基于血液炎症指标的食管鳞状上皮内瘤变风险预测模型的构建与应用

Construction and application of a predictive model for esophageal squamous intraepithelial neoplasia based on inflammation-related indicators

目的探究食管鳞状上皮内瘤变发生风险的预测因素,构建列线图预测模型。方法回顾性收集2016年01月至2021年12月在扬州大学附属医院诊断为食管鳞状上皮内瘤变患者126例,以及同期于健康管理中心进行体检者344例,收集患者一般临床资料、血常规检测、肿瘤标志物、病变资料。确定食管鳞状上皮内瘤变的独立预测因子,构建列线图预测模型、绘制受试者操作特征(receiver operating characteristic,ROC)曲线,计算得出Harrell一致性指数(C-index)并采用Bootstrap自抽样法,对模型进行内部验证并绘制校准曲线及决策曲线。结果为构建食管上皮内瘤变的风险预测模型,经PSM法匹配后,本研究共纳入96例食管鳞状上皮内瘤变患者及96例健康对照。多因素Logistic回归分析显示,血红蛋白(HGB)计数、血小板(PLT)计数、血小板分布宽度(PDW)、NLR水平是食管鳞状上皮内瘤变的独立预测因子。根据以上四项指标构建的列线图模型曲线下面积(area under curve,AUC)为0.787。Bootstrap内部验证的C-index值为0.771,提示该食管鳞状上皮内瘤变风险预测模型的识别能力、一致性和临床净获益良好。结论基于HGB、PLT、PDW及NLR建立的列线图预测模型可有效鉴别有无食管鳞状上皮内瘤变病变人群,对诊断食管癌前病变有辅助价值。

Objective To investigate the predictors of the risk of esophageal squamous intraepithelial neoplasia and to construct a columnar graph prediction model.Methods 126 patients diagnosed with esophageal squamous intraepithelial neoplasia at the Affiliated Hospital of Yangzhou University from January 2016 to December 2021 and 344 patients who underwent physical examination at the health management center during the same period were retrospectively collected,and general clinical data,routine blood tests,tumor markers and lesion data were collected.The independent predictors of esophageal squamous intraepithelial neoplasia were identified,a line graph prediction model was constructed,receiver operating characteristic(ROC)curves were plotted,the Harrell consistency index(C-index)was calculated and the model was internally validated using Bootstrap self-sampling.The calibration curve and decision curve analysis were performed.Results To construct a risk prediction model for esophageal intraepithelial neoplasia,a total of 96 patients with esophageal squamous intraepithelial neoplasia and 96 healthy controls were included in this study after matching by the PSM method.Multi-factor logistic regression analysis showed that hemoglobin(HGB)count,platelet(PLT)count,platelet distribution width(PDW)and NLR levels were independent predictors of esophageal squamous intraepithelial neoplasia.The AUC of the column line graph model constructed based on these four indicators was 0.787.The C-index value validated within Bootstrap was 0.771,which suggested that it was a good discrimination,consistency and net clinical benefit of the esophageal intraepithelial neoplasia risk prediction model.Conclusion The columnar line graph prediction model based on HGB,PLT,PDW and NLR can effectively identify people with or without esophageal squamous intraepithelial neoplasia lesions and is of value in the diagnosis of pre-cancerous esophageal lesions.