Our study established a nomogram based on baseline characteristics to predict tumor regression patterns after NAC in breast cancer. Five clinicopathological factors and laboratory indexes, including T stage, N stage, menopausal status, ER status, and LMR, were included in the nomogram. Patients who were premenopausal or ER-negative and had higher LMR and later T/N stages had a higher probability of type 1 tumor regression. Such patients had a relatively lower probability of type 2 regression, which means they are not favored candidates for BCS. Consistent with the findings of previous literature[13,14], large tumor size, later nodal stage and lower ER expression are associated with a nonconcentric regression of the tumor after NAC.
Tumor regression patterns have been explored by various institutes. Molecular subtypes have been utilized to correlate with tumor regression patterns, but no consensus has been reached[15,16]. In our study, HR+/HER2- breast cancers had a significantly higher rate of type 2 regression (24.3%), which could be explained by their lower aggressiveness in nature and lower sensitivity to chemotherapy. HR-/HER2+ and HR-/HER2- breast cancers were found to have a better response to presurgical chemotherapy, especially in HR-HER2+ cases when HER2-targeted therapy was added. In the present study, the rates of type 2 regression in HR-HER2+ and HR-HER2- cases were only 11.4% and 12.9%, respectively. Similarly, a meta-analysis revealed that different breast cancer subtypes showed different responses to NAC. Molecular subtypes could, therefore, be considered a predictor for tumor regression patterns. However, it was not incorporated into our nomogram, considering that it is decided by ER/PR/HER2 status and Ki67 index and might cause confounding bias if they were integrated together. In our study, the BCS rate was 38.2% (102/267) and 13.2% (7/53) in patients with type 1 and type 2 regression, respectively. For patients who desire BCS, a consultation with their surgeons immediately before the initial preoperative systemic treatment based on the predicted result of the nomogram would be helpful.
Presurgical breast MRI is the favored modality for evaluating residual cancer after NAC[19,20]. Due to its high sensitivity and low specificity, determining the tumor regression pattern by MRI may not be adequately accurate[21,22]. Furthermore, the concordance rate between MRI and pathology results was barely satisfactory[23,24]. Plana et al. reported that presurgical MRI detected additional breast diseases and prompted conversion from local excision to more extensive surgery in 12.8% of women and nearly half of these were inappropriate due to the incidence of false positive. Thus, the tumor regression patterns are determined by pathology results. The presurgical breast MRI does not provide information on the possible tumor regression model at the early phase. Our nomogram utilized the baseline characteristics of breast tumors, which allows the early prediction of tumor response to NAC, and the treatment strategy may be modified in the early stage.
Laboratory indexes, including peripheral immune cell counts and nutritional indicators, are inexpensive and easily accessible. In our study, 4 laboratory indexes, including LMR, PLR, fibrinogen and D-dimer, were explored. Higher levels of D-dimer and fibrinogen are associated with advanced malignancy in breast cancers, and these parameters are linked to homeostasis and tumor progression. PLR has been recognized as a predictor for chemotherapy response and a prognostic factor for better DFS outcomes[26,27]. LMR has been studied in various types of cancers, especially in alimentary system tumors[26,28]. Our study also identified LMR as an independent factor for predicting tumor regression patterns. A higher LMR was associated with a higher probability of type 1 regression, which could be explained by the stronger aggressiveness in tumors with a higher LMR[29,30] and increased sensitivity to chemotherapy. LMR can be recognized as a promising predictive factor for the tumor response to NAC. To the best of our knowledge, no other studies have reported the correlation between LMR and tumor regression patterns.
Our study had a retrospective design, and we could only incorporate accessible factors in clinical practice. Other variables, e.g., from gene levels or unknown aspects, may need further exploration. The AUC of our model needs to be improved possibly by incorporating other types of variables. A radiomic signature combining multiparametric MRI features would be promising. Meanwhile, type 1 regression is encountered much more frequently than type 2 regression in clinical observation, which leads to an imbalance in the number of cases with the two regression patterns. This situation is consistent with our clinical practice since most tumors have a good response to chemotherapy (and HER2-targeted therapy, e.g., trastuzumab in HER2-positive cases). Additionally, the nomogram established based on single-center data needs further external validation in other cohorts.
In conclusion, HR+/HER2- breast cancers are more likely to have type 2 regression after neoadjuvant chemotherapy. Five baseline factors, including clinicopathological factors and laboratory indexes, were incorporated to establish a nomogram, which exhibited a sufficient discriminatory ability for predicting different patterns of tumor regression.