目的 建立一个对首次发作急性胰腺炎(AP)病情严重程度有早期预测价值的可视化的列线图.方法 收集2013年1月至2016年1月间温州医科大学附属第一医院收治的发病72 h内入院的首次发作AP患者706例,依据2012年亚特兰大分类标准将AP患者分为非重症胰腺炎(NSAP,即MAP+MSAP)和重症胰腺炎(SAP)两组,统计并分析患者的一般资料(年龄、体重指数和入院时间等)、实验室检查(血淀粉酶、血糖、白蛋白、白细胞、肌酐、尿素氮)结果 .对纳入的相关临床指标进行Logistic单因素及多因素回归分析,根据有统计学差异的指标得出回归方程式,利用R语言软件可视化处理逻辑回归(LR)模型获得列线图,并通过受试者工作特征(ROC)曲线分析验证.结果 单因素Logistic回归分析结果显示,NSAP和SAP两组间血糖、入院时肌酐、入院24 h肌酐、入院时尿素氮、入院24 h尿素氮、白细胞及白蛋白的OR(95%CI)值分别为1.132(1.080~1.186),1.019(1.013~1.025),1.026(1.020~1.033),1.066(1.035~1.099),1.333(1.241~1.432),1.083(1.032~1.136)和0.853(0.811~0.889),差异均有统计学意义(P值均<0.01).经多因素Logistic回归分析LR模型的回归方程式为Y=-2.657~-0.116×白蛋白(g/L)+0.082×白细胞(×109/L)+0.118×血糖(mmol/L)+0.022×入院24 h肌酐(μmol/L).列线图总分超过60分有发生SAP的可能,总分超过130分发生SAP可能将高达14%以上.进一步通过ROC曲线分析验证,本研究建立的LR模型的曲线下面积(AUC)预测SAP发生的灵敏度、特异度均优于尿素氮、肌酐、BISAP评分单独预测.结论 本列线图可能是预测首次AP病情严重程度的有效临床工具.
Objective To establish a visualized nomogram with early predictive value for the severity of first-onset acute pancreatitis ( AP ) . Methods 706 cases of first-onset AP patients admitted to the First Affiliated Hospital of Wenzhou Medical University within 72 hours from January 2013 to January 2016 were collected. According to the revised Atlanta classification of AP in 2012, AP patients was divided into non-severe pancreatitis ( NSAP, also called mild acute pancreatitis and moderately severe acute pancreatitis) group and severe acute pancreatitis ( SAP) group. The demographic data ( age, body mass index and admission time, etc) and laboratory tests (serum amylase, blood sugar, albumin, white blood cells, creatinine, urea nitrogen) were collected and statistically analyzed. Logistic univariate and multivariant regression analysis were performed based on the relevant clinical indicators. The statistically significant indicators were used to obtain regression equations. The R-language software was used to obtain the visualized nomogram via LR model, which was further validated by ROC curve analysis. Results In univariate logistic regression analysis, the OR ( 95%CI) values of blood glucose, creatinine at admission and 24 h after admission, urea nitrogen at admission and 24 h after admission, white blood cell, albumin in NSAP group and SAP group were 1. 132(1. 080-1. 186), 1. 019(1. 013-1. 025), 1. 026(1. 020-1. 033), 1. 066(1. 035-1. 099), 1. 333(1. 241-1. 432), 1. 083 (1.032-1.136), and 0.853(0.811-0.889), and all the differences were statistically significant (all P values <0. 05). Multivariate logistic regression analysis showed that the regression equation of the LR model was Y= -2. 657-0. 116 × albumin(g/L) +0. 082 × white blood cell( × 109/L) +0. 118 × glycemia(mmol/L) +0. 022 × 24 h after admission creatinine (μmol/L). A total score of more than 60 points on the nomogram predicted the possibility of SAP. If the total score exceeded 130, the possibility of SAP may be up to 14% or more. Furthermore, the ROC curve analysis confirmed that the sensitivity and specificity of LR model established in this study for predicting of SAP were superior to those of urea nitrogen, creatinine and BISAP score alone by AUC, respectively. Conclusions This nomogram may be a useful clinical tool for predicting the severity of the first-onset acute pancreatitis.