淋巴结阳性胰腺癌患者术后生存预测模型的构建

Construction of a postoperative survival prediction model for pancreatic cancer patients with positive lymph nodes

目的研究淋巴结阳性胰腺癌患者的预后危险因素,建立预测模型对患者的预后进行评估并检验模型的准确性。方法使用SEERStat软件下载2010年至2015年诊断为胰腺癌并进行手术治疗的患者的临床病理学数据,经过筛选将5816例患者随机分成建模组和验证组并进行组间资料的比较,使用SPSS 22.0软件对建模组数据进行单因素及多因素COX回归分析,使用Kaplan-Meier法进行生存分析并绘制生存曲线,使用R4.0.0软件将预后独立影响因素构建列线图模型,并绘制受试者工作特征曲线(ROC)、校准曲线。结果纳入研究的5816例患者的中位生存时间为19个月,随访时间为1~72个月,将建模组中多因素COX回归分析P<0.05的因素纳入列线图构建,结合单因素及多因素COX回归结果,将年龄、肿瘤位置、组织学类型、分化程度、放疗、婚姻状况、肿瘤分期、T分期纳入列线图模型,模型经过准确性评估发现,建模组和验证组的C-index分别为0.627、0.639,ROC曲线下面积均>0.65,校准曲线均贴合良好,证明该模型具有良好的准确性。结论针对淋巴结阳性患者构建的预测模型,能够很好地预测患者的术后1年、3年及5年生存率,对于独立的影响因素,可以指导临床给予针对性的治疗。

Objective To study the prognostic risk factors of pancreatic cancer patients with positive lymph nodes,and to establish a prediction model to evaluate the prognosis of patients and test the accuracy of the model.Methods SEERStat software was used to download clinicopathological data of patients diagnosed with pancreatic cancer and treated surgically from 2010 to 2015 in The First Affiliated Hospital of Wannan Medical College,and after screening,5816 patients were randomly divided into the modeling group and the validation group.The data were compared between the groups.SPSS 22.0 software was used to perform univariate and multivariate COX regression analysis on data of the modeling group,and Kaplan-Meier method was used for survival analysis and survival curve drawing.Software R4.0.0 was used to construct a nomogram model of independent prognostic influences,and draw receiver operating characteristic curve(ROC)and calibration curve.Results The median survival time of 5816 patients included in the study was 19 months,and the follow-up time was from 1 to 72 months.The factors with P<0.05 in the multivariate COX regression analysis of the modeling group were included in the construction of the nomogram model.Combined with univariate and multivariate COX regression results,the age,tumor location,histological type,differentiation degree,radiotherapy,marital status,tumor stage and T stage were included in the nomogram model.After accuracy evaluation of the model,it was found that the C-index of the modeling group and the validation group were 0.627 and 0.639 respectively,the area under ROC curve was greater than 0.65,and the calibration curve was well fitted,which proved that the model had good accuracy.Conclusion The prediction model for patients with positive lymph nodes can well predict the 1-year,3-year and 5-year postoperative survival rates of patients,and can guide the clinical treatment for independent influences factors.