基于中西医诊疗信息的食管癌预后模型的构建与验证

Development and validation of a prediction model for patients with esophageal cancer based on clinical data

目的基于临床数据对可能影响食管癌患者预后的中西医诊疗信息进行分析,探讨食管癌不良预后的危险因素,并构建列线图对其进行个体化预测。方法本研究为横断面研究和回顾性研究。收集2018年1月1日-2020年12月31日河南省中医院肿瘤科177例经病理诊断为食管癌患者的中西医诊疗信息,按7∶3比例分为建模数据集124例与验证数据集53例。对建模数据集的病历资料进行单因素及多因素Cox回归分析,以确定与食管癌患者死亡预后相关的独立危险因素。基于独立危险因素构建列线图预后模型,对食管癌患者1、2、3年的生存概率进行预测,并利用验证数据集数据对其区分度、校准度与临床适用度进行内部验证。结果多因素Cox回归分析结果显示,中医辨证分型、中药内服时长、肿瘤位置、临床分期及D-二聚体水平是食管癌患者不良预后的独立危险因素(P<0.05)。就区分度而言,建模数据集与验证数据集的C-指数分别为0.751、0.643。校准图结果显示,校准曲线与参考线拟合良好。决策曲线分析结果表明,食管癌患者应用该列线图模型进行治疗决策在大部分阈值上可取得一定的净获益。结论中医诊疗对食管癌患者的不良预后具有一定参考意义,但今后仍需前瞻性、大样本、多中心研究以探索其预测价值。

Objective To analyze the diagnosis and treatment information of traditional Chinese and Western medicine that may impact on the prognosis of patients with esophageal cancer,and the risk factors of esophageal cancer based on clinical data.We constructed intuitive and convenient tools to individualize the prediction of patients'prognosis with esophageal cancer.Methods This is a cross-sectional and retrospective study.To collect the diagnosis and treatment information of 177 patients with pathological diagnosis of esophageal cancer who were hospitalized in the Oncology Department of Henan Provincial Hospital of Traditional Chinese Medicine from January 1,2018 to December 31,2020.The patients who were screened by inclusion and exclusion criteria,and were randomly divided into training dataset and validation dataset in a ratio of 7:3.Univariate and multivariate Cox regression analysis were analyzed on the training dataset to identify independent risk factors of esophageal cancer.Then using independent risk factors to develop a nomogram for predicting the survival probability of esophageal cancer patients at 1 year,2 years and 3 years.The discrimination,calibration and clinical applicability of the prognostic model were evaluated by C-index,calibration plot and decision curve analysis,respectively.Results The results of multivariate Cox regression analysis showed that the risk factors with P value less than 0.05 included syndrome differentiation classification,duration of internal administration of Chinese medicine,tumor location,clinical stage and D-dimer level(P<0.05).In terms of differentiation,the C-index of the training dataset and the validation dataset was 0.751 and 0.643,respectively.The calibration plot shows that almost all of the calibration curves can fit the calibration line well.The results of decision curve analysis show that patients with esophageal cancer who use this nomogram model for treatment decisions have a net benefit at most thresholds.Conclusion The diagnosis and treatment information of Traditional Chinese Medicine has certain reference significance for the poor prognosis of esophageal cancer patients,but there is need to conduct more prospective,large-sample,multi-center studies to explore its predictive value in the future.