Establishment And Validation Of The Predictive Model For Risk Factors Of Axillary Lymph Node Metastasis In Triple-negative Breast Cancer Evaluated By Ultrasound Combined With Clinical Pathology

ObjectiveTo investigate the correlation among the clinicopathological features,ultrasound imaging features and axillary lymph node metastasis in triple-negative breast cancer patients.Establishing a prediction model,which can predict axillary lymph node metastasis,so as to serve an accurate evaluation of lymph node status in TNBC patients for clinicians.MethodsA total of 238 patients with triple-negative breast cancer were enrolled in this study retrospectively.Enrolled patients are randomly assigned to two groups with a ratio of 2:1,of which 159 cases were enrolled in modeling group,79 cases were enrolled in validation group,each group was divided into metastasis group and non-metastasis group according to the axillary lymph node status.Collecting the clinical pathological characters and sonographic features were analyzed by univariate in both modeling group and the validation group respectively,and selecting the risk factors related to axillary lymph node metastasis by SPSS 26.0software.R 4.2.1 software was applied to binomial logistic regression for multivariate analysis,established a binary logistic regression model,and construct a nomogram.Finally,evaluating the validity and clinical application of the model.ResultsAmong the 238 patients with TNBC,the onset age is 23~80 years old,and81 patients(34%)had axillary lymph node metastasis.There was no significant difference in basic descriptive characteristics between the modeling group and the validation group(P>0.05),this indicates that the data are comparable between the two groups.Univariate analysis showed that: There were statistically significant differences between groups in low expression of HER-2,the higher index of Ki-67,tumor diameter,unclear tumor boundary and tumor blood flow grade(P<0.05).There were no statistically significant differences in age,BMI,menopause,familial medical history,P53,CK5/6,pathological type,tumor histological grade,tumor location,shape,internal echo,posterior echo and calcification(P>0.05).Subsequent multivariate analysis screens that low expression of HER-2,Ki-67 index,tumor diameter,unclear tumor boundary and tumor blood flow grade were the risk factors of ALNM in TNBC patients(P<0.05).The nomogram model,established on the basis of the analysis result above,the areas under the curve(AUC)of the modeling group was 0.779[95%CI(0.71,0.85)],the cut-off value was 0.28,and the specificity was 85.5%.The newly established logistic regression model was internally validated by validation group data,the AUC was 0.768[95% CI(0.65,0.89)],the positive predictive value of the model was 77.0%,the negative predictive value of the model was 90.5%.The calibration curve and decision curve analysis show that the model was of great stability and clinical practicability.ConclusionsLow expression of HER-2,higher Ki-67 index,tumor diameter,unclear tumor boundary and tumor blood flow grade were the risk factors of ALNM in TNBC patients,the prediction model of lymph node metastasis which combined with the clinicopathological and ultrasound features of the primary tumor can provide a reference of individualized treatment.