Abstract: Objective　To establish an intraoperative predictive model to assess the risk of non-sentinel lymph node (NSLN) metastasis in breast cancer patients with positive sentinel lymph nodes and total mastectomy. Methods　Data of 601 breast cancer patients who underwent total mastectomy in China were retrospectively collected, which including 221 cases in the modeling group, 189 cases in the internal validation group, and 191 cases in the external validation group. Logistic regression analysis was used to mine the risk factors related to NSLN metastasis in patients in the modeling group, and a multi-factor model was established and verified internally and externally. At the same time, the application effect of three existing prediction models after NSLN metastasis (MSKCC model, Tenon scoring system and MDA model) in population of this study was compared. Results　The positive NSLN rates were 32.6%, 32.3% and 50.3% in the modeling group, the internal validation group and the external validation group, respectively. Multivariate analysis showed that age (OR=0.968, 95%CI: 0.941-0.996), PR status (OR=0.484, 95%CI: 0.247-0.951), tumor size (OR=1.491, 95%CI: 1.151-1.932), the number of positive SLN (OR=1.868, 95%CI: 1.278-2.730) and the number of negative SLN (OR=0.763, 95%CI: 0.631-0.922) were the influencing factors of NSLN metastasis. The calibration degree (Hosmer-Lemeshow χ2=8.309, P=0.404) and differentiation degree (AUC=0.752, 95%CI: 0.685-0.819) of the NSLN metastasis prediction model based on the above indicators were good in the modeling group. The AUC values of the internal validation group and the external validation group were 0.751 (95%CI: 0.676-0.825) and 0.681 (95%CI: 0.606-0.757), respectively. The predictive value of this model was similar to that of MSKCC model and TENON scoring system, and slightly better than MDA model. Conclusion　The intraoperative prediction model of NSLN metastasis established for SLN positive patients with early breast cancer undergoing total mastectomy has good internal and external validation results, and its accuracy is comparable to existing postoperative prediction models.

Key words: breast neoplasms, sentinel lymph node biopsy, lymphatic metastasis, intraoperative prediction model