[A nomogram to predict non-sentinel lymph node metastasis for breast cancer patients with positive axillary sentinel lymph node].

Objective: To identify the risk factors of non-sentinel lymph node (nSLN) metastasis in breast cancer patients with 1~2 positive axillary sentinel lymph node (SLN) and construct an accurate prediction model. Methods: Retrospective chart review was performed in 917 breast cancer patients who underwent surgery treatment between 2002 and 2017 and pathologically confirmed 1-2 positive SLNs. According to the date of surgery, patients were divided into training group (497 cases) and validation group (420 cases). A nomogram was built to predict nSLN metastasis and the accuracy of the model was validated. Results: Among the 917 patients, 251 (27.4%) had nSLN metastasis. Univariate analysis showed tumor grade, lymphovascular invasion (LVI), extra-capsular extension (ECE), the number of positive and negative SLN and macro-metastasis of SLN were associated with nSLN metastasis (all P<0.05). Multivariate Logistic regression analysis showed the numbers of positive SLN, negative SLN and macro-metastasis of SLN were independent predictors of nSLN metastasis (all P<0.05). A nomogram was constructed based on the 6 factors. The area under the receiver operating characteristic curve was 0.718 for the training group and 0.742 for the validation group. Conclusion: We have developed a nomogram that uses 6 risk factors commonly available to accurately estimate the likelihood of nSLN metastasis for individual patient, which might be helpful for radiation oncologists to make a decision on regional nodal irradiation.目的： 探讨腋窝前哨淋巴结(SLN)1～2枚阳性的乳腺癌患者，腋窝非前哨淋巴结(nSLN)转移的危险因素，预测nSLN转移风险。 方法： 选择2002—2017年手术治疗的、SLN 1～2枚阳性的乳腺癌患者917例，收集其临床病理资料。根据手术时间将患者分为建模组(497例)和验证组(420例)。以nSLN是否转移为因变量，建立列线图模型，并对模型的预测效果进行验证。 结果： 917例患者的nSLN转移率为27.4%。单因素分析显示，病理分级、原发肿瘤脉管瘤栓、淋巴结包膜外受侵、SLN阳性数、SLN阴性数、SLN是否宏转移与乳腺癌患者腋窝nSLN转移有关(均P<0.05)。多因素Logistic回归分析显示，SLN阳性数、SLN阴性数和SLN是否宏转移为乳腺癌患者腋窝nSLN转移的独立影响因素(均P<0.05)。以有统计学意义的变量，建立了列线图模型。建模组预测nSLN转移风险的受试者工作特征(ROC)曲线下面积为0.718，验证组预测nSLN转移风险的ROC曲线下面积为0.742。 结论： 利用6个危险因素建立的nSLN转移风险预测模型有较高的准确性，可能有助于临床医师进行临床决策。.