Changes in Management Strategy and Impact of Neoadjuvant Therapy on Extent of Surgery in Invasive Lobular Carcinoma of the Breast: Analysis of the National Cancer Database (NCDB)

Background Given reports of low response rates to neoadjuvant chemotherapy (NAC) in invasive lobular carcinoma (ILC), we evaluated whether use of alternative strategies such as neoadjuvant endocrine therapy (NET) is increasing. Additionally, we investigated whether NET is associated with more breast conservation surgery (BCS) and less extensive axillary surgery in those with ILC. Patients and Methods We queried the NCDB from 2010 to 2016 and identified all women with stage I–III hormone receptor positive, human epidermal growth factor receptor-2 negative (HR+/HER2−) ILC who underwent surgery. We used Cochrane–Armitage tests to evaluate trends in utilization of the following treatment strategies: NAC, short-course NET, long-course NET, and primary surgery. We compared rates of BCS and extent of axillary surgery stratified by clinical stage and tumor receptor subtype for each treatment strategy. Results Among 69,312 cases of HR+/HER2− ILC, NAC use decreased slightly (from 4.7 to 4.2%, p = 0.007), while there was a small but significant increase in long-course NET (from 1.6 to 2.7%, p < 0.001). Long-course NET was significantly associated with increased BCS in patients with cT2–cT4 disease and less extensive axillary surgery in clinically node positive patients with HR+/HER2− tumors. Conclusions Primary surgery remains the most common treatment strategy in patients with ILC. However, NAC use decreased slightly over the study period, while the use of long-course NET had a small increase and was associated with more BCS and less extensive axillary surgery.

ABSTRACT Background. Given reports of low response rates to neoadjuvant chemotherapy (NAC) in invasive lobular carcinoma (ILC), we evaluated whether use of alternative strategies such as neoadjuvant endocrine therapy (NET) is increasing. Additionally, we investigated whether NET is associated with more breast conservation surgery (BCS) and less extensive axillary surgery in those with ILC. Patients and Methods. We queried the NCDB from 2010 to 2016 and identified all women with stage I-III hormone receptor positive, human epidermal growth factor receptor-2 negative (HR?/HER2-) ILC who underwent surgery. We used Cochrane-Armitage tests to evaluate trends in utilization of the following treatment strategies: NAC, short-course NET, long-course NET, and primary surgery. We compared rates of BCS and extent of axillary surgery stratified by clinical stage and tumor receptor subtype for each treatment strategy. Results. Among 69,312 cases of HR?/HER2-ILC, NAC use decreased slightly (from 4.7 to 4.2%, p = 0.007), while there was a small but significant increase in long-course NET (from 1.6 to 2.7%, p \ 0.001). Long-course NET was significantly associated with increased BCS in patients with cT2-cT4 disease and less extensive axillary surgery in clinically node positive patients with HR?/HER2tumors.
Conclusions. Primary surgery remains the most common treatment strategy in patients with ILC. However, NAC use decreased slightly over the study period, while the use of long-course NET had a small increase and was associated with more BCS and less extensive axillary surgery.
Use of systemic therapy prior to surgical resection can improve outcomes and provide prognostic information for many women with breast cancer. Neoadjuvant chemotherapy (NAC) increases breast conservation rates, and achieving a pathologic complete response (pCR) after NAC is associated with improved disease-free survival. In recent years, a newer approach of using neoadjuvant endocrine therapy (NET) has emerged as a treatment strategy for hormone receptor (HR)-positive tumors. 1 When given for 3-6 months, NET can potentially downstage the tumor in the breast and/or axilla, affording the opportunity to decrease the extent of surgical intervention. 2 When short courses of NET (2-4 weeks) are used, investigators have shown that changes in biomarkers such as the proliferation marker Ki67 can predict improved long-term outcomes and potentially guide adjuvant therapy decisions. 3,4 The POETIC trial (perioperative endocrine therapy-individualizing care) is evaluating the relationship between change in Ki67 after 2 weeks of endocrine therapy and long-term outcomes in over 4000 post-menopausal women with HR-positive breast cancer with the goal of identifying patients who would benefit from additional adjuvant therapy. [5][6][7] Of these strategies, selecting the optimal treatment approach (surgery first, NAC, short-course, or long-course NET) for an individual patient, especially those with HR? disease, remains a clinical challenge. 8,9 For women with invasive lobular carcinoma (ILC) in particular, limited data exist to guide optimal management. 10 ILC is the second most common subtype of breast cancer, representing approximately 10-15% of all breast cancers. ILC has unique features, and is characterized by its diffuse growth pattern in so called ''single file lines,'' resulting from the absence of the adhesion protein E-cadherin. 11 It differs from the more common invasive ductal carcinoma in its mutational profile, appearance on imaging, surgical outcomes, and timing/pattern of recurrence. [11][12][13][14] Many studies have shown poor responses to NAC in ILC with low pCR rates. [15][16][17] Given that most ILC are estrogen receptor (ER)positive and human epidermal growth factor receptor-2 (HER2)-negative tumors, the poor overall response to NAC has led to interest in the use of NET in this tumor type. 18 But, whether this approach is being utilized (either in the short or long durations described) in ILC is unknown. We sought to evaluate the primary management strategies used for ILC, including duration of NET, and how these have changed over time. Additionally, we investigated which surgical procedures were associated with each management strategy.

