基于SEER数据库构建胆囊癌病人术后生存列线图研究

Construction of prognostic nomogram of gallbladder cancer based on SEER database

目的构建可预测胆囊癌病人术后生存的列线图。方法从美国国立癌症研究院"监测、流行病学和结果"(surveillance,epidemiology,and end results,SEER)数据库提取1975—2016年诊断为胆囊癌的3616例病人数据,随机分为训练集和验证集,使用机器学习和Cox模型构建列线图,并评估列线图的判别能力和预测精度。结果机器学习方法和多因素Cox风险回归模型结果表明,年龄、肿瘤分级、T分期、M分期、肿瘤分期和淋巴结清扫数目是胆囊癌病人预后的影响因素(P<0.05),以此建立列线图。ROC曲线下面积为0.825(95%CI 0.810~0.841)。验证集ROC曲线下面积为0.844(95%CI 0.819~0.868)。验证校正曲线提示,观察值与预测值之间有良好的一致性。分层分析显示,在T3、T4期胆囊癌病人中,淋巴结清扫数目≥4枚与1~3枚对预后的影响差异无统计学意义(P=0.538);在T1、T2期胆囊癌病人中,淋巴结清扫数目≥4枚的病人较清扫1~3枚者有更好的预后,差异有统计学意义(P<0.01)。结论列线图预后模型有较高的准确性,可为评估胆囊癌病人术后生存提供参考;T1、T2期胆囊癌病人增加淋巴结清扫数目相较于T3、T4期病人获益更多。

Objective To construct a nomogram which can predict the survival rate of postoperative patients withgallbladder cancer.MethodsThe data of 3616 patients diagnosed with gallbladder cancer from 1975 to 2016 wereextracted from SEER database of National Cancer Institute.They were randomly divided into training group andvalidation group.The nomogram was established by machine learning and Cox model.The discriminant ability andprediction accuracy of nomogram were evaluated by receiver operator characteristic(ROC)curve and calibration chart.ResultsMachine learning and multivariate Cox risk regression model showed that age,tumor grade,T stage,M stage,tumor stage and the number of lymph node dissection were the prognostic factors of patients with gallbladder cancer(P<0.05).Nomogram was established.The area under the ROC curve was 0.825(95%CI 0.810-0.841).In the validationgroup,the area under the ROC curve was 0.844(95%CI 0.819-0.868).The C-index of internal verification and thecalibration chart indicated that the prediction model was moderately accurate and the prediction results is reliable.Inpatients with stage T3 and T4 gallbladder carcinoma,there was no significant difference in the effect of dissection of≥4 lymph nodes and 1-3 lymph nodes on prognosis(P=0.538).In patients with stage T1 and T2 gallbladder carcinoma,theprognosis of≥4 lymph nodes dissected was better than that of 1-3 lymph nodes dissected,and the difference wasstatistically significant(P<0.01).ConclusionThe nomogram prognostic model has been proved to have high accuracyand can provide a reference for evaluating the prognosis and survival rate of patients with gallbladder cancer.Patientswith stage T1 and T2 gallbladder cancer have more benefits than patients with stage T3 and T4 by increasing the numberof lymph node dissection.