Drug-resistant Gram-negative bacterial (DR-GNB) infections, particularly those caused by carbapenem-resistant organisms (CRO), currently pose one of the most serious challenges to global public health (Holmes et al., 2021). It is crucial to note that infections caused by CRO, such as pneumonia, sepsis, and complicated intra-abdominal infections, are associated with high prevalence and significant mortality rates (Gurjar et al., 2013; Shah and Shah, 2015). This global crisis is exacerbated by the fact that CROs exhibit resistance to the majority of currently available antibiotics, severely limiting therapeutic options. In this context, colistin has emerged as the last-line therapeutic option for combating CRO infections (Lu et al., 2022). However, the clinical utility of colistin is now being challenged by the emergence and global dissemination of plasmid-mediated colistin resistance genes (mcr). These genes have been identified in isolates from both animal and environmental sources across multiple countries, and clinical isolates harboring mcr genes are being increasingly reported (Mohsin et al., 2018; Papst et al., 2018). To date, more than 11 mcr gene subtypes have been identified (Li et al., 2017), highlighting the urgent need to strengthen prevention and control measures against polymyxin antimicrobial resistance.

Colistin is a concentration-dependent, cationic polypeptide antibiotic that lacks a post-antibiotic effect (Gunderson et al., 2003). It is administered intravenously in the form of colistin methanesulfonate sodium (CMS), which exerts its antibacterial activity by binding to bacterial lipopolysaccharide and disrupting the integrity of the outer membrane. CMS is primarily eliminated through renal excretion and may undergo renal tubular reabsorption (Li et al., 2004; Kristoffersson et al., 2020). The clinical effectiveness of colistin is believed to be influenced by a variety of factors, including dosing strategy, co-therapy with other antimicrobial agents, adjunctive aerosolized therapy, and baseline clinical characteristics, among others. However, existing studies have not demonstrated consistent findings (Nation et al., 2017; Ahmadpour et al., 2022). Moreover, due to its significant nephrotoxicity, specifically AKI, colistin was once regarded as a suboptimal choice. Whether colistin-associated AKI has clinical importance remains a matter of debate (Dalfino et al., 2012; Gurjar et al., 2013). While early clinical and preclinical studies reported significant rates of nephrotoxicity and neurotoxicity following systemic administration, more recent data suggest that colistin may be safer than previously thought, with a lower incidence of AKI in some cohorts (Falagas and Kasiakou, 2005). Prior studies indicate a 20%–60% incidence of colistin-associated AKI, which is likely influenced by a multitude of factors including baseline glomerular filtration rate, severity of infection, hemodynamic status, colistin dose, and concomitant use of nephrotoxic drugs. In addition, some studies have shown that colistin-associated AKI occurs early, is not severe, and does not lead to discontinuation of colistin (Gurjar et al., 2013).

To address these controversies, this meta-analysis aims to comprehensively analyze the factors influencing the effectiveness and safety of colistin in treating drug-resistant Gram-negative bacterial infections. We take into account three aspects: drugs, microorganisms, and patient-related factors. From the drug perspective, we focus on the daily dose, combination drug therapy, and adjuvant aerosolized administration; and for microorganisms, emphasis is placed on pathogenic bacteria species, for patient-related factors; key considerations include age and comorbidities.

Our meta-analysis was conducted in strict accordance with the PRISMA 2020 guidelines (Rethlefsen and Page, 2021), and has been registered in PROSPERO (Registration No.: CRD420250655507).

This meta-analysis systematically evaluated the factors influencing the efficacy and safety of colistin in treating drug-resistant Gram-negative bacterial infections from three dimensions: drugs, host-related factors, and microbial characteristics.

Our findings revealed that the daily dose of colistin showed no statistically significant differences in efficacy outcomes including mortality rates and clinical response rates, which aligns with previous studies (Katip et al., 2017; Alp et al., 2017). On the one hand, pharmacokinetic studies have demonstrated that the 24-h area under the free drug plasma concentration-time curve to minimum inhibitory concentration ratio (fAUC/MIC) serves as the best correlate of antibacterial activity for colistin following intravenous administration of CMS (Couet et al., 2012). However, routine monitoring of fAUC remains clinically impractical due to technical limitations. Notably, Plachouras et al., 2009 found that when maintaining a constant daily CMS dose of 9 MIU, there was no significant difference in peak CMS concentrations between the three-dose regimen (3 MIU every 8 h) and the twice-daily regimen (4.5 MIU every 12 h). In contrast, a single daily administration of the equivalent total dose resulted in a statistically significant elevation of peak plasma concentrations compared to divided-dose regimens. Additionally, the study revealed that colistin exhibited a flatter plasma concentration-time profile compared to CMS, which was consistent with its longer elimination half-life (Plachouras et al., 2009). These findings suggest that, unlike polymyxin B sulfate, the daily dose of CMS shows no statistically significant correlation with clinical outcomes.

