Logistic回归分析乳腺癌常规超声征象与腋窝淋巴结转移的关系

Relationships of conventional ultrasonic manifestations and axillary lymph node metastasis of breast cancer

目的采用Logistic回归分析乳腺癌常规超声征象与腋窝淋巴结转移的关系。方法收集820例女性乳腺癌共874个乳腺原发病灶的超声声像图,根据发生腋窝淋巴结转移与否分为转移组(n=334)和未转移组(n=540);对比2组乳腺癌病灶的超声特征(位置、形态、边界、边缘、有无钙化、最大径、纵横比、内部回声、后方回声及Alder血流分级);采用组间差异有统计学意义的参数建立Logistic回归模型,分析其预测淋巴结转移的价值。结果转移组与未转移组乳腺癌病理类型及原发病灶最大径、纵横比、内部回声、边界和Adler血流分级差异均有统计学意义(P均<0.05)。根据原发病灶最大径>2 cm、纵横比≤1、内部回声不均、边界不清及Adler血流Ⅱ~Ⅲ级预测乳腺癌腋窝淋巴结转移的AUC分别为0.571、0.571、0.512、0.588及0.579;联合以上5个超声特征构建预测模型,以0.407为截断值时,AUC为0.696,大于单一特征(P均<0.05),且与病理学结果的一致性尚可(Kappa=0.702)。结论根据乳腺癌常规超声征象构建的Logistic联合模型可预测乳腺癌腋窝淋巴结转移;原发病灶最大径>2 cm、纵横比≤1、内部回声不均匀、边界不清晰及Adler血流Ⅱ~Ⅲ级提示转移风险较高。

Objective To analyze the relationships of conventional ultrasonic manifestations of primary lesions of breast cancer and axillary lymph node metastasis with Logistic regression.Methods Data of ultrasound of 820 female breast cancer patients with 874 primary breast lesions were collected.The patients were divided into metastases group(n=334)and non-metastases group(n=540)according to axillary lymph node metastasis or not.The ultrasonic features of primary breast lesions(location,shape,boundary,edge,calcification,maximum diameter,aspect ratio,internal echo,posterior echo and Alder blood flow grade)were observed and compared between groups.Logistic regression model was established according to indexes being statistically different to analyze the value of the above features for predicting lymph node metastasis of breast cancer.Results There were significant differences of pathological type,maximum diameter,aspect ratio,internal echo,boundary and Adler blood flow grade of breast cancer between groups(all P<0.05).The area under the curve(AUC)of the maximum diameter of primary breast lesion>2 cm,aspect ratio≤1,uneven internal echo,clear boundary and Adler blood flow levelⅡ—Ⅲfor predicting axillary lymph node metastasis of breast cancer was 0.571,0.571,0.512,0.588 and 0.579,respectively.Taken 0.407 as the cut-off value of combination of all the features,the AUC of the prediction model was 0.696,higher than that of all features alone(all P<0.05),with acceptable consistency with pathological results(Kappa=0.702).Conclusion Logistic model established based on combination of breast cancer conventional ultrasonic manifestations could be used to predict axillary lymph node metastasis.The maximum diameter of primary breast lesion>2 cm,aspect ratio≤1,uneven internal echo,not clear boundary and Adler blood flow levelⅡ—Ⅲindicated high risks for axillary lymph node metastasis of breast cancer.