The Value Of Preoperative Ultrasound Combined With Clinical Multivariate Regression Model To Predict Axillary Lymph Node Metastasis In PN1Breast Cancer

Objective:To investigate relationship of ultrasonographic multiparameters and axillary lymph node metastasis in pNl breast cancer, and to build a multivariate regression model to portend pNl axillary lymph node involvement by ultrasonographic and clinical-pathological factors.Methods:A total of80patients with invasive breast cancer who were underwent axillary lymph node dissection or sentinel lymph node biopsy were confirmed by pathology with pN1and pN0and were examined carefully in primary breast tumors and axillary lymph nodes by high frequency liner-array probes (5-12MHz) of PHILIPS HDII. They were analyzed that22parameters of clinical and pathological information and tumor ultrasonographic characteristics and6parameters of axillary ultrasonographic features came from80patients and92axillary lymph nodes. Univariate and multivariate analysis were performed by SPSS16.0statistical software. Receiver operating characteristic curve (ROC curve) was used to test the efficiency of regression model. p-value<0.05was considered statistically significant.Restults:By univariate analysis, the incidence of pNl axillary lymph node involvement was significantly associated with age of primary diagnosis, tumor location, classification of blood supply in the primary tumor, spiculate margin, posterior echo, the fat layer echo in front of tumor, retromammary space, histologic grade, cortical thickness, the border between cortex and medulla, nonuniform or uniform cortex, cortex-hilum thickness ratio. By multivariate analysis, pNl axillary lymph node involvement was significantly associated with the presence younger than40of primary diagnosis, classification Ⅲ of blood supply in the primary tumor, enhance of the fat layer echo in front of tumor, retromammary space disappearing, nonuniform cortex, cortex-hilum thickness ratio>1, which Odd Ratio (OR) was respectively17.78(95%CI2.50～126.54),2.47(95%CI0.16～32.97),4.57(95%CI1.13～18.53),4.4(95%CI0.87～22.18),5.1(95%CI1.68～15.44),3.53(95%CI1.16～10.75). ROC curves were created with the fat layer echo in front of tumor, the combination of these risk factors(including age, classification of blood supply in the primary tumor, the fat layer echo in front of tumor, retromammary space), nonuniform or uniform cortex, cortex-hilum thickness ratio, the combination of these risk factors (including nonuniform or uniform of cortex, cortex-hilum thickness ratio), union of total six parameters (including age, classification of blood supply in the primary tumor, the fat layer echo in front of tumor, retromammary space, nonuniform or uniform cortex, cortex-hilum thickness ratio). Area Under the Curve (AUC) were respectively0.679(95%CI0.542～0.816),0.824(95%CI0.709～0.940),0.696(95%CI 0.569-0.823),0.670(95%CI0.541～0.799),0.757(95%CI0.640～0.873),0.909(95%CI0.825～0.994). The sensitivity and specificity of preoperative ultrasound combined with clinical multivariate regression model to predict axillary lymph node metastasis in pNl breast cancer were72.7%and98.2%. In our study, the sensitivity and specificity of ultrasound assessment of pNl axillary lymph node metastasis were91.67%and32.14%, positive predictive value (PPV) and negative predictive value (NPV) were46.48%and85.71%.Conclusion:There was relatively accurate (AUC=0.82) for predicting pNl ALNM by primary tumor ultrasonographic characteristics combined clinical information which provide important information for axillary ultrasound examination. Nonuniform cortex was an early feature of ALNM, which combined cortex-hilum thickness ratio>1provide a fairly accurate predicted (AUC=0.76). The combination of primary tumor and axillary ultrasonographic characteristics and clinical information was the most accurate for predicting pN1ALNM (AUC=0.909).