Conversely, a loading dose of colistin enables rapid achievement of steady-state concentrations (Css), which may exert a more substantial influence on therapeutic efficacy. Our results, in line with multiple prior studies (Li et al., 2004; Landersdorfer and Nation, 2015; Nazer and Anabtawi, 2017), support the use of an intravenous loading dose in critically ill patients with CRO infections. The study by Martinez et al. (2012) demonstrated that adequate drug exposure at the initiation of therapy is essential to optimizing infection-related clinical outcomes (Martinez et al., 2012). While Kumar et al. (2006) demonstrated in animal models that delayed antimicrobial administration significantly increases mortality (Kumar et al., 2006). It is worth noting that colistin is administered as the prodrug CMS, which predominantly requires renal metabolic activation before being reabsorbed into the systemic circulation to exert its bactericidal effects. This pharmacokinetic pathway suggests that patients’ baseline glomerular filtration rate (GFR) may substantially influence optimal colistin dosing regimens (Gautam et al., 2018). This may explain the absence of a statistically significant difference in therapeutic efficacy between high- and low-dose subgroups observed in our meta-analysis. Other prior studies have demonstrated that colistin exhibits significantly enhanced therapeutic efficacy when administered at elevated doses. Compared with our study, these studies have predominantly enrolled homogeneous patient cohorts restricted to specific infection sites such as burn injuries and HAP/VAP, consequently limiting the generalizability of their findings to broader patient populations (Falagas et al., 2010). In clinical practice, colistin is typically administered using fixed or weight-based dosing regimens without a defined target mean Css. As a result, interpatient variability in Css is substantial (Nation, Velkov, and Li, 2014), and the fractional conversion of CMS to colistin adds another layer of unpredictability (Grégoire et al., 2014).

Similar to findings on efficacy outcome, existing research remains inconclusive regarding the association between daily colistin dosing and AKI incidence. Our analysis demonstrated no statistically significant correlation between colistin daily dosing and AKI occurrence (p = 0.74), which is consistent with previous reports (Dalfino et al., 2012; Kwon et al., 2015; Wilkinson et al., 2017). However, conflicting evidence exists as several studies suggested a dose-dependent increase in AKI risk with high-dose colistin regimens (Vicari et al., 2013; Benattar et al., 2016; Omrani et al., 2015). These discrepancies likely stem from heterogeneity in patient populations, inconsistencies in AKI diagnostic criteria (KDIGO vs. RIFLE), and variation in controlling for confounders such as concurrent nephrotoxic medications. Furthermore, a significant increase in the incidence of AKI is observed only when the average Css of colistin exceeds 2 mg/L. However, achieving colistin Css levels above this threshold proves clinically challenging in most patients, as such concentrations are rarely attained without administering an initial loading dose (Couet et al., 2012). The heterogeneity in patients’ baseline renal function partially contributes to the occurrence of colistin-associated AKI. The current international consensus recommends maintenance doses based on individuals with normal renal function; therefore, our subgroup definitions for high- and low-dose were grounded on this baseline. In patients with impaired renal function, unchanged daily dosing may inadvertently result in plasma concentrations exceeding 2 mg/L, which might enhance bactericidal activity but also increase nephrotoxicity. Therefore, in our low-dose subgroup, the actual Css in renally impaired patients may have been higher than theoretical predictions, confounding the observed associations. Therapeutic drug monitoring (TDM) may help optimize exposure and minimize toxicity, but more research is needed to determine the most appropriate patient populations and implementation strategies for TDM in colistin therapy. Furthermore, concomitant medication regimens also partially contribute to the occurrence of colistin-associated AKI. In conclusion, the loading dose of colistin is associated with both therapeutic efficacy and the risk of colistin-associated AKI, highlighting the delicate balance clinicians must navigate between maximizing antibacterial activity and minimizing nephrotoxicity.

