Comparative Effectiveness of Colistin Sodium versus Polymyxin B Sulfate Against CR-GNB: A Retrospective Cohort Study

To compare the clinical effectiveness of colistin sodium (CS) and polymyxin B sulfate (PMB) in the treatment of carbapenem-resistant Gram-negative bacteria (CR-GNB) infections.

Adult patients receiving intravenous CS or PMB ≥ 72 hours between October 2021 and November 2022 were retrospectively enrolled. Demographic information and pertinent clinical laboratory data were collected for each patient. The primary outcome was 28-day all-cause mortality, and the secondary outcomes consisted of microbial clearance, clinical success, and normalization of inflammatory marker. Inverse probability of treatment weighting (IPTW) was applied to address potential confounding. Cox regression with IPTW was used to evaluate the association between polymyxin type and mortality.

A total of 93 CS patients and 118 PMB patients were included in this study. After IPTW, a significant difference in 28-day all-cause mortality was observed between the CS and PMB groups (11.0% vs 26.7%, P = 0.016). However, multivariable Cox regression analysis showed that CS therapy was not independently associated with 28-day all-cause mortality (HR = 0.51, 95% CI: 0.06-4.64, P = 0.547). No significant differences in microbial clearance rate or clinical success were found between the two groups either before or after IPTW (all P > 0.05). The rate of procalcitonin normalization was significantly higher in the CS group than in the PMB group prior to IPTW (43.0% vs 28.0%, P = 0.033). However, this difference did not persist after weighting (39.6% vs 30.0%, P = 0.177). No between-group differences were observed for other inflammatory markers.

Although IPTW-adjusted analysis revealed a significant difference in mortality between CS and PMB groups, multivariable Cox regression incorporating additional covariates did not confirm an independent protective effect of CS. This suggests that the observed survival benefit may be attributable to confounding, and the finding warrants validation in larger prospective multicenter studies.