Establishment and validation of the prediction model based on lymphocyte subsets for acute kidney injury in sepsis patients

Leyi Wang¹Qi Liu¹Changdong Wu¹Ming Hou^1*Zhiwei Li^2*

• ¹Department of Nursing, People's Hospital of Xinjiang Uygur Autonomous Region, Urumqi, China
• ²Clinical Laboratory Center, People's Hospital of Xinjiang Uygur Autonomous Region, Urumqi, China

Purpose: This study aimed to construct a risk predictive model for acute kidney injury in sepsis based on peripheral blood lymphocyte subsets. Methods: This prospective study included patients with sepsis admitted to the ICU from March to August 2024 (483 for training and 146 for validation), and 125 patients from September to December 2024 as the external test cohort. Clinical data and peripheral blood samples on days 1 and 3 were collected after ICU admission. Lymphocyte subsets were analyzed using flow cytometry, covering T cell, B cell, NK cell populations. Differences in clinical variables and lymphocyte subsets between AKI and non-AKI groups were analyzed. A predictive model was developed using LASSO and multivariate logistic regression and validated internally (5-fold cross-validation) and externally. Model performance was assessed using ROC curves, calibration plots, and decision curve analysis (DCA). A nomogram was constructed for clinical applications. Results: Among the 483 patients, the incidence of AKI was 54.66%. Compared to non-AKI patients, the AKI group had significantly higher SOFA and APACHE II scores and lower GCS scores. Laboratory findings showed higher neutrophil and monocyte counts, and elevated serum creatinine in the AKI group. On day 1, several lymphocyte subsets were significantly altered in the AKI group, including increased CD4⁺CD38⁺T%, CD8⁺CD38⁺T%, CD155⁺T%, CD4⁺TeM⁺T%, CD8⁺TIGIT⁺T%, and M-MDSC, and decreased CD4⁺LAG3⁺T%, CD4⁺TN⁺T%, and Th17 cells. On day 3, AKI patients exhibited further distinct changes in NK cells and T cell activation/exhaustion markers. A predictive model incorporating key clinical (APACHE II and creatinine) and lymphocyte subsets (CD15⁺T%_1st, CD4⁺LAG3⁺T%_1st, Th17_1st, CD8⁺PD1⁺T%_3rd, CD8⁺TIGIT⁺T%_3rd, E_MDSC_3rd, CD8⁺CCR7⁺CD45RA⁺T%_3rd, CD4⁺CTLA4⁺T%_3rd, CD4⁺TIM3⁺T%_3rd, PMN_MDSC_3rd, and M_MDSC_3rd) achieved high accuracy, with an AUC of 0.989 in the training set, 0.895 in the validation set, and 0.906 in the test set. Calibration curves and DCA confirmed the model's reliability and clinical utility. Conclusion: Peripheral blood lymphocyte subsets are significantly altered in patients who develop SA-AKI and can serve as potential early biomarkers. The developed predictive model based on clinical and immunological parameters demonstrated robust performance in identifying patients at high risk of SA-AKI, offering a practical tool for early warning and clinical decision-making.