83 (95% CI 0.28–2.53; p = 0.75), indicating that ceftolozane–tazobactam therapy was associated with 17% lower odds of mortality compared to colistin, though the difference was not statistically significant, suggesting comparable survival outcomes between the two groups.^66–68 The certainty of evidence was rated very low due to the observational nature of the studies, serious inconsistency, and suspected publication bias (Fig. 22).

Fig. 22.

Forest plot comparing all-cause mortality in critically ill patients with CRPA infections treated with polymyxin B/colistin vs non-polymyxin B/non-colistin regimens in observational studies

CI, confidence interval; CRPA, carbapenem-resistant Pseudomonas aeruginosa; IV, independent variable; M-H, Mantel-Hanszael

AE Rate

The safety profile of colistin compared to ceftolozane–tazobactam in Pseudomonas aeruginosa infections was assessed across three observational studies (n = 281). Adverse events occurred in 42.1% (69/164) of patients receiving colistin vs 17.9% (21/117) in those treated with ceftolozane–tazobactam, yielding an pooled OR 2.89 (95% CI: 0.93–9.04; p = 0.07), indicating that patients treated with colistin had nearly threefold higher odds of experiencing AEs compared to those receiving ceftolozane–tazobactam, with a trend favoring ceftolozane–tazobactam for better safety, though the difference was not statistically significant.^66–68 The certainty of evidence was rated very low due to serious concerns regarding RoB, inconsistency, and imprecision, underscoring the need for cautious interpretation in critically ill patients (Fig. 23).

Fig. 23.

Forest plot depicting adverse event rates among critically ill patients with CRPA infections treated with polymyxin B/colistin vs non-polymyxin B/non-colistin therapy in observational studies

CI, confidence interval; CRPA, carbapenem-resistant Pseudomonas aeruginosa; M-H, Mantel-Hanszael

Experts Recommendation

Weak recommendation to use non-polymyxin-based therapy (Ceftolazone-Tazobactam or other available therapy) in comparison to colistin/polymyxin-based therapy in critically ill patients with CRPA infection.

Certainty of evidence: Very low (Tables 5 to 8)

Rationale for Weak Recommendation

Significantly increased adverse effect with colistin/polymyxin B-based therapy with similar clinical cure, microbiological cure, and mortality compared to non-polymyxin-based therapy (Ceftolazone-Tazobactam and other available therapy). All available evidence for this comparison was derived from observational studies, with no RCTs identified. Accordingly, the certainty of evidence was rated as very low across outcomes, and findings should be interpreted with caution.

Discussion

Our guideline synthesizes and grades the evidence for therapeutic strategies targeting CRGN infections in critically ill settings. Notably, we found that novel BL/BLI combinations consistently outperformed polymyxin B or colistin-based regimens for CRE and many CRPA infections, while sulbactam-based therapies (including sulbactam–durlobactam) are emerging as preferred options in CRAB. These recommendations are underpinned by clinical, microbiological, and pharmacological data, with explicit consideration for Indian epidemiology.

CRE Infections

This systematic analysis of CRE infections encompassed 36 studies evaluating polymyxin/colistin-based therapy, demonstrating clinical cure rates of 58.7% with CZA compared to 48.9% with colistin/polymyxin-based regimens. Microbiological cure rates consistently favored CZA across observational studies, though statistical significance was limited by small sample sizes. AE rates were lower with novel agents compared to colistin-based regimens, though the certainty of evidence remained very low due to methodological limitations and RoB. Our GRADE recommendations for CRE infections are conditional, based on very low to low certainty evidence.

