This study integrated the clinicopathological characteristics of female patients with breast cancer and positive ipsilateral ALNs, who were treated with NAT and surgery. Statistical methods were used to screen for predictors of apCR and design a predictive model to predict the apCR rate post-NAT. The results of univariate analysis showed that apCR was significantly associated with the pathological grading of aspiration specimens, molecular typing, STR, clinical N staging, and response to chemotherapy (all P <0.05). A multivariate binary logistic regression equation was established, and a forward stepwise selection method was adopted to enter variables into the model. The findings demonstrate a greater possibility of apCR post-NAT in cases of histological grade III, low stroma, clinical N1 stage, and PR or CR to chemotherapy. A predictive model for apCR post-NAT was developed using statistically significant variables in the logistic regression equation. The predictive model consisted of four independent predictors: pathological grading of the aspiration specimen, STR, clinical N staging, and response to chemotherapy. The predictive model had good accuracy and demonstrated good clinical utility when both ROC curves and internal validation were performed, which contributed to the option of individualized de-escalation of axillary surgery.
With the concept of multimodal treatment for breast cancer, NAT transforms locally advanced, inoperable patients into operable ones; i.e., it results in downstaging of the breast cancer focus and ALNs.14 This study focused on ALN management after NAT. According to previous studies, the effective use of NAT and targeted therapy has enabled the post-NAT apCR rate to reach 30–63 %.5–6 The National Comprehensive Cancer Network guidelines previously recommended that ALND remain the standard of care in patients with positive clinical lymph nodes after NAT, regardless of lymph node outcome; recently, however, the feasibility of replacing ALND with sentinel lymph node biopsy in initially cN+ patients turning to negative ALN and undergoing NAT, has been confirmed. Based on the results of the Z1071 and SENTINA trials,15–16 for initial cN+ patients who achieved ycN0 after NAT, follow-up ALND may not be considered when a combined tracer detected >2 negative sentinel lymph nodes. Clinicians therefore need to be cautious when assessing if the patient has achieved apCR preoperatively; incorrect clinical judgment may otherwise result in residual localized metastatic lymph nodes. The predictive model for apCR post-NAT constructed in this study indicated that patients with pathological grade 3, low stroma, clinical stage N1, and CR/PR to chemotherapy achieved apCR more easily.
While the most commonly used noninvasive tool for clinical judgment of the status of ALNs after NAT is imaging, imaging data only assesses tumor burden and not the biological behavior of the tumor or its response to chemotherapeutic agents; therefore, imaging alone cannot predict the probability of pathological complete response of ALN with maximum accuracy. The predictive model of ALN status in breast cancer was first developed by the Memorial Sloan-Kettering Cancer Center in the USA in 2003.17 Scholars at the center incorporated nine types of clinical and pathological data into the model, including tumor size, histological grade, whether it was a rapid pathological section, number of positive sentinel lymph nodes, number of negative sentinel lymph nodes, detection method of sentinel lymph nodes, estrogen receptors, vascular cancer thrombi, and multifocal tumors. The AUC values obtained from the validation of this model at different institutions varied considerably from 0.58–0.86.18, 19 Other institutions have also attempted to build different predictive models, such as the SCH model, constructed by Fudan University Shanghai Cancer Hospital in China20; the Olga predictive model, constructed using data from the US National Cancer Database21; the Tenon score in France22; the M.D. Anderson score23; the Cambridge score in the UK.24 These scoring systems have shown good accuracy regarding internal validation; however, they remain to be tested regarding external validation.
In this study, we included STR—the ratio of tumor cells to stroma in the tumor tissue, which indirectly reflects the relationship between the tumor cells and microenvironment—for the first time. Sophie et al.,25 demonstrated that patients with high-stroma, HER2-negative, early breast cancer have a lower rate of pCR and MP response after NAT. Similar to primary tumor cells in the breast, tumor cells in ALNs are affected by STR. This study revealed that the high-stroma group had a lower apCR rate than the low-stroma group, suggesting that STR may also be a predictor of post-NAT apCR. The significantly higher apCR rate after NAT indicated that ALND can be selectively avoided after NAT in some patients with initially positive ALNs, especially those with pathological grade 3, low stroma, clinical stage N1, and CR/PR to chemotherapy when >2 negative sentinel lymph nodes were detected. Still, for patients in the initial cN2-3 group, the apCR rate was only 26% due to the high tumor burden, and ALND was recommended. The nomogram predictive model constructed in this study can predict the possibility of apCR based on patient and disease characteristics, which is useful for guiding both physicians and patients in selecting further treatment options; however, the predictive model is not intended to replace the clinician’s judgment but assist the physician’s decision by providing more objective probability estimates to complement other relevant clinical data.
This study has some limitations. First, this was a single-center study, and further analysis and modeling with a larger data are needed to refine the model and its application. Second, the predictive model is based on the treatment modality of the disease over a certain period; with the development of medical standards and treatment techniques, the predictive model becomes inaccurate over time and needs to be updated.