下咽癌预后差异及Nomogram预测模型的建立和验证

目的应用surveillance, epidemiology and end results (SEER) 数据库中下咽癌患者的公共数据，比较不同转移类型和治疗方式的预后差异，并建立和验证下咽癌Nomogram预测模型。方法Windows系统计算机下载SEER官方统计软件，检索第一原发部位为下咽部的癌症患者资料，诊断年份为2007—2016年。采用Kaplan Meier 生存曲线法比较不同转移类型和治疗方式分组患者的预后情况，采用单因素Cox和多因素Cox回归模型筛选出与预后相关的影响因素。将下咽癌患者数据按7∶3随机分为训练集和验证集，将筛选出的影响因素应用训练集建立Nomogram模型并计算C指数、分析校准曲线；应用验证集评估模型预测效能。结果最终纳入下咽癌患者1 008例生存率分析结果显示，下咽癌患者1、3年和5年生存率分别为70.1%、47.6%和40.4%，中位生存时间为28个月。不同治疗方式比较发现，临床早期（I、II）患者中，非手术放（化）疗组与手术+术后放疗组患者的预后比较无统计学意义（P>0.05）；而临床中晚期（III、IV）患者中，手术+术后放疗组的预后优于非手术放（化）疗组（P<0.000 1）。Cox回归模型分析发现年龄、临床分期、T分期、M分期及治疗方案是下咽癌预后的影响因素，基于此5项影响因素成功构建了下咽癌的Nomogram模型，效能分析表明预测模型训练集C指数为0.749，验证集C指数为0.711。结论手术+术后放疗对中晚期下咽癌具有更好的远期生存疗效，依据下咽癌多项临床参数建立的Nomogram肿瘤模型有望成为一种下咽癌治疗效果的预测工具。

Establishment and validation of nomogram prediction model of hypopharyngeal carcinoma related to prognostic clinical features

ObjectiveTo compare the prognosis of hypopharyngeal carcinoma (HPC) of different metastatic types and treatment methods, and to establish and verify the HPC prognosis Nomogram predicting model by applying public data of HPC patients in surveillance, epidemiology and end results (SEER) database.MethodsSEER official statistical software was downloaded on a Windows system computer, and the data of patients with primary HPC were retrieved. The diagnosis date ranged from 2007 to 2016. Kaplan Meier survival curve method was used to compare the prognosis of HPC patients with different metastasis or treatment subgroups, and single factor Cox regression model and multi factor Cox regression model were used to screen clinical factors with prognostic significance. The HPC patient data were randomly divided into a training set and a verification set as a ratio of seven to three. Based on the multi factor Cox regression results, the training set was applied to establish the Nomogram model, the C index and the calibration curve were analyzed. The verification set was utilized to evaluate the model prediction efficiency.ResultsA total of 1 008 HPC patients were recruited. Survival analysis showed that the 1 year, 3 year and 5 year survival rates of patients with HPC were 70.1%, 47.6% and 40.4% respectively, and the median survival time was 28 months. Comparison of different treatment methods showed an equivalent survival outcome between surgical treatment followed by irradiation and chemoradiation management for the early stage patients (P>0.05), while better five year overall survival was found in surgical treatment followed by irradiation group for advanced stage patients (P<0.000 1).Cox regression model analysis demonstrated that age, clinical stage, T stage, M stage and treatment method were the influencing factors for the prognosis of HPC. Based on these five influencing factors, the Nomogram model of HPC was successfully constructed. The effectiveness analysis of this model showed that the C index for training set was 0.749 and that for verification set was 0.711.ConclusionsSurgical treatment followed by irradiation has a better overall survival for advanced HPC. The Nomogram tumor model based on multiple clinical parameters of HPC is expected to become an accurate predictive tool for the treatment of this tumor.