Diagnostic Biomarkers and Metabolic Mechanisms of Polymyxin B-Induced Acute Kidney Injury

Background/Objectives: The clinical utility of polymyxin B (PB), a critical last–line antibiotic for multidrug–resistant Gram–negative infections, is severely compromised by acute kidney injury (AKI). However, the lack of specific diagnostic biomarkers and the incomplete understanding of pathogenic mechanisms pose substantial challenges to clinical management of PB–associated AKI. Methods: We enrolled 83 patients receiving PB regimens (23 AKI, 60 non–AKI) and quantified the plasma concentrations of five PB components to identify the optimal therapeutic drug monitoring (TDM) metric for nephrotoxicity. A rat model of PB–induced AKI was established to map renal PB distribution via mass spectrometry imaging and screen targets for tubular accumulation virtually. Untargeted metabolomics was performed on clinical and rat samples to identify metabolic dysregulation, with key pathways validated by reverse transcription quantitative real–time PCR (RT–qPCR). Diagnostic models were constructed using the eXtreme Gradient Boosting (XGBoost) algorithm and evaluated using Area Under the Curve (AUC). Shapley Additive exPlanations (SHAP) analysis was applied to interpret the models, and decision curve analysis (DCA) was used to assess clinical utility. Results: A novel PB–component model outperformed the conventional TDM index (AUC 0.703 vs. 0.561), and a 6–metabolite exploratory classifier achieved 0.95 accuracy and AUC 1.000 for PB–AKI discrimination on an independent test set. Mechanistically, high PB concentrations might saturate P–glycoprotein (P–gp), leading to PB accumulation in renal tubules and subsequent metabolic dysregulation, manifested as upregulation of fatty acid pathways and downregulation of the TCA cycle. Conclusions: These findings advance our understanding of metabolic dysregulation in PB–induced AKI and support the potential utility of responsive metabolic features in the clinical diagnosis of PB-induced AKI.