基于SEER数据库亚洲及太平洋岛民人群中结肠癌患者的预后模型构建

Construction of a prognostic model for colon cancer patients in Asian and Pacific islander populations based on the SEER database

目的构建基于SEER数据库亚洲及太平洋岛民人群中结肠癌患者的预后模型。基于SEER数据库最新公布的数据,目标人群为亚洲及太平洋岛民,构建该人群中结肠癌预后的nomogram模型。方法运用SEER*Stat软件提取SEER数据库中2004年至2020年间诊断为结肠癌的患者信息。将所有病例按7∶3的比例随机分配为建模队列和验证队列并统计基线资料。将临床上考虑及单因素COX分析有统计学差异的因素纳入多因素COX回归分析,绘制生存曲线,构建nomogram预测模型,通过验证队列对模型的区分度及一致性进行验证。结果提取人群总病例数为26151例,随机分为建模队列(n=18394)和验证队列(n=7757)。该人群中结肠癌主要原发部位为乙状结肠(38.94%),中分化(Ⅱ级,63.68%)及T3期(43.88%)肿瘤居多,且多为单发(77.56%)。年龄、性别、肿瘤原发部位、组织学类型、分级、肿瘤大小、T分期、N分期、M分期、原发部位是否手术、切除送检淋巴结个数、肿瘤数量为影响患者生存的独立危险因素。构建的nomogram模型C-Index为0.82(建模队列)及0.85(验证队列),Bootstrap抽样1000次检验预测模型3年、5年预测的准确度,结果显示一致性较好。结论通过SEER数据库最新纳入的病例信息构建了一个新的列线图,用于预测亚洲及太平洋岛民人群中结肠癌预后及总体生存率。经过验证队列验证,显示出相对满意的预测性能,并有助于未来相关人群的预测及临床实践。

Objective To construct a prognostic nomogram for colon cancer in Asian and Pacific islanders based on the latest released data in the SEER database.Methods The information of patients diagnosed with colon cancer from 2004-2020 was extracted from the SEER database with the SEER*Stat software.All cases were randomly assigned to a training or a validation cohort at a ratio of 7∶3,then baseline data were collected.Factors with clinical or statistical significance in uni-variate COX analysis were included in the multivariate COX regression analysis,the construction of a prognostic nomogram model as well as survival curves.The validation cohort was used to validate the discrimination and calibration of the model.Results A total cohort of 26151 cases was selected form the database,among which,18394 cases were included in the training cohort whereas 7757 cases in the validation cohort.The main primary site of colon cancer in this population was sigmoid colon(38.94%),and most of the tumors were moderately differentiated(gradeⅡ,63.68%)and T3(43.88%).Moreover,most of the tumors were solitary tumor(77.56%).Age,sex,primary tumor site,histological type,grade,tumor size,T stage,N stage,M stage,whether the primary site was operated,the number of lymph nodes resected and examined,and the number of tumors were independent risk factors affecting the survival of patients.The C-index of the nomogram model was 0.82(training cohort)and 0.85(validation cohort),and Bootstrap sampled 1000 times to test the accuracy of 3-year and 5-year prediction of the model,and the results showed good calibration.Conclusion We constructed a novel model using the latest case information included in the SEER Database to predict prognosis and overall survival for colon cancer in Asian and Pacific islanders.The vali-dation cohort showed relatively satisfactory predictive per-formance,the research might contribute to the prediction and clinical practice of the relevant population in the future.