Indian surveillance data strongly corroborate our conservative estimates, demonstrating a marked decline in carbapenem effectiveness against E. coli and K. pneumoniae over time. Indian ICU studies highlight a substantial burden of CRE colonization with high progression to infection, increased mortality, and prolonged ICU stay.⁶⁹ This finding is further supported by Indian data from a study by Prayag et al., which reported improved clinical cure and survival with ceftazidime–avibactam compared to polymyxin-based regimens, reinforcing the limited effectiveness and higher toxicity burden associated with polymyxins.⁷⁰

Importantly, these findings are highly relevant in the Indian context, where OXA-48-like carbapenemases constitute a major proportion of CRE isolates. Ceftazidime–avibactam retains activity against OXA-48–producing organisms, which likely contributes to the improved clinical and microbiological outcomes observed.⁷¹ Mechanistically, BL/BLI combinations offer more favorable PK/PD profiles, improved tissue and lung penetration, and targeted inhibition of serine β-lactamases such as OXA-48, in contrast to polymyxins, which have suboptimal PK/PD and higher toxicity.⁷²

International data reflect similar profiles. The EURECA study reported 23.8% 30-day mortality, and global studies of CZA show clinical success rates between 59 and 77.3%, higher than our pooled estimate of 58.7%.^73–75 The TANGO II trial reported meropenem-vaborbactam success rates of 65.6% vs 33.3% for best available therapy, while the recent meta-analysis of real-world cohorts shows 60–75% cure.^76,77 International cohorts receiving CZA had 76% 30-day survival, 59% clinical success, and lower AKI rates (10%) than carbapenem-colistin combinations (30%).⁷⁴

However, the high prevalence of NDM-producing Enterobacterales in India poses a major limitation for CZA monotherapy, as it lacks activity against metallo-β-lactamases.^6,78 In this context, aztreonam–avibactam has emerged as a promising option, demonstrating approximately 80% susceptibility in NDM-positive isolates in CRE infections.⁷⁹ Moreover, the combination of CZA with aztreonam has shown good synergistic activity in MBL-producing Enterobacterales.^80,81 An Indian study conducted by Vijaykumar et al. (2023) reported a clinical success rate of 43.3% with colistin, compared to 59.3% with the combination of CZA and aztreonam, indicating improved outcomes with the latter regimen. These novel strategies offer a potential path forward in Indian ICUs, where treatment options for NDM-mediated resistance remain limited.

CRAB Infections

In the management of CRAB infections, our systematic analysis highlights the superior clinical efficacy and safety profile of sulbactam-based regimens compared to polymyxin-based therapies. The clinical cure rates range from 48.7% with polymyxin-based therapy to higher rates with sulbactam-based treatments. Notably, sulbactam possesses intrinsic antibacterial activity against Acinetobacter baumannii, mediated through high-affinity binding to penicillin-binding proteins (particularly PBP1 and PBP3), independent of its β-lactamase inhibitory effect. This unique mechanism likely explains the consistent microbiological advantage observed with sulbactam-based regimens across studies.⁸² Incorporating the ATTACK trial, we found sulbactam-durlobactam to be superior to colistin, achieving cure rates of (61.9 vs 39.7%), with a significantly lower nephrotoxicity burden (13 vs 38%).³⁰ This high-certainty evidence supports sulbactam-durlobactam as the preferred agent, in contrast to the very low certainty associated with polymyxin-based regimens, which had AE rates approaching 29.5% in our pooled analysis. Our GRADE recommendations strongly favor sulbactam-based regimen over colistin/polymyxins where available. A dissociation between microbiological and clinical outcomes is evident in CRAB infections, where improved eradication rates with sulbactam-based therapies do not consistently translate into better clinical cure or survival. This highlights the need to interpret microbiological endpoints with caution as surrogate markers in MDR infections, where host factors, infection site, and timing of appropriate therapy critically influence outcomes.⁸³

Indian CRAB surveillance shows carbapenem resistance as high as 87.5%, with recent data confirming that up to 97% of A. baumannii isolates in Indian ICUs are carbapenem resistant.⁸⁴ Mortality rates reported in Indian studies are substantial and align with our findings, particularly among NDM-positive isolates, where adverse outcomes are driven by prolonged ICU stay, prior antimicrobial exposure, and inadequate therapy.⁸⁵ Polymyxin-associated nephrotoxicity in Indian studies ranged from 37.5 to 59.3%, exceeding our pooled estimate (i.e., 29.5%).^86,87 Despite their widespread use, even combination strategies (e.g., polymyxin plus meropenem) have not demonstrated consistent mortality benefit.⁸⁴

