肝硬化再代偿的临床特征、影响因素及其预测效能分析

Clinical Features, Influencing Factors and Predictive Efficacy Analysis of Decompensated Cirrhosis

摘要: 目的：分析肝硬化失代偿期患者发生再代偿的临床特征及影响再代偿发生的相关因素并分析预测效能。方法：回顾性纳入2018年~2023年初次确诊的失代偿期肝硬化住院患者120例，根据再代偿情况，分为再代偿组53例和持续失代偿组67例。收集两组人口学资料、实验室指标、影像学结果、内镜结果及合并症情况，计算临床评分，其中包括终末期肝病模型评分(MELD评分)、Child-Pugh评分、FIB-4指数、谷草转氨酶/血小板比值指数(APRI)、King评分、Lok评分和白蛋白–胆红素评分(ALBI)。采用方差膨胀因子(VIF)诊断变量多重共线性，结合单因素分析筛选变量，通过二元Logistic回归分析肝硬化失代偿期患者发生再代偿的影响因素，采用受试者工作特征(ROC)曲线评价预测效能并分析其预测效能。结果：再代偿组患者年龄、天冬氨酸氨基转移酶(AST)、直接胆红素(DBil)、肌酐(Scr)、MELD评分、Child-Pugh分级、FIB-4指数、King评分、Lok评分、白蛋白–胆红素评分(ALBI)低于持续失代偿组(P < 0.05)；再代偿组患者血红细胞、白蛋白、丙氨酸氨基转移酶(ALT)、血钙和甲胎蛋白高于持续失代偿组(P < 0.05)。经共线性诊断显示ALBI、Lok评分与白蛋白、Child-Pugh评分存在强共线性(VIF > 5)，二元Logistic回归分析显示，白蛋白、ALT、AST、Child-Pugh评分、MELD评分是影响患者实现再代偿的独立因素(P均<0.05)；ALBI、Lok评分因强共线性出现回归系数方向异常，该异常结果具有统计学与临床探讨价值，予以保留并深入分析。ROC曲线结果分析显示，白蛋白、ALT、AST、Child-Pugh评分、MELD评分值联合预测，AUC提高至0.832 (95% CI: 0.761~0.903)，此时模型灵敏度为0.925、特异性0.642。结论：白蛋白、ALT、AST、Child-Pugh评分、MELD评分值是肝硬化失代偿期患者发生再代偿的独立影响因素；联合白蛋白、ALT、AST、Child-Pugh评分、MELD评分值为预测肝硬化失代偿期患者再代偿发生提供了新的证据。

Abstract: Objective: To analyze the clinical characteristics of decompensated cirrhosis patients undergoing recompensation, investigate the factors influencing recompensation, and analyse the predictive efficacy of the related factors. Methods: A retrospective study was conducted on 120 hospitalized patients with decompensated cirrhosis diagnosed for the first time between 2018 and early 2023. Based on recompensation status, the patients were divided into a recompensation group (n = 53) and a persistent decompensation group (n = 67). Demographic data, laboratory parameters, imaging findings, endoscopic results, and comorbidities were collected from both groups. Clinical scores, including the end-stage liver disease model score (MELD score), Child-Pugh score, FIB-4 index, aspartate aminotransferase/platelet ratio index (APRI), King score, Lok score, and albumin-bilirubin index (ALBI), were calculated. The variance inflation factor (VIF) was used to diagnose multicollinearity of variables, and single factor analysis was employed to screen variables. Binary Logistic regression analysis was conducted to investigate the influencing factors of decompensated cirrhosis patients’ recompensation. Receiver operating characteristic (ROC) curves were used to evaluate predictive performance, and a predictive model was constructed. Results: The recompensation group exhibited lower levels of age, aspartate aminotransferase (AST), direct bilirubin (DBil), creatinine (Scr), MELD score, Child-Pugh grade, FIB-4 index, King score, Lok score, and ALBI compared to the persistent decompensation group (P < 0.05). The recompensation group also showed higher levels of hemoglobin, albumin, alanine aminotransferase (ALT), serum calcium ions, and alpha-fetoprotein than the persistent decompensation group (P < 0.05). Collinearity diagnosis showed strong collinearity between ALBI, Lok score and albumin, Child-Pugh score (VIF > 5). Binary Logistic regression analysis revealed that albumin, ALT, AST, Child-Pugh score, and MELD score were independent factors influencing patients’ achievement of recompensation (all P < 0.05). that albumin, ALT, AST, Child-Pugh score and MELD score were independent factors influencing recompensation (all P < 0.05). The abnormal direction of regression coefficients of ALBI and Lok score was caused by strong collinearity, which was of statistical and clinical significance and retained for in-depth analysis. ROC curve analysis revealed that the combined prediction of albumin, ALT, AST, Child-Pugh score, MELD score increased the AUC to 0.832 (95% CI: 0.761~0.903), with a model sensitivity of 0.925 and specificity of 0.642. Conclusion: Albumin, ALT, AST, Child-Pugh score and MELD score are independent factors for recompensation in patients with decompensated cirrhosis. The combined model of these indicators provides new evidence for predicting recompensation in decompensated cirrhosis patients.