The influence of patient’s comorbidities and physiological status on the therapeutic efficacy of colistin underscores the importance of individualized treatment strategies in clinical practice. Multiple studies have demonstrated that advanced age and obesity are independent risk factors influencing both the efficacy and safety of colistin therapy, although the specific thresholds for age and body weight associated with heightened risk vary across studies (Gauthier et al., 2012; Tuon et al., 2014). Nevertheless, the high proportion of missing values in recorded body weights among the analyzed cohort precluded further investigation into potential weight-associated effects. Renal impairment may significantly hinder colistin metabolism and clearance, increasing the risk of adverse effects, particularly nephrotoxicity (Bk et al., 2021). Additionally, immunocompromised conditions resulting from diabetes mellitus and hematologic malignancies may impair host immune responses and consequently reduce antimicrobial efficacy (Gunay et al., 2020). Based on prior studies, we employed the ACCI to comprehensively evaluate both chronological age and comorbidity burden of patient baseline status. Our findings revealed statistically significant differences in overall mortality and clinical efficacy rates between the low-ACCI group (score <5), moderate-ACCI group (score from 5 to 6), and high-ACCI group (score >6) (p < 0.01), supporting ACCI as a risk factor affecting the efficacy and safety of colistin.

Previous studies have demonstrated that the prevalence of heteroresistance to polymyxins exceeds 14% (Hawley et al., 2007), and combination therapy may mitigate bacterial heteroresistance through synergistic mechanisms. Results of our study demonstrated that the combination of colistin with carbapenems and quinolones significantly reduced mortality rates, improving clinical response and bacterial eradication rates. For infections caused by CRAB, including HAP/VAP and bloodstream infections, combination regimens of colistin with rifampicin (Durante-Mangoni et al., 2013), fosfomycin (Sirijatuphat and Thamlikitkul, 2014), or sulbactam (Ungthammakhun et al., 2024) have shown favorable therapeutic efficacy. These combination therapies exhibited significant advantages over colistin monotherapy, particularly in terms of bacterial eradication rates. In the context of intra-abdominal infections, the combination therapy of colistin and fosfomycin demonstrates comparable therapeutic advantages. Similarly, in the context of UTI and skin infections, colistin combined with fosfomycin has demonstrated comparable therapeutic advantages, which aligns consistently with the findings of our study (Durante-Mangoni et al., 2013). In our study, the colistin-fosfomycin subgroup demonstrated a notably elevated bacterial eradication rate (98.17%, 95%CI: 96.07%–100.00%), which potentially attributed to the inclusion of urinary tract infection patients where both colistin and fosfomycin achieve high urinary concentrations (Parker et al., 2015). Whereas, the results of the largest RCT to date (the AIDA study) reported no statistically significant differences between the colistin monotherapy group and the colistin-meropenem combination therapy group in clinical failure rates (79% vs. 73%; p = 0.17), 28-day mortality (43% vs. 45%; p = 0.78), or bacterial failure rates (31% vs. 35%; p = 0.49; Paul et al., 2018). Despite such discrepancies, current international guidelines recommend the use of colistin in combination with other antimicrobials for the treatment of multidrug-resistant Gram-negative infections (Tsuji et al., 2019).

On the other hand, our findings demonstrate that colistin administration in combination with nephrotoxic agents, such as glycopeptides, aminoglycoside antibiotics, and loop diuretics, significantly increases the incidence of AKI. These concomitant medications were identified as independent risk factors for colistin-associated AKI (Petrosillo et al., 2014; Temocin et al., 2015). Although most studies in this subgroup were observational and methodologically limited, the growing body of consistent evidence has strengthened the overall quality of this association. This conclusion is further supported by a comprehensive meta-analysis of 237 studies involving more than 35,000 patients (Wagenlehner et al., 2021). While prior studies primarily focused on the safety profile of colistin-based combination regimens, further research is warranted to assess the impact of specific combinations on the development of colistin resistance. Furthermore, future investigations should emphasize exploring how different anatomical sites of infection influence therapeutic outcomes in combined antimicrobial interventions.

