n-Ranson评分列线图预测模型的构建及其对急性胰腺炎重症化的预测价值

To construct a nomogram prediction model of n-Ranson score and its predictive value for severe acute pancreatitis

目的构建急性胰腺炎(AP)患者进展为重症AP(SAP)的列线图预测模型,并探讨其在预测AP重症化进程中的应用价值。方法根据2012年亚特兰大新分类标准,将264例AP患者分为SAP组(96例)和非SAP(NSAP)组(168例)。比较两组患者临床指标、实验室数据和Ranson评分。采用LASSO回归方法筛选预测因子,并构建列线图(n-Ranson评分)预测SAP疾病进展。采用C-index、校准曲线和临床决策曲线分析(DCA)评估n-Ranson评分的性能、拟合度和临床实用性。结果SAP组WBC计数、PLT平均容积(MPV)、血清淀粉酶(AMY)、血清脂肪酶(LIPA)、C反应蛋白(CRP)、PLT压积(PCT)、中性粒细胞与淋巴细胞比值(NLR)、血小板与淋巴细胞比值(PLR)、谷氨酰转移酶(GGT)、血糖(Glu)、AST及年龄>60岁患者比例均高于NSAP组,RBC计数、红细胞压积(HCT)、Hb、血清白蛋白(Alb)、估算的肾小球滤过率(eGFR)及饮酒史患者比例均低于NSAP组;两组Ranson评分不同分数患者构成比比较差异有统计学意义(P<0.05)。LASSO回归交叉验证结果显示,NLR、eGFR、Ranson评分是SAP严重程度的独立预测因子(P<0.05)。基于以上3个变量构建n-Ranson评分预测模型,结果显示n-Ranson评分的C-index为0.911(95%CI 0.873~0.950),校正C-index为0.908,提示其具有较精准的预测能力。经Bootstrap法内部验证,n-Ranson评分预测与临床实际发生的AP严重程度曲线具有较高的一致性,且预测效果优于Ranson评分。结论基于NLR、eGFR、Ranson评分构建的n-Ranson评分列线图预测模型对AP严重程度具有一定预测价值。

Objective To construct a nomogram prediction model for the progression of patients with acute pancreatitis(AP)to severe acute pancreatitis(SAP),and to explore its application value in predicting the progression of severe AP.Methods According to the 2012 Atlanta classification criteria,264 patients with AP were divided into SAP group(96 cases)and non⁃SAP(NSAP)group(168 cases).Clinical indicators,laboratory data and Ranson scores were compared between the two groups.LASSO regression method was used to screen predictors,and a nomogram(n⁃Ranson score)was constructed to predict the development of SAP.C⁃index,calibration curve and clinical decision curve analysis(DCA)were used to evaluate the performance,fit and clinical practicability of the n⁃Ranson score.Results WBC count,mean platelet volume(MPV),serum amylase(AMY),serum lipase(LIPA),C⁃reactive protein(CRP),platelet specific volume(PCT),neutrophil⁃to⁃lymphocyte ratio(NLR),platelet⁃to⁃lymphocyte ratio(PLR),glutamyltransferase(GGT),blood glucose(Glu),AST and the proportion of patients>60 years old in SAP group were higher than those in NSAP group,while RBC count,hematocrit(HCT),Hb,serum albumin(Alb),estimated glomerular filtration rate(eGFR)and the proportion of patients with drinking history were lower than those in NSAP group;there was statistically significant difference in the constituent ratio of patients with different Ranson scores between the two groups(P<0.05).The results of LASSO regression cross⁃validation showed that NLR,eGFR and Ranson score were independent predictors of SAP(P<0.05).Based on the above three variables,the n⁃Ranson sccore prediction model was constructed.The results showed that the C⁃index of the n⁃Ranson score was 0.911(95%CI 0.873~0.950),and the corrected C⁃index was 0.908,indicating that it had a relatively accurate prediction ability.The Bootstrap method internally verified that the n⁃Ranson score had a high consistency with the actual clinical AP severity curve,and the prediction effect was better than that of the Ranson score.Conclusion N⁃Ranson score nomogram prediction model based on NLR,eGFR,and Ranson score has a certain predictive value for the severity of AP.