Contemporary Trends in Axillary Surgery for ER-Positive, HER2-Negative Breast Cancer Stratified by Neoadjuvant Endocrine Therapy, Neoadjuvant Chemotherapy, or Upfront Surgery.

Methods
Female patients age ≥50 years with a diagnosis of clinical T1-4c, N0-1, ER+/HER2– breast cancer from 2012 to 2021 were selected from the NCDB (fall 2023 data release). The study excluded patients with inflammatory breast cancer (clinical T4d) or metastatic disease (clinical M1 or pathologic M1), as well as those who did not undergo lumpectomy or mastectomy. Patients with missing or unknown breast surgery, axillary surgery, pathologic T or N category, time from diagnosis to definitive breast surgery, chemotherapy receipt and timing, or endocrine therapy receipt and timing also were excluded (Fig. 1).
The study defined NET as receipt of endocrine therapy before definitive breast surgery and NAC as receipt of chemotherapy before definitive surgery. Patients recorded as receiving both NET and NAC were excluded. Time to treatment was defined as the time from diagnosis to chemotherapy for patients who received NAC, as time fom diagnosis to endocrine therapy for patients who received NET, and as time from diagnosis to definitive breast surgery for those who had surgery first. Axillary surgery was defined as described in the NCDB data dictionary, variable RX_SUMM_SCOPE_REG_LN_2012, as follows: code 0 (none), code 2 (SLNB alone), codes 3 to 5 (ALND alone), and codes 6 and 7 (SLNB + ALND).
The axillary surgery group was designated as none, SLNB, or ALND (ALND alone or SLNB + ALND). Targeted axillary dissection is not coded separately in the NCDB and likely would be classified under SLNB unless a completion lymph node dissection was subsequently performed, in which case it likely would be classified as SLNB + ALND.
Patient characteristics were summarized as number (%) for categorical variables and as median (interquartile range [IQR]) for continuous variables. Differences were tested with chi-square for categorical variables and with analysis of variance (ANOVA) for continuous variables. Logistic regression was used to identify factors associated with undergoing ALND versus not undergoing ALND (SLNB alone or none).
This model was built in the generalized estimating equations framework and included an exchangeable covariance structure to account for the correlation of patients treated at the same facility. The following covariates were included in the logistic model: age, Charlson-Deyo comorbidity score, clinical T category, clinical N category, facility type, facility location, insurance type, race and ethnicity, year of diagnosis, LVI, and treatment group.
Odds ratios (ORs) and 95 % confidence intervals (CIs) are reported. Unadjusted overall survival was estimated with the Kaplan-Meier method, and log-rank tests were used to compare survival between groups. Follow-up time was estimated with the reverse Kaplan-Meier method.
No adjustments were made for multiple comparisons. Only patients with complete data were included in the logistic regression model, and the model sample size is reported. All statistical analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC, USA). This project was deemed exempt from institutional review board (IRB) review because it used only de-identified, publicly available data from the NCDB in accordance with institutional policies.

Results
The study identified 792,581 female patients age ≥50 years with a diagnosis of clinical T1-4c, N0-1, ER+/HER2– breast cancer from 2012 to 2021 who met all the other inclusion and exclusion criteria (Fig. 1). Of these patients, 94.3 % (747,659) underwent surgery first, 2.6 % (20,704) received NAC, and 3.1 % (24,218) received NET (Table 2). Regarding axillary surgery, 74.5 % (n = 590,827) underwent SLNB, 23.5 % (n = 186,457) underwent ALND, and 1.9 % (n = 15,297) had no axillary surgery. The overall rate of ALND decreased from 35.0 % in 2012 to 13.5 % in 2021. Stratification by clinical nodal status showed that ALND use declined from 30.3 to 10.6 % among the patients with cN0 disease and from 84.7 % to 69.6 % among the patients with cN1 disease. The rate of ALND at academic centers was 23.2 % versus 27.8 % at community centers. The rate of ALND was 78.5 % for cN1 disease compared with 19.6 % for cN0 disease and 44.5 % if LVI was present compared with 19.5 % if LVI was absent (p < 0.001).