The variation in treatment outcomes across microbial species could potentially stem from differential intrinsic resistance profiles to colistin among microorganisms. The results of our study demonstrated a statistically significant difference in overall mortality rates across various bacterial infections (p < 0.01). Specifically, CRAB infections were associated with the highest case fatality rate (43.75%, 95%CI: 37.67%–49.83%), significantly higher than the morality (28.69%, 95%CI: 21.11%–36.27%) observed in CRE infections. Furthermore, the bacterial clearance rate was significantly higher in the CRE subgroup compared to other subgroups, which may be attributed to their lower resistance to colistin. However, the limited number of studies in this subgroup, all observational studies focusing on BSI and UTI, has resulted in a downgrading of the evidence quality. Results from a large-scale international RCT demonstrated that infections caused by CRAB were associated with high mortality and clinical failure rates (Kaye et al., 2022). In a large systematic review on urinary tract infections, the resistance rate of P. aeruginosa to colistin (3.8%) was lower than that of A. baumannii (4.6%), and in this study, the resistance rate of P. aeruginosa to colistin (7.2%) was also lower than that of A. baumannii (10.0%) in a cross-sectional study of local district hospitals (Requena-Cabello et al., 2024). These studies might explain the lower mortality rate of P. aeruginosa than that of A. baumannii in our study. Compared with P. aeruginosa, A. baumannii tends to exhibit higher antimicrobial resistance, which may be related to its outer membrane protein expression, biofilm formation ability, and the presence of plasmid-mediated multidrug resistance genes. The primary bactericidal mechanism of colistin involves the interaction of its cationic nonribosomal lipopeptide with lipid A, destabilizing the outer membrane, increasing permeability, and ultimately leading to cell death. In A. baumannii, a distinctive resistance mechanism—uncommon among other Gram-negative bacteria—involves spontaneous mutations in the lipid A biosynthesis genes (lpxA, lpxC, or lpxD), which abolish lipopolysaccharide (LPS) or lipid A production and thereby remove the drug’s target. Reduced availability of essential cofactors for LPS synthesis, such as biotin, can further diminish susceptibility to polymyxins (Kyriakidis et al., 2021). Mutations in the PmrA–PmrB two-component regulatory system also contribute to resistance: Adams et al. (2009) identified alterations in both pmrA and pmrB in resistant strains (Adams et al., 2009), while Sun et al. (2020) reported two novel mutations, pmrA ^I13M and pmrB ^Q270P, associated with colistin resistance in A. baumannii (Sun et al., 2020).

Furthermore, the site of infection, geographical variations, disease severity, and baseline patient characteristics substantially influence clinical outcomes (Papst et al., 2018; Wang et al., 2023). CRE infections are frequently observed in patients with BSI and complicated UTI, whereas CRAB is more commonly associated with pneumonia. Bacterial clearance rates in invasive BSI and complicated UTI are typically higher than those in pulmonary or intracranial infections. Furthermore, the average Css of colistin is significantly elevated in serum compared to its concentrations in pulmonary and cerebral tissue (Boisson et al., 2014). Additionally, existing studies have largely failed to adequately distinguish between colonizing bacteria and invasive bacteria, which constitutes another critical contributing factor to the observed differences in treatment efficacy among various drug-resistant bacterial strains. Our findings also demonstrated a significantly lower incidence of AKI in the P. aeruginosa infection subgroup compared to other subgroups. However, due to substantial heterogeneity among studies within this subgroup, the certainty of evidence remains limited. Although several pathogen-focused RCTs have been conducted in recent years (Dl et al., 2020), these investigations pose considerable challenges, with most studies demonstrating limited patient recruitment and substantial heterogeneity across trial designs.

Due to suboptimal distribution characteristics, colistin concentrations in pulmonary tissue may be inadequate. Aerosolized administration has been shown to enhance drug concentration in the epithelial lining fluid of the lungs (Boisson et al., 2014). However, the current research remains controversial, with some studies demonstrating that intravenous adjunctive aerosolized colistin may provide additional therapeutic benefits for patients with CRO-associated pneumonia (Pérez-Pedrero et al., 2011; Moradi Moghaddam et al., 2019). Other studies have reported lower mortality rates in the intravenous adjunctive aerosolized group, while no statistically significant difference in mortality was observed between the two groups (Almangour et al., 2021; Kofteridis et al., 2010). Several clinical guidelines currently recommend intravenous adjuvant aerosolized for the treatment of CRO-associated pneumonia, although as a weak recommendation supported by low-quality evidence (Rello et al., 2017; Tsuji et al., 2019). Further studies have indicated there is no significant difference in hospital mortality (p = 0.63) or ICU mortality (p = 0.92) between combination intravenous-aerosolized therapy and intravenous infusion alone, based on multivariate regression analysis adjusting for age and comorbidities (Kofteridis et al., 2010; Korbila et al., 2010). Despite discrepancies in mortality outcomes, most clinical studies have demonstrated improved bacterial eradication and clinical response rates with adjunctive aerosolized therapy. In our analysis, bacterial clearance and clinical response rates were higher in the combination therapy group (56.02% vs. 45.56% and 62.71% vs. 48.19%, respectively), although these differences did not reach statistical significance. Most studies incorporated into this subgroup analysis consisted of observational designs with limited sample sizes and significant inter-study variability.

