Construction and validation of a model for predicting axillary lymph node metastasis in breast cancer based on multiparameter ultrasound and clinicopathological factors
CHEN Shunjun
XU Bin
QIN Shaojie
LI Zhi
LIN Dewei
ZHANG Yi
Abstract：Objective To analyze the relationship between ultrasound image features,clinicopathological features and axillary lymph node(ALN)metastasis of patients with breast cancer,and construct its regression equation.Methods Data of 396 breast cancer patients diagnosed by surgery and pathology in the First Affiliated Hospital of Henan University of Science and Technology from June 2020 to January 2024 were retrospectively collected.All patients received ultrasound examination before surgery.A study used propensity score matching to divide 126 patients with ALN metastasis into metastasis groups Ⅰ and Ⅱ(63 cases each),and 270 non-metastasis patients into non metastasis groups Ⅰ and Ⅱ(135 cases each).Among them,metastasis group Ⅰ and non-metastasis group Ⅰ were used as the training set(198 cases),and metastasis group Ⅱ and non-metastasis group Ⅱ were used as the validation set(198 cases).By comparing baseline characteristics,clinicopathological features,and ultrasound imaging findings between the two patient groups(with different datasets),multivariate analysis was performed to identify independent risk factors for ALN metastasis.A logistic regression prediction model was subsequently developed,and its predictive performance was evaluated.Results Clinical pathological analysis revealed significant correlations between metastasis and several factors,including vascular infiltration(training set:28.57%metastatic vs.8.15%non-metastatic,P<0.001;validation set:22.22%vs.9.63%,P=0.016),the pathological type of"other invasive cancers"(training set:19.05%vs.0.74%,validation set:17.46%vs.0.74%,all P<0.001),and low differentiation degree(training set:71.43%vs.33.33%;validation set:65.08%vs.32.59%,all P<0.001).Additionally,the validation set identified a higher proportion of triple-negative type in the metastasis group(7.94%vs.2.22%,P=0.030).Ultrasound features further distinguished metastatic cases,with higher proportions of tumor maximum diameter greater than 2 cm,irregular edges,heterogeneous internal echo,hyperechoic halo,posterior echo attenuation,Alder blood flow grade Ⅱ-Ⅲ,and cortical thickening greater than or equal to 3 mm(all P<0.05).Multivariate logistic regression identified vascular infiltration(OR=1.645,P=0.028),poor differentiation(OR=2.061,P=0.018),tumor diameter greater than 2 cm(OR=1.826,P=0.012),hyperechoic halo(OR=1.680,P=0.030),Alder grade Ⅱ-Ⅲ(OR=1.711,P=0.025),and cortical thickening greater than or equal to 3 mm(OR=1.842,P=0.002)as independent risk factors for ALN metastasis.The resulting prediction model,logit(P)=-9.762+0.498×vascular infiltration+0.723×differentiation degree+0.602×maximum tumor diameter+0.519×hyperechoic halo+0.537×Alder grading+0.611×cortical thickness,demonstrated excellent performance(area under the curve=0.899,95%CI:0.848-0.937)with sensitivity of 87.30%and specificity of 84.44%,indicating strong discriminatory ability and clinical utility.Conclusion The independent risk factors for ALN metastasis of breast cancer include vascular invasion,poor differentiation,tumor maximum diameter greater than 2 cm,hyperechoic halo ring,Alder blood flow Ⅱ-Ⅲ grade and cortical thickening greater than or equal to 3 mm.The prediction model based on these factors has excellent differentiation ability and can provide quantitative basis for clinical decision-making.
Keywords：Breast cancerAxillary lymph node metastasisUltrasonic image characteristicsClinicopathologyRegression equationPredicted value
Publication Date：2025-07-28
Online Publishing Date：2025-10-17(First online date of this platform, not the publication date of the document)
Pages：8( 658-665 )