Study On The Risk Prediction Model Of Non-sentinel Lymph Node Metastasis In Breast Cancer Patients With Positive Sentinel Lymph Node

Purpose: 40%-70% of sentinel lymph node(SLN)metastases early breast cancer patients are limited to SLN.Developing a non-sentinel lymph node(NSLN)metastases risk prediction model can screen out low-risk patients with NSLN metastases and make patients exempt from axillary lymph node dissection(ALND)or regional nodal irradiation(RNI).Methods: Our study included 374 early breast cancer women with positive SLNs and further ALND who were treated in Liaoning Cancer Hospital from 2013 to 2019.Based on the clinical and pathological features of 245 patients in training cohort,the independent risks for NSLN metastases obtained by multivariate analysis were used to construct the NSLN metastases risk prediction model.we tested the model’s calibration and discrimination in training and validation cohort and compared it with Memorial Sloan Kettering Cancer Center(MSKCC)model.Results: The pathological size,lymphovascular invasion,neural invasion,the size of SLN metastases,the number of positive SLNs,the number of negative SLNs,Ki67 status,and the percentage of positive SLNs were statistically significant in univariate analysis(P<0.05).Lymphovascular invasion,neural invasion,the number of positive SLNs,the size of SLN metastases,and Ki67 status were statistically significant in multivariate analysis(P <0.05),and were independent risks for NSLN metastases.The MSKCC model was tested by the patients from training cohort,and areas under the receiver operating characteristic curve(AUC)=0.674(95%CI:0.595～0.753)was smaller than the new model which AUC=0.781(95%CI:0.719～0.843).In the validation cohort,AUC=0.764(95%CI:0.676～0.852),and he calibration of the new model performed well.The false-negative rates of the new nomogram were 2.5%,10.1%,and 25.3%,respectively for the predicted probability cut-off values of 10%,20% and 30 % when applied to screen out low-risk patients.Conclusions: The new model included five variables: lymphovascular invasion,neural invasion,number of positive SLNs,size of SLN metastases,and Ki67 status.Reference to our model can screen out low-risk patients with NSLN metastases more accurately to guide clinical practice with cut-off value of 20%.