Internationally, the ATTACK trial offers robust data across 59 centers with mostly ICU patients, confirming sulbactam-durlobactam's clinical advantage and safety. In that mortality was 19% with sulbactam-durlobactam vs 32.3% with colistin.⁵² Polymyxin nephrotoxicity varies geographically—39.9% in South America, 36.4% in Europe, 34.0% in Asia—and remains high in ICU cohorts (34.8%).⁸⁸ These figures reinforce the nephrotoxicity concerns associated with polymyxins. The resistance surveillance in Greece shows 87.9% of CRAB isolates remain susceptible to sulbactam-durlobactam.⁸⁹ Similarly, the IDSA 2024 guidelines recommend sulbactam-durlobactam, in combination with meropenem or imipenem-cilastatin, as the preferred agent for CRAB infections. High-dose ampicillin-sulbactam (27 gm/day) is now considered an alternative when sulbactam-durlobactam is unavailable.⁹⁰

CRPA Infections

Carbapenem-resistant Pseudomonas aeruginosa remains one of the most challenging pathogens in Indian ICUs due to its high-level resistance, rapid disease progression, and the absence of standardized, high-certainty treatment options. Our systematic analysis highlights poor clinical outcomes with traditional polymyxin-based regimens, where cure rates often fall below 50% and 30-day mortality in severe infections frequently exceeds 35%.

National surveillance data from 2023 reports carbapenem resistance rates of 38.5% for imipenem and 34.5% for meropenem.⁹¹ Globally, CRPA outcomes vary significantly, with 30-day mortality ranging from 6% in China to 29% in the Middle East—reflecting disparities in resistance mechanisms, healthcare infrastructure, and access to novel therapies.⁹²

Beyond resistance patterns, there is a clear lack of Indian cohort data reporting isolate-specific treatment outcomes, underscoring the urgent need for prospective, pathogen-focused research in ICU settings to guide optimized therapy and improve patient survival. The safety profile of polymyxins further limits their utility, with nephrotoxicity affecting up to 34.8% of critically ill patients and approximately 28% requiring renal replacement therapy.^88,93–95 Given the limited efficacy and high toxicity associated with polymyxins, there is a critical unmet need for alternative therapies that are both safer and clinically effective in the Indian critical care setting.

The molecular epidemiology of CRPA in India is characterized by widespread metallo-β-lactamase (MBL) production, particularly NDM variants, with reported prevalence ranging from 17.7% to as high as 90% across regions.^93–95 This is clinically significant, as MBLs confer resistance to most BL/BLI combinations, including newer agents, thereby narrowing treatment options. Additionally, high genetic diversity among Indian CRPA isolates, with international surveillance identifying up to 46 distinct sequence types, indicates polyclonal evolution rather than clonal spread. The emergence of high-risk clones, such as ST357 and ST1971, suggests a convergence of multidrug resistance and enhanced virulence.⁹⁶

Amid this complex resistance landscape, global experience has demonstrated the therapeutic potential of novel agents. Ceftazidime-avibactam has emerged as a cornerstone in CRPA treatment, with clinical cure rates ranging from 72 to 89% and 30-day mortality as low as 15–24%.^97,98 In a Chinese propensity-matched study, bacterial clearance was significantly higher with CZA compared to polymyxin B, highlighting the superior clinical efficacy of CZA.⁹⁹ Similarly, ceftolozane-tazobactam has shown consistent efficacy, with reported mortality rates between 17 and 21%.^100,101

Cefiderocol—a siderophore cephalosporin with a unique iron-transport-mediated uptake mechanism marks a paradigm shift in targeting MBL-producing pathogens, with real-world data demonstrating favorable clinical outcomes and mortality profiles.^102,103 Aztreonam–avibactam, specifically designed to overcome MBL-mediated resistance, has emerged as another promising option.¹⁰⁴ In the phase 3 ASSEMBLE trial, aztreonam–avibactam demonstrated improved clinical outcomes compared to best available therapy in severely ill patients, suggesting its potential to address a critical therapeutic gap in MBL-endemic settings such as India.¹⁰⁴

Newer therapies, including BL/BLI combinations (such as CZA, meropenem-vaborbactam, imipenem-relebactam), cefiderocol, and sulbactam-based regimens, offer meaningful clinical promise, delivering higher cure rates and improved safety profiles when guided by local susceptibility patterns and patient-specific considerations.^105–107 However, in India, their real-world potential is limited by persistent challenges: Inconsistent availability, cost barriers, delays in microbiological diagnosis, and the absence of unified national treatment guidelines continue to restrict their uptake and clinical benefit.^108,109