Although aerosolized colistin appears to be a better option for adjunctive therapy, there is a relative lack of high-quality RCTs investigating the different administration routes of colistin, and thus its exact role in treating CRO-induced pulmonary infections remains unclear (Tumbarello et al., 2013). In terms of safety assessment, existing evidence from multiple clinical studies suggests that intravenous adjunctive aerosolized colistin is not associated with an elevated risk of nephrotoxicity (specifically colistin-induced AKI), a conclusion corroborated by our observational data (Choe et al., 2019; Almangour et al., 2021). The relatively low dose of aerosolized colistin administration demonstrated minimal impact on the average Css of colistin (Karnik et al., 2013; Boisson et al., 2014), and consequently, no significant difference in colistin-associated AKI incidence was observed between the two administration methods (Gkoufa et al., 2022). Furthermore, although standalone aerosolized colistin monotherapy may improve local microbiological outcomes, it does not significantly enhance clinical prognosis in patients with HAP/VAP. Conversely, this approach has been associated with a marked increase in the incidence of tracheal spasms. Therefore, colistin is not recommended for standalone aerosolization in clinical practice (Zhang et al., 2023). Future research should prioritize investigations into the pharmacokinetic characteristics of colistin in pulmonary tissues to optimize dosing strategies.

This study has several limitations: (1) Significant heterogeneity exists among the included studies, encompassing both observational studies and RCTs. Variations in sample sizes and experimental designs may compromise the robustness of our conclusions. Subgroup analyses did not always adjust for baseline differences, potentially confounding intergroup comparisons. (2) Incomplete data on certain influencing factors limited to more in-depth analyses. Specifically, missing weight records for some patients constrained our ability to comprehensively assess weight-related impacts on pharmacokinetic parameters. Although we implemented body weight-based dose conversions and minimized weight-related variations in plasma drug concentrations through normalization procedures, residual uncertainties persist regarding the precise characterization of concentration-efficacy and concentration-safety relationships. Contrary, this conversion approach may introduce additional measurement errors during dose equivalency calculations. Furthermore, As the original studies did not systematically report drug-resistance gene data, this meta-analysis could not assess this factor. Future studies should prioritize analyzing resistance mechanisms. (3) The included studies exhibited heterogeneity in the definitions of outcome measures, which may have influenced our analytical outcomes to some extent. For example, diagnostic criteria for AKI varied significantly across studies, and the methodological adjustments required to reconcile these differing definitions could potentially introduce measurement errors. Similarly, inconsistencies were observed in mortality assessments, with studies reporting disparate endpoints such as 14-day, 28-day, or 30-day mortality rates. These variations in temporal parameters for mortality evaluation further complicated comparative analyses.

Despite certain limitations, this meta-analysis highlights the importance of a comprehensive, individualized approach that integrates host factors, microbial characteristics, and pharmacological considerations in colistin therapy. To the best of our knowledge, this study represents the most comprehensive investigation to date into the factors influencing the efficacy and safety of colistin. Compared with previous meta-analyses, our study further enhanced the methodological rigor by excluding case reports during study selection, thereby elevating the quality hierarchy of incorporated evidence. Furthermore, our findings highlight the necessity for personalized therapeutic strategies and rigorous renal function monitoring in clinical practice, providing evidential support for optimizing colistin-based regimens and informing future clinical guidelines.

Our results suggest that the colistin loading dose, ACCI score, combination therapy, and microbial factors may significantly influence its efficacy in treating drug-resistant Gram-negative infections. Specifically, administration of a colistin loading dose, lower ACCI scores, and combination therapy with carbapenem antibiotics were associated with improved treatment success rates, and infections caused by CRE exhibited higher bacterial clearance rates. Conversely, infections caused by A. baumannii might adversely affect therapeutic outcomes. Loading dose, concomitant administration of nephrotoxic medications, ACCI, and microbial factors were identified as risk factors for colistin-associated nephrotoxicity. These findings offer preliminary insights into the contentious factors influencing colistin therapy, warrant cautious interpretation, and underscore the need for future studies to better define patient subgroups, standardize treatment regimens, and evaluate outcomes to clarify these interactions and inform individualized treatment.