双模态影像联合免疫组化构建模型预测cN1期乳腺癌新辅助化疗后腋窝淋巴结状态

Combining bi-modal imaging and immunohistochemical indicators to predict axillary lymph nodes status in cN1 breast cancer patients receiving neoadjuvant chemotherapy

目的探讨超声、乳腺X线双模态影像及免疫组化指标构建模型预测临床淋巴结Ⅰ期(cN1期)乳腺癌新辅助化疗(NAC)后腋窝淋巴结(ALN)病理状态的价值。方法回顾性选取2020年7月至2022年7月于哈尔滨医科大学附属第二医院及附属肿瘤医院就诊、且经穿刺病理证实为ALN转移的cN1期乳腺癌患者134例, 根据NAC后ALN手术病理结果分为两组:完全缓解(pCR)组(67例)和非完全缓解(npCR)组(67例)。对两组的超声、乳腺X线图像特征及免疫组化指标进行组间比较, 筛选出与ALN pCR相关的差异特征, 基于Logistic回归分别构建模型A (超声)、模型B(超声+乳腺X线)、模型C(超声+免疫组化)、模型D(超声+乳腺X线+免疫组化)4个预测模型。绘制ROC曲线并通过Delong检验评估、比较不同模型的诊断效能, 对预测效能最佳的模型以列线图形式展示, 得到较为直观的结果。结果①ALN短径、ALN长径/短径、淋巴门结构情况、淋巴门血流分布、乳腺X线毛刺、ER、PR、HER-2分子表达水平在两组间比较差异有统计学意义(均P<0.05)。②模型A中ALN长径/短径、ALN淋巴门结构情况是乳腺癌患者NAC后ALN病理状态的独立预测因素;模型B、模型C的独立预测因素则分别在模型A的基础上增加了乳腺X线毛刺、免疫组化指标(ER、PR);模型D中ALN长径/短径、ALN短径、ALN淋巴门结构情况、乳腺X线毛刺、免疫组化指标(ER、PR)是NAC后ALN病理状态的独立预测因素。③模型A、B、C、D预测cN1期乳腺癌NAC后ALN状态的ROC曲线下面积(AUC)分别为0.78、0.84、0.84、0.89, 其敏感性分别为0.71、0.80、0.78、0.86, 特异性分别为0.76、0.74、0.76、0.80, 准确性分别为0.73、0.76、0.77、0.83。④Delong检验结果显示, 模型D的预测效能大于模型A、B、C(0.89比0.78、0.84、0.84, 均P<0.05)。结论超声、乳腺X线双模态影像学特征联合免疫组化指标的预测模型对cN1期乳腺癌NAC后ALN病理状态有较好的预测价值, 可为cN1期乳腺癌患者制定合理的治疗方案提供影像学依据。

Objective To investigate whether ultrasound features,mammographic features and immunohistochemical indicators show any association with rates of axillary pathologic complete response(pCR)in cN1 breast cancer patients receiving neoadjuvant chemotherapy(NAC),and to construct prediction models of axillary pCR to predict axillary lymph nodes(ALN)status,so as to select suitable patients for less invasive axillary surgery after NAC.Methods This retrospective study evaluated 134 consecutive cN1 breast cancer patients with ALN metastasis who underwent NAC in the Second Affiliated Hospital and Tumor Hospital of Harbin Medical University from July 2020 to July 2022.According to the pathological results of ALN surgery after NAC,the cases were divided into pCR and non pathologic complete respose(npCR)groups.The ultrasound images,mammographic images and immunohistochemical indicators of the two groups were compared.In terms of logistic regression algorithm,the model A(the ultrasound model),the model B(the ultrasound combined with mammography model),the model C(the ultrasound combined with immunohistochemistry model)and the model D(the ultrasound combined with mammography and immunohistochemistry model)were respectively established for predicting the pathological state of axillary lymph nodes in breast cancer patients,ROC curves were plotted to evaluate the performance of the models,and the diagnostic efficiency of different models was compared by Delong′s test.The model with the best predictive performance was shown in a nomogram.Results①The P values between two groups of the short diameter of ALN,the ratio of long/short diameter of ALN,fatty hilum and central hilar vascularity,mammographic spiculation,estrogen receptor(ER),progesterone receptor(PR),human epidermal growth factor receptor 2(HER2)were<0.05 by the t test andχ2 test analysis.②The ratio of long/short diameter and fatty hilum in the model A were independent factors for predicting the pathological status of ALN after NAC.The independent predictors of model B and Model C were respectively added with mammographic spiculation and immunohistochemical indicators(ER,PR)on the basis of model A.In the model D,the ratio of long/short diameter,short diameter,fatty hilum,mammographic spiculation,and immunohistochemical indicators(ER,PR)remained significant independent predictors associated with axillary pCR.③The area under ROC curve(AUC)of the model A,B,C,D was 0.78,0.84,0.84 and 0.89,respectively.The sensitivity was 0.71,0.80,0.78 and 0.86,the specificity was 0.76,0.74,0.76 and 0.80,and the accuracy was 0.73,0.76,0.77 and 0.83,respectively.④Delong′s test showed the model D had an improved AUC of 0.89(0.89 vs 0.78,0.84,0.84,all P<0.05).Conclusions The prediction models combining bi-modal imaging and immunohistochemical indicators show good prediction ability and can provide reference for selecting suitable patients for less invasive axillary surgery after NAC.