Graded nomograms based on perioperative parameters for predicting New-Onset severe acute kidney injury following liver transplantation in patients with normal preoperative renal function: the SALT scale

This study aimed to develop a predictive model and construct a graded nomogram to estimate the risk of severe acute kidney injury (AKI) in patients without preexisting kidney dysfunction undergoing liver transplantation (LT). Patients undergoing LT between January 2022 and June 2023 were prospectively screened. Severe AKI was defined as Kidney Disease: Improving Global Outcomes stage 3. Using the least absolute shrinkage and selection operator (LASSO) analytics, we identified the preoperative, intraoperative, and postoperative factors associated with severe AKI. Machine learning were employed to develop predictive models, and the most suitable model was selected for further analysis. The Shapley Additive Explanation was utilized to construct graded nomograms, forming the Severe AKI post-LT (SALT) scale. Among the 405 patients, 44 had AKI stage 3 (severe AKI). The Model for End-Stage Liver Disease (MELD) score, estimated blood loss, alanine aminotransferase, D-dimer, and thromboelastography reaction time within 24 h post-LT were identified as risk factors. The logistic regression model achieved the highest area under the receiver operating characteristic curve (AUROC) of 0.885. The graded SALT scale, based on the logistic regression model, achieved AUROCs of 0.751, 0.826, and 0.894. The AUROCs for the testing cohort is 0.791. This preliminary study provides a SALT scale for assessing the occurrence of severe AKI after LT. Although additional data are needed to externally validate our model before applying it to patient care, our findings suggest that the SALT scale may be a feasible bedside tool for assessing the risk of AKI after LT.

本研究旨在构建预测模型并建立分级列线图，以估算无基础肾功能不全的肝移植（liver transplantation, LT）患者发生严重急性肾损伤（severe acute kidney injury, AKI）的风险。
本研究前瞻性筛选了2022年1月至2023年6月期间接受LT的患者。严重AKI定义为改善全球肾脏病预后组织（Kidney Disease: Improving Global Outcomes, KDIGO）3期病变。采用最小绝对收缩和选择算子（least absolute shrinkage and selection operator, LASSO）分析，筛选出与严重AKI相关的术前、术中及术后危险因素。
采用机器学习方法开发预测模型，并选取最优模型开展后续分析。利用Shapley可加解释（Shapley Additive Explanation）构建分级列线图，形成肝移植术后严重AKI（Severe AKI post-LT, SALT）评分量表。共纳入405例患者，其中44例发生KDIGO 3期AKI（即严重AKI）。
最终筛选得到的危险因素包括终末期肝病模型（Model for End-Stage Liver Disease, MELD）评分、预估失血量、丙氨酸氨基转移酶（alanine aminotransferase, ALT）、D-二聚体（D-dimer）以及肝移植术后24小时内的血栓弹力图反应时间（thromboelastography reaction time）。Logistic回归模型的受试者工作特征曲线下面积（area under the receiver operating characteristic curve, AUROC）最高，达0.885。基于该Logistic回归模型构建的分级SALT量表，其AUROC分别为0.751、0.826和0.894；测试队列的AUROC为0.791。
本初步研究构建了用于评估LT术后严重AKI发生风险的SALT量表。尽管在将该模型应用于临床诊疗前尚需外部队列数据进行外部验证，但本研究结果提示，SALT量表或可成为一种可行的床旁工具，用于评估LT术后AKI的发生风险。