The baseline median tumor size for the patients undergoing ALND was 1.9 cm (IQR, 1.2–3.0 cm) versus 1.3 cm (IQR, 0.9–2.0 cm) for the patients undergoing SLNB and 1.2 cm (IQR, 0.8–2.0 cm) for the patients not undergoing axillary surgery (Table S1). The median time from diagnosis to treatment was shorter in the NET group (0.95 months; IQR, 0.53–1.68 months) than in the NAC group (1.18 months; IQR, 0.82–1.68 months) or the surgery-first group (1.18 months;, IQR, 0.79–1.71 months) (p < 0.001) The patients who received NAC were more likely to receive adjuvant radiation (71.4 %) than surgery first (61.2 %) or NET (51.6 %) (p < 0.001; Table 2).
Of the patients who received NAC, 63.0 % underwent ALND compared with 29.7 % of those receiving NET and 22.2 % of those undergoing surgery first (p < 0.001; Table 2). The patients receiving NAC had the highest rates of ALND in both cN0 and cN1 disease (39.6 % and 80.7 %, respectively). For those with clinically node-positive disease at presentation, the rates of ALND after NET did not differ from that for surgery first (77.9 % for both; Fig. 2).
In addition to treatment group, several other factors were significantly associated with receipt of ALND. Larger tumor size (T2-4) and clinically node-positive disease were strongly associated with higher odds of receiving ALND. The presence of LVI also significantly increased the likelihood of ALND (OR, 2.08; 95 % CI, 2.03–2.13; p < 0.001). Younger age, black race, and higher comorbidity scores were associated with increased ALND rates. Conversely, treatment at an integrated network or at institutions in the West or Northeast were associated with lower odds of ALND. After adjustment for such variables affecting ALND rates, the odds of undergoing ALND were higher for those receiving NAC (OR, 1.18; 95 % CI, 1.12–1.24) and, to a lesser degree, for those receiving NET (OR, 1.10; 95 % CI, 1.05–1.14) compared with those receiving surgery first (p < 0.001; Table 3).

Among the clinically node-positive patients, 19.6 % were found to be pathologically node-negative after receiving NAC, whereas this was true for only 8.3 % of those receiving NET (p < 0.001). An ALND alone was performed for 55.4 % of the surgery-first group, 54.9 % of the NAC group, and 56.4 % of the NET group. Interestingly, of the clinically node-positive patients who underwent SLNB and were found to have residual nodal disease, 70.5 % in the NAC group went on to completion ALND compared with 57.5 % in the NET group (p < 0.001; Table 4).

After 10 years of follow-up evaluation, unadjusted overall survival varied significantly by treatment group (p < 0.001, log-rank). Among the patients with cN0 disease, the highest survival rate was in the surgery-first group (76.1 %; 95 % CI, 75.9–76.4 %) compared with the NAC group (69.8 %; 95 % CI, 67.1–72.4 %) and the NET group (66.6 %; 95 % CI, 64.4–68.8 %). Among the patients with cN1 disease, the highest survival rate was in the surgery-first group (62.0 %; 95 % CI, 60.9–63.1 %) followed closely by the NAC group (62.0 %; 95 % CI, 59.8–64.1 %) and more distantly by the NET group (43.4 %; 95 % CI, 36.8–49.9 %) (Fig. 3).
Unadjusted overall survival also varied significantly by axillary surgery type (p < 0.001, log-rank). At 10 years, those who underwent SLNB had the highest survival rate (77.8 %; 95 % CI, 77.5–78.1 %) followed by those who underwent ALND (69.4 %; 95 % CI, 69.0–69.9 %) and finally by those who had no axillary surgery (48.6 %; 95 % CI, 46.4–50.8 %) (Fig. 4).

