Background
Whether there is axillary lymph node metastasis is crucial for formulating the treatment plan for breast cancer. Currently, invasive methods are still used for preoperative evaluation of lymph nodes. If non-invasive preoperative evaluation can be achieved, it will effectively improve the treatment plan.
Objective
Constructed a predict model based on ultrasound examination, which forest axillary lymph node metastasis in breast cancer, and validated this model.
Method
Patients admitted to Xiamen First Hospital from April 2018 to August 2021 with complete case data were included in this study. Patients who had undergone breast cancer resection and axillary lymph node dissection or sentinel lymph node biopsy were divided into a training and validation cohort in a 7:3 ratio. In the training cohort, patients were divided into metastatic and non-metastatic groups based on whether axillary lymph nodes had metastasis. The parameters of the two groups were compared, and statistically significant parameters were included in multivariate analysis. Then, a Nomogram model was constructed, named Lymph metastasis predict model (LMPM). Calibration curves, receiver operating curve (ROC), and decision curve analysis (DCA) were plotted between the training and validation cohort, calculate the risk score of each patient, identify the optimal cutoff value, and test the predictive efficacy of LMPM.
Result
Two hundred seventy-three patients were enrolled in final study, the average age 49.7 ± 8.7, training cohort included 191 patients, the diameter of breast cancer, the lymph node peak systolic flow velocity (LNPS) and the cortex area hilum ratio (CH) of lymph node were exist significant difference in metastatic and non-metastatic group. Multivariate analysis showed cancer diameter, LNPS and CH included in LMPM, the cutoff value was 95, the calibration curve, ROC, DCA in training and validation cohort show satisfactory result.
Conclusion
The predict model-LMPM, can predict axillary lymph node metastasis in breast cancer, which is useful for developing personalized treatment plans. However, further validation of the model is required by incorporating a larger number of patients.