Host inflammation drives low-exposure nephrotoxicity in critically ill patients receiving polymyxin B: A prospective...

International guidelines recommend a polymyxin B (PMB) steady-state area under the curve (AUC_ss,24h) target of 50.0-100.0 mg·h/L to balance efficacy and safety. However, the real-world predictive value of this exposure threshold for PMB-associated acute kidney injury (PMB-AKI) remains controversial.

In this prospective cohort study of 96 critically ill patients, we evaluated clinical efficacy and nephrotoxicity across stratified PMB exposure levels. Predictors of PMB-AKI were identified using multivariable logistic regression. Non-linear relationships were modeled using restricted cubic splines (RCS). Predictive performance of thresholds was assessed using ROC curves and F1-scores, supplemented by a meta-analysis pooling our cohort with eight published studies (N = 1053).

PMB-AKI occurred in 49.0% (47/96) of patients. Maintaining target exposure (50.0-100.0 mg·h/L) achieved significantly higher clinical success compared to underexposure (<50.0 mg·h/L; 68.0% vs. 45.8%) without increasing AKI risk. Multivariable analysis identified AUC_ss,24h, septic shock, and concomitant vancomycin as independent risk factors for PMB-AKI. While concentration-driven control is essential for efficacy, AKI risk within the target window was primarily driven by baseline white blood cell (WBC) count and concomitant vancomycin (AUROC = 0.734). RCS and predictive margin analyses identified a WBC threshold of 7.85 × 10⁹/L as a key risk differentiator (specificity: 83.3%; F1-score: 0.805). A prognostic nomogram combining these variables demonstrated distinct net clinical benefit across decision threshold probabilities of 20% to 70%.

Mitigating PMB-AKI requires a bimodal surveillance strategy: while exposure control remains essential, clinical monitoring must pivot toward host-centered inflammatory dynamics and polypharmacy when drug exposure is maintained within or below the conventional target window.