Discussion
Neoadjuvant endocrine therapy can be an attractive alternative treatment for HR+/HER2– invasive breast cancer for select patients. Evidence exists to show that NET can provide similar efficacy in terms of tumor downstaging with decreased toxicity compared with NAC.14,21,22 This treatment strategy has been shown to improve rates of BCS for patients who would otherwise be ineligible.23 Additionally, we found that the median time from diagnosis to treatment was shortest in the NET group, likely due to ease of administration and less coordination required. This expedited treatment initiation can be advantageous for various reasons. Previous studies also have demonstrated that NET can be safely used as a bridge to surgery, such as during the COVID-19 pandemic.24 Despite these benefits, no clear guidelines on axillary management after NET currently exist, and this paucity of data was the target of our examination of recent data in the NCDB (Table 1).
We found that evidence-based efforts toward de-escalation of axillary surgery have resulted in lower rates of ALND over time regardless of neoadjuvant treatment type or upfront surgery. This is especially true at academic centers, which may be quicker to implement newer evidence-based findings and guidelines.25 For example, clinically node-positive disease historically required ALND, but currently, SLNB is recommended for select patients who can be clinically downstaged before surgery.11,16 Additionally, for postmenopausal women with up to three positive nodes, the incorporation of oncotype DX testing also may have contributed to selective de-escalation of systemic treatment with reduced emphasis on extensive axillary surgery as an extension of this treatment approach.26
We found that a much higher percentage of patients in the NAC group underwent ALND than in the NET and surgery-first groups. We hypothesized that this was due to what would be considered traditionally more aggressive disease (larger tumor size, increased nodal burden) treated with NAC followed by NET then surgery first (Table 2). Interestingly, even after adjustment for other meaningful variables, NAC still was associated with increased odds of treatment with ALND after reception of NAC compared with NET.
Until we have the results of Alliance 11202 and/or the TAXIS trial, patients with ypN+ disease are currently still advised to undergo completion ALND.27–29 However, we found that not all patients with residual nodal disease after SLNB went on to receive ALND. Among the clinically node-positive patients in this study, 19.6 % achieved nodal pCR after receiving NAC, compared to 8.3 % of those who received NET. Of those who did not achieve nodal pCR, only 70.5 % in the NAC group and 57.5 % in the NET group went on to completion ALND. Although findings have shown pCR to be a reasonable surrogate for overall survival in triple-negative and HER2+ breast cancer, it is a less appropriate surrogate marker for HR+/HER2– tumors, especially luminal A type tumors.30,31 Therefore, lack of pCR may not have as much influence in the decision to proceed with completion ALND, especially for patients treated with NET who receive only a small proportion of their systemic therapy upfront. Additionally, in the NET setting, tools such as the modified Preoperative Endocrine Prognostic Index (PEPI) score are increasingly used to guide treatment decisions given that PCR is a less reliable surrogate for long-term outcomes in HR+/HER2– breast cancer.31–33 This reasoning may explain such a discrepancy in the care observed in this study compared with the current guidelines.
Survival analysis indicated higher survival rates for patients undergoing surgery first for both cN0 and cN1 disease, likely related to patient selection and more readily operable tumors. Interestingly, the survival rates between those receiving NAC and those receiving NET were similar for cN0 disease, but for the individuals with cN1 disease, those receiving NAC had survival rates similar to those for the surgery-first patients, and those receiving NET had the worst overall survival. Although causation cannot be established from this retrospective analysis, this may reflect the selection of patients for NET who were older or had more comorbidities making them less likely to tolerate NAC.
Survival curves also differed significantly when stratified by axillary surgery, with the lowest survival rates for individuals not undergoing axillary surgery, likely because omission of axillary surgery was more common for older individuals and those with comorbidities. Survival was higher for the individuals undergoing SLNB than for those undergoing ALND, which is most likely representative of increased nodal disease in the latter.
The limitations to this study included its retrospective design with extraction of data from a large database representative of Commission on Cancer (CoC)-accredited intuitions, which may limit generalizability to other practices. The CoC accreditation focuses on meeting standards of multidisciplinary care, data collection, and quality improvement. As a result, the implementation of NET may be over-represented, and the implementation of axillary management after treatment may be different. The database also lacks certain granular details on the reason behind treatment selection, including patient preferences that may influence treatment decision-making. The NCDB also does not capture the use of specific agents, such as CDK4/6 inhibitors, and the number of individuals treated with such agents could not be accounted for or excluded. Finally, this observational study lacked the ability to draw causal inferences from the patterns elucidated in this analysis.
Collectively, however, the information in this report helps to inform the current state of the art of axillary management for NET and highlights areas for improvement in standardizing care guidelines. Despite the lower nodal PCR rates with NET, omission of completion ALND remains common in this group. This may be reflective of a tendency to treat this group similarly to the surgery-first group given that a much smaller proportion of systemic therapy is administered before surgery than with NAC. However, definitive conclusions as to the reason cannot be drawn, underscoring the need for evidence-based NET-specific guidelines for axillary management.

Supplementary Information
Below is the link to the electronic supplementary material.