A key limitation of this analysis is the predominant reliance on observational data, with limited availability of RCTs for several comparisons, resulting in overall low to very low certainty of evidence. Additionally, clinical heterogeneity across included studies, particularly with respect to infection sites (e.g., bloodstream infections, pneumonia, intra-abdominal infections), resistance mechanisms, and treatment regimens, may limit the generalizability of findings and introduce indirectness in the evidence base.

The systematic search was restricted to PubMed, which may have led to the omission of relevant studies indexed in other databases such as Embase, the Cochrane Library, or regional repositories like IndMED. Furthermore, significant clinical heterogeneity among the included studies, stemming from differences in infection sites, resistance mechanisms, and treatment regimens, limited the ability to derive highly specific conclusions. Although the GRADE methodology provides a structured and transparent approach to assessing the certainty of evidence, it may not fully capture critical contextual factors such as local AMR patterns, MIC breakpoints, or PK/PD-based dosing strategies -factors that are particularly relevant in the Indian ICU setting. Finally, limited availability and accessibility of newer antimicrobial agents in India, driven by cost and regulatory factors, may impact the real-world applicability and implementation of these recommendations despite favorable efficacy and safety profiles.

Despite these limitations, this study represents the first GRADE-based initiative to offer evidence-informed, context-specific treatment recommendations for MDR infections in Indian critical care. Aligning available clinical data with the country's AMR landscape, it offers a pragmatic and locally relevant framework to support treatment decision-making in the ICU. Importantly, the analysis also highlights persistent evidence gaps and reinforces the urgent need for Indian patient populations to be more comprehensively represented in future antimicrobial clinical trials.

Conclusion

This systematic review, using the GRADE framework, evaluated current and emerging treatment options for CRE, CRAB, and CRPA in critically ill Indian patients. It identified the continued limitations of polymyxin-based regimens and highlighted newer agents with better efficacy and safety where susceptibility allows. The findings offer an evidence-based reference to support treatment choices in Indian ICUs. Importantly, these recommendations represent the best available evidence in a high-mortality setting with limited therapeutic options, rather than definitive standards of care. Incorporating these insights into clinical practice may improve patient outcomes. Ongoing surveillance and India-specific data will be vital to refine these recommendations over time.

Declaration on the Use of Artificial Intelligence (AI)

The authors declare that AI tools were used solely to assist with language refinement, grammar correction, and improving the clarity of the manuscript.

Acknowledgments

The authors would like to thank the medical writing and editorial assistance provided by Dr Alok Nachane, Director, VeritasQ Research and Analytics Pvt. Ltd., in the preparation of this manuscript.

Orcid

Subhash K Todi https://orcid.org/0000-0003-2306-6080

Bharat G Jagiasi https://orcid.org/0000-0002-3068-1201

Deven Juneja https://orcid.org/0000-0002-8841-5678

Justin A Gopaldas https://orcid.org/0000-0002-0420-6925

Kapil D Soni https://orcid.org/0000-0003-1214-4119

Sunil Karanth https://orcid.org/0000-0003-3597-4473

Ritu Singh https://orcid.org/0000-0001-9855-456X

Mohit Kharbanda https://orcid.org/0000-0002-9625-1909

Ashutosh Bhardwaj https://orcid.org/0000-0002-8940-3898

Tapas K Sahoo https://orcid.org/0000-0001-8174-5992

Shilpushp Bhosale https://orcid.org/0000-0002-0290-0526

Sharmili Sinha https://orcid.org/0000-0001-5242-9405

Manish Bharti https://orcid.org/0009-0009-7310-3266

Anuj M Clerk https://orcid.org/0000-0001-8944-0077

Amit Kumar https://orcid.org/0000-0002-5064-2859

Bram Rochwerg https://orcid.org/0000-0002-8293-7061

Vinod K Singh https://orcid.org/0000-0002-1714-6466

Pradip K Bhattacharya https://orcid.org/0000-0002-0219-385X