Unveiling the immuno-inflammatory endotype of sepsis-associated acute kidney injury: a machine learning-based approach.

Abstract

Background Sepsis-associated acute kidney injury (SA-AKI) is a frequent and high-mortality complication in critically ill patients. Current diagnostic criteria rely on functional markers that often lag behind the onset of renal injury. While immune dysregulation is central to SA-AKI pathogenesis, few prediction models systematically integrate multidimensional immune phenotypes. Methods The present analysis was based on a multicenter prospective cohort study involving 1,551 septic patients. The study was carried out across five tertiary hospitals in Beijing. Clinical data and immuno-inflammatory biomarkers (including humoral, complement, and T lymphocyte subsets) were collected within 24 hours of sepsis diagnosis. An Extreme Gradient Boosting (XGBoost) model was then developed to predict SA-AKI. Results New-onset SA-AKI occurred in 44.8% of the cohort and was associated with increased mortality. We identified a distinct high-risk immuno-inflammatory endotype characterized by elevated immunoglobulin A and procalcitonin (PCT), consumptive complement component 3 depletion, and a "complex and divergent cellular immune pattern"-manifested as depleted naïve CD4+ T cells coexisting with expanded CD28+CD4+ T cells. Integrating these immune features with routine clinical variables (Acute Physiology and Chronic Health Evaluation II (APACHE II), the Sequential Organ Failure Assessment [SOFA], PCT, prothrombin time [PT], sex) yielded an XGBoost-based clinical-immune model with superior discrimination (area under the curve [AUC], 0.914), significantly outperforming the clinical model (AUC, 0.788). Decision curve analysis further confirmed greater net clinical benefit for the integrated model. Conclusion Immune dysregulation plays an important role in developing AKI in sepsis patients. Immune biomarkers could significantly improve early identification of high SA-AKI risk patients.