Early Prediction of Acute Kidney Injury Following Liver Transplantation: Development and Validation of a Clinical Risk Model

Abstract

Background

The first 48 h following liver transplantation (LT) represent a critical therapeutic window. Early identification of patients who are at high risk of developing acute kidney injury (AKI) can optimize treatment strategies and improve patient outcomes. This study aimed to develop and validate a clinical risk prediction model for AKI within 48 h following LT by utilizing preoperative and intraoperative parameters.

Methods

A total of 453 adult LT recipients treated at the Beijing Tsinghua Changgung Hospital between January 2018 and October 2022 were enrolled. Patients were randomly assigned to a development cohort and a validation cohort at a 6:4 ratio. AKI was diagnosed using the 2012 Kidney Disease Improving Global Outcomes (KDIGO) criteria. Univariate and multivariate logistic regression analyses identified clinical factors associated with early AKI. A predictive model was constructed and internally validated. Additionally, stages 2 and 3 AKI, as defined by the KDIGO criteria, were classified as severe AKI. Independent risk factors for severe AKI within 48 h following LT were similarly identified using logistic regression analyses.

Results

At 48 h following LT, 125 (46%) patients developed AKI. Univariate analysis identified 17 potential predictive factors for AKI, including preoperative hepatic encephalopathy (HE), a history of alcohol-associated cirrhosis, body mass index ≥28 kg/m², and a prognostic nutritional index > 43 (P < 0.1). A backward stepwise regression model was utilized to develop a clinical risk prediction model incorporating the following variables: HE, alcohol-associated cirrhosis, preoperative albumin–bilirubin score ≥ −1.78, operation time ≥560 min, and intraoperative fresh frozen plasma transfusion volume (per 1000 mL). The model achieved an area under the curve (AUC) of 0.760 (P < 0.05) in the development cohort and 0.759 (P < 0.05) in the validation cohort. The calibration curve indicated excellent agreement between predicted and observed probabilities of early AKI (P > 0.05). Multivariate logistic regression analysis identified the preoperative model of end-stage liver disease score ≥14, operation time ≥560 min, intraoperative blood loss ≥1000 mL, intraoperative urine output <1000 mL, and elevated lactic acid level as independent risk factors for severe AKI.

Conclusion

The proposed predictive model could promote the identification of high-risk LT recipients immediately following surgery, enabling clinicians to intervene early to mitigate the risk of developing AKI within 48 h postoperatively. This approach has the potential to improve patient prognosis by supporting timely and targeted management strategies.