ER/PR^+和HER2^-型乳腺癌患者免疫组化指标的列线图预后模型

A prognostic nomogram for ER/PR^+ and HER2^- breast cancer patients based on classical immunohistochemical index

目的:探讨改良雌激素受体(estrogen receptor,ER)/孕激素受体(progesterone receptor,PR)阳性(^+)及人类表皮生长因子受体2(human epidermal growth factor receptor 2,HER2)阴性(-)(ER/PR^+、HER2-)型乳腺癌患者的传统预后模型,满足目前的临床实际需求。方法:选取了2009年1月至2009年12月上海市黄浦区中心医院乳腺外科收治的335例ER/PR^+、HER2-型乳腺癌患者。将97个变量纳入模型,采用SCAD变量选择的方法,在充分考虑协变量是否存在对数线性关系、非对数线性关系(分段线性关系)临界值的合理确定、共线问题后,构建一个新的ER/PR^+、HER2-型乳腺癌患者传统免疫组化指标的Cox回归预后模型,并进一步建立其列线图模型;在此基础上建立了术后1、3和5年生存概率的列线图;并通过比较模型的区分度(discrimination)和校准度(calibration)来评价模型的预测能力。结果:通过乳腺癌预后建立Cox回归模型结果显示,患者的预后与组织级别、淋巴结转移、Ki67、PR和年龄等因素有关;其中组织级别和淋巴结转移对风险比的影响呈对数线性关系,Ki67、PR和年龄对风险比的影响呈非对数线性关系,其合理临界值分别为Ki67(60%)、PR(20%)和年龄(55岁)。该模型术后1、3和5年的ROC曲线下面积(AUC值)均高于0.85,说明该Cox回归模型具有较高的区分度。该模型Gr?nnesby-Borgan拟合优度检验统计量值为1.37、对应的P值为0.5,说明该Cox回归模型有较好的校准度。结论:通过改良ER/PR^+和HER2-型乳腺癌患者的传统预后模型,建立患者术后1、3和5年生存概率的列线图,能准确、直观、有效地预测患者的生存概率,对乳腺癌患者临床治疗有较好的指导意义。

Objective： To modify traditional prognostic model for patients with ER/PR^＋, HER2-breast cancer to meet the actual requirements in current clinical practice. Methods： 335 patients with ER/PR^＋, HER2-breast cancer, who were admitted in Department of Breast Surgery, Shanghai Huangpu Center Hospital from January 2009 to December 2009, were enrolled in this study. 97 variables were incorporated into the model, using SCAD variable selection method, after fully considering whether covariates existing a log-linear relationship, reasonable determination of the cut-off value of the covariates in non-logarithmic linear relationship（piecewise linear relationship） and collinear and interaction, then we set up a new Cox regression prognostic model for traditional ER/PR^＋, HER2-type breast cancer patients with traditional immunohistochemical indicators, and further establish its nomogram model. On this basis, a nomogram of the survival probability of 1-, 3-, and 5-years after surgery was established; The discrimination and calibration of model were compared to evaluate the predictive ability of the model. Results： The Cox regression model shows that the prognosis of patients are associated with the histologic grade, lymph node metastasis, Ki67, PR and age etc. Among them, the histologic grade and lymph node metastasis have log-linear relationship with prognosis; Ki67, PR and age have non-log-linear relationship with prognosis and the reasonable cut-off values are Ki67（60%）,PR（20%）and age（55 years old）. Area under the receiver operating characteristic（ROC）curve（AUC）of this Cox model for 1-, 3-and 5-year survival after surgery are all above 0.85, indicating high discrimination. TheGrθnnesby-Borgan goodness-of-fit test statistics of this model is 1.37 with P〈0.05, indicating good calibration. Conclusion： The modified nomogram.could accurately, directly and effectively predict the survival probability of patients, which may exert good guidance for the clinical practice for patients with breast cancer.