男性乳腺癌特异性生存率预测模型的构建

A nomogram model for predicting breast cancer-specific survival rate of male breast cancer

目的:构建男性乳腺癌特异性生存(BCSS)生存率预测模型。方法:选取SEER数据库中诊断时间为2010年1月1日至2015年12月31日的2226名男性乳腺癌病例,按7∶3的比例随机分为训练集(n=1558)和验证集(n=668)。在训练集中,使用Cox比例风险回归模型筛选与BCSS相关的变量,构建BCSS生存率多因素预测模型并绘制列线图;同时构建单独临床分期预测模型并绘制列线图。使用验证集数据对模型进行内部验证。用2012年1月1日至2016年5月31日于郑州大学第一附属医院确诊的53例男性乳腺癌病例进行外部验证,随访终点为2021年5月31日。结果:多因素预测模型包括婚姻、医疗保险、组织学分级、雌激素受体表达、孕激素受体表达、临床分期、手术方式、化疗情况等8个变量。在训练集、内部验证集和外部验证集中,多因素列线图预测的C指数(95%CI)分别为0.830(0.790~0.870)、0.865(0.825~0.905)、0.966(0.938~0.994),临床分期列线图预测的C指数(95%CI)分别为0.764(0.718~0.810)、0.740(0.653~0.828)、0.830(0.653~0.976),多因素列线图的C指数均高于临床分期。在内部验证集中,多因素列线图预测1、3、5 a BCSS生存率的ROC的AUC(95%CI)分别为0.910(0.860~0.960)、0.852(0.778~0.926)、0.882(0.826~0.938),临床分期列线图的结果分别为0.777(0.595~0.958)、0.723(0.621~0.824)、0.800(0.730~0.870)。在外部验证集中,多因素列线图预测3、5 a BCSS生存率的ROC的AUC(95%CI)分别为0.975(0.904~0.985)、0.973(0.951~0.994),临床分期列线图的结果分别为0.849(0.707~0.990)、0.824(0.655~0.992)。校正曲线和决策曲线分析结果表明多因素列线图预测效能优于临床分期。结论:本研究建立的多因素列线图有可能成为男性乳腺癌预后评估的有效工具。

Aim:To establish a nomogram model for prediction of breast cancer-specific survival(BCSS)rate of male breast cancer.Methods:A total of 2226 male breast cancer cases diagnosed between January 1,2010 and December 31,2015 from SEER database were selected,and were randomly divided into training set(n=1558)and validation set(n=668)according to the ratio of 7∶3.In the training set,Cox regression model was used to screen the variables related to BCSS,and the BCSS prediction model was constructed and the nomogram was drawn.At the same time,the BCSS prediction model and the nomogram based on clinical stage was constructed.Validation set data were used for internal validation of the model.The data from 53 male breast cancer cases diagnosed at the First Affiliated Hospital of Zhengzhou University from January 1,2012 to May 31,2021,with a follow-up endpoint of May 31,2016,were used for the external validation.Results:In the training set,the internal validation set and the external validation set,the C index(95%CI)of multi-factor nomogram were 0.830(0.790-0.870),0.865(0.825-0.905),0.966(0.938-0.994),respectively,those of single clinical stage were 0.764(0.718-0.810),0.740(0.653-0.828)and 0.830(0.653-0.976),respectively.The C index of multi-factor nomogram was higher than that of clinical stage.In the internal validation set,the AUC(95%CI)of 1,3 and 5 years BCSS predicted by multi-factor nomogram were 0.910(0.860-0.960),0.852(0.778-0.926),0.882(0.826-0.938),respectively,those of clinical stage were 0.777(0.595-0.958),0.723(0.621-0.824)and 0.800(0.730-0.870),respectively.In the external validation set,the AUC(95%CI)of 3 and 5 years BCSS predicted was 0.975(0.904-0.985)and 0.973(0.951-0.994),and those of clinical stage were 0.849(0.707-0.990)and 0.824(0.655-0.992),respectively.The calibration curve and decision curve analysis also showed that the multi-factor nomogram was superior to the clinical stage.Conclusion:The multi-factor nomogram may be an effective tool for assessing the prognosis of male breast cancer.