Table_1_A nomogram based on genotypic and clinicopathologic factors to predict the non-sentinel lymph node metastasis in Chinese women breast cancer patients.xlsx

BackgroundSentinel lymph node biopsy (SLNB) is the standard treatment for breast cancer patients with clinically negative axilla. However, axillary lymph node dissection (ALND) is still the standard care for sentinel lymph node (SLN) positive patients. Clinical data reveals about 40-75% of patients without non-sentinel lymph node (NSLN) metastasis after ALND. Unnecessary ALND increases the risk of complications and detracts from quality of life. In this study, we expect to develop a nomogram based on genotypic and clinicopathologic factors to predict the risk of NSLN metastasis in SLN-positive Chinese women breast cancer patients.
MethodsThis retrospective study collected data from 1,879 women breast cancer patients enrolled from multiple centers. Genotypic features contain 96 single nucleotide polymorphisms (SNPs) associated with breast cancer susceptibility, therapy and prognosis. SNP genotyping was identified by the quantitative PCR detection platform. The genetic features were divided into two clusters by the mutational stability. The normalized polygenic risk score (PRS) was used to evaluate the combined effect of each SNP cluster. Recursive feature elimination (RFE) based on linear discriminant analysis (LDA) was adopted to select the most useful predictive features, and RFE based on support vector machine (SVM) was used to reduce the number of SNPs. Multivariable logistic regression models (i.e., nomogram) were built for predicting NSLN metastasis. The predictive abilities of three types of model (based on only clinicopathologic information, the integrated clinicopathologic and all SNPs information, and integrated clinicopathologic and significant SNPs information) were compared. Internal and external validations were performed and the area under ROC curves (AUCs) as well as a series of evaluation indicators were assessed.
Results229 patients underwent SLNB followed by ALND and without any neo-adjuvant therapy, 79 among them (34%) had a positive axillary NSLN metastasis. The LDA-RFE identified the characteristics including lymphovascular invasion, number of positive SLNs, number of negative SLNs and two SNP clusters as significant predictors of NSLN metastasis. Furthermore, the SVM-RFE selected 29 significant SNPs in the prediction of NSLN metastasis. In internal validation, the median AUCs of the clinical and all SNPs combining model, the clinical and 29 significant SNPs combining model, and the clinical model were 0.837, 0.795 and 0.708 respectively. Meanwhile, in external validation, the AUCs of the three models were 0.817, 0.815 and 0.745 respectively.
ConclusionWe present a new nomogram by combining genotypic and clinicopathologic factors to achieve higher sensitivity and specificity comparing with traditional clinicopathologic factors to predict NSLN metastasis in Chinese women breast cancer. It is recommended that more validations are required in prospective studies among different patient populations.

研究背景
临床腋窝淋巴结阴性的乳腺癌患者，前哨淋巴结活检（Sentinel lymph node biopsy, SLNB）是标准治疗方案。然而，对于前哨淋巴结（SLN）阳性的患者，腋窝淋巴结清扫术（axillary lymph node dissection, ALND）仍是标准治疗手段。临床数据显示，接受ALND后的患者中，约40%~75%并未发生非前哨淋巴结（NSLN）转移。不必要的ALND会增加并发症风险，降低患者生活质量。本研究旨在基于基因型与临床病理因素构建列线图，用于预测中国SLN阳性女性乳腺癌患者的非前哨淋巴结转移风险。
研究方法
本项回顾性研究收集了多中心纳入的1879例女性乳腺癌患者的临床数据。基因型特征包含96个与乳腺癌易感性、治疗及预后相关的单核苷酸多态性（single nucleotide polymorphisms, SNPs）。采用定量PCR检测平台完成SNP基因分型。基于突变稳定性将遗传特征划分为两个聚类簇。采用标准化多基因风险评分（polygenic risk score, PRS）评估每个SNP聚类簇的联合效应。采用基于线性判别分析（linear discriminant analysis, LDA）的递归特征消除（recursive feature elimination, RFE）筛选最优预测特征，并采用基于支持向量机（support vector machine, SVM）的RFE缩减SNP数量。构建多变量logistic回归模型（即列线图）以预测NSLN转移风险。对比三类模型的预测性能：仅基于临床病理信息的模型、整合临床病理与全部SNP信息的模型，以及整合临床病理与筛选出的显著SNP信息的模型。开展内部与外部验证，并评估受试者工作特征曲线下面积（AUC）及一系列评估指标。
研究结果
本研究纳入229例先接受SLNB、后续行ALND且未接受任何新辅助治疗的患者，其中79例（34%）发生腋窝NSLN转移。基于LDA的RFE筛选出淋巴管侵袭、阳性SLN数量、阴性SLN数量及两个SNP聚类簇作为NSLN转移的显著预测因子。此外，基于SVM的RFE筛选出29个用于NSLN转移预测的显著SNP。内部验证中，临床+全部SNP联合模型、临床+29个显著SNP联合模型以及单纯临床模型的中位AUC分别为0.837、0.795和0.708。外部验证中，三类模型的AUC分别为0.817、0.815和0.745。
研究结论
本研究构建了一款整合基因型与临床病理因素的新型列线图，相较于传统临床病理因素模型，其预测中国女性乳腺癌患者NSLN转移的灵敏度与特异度更优。建议在不同患者群体中开展前瞻性研究以进一步验证该模型的性能。