Urinary microRNAs in sepsis-associated acute kidney injury: A LASSO-logistic regression approach for risk prediction...

To develop and validate a urine microRNA (miR)-based risk model using LASSO-logistic regression for early prediction of sepsis-associated acute kidney injury (SA-AKI).

We enrolled 210 septic patients admitted between June 2023 and May 2025. Patients were stratified into AKI (n = 72) and non-AKI (n = 138) groups based on AKI onset within 7 days of admission. Clinical variables and urinary miRNA levels (miR-16, miR-21, miR-142, miR-30c, miR-31) were analyzed by RT-qPCR. LASSO regression identified key predictors; multivariate logistic modeling generated a nomogram for SA-AKI risk prediction. Performance was assessed using ROC analysis and calibration plots.

SA-AKI patients had lower MAP, reduced first 24-hour urine output, higher lactate levels, and elevated SOFA scores (P < 0.05). LASSO selected seven predictors: miR-21, miR-142, miR-16, Lactate, MAP, SOFA score, and urine output. Multivariate analysis confirmed mir-16, miR-21, miR-142, Lactate, and SOFA as independent risk factors (P < 0.05), while higher urine output in first 24 hour and MAP were protective (P < 0.05). The nomogram achieved an AUC of 0.993 (95% CI: 0.986-0.999) with near-perfect calibration (c-index = 0.992, 95% CI: 0.987 - 0.998).

A urine miRNA and clinical parameter-based LASSO-logistic model offers exceptional accuracy in predicting early SA-AKI, enabling timely risk stratification.