Study Design
The National Cancer Database (NCDB) is a deidentified national database sponsored by the American College of Surgeons and the American Cancer Society, containing clinicopathologic data and outcomes collected from 1500 accredited cancer centers. The data therein represent approximately 70% of newly diagnosed cancer cases in the USA.
With institutional review board exemption provided by both participating institutions, we queried the NCDB (2016 PUF) from 2010 to 2016 and identified all patients with clinical stage I-III HR?/HER2-ILC who underwent surgery. Patients with HER2? and triple-negative ILC were excluded from our primary analysis sample since they make up a small minority of all ILC cases, but these patients were described separately. Cases with prior cancer history (breast or otherwise), those receiving no treatment at the reporting facility, those with male gender, those with mixed invasive ductal/lobular histology, and those under the age of 18 years were excluded. Primary treatment strategy was classified as primary surgery, NAC, shortcourse NET, or long-course NET by comparing the timing of surgery relative to chemotherapy and hormone therapy start times. Cases were categorized as undergoing primary surgery if they had no systemic treatment prior to surgery. Patients who received chemotherapy starting 31-365 days before surgery constituted the NAC cohort, those who received endocrine therapy for 7-30 days before surgery constituted the short-course NET cohort, and those who received endocrine therapy for 31-365 days before surgery constituted the long-course NET cohort. Patients for whom primary treatment strategy could not be fully defined due to missing data regarding timing were also excluded ( Fig. 1). Within the long-course NET cohort, we analyzed the association between the following treatment duration groups and surgical therapy received based on prior reports studying NET duration: 1-3 months, 3-6 months, 6-9 months, and [ 9 months. 19 Patients ER and progesterone receptor (PR) immunohistochemical staining of C 1% was classified as positive. Tumors that were ER? and/or PR? were classified as hormone receptor positive; tumors that were both ER-and PRwere classified as HR negative.
Our primary study questions were (1) whether the primary treatment strategy for ILC changed over time, (2) whether a particular strategy was associated with higher rates of breast conservation surgery (BCS), and (3) association of management strategy with extent of axillary surgery.

Statistical Analysis
Cochrane-Armitage tests were used to assess treatment trends across time. Variables associated with primary treatment strategy were assessed in multivariable analysis using logistic regression. The relationship of primary treatment strategy with breast operation (BCS versus mastectomy) was assessed by clinical T category using chisquare tests for univariate analysis and with logistic regression models including a treatment by clinical T category interaction for multivariable analysis. Impact of primary treatment strategy on extent of axillary surgery (1-5 vs. [ 5 nodes removed) was assessed similarly using clinical N status (positive or negative) as a stratification factor. Additional variables included in the multivariable model were year of diagnosis, age (years), tumor grade, tumor multicentricity, Charlson-Deyo comorbidity score, race, ethnicity, and insurance status. We included all standard clinicopathologic variables with significant association with treatment strategy on univariate analysis in the multivariable model. Finally, we conducted exploratory analyses on the small subset of patients with triple-negative (ER negative, PR negative, HER2 negative) or HER2overexpressing ILC. Results are reported as odds ratios (OR) with 95% confidence intervals (CI). Analysis was performed using SAS (Version 9.4, SAS Institute Inc., Cary, NC). p-values \ 0.05 were considered statistically significant.

Patient and Tumor Characteristics
Of the 69,312 cases of HR?/HER2-ILC from 2010 to 2016, the average age was 63.1 years (range 22-90 years). Most patients had clinical stage I disease (60.5%). Additional patient and tumor characteristics are presented in Table 1.

Trends in Primary Treatment Strategy
Across the study period, the percentage of HR?/ HER2-ILC patients treated with primary surgery did not change significantly at 93.2% in 2010, to 92.5% in 2016 (p = 0.20). There was a small decrease in the use of NAC (from 4.7 to 4.2%, p = 0.007), and a small but statistically significant increase in the use of long-course NET (1.6% to 2.7%, p \ 0.001, Fig. 2). The use of short-course NET was very low and did not change significantly during the time period analyzed (0.4% in 2010 to 0.6% in 2016). Primary surgery remained the most common treatment modality in ILC across the time period of the study.
Examined separately by clinical T category (Fig. 2), the most notable trend was in patients with cT2 disease, where the proportion undergoing primary surgery decreased significantly from 91.8 in 2010 to 88.6% in 2016 (p \ 0.001), driven mostly by an increase in the use of long-course NET from 2.1 in 2010 to 5.0% in 2016 (trend p \ 0.001). In these cT2 patients, the use of NAC remained stable at 5-6% over the entire period, and the use of short-course NET was consistently \ 1%. No significant changes over time were seen in the management of cT3/cT4 disease, where 68.0% underwent primary surgery, 24.2% underwent NAC, 6.8% underwent long-course NET, and 0.9% underwent short-course NET overall. In cT1 disease, 98.2% had primary surgery and 1.8% received neoadjuvant therapy overall with no large changes over time; however, very small but statistically significant changes were observed in the use of short-course NET (0.2% in 2010 to 0.5% in 2016, p = 0.005) and long-course NET (0.5% in 2010 to 1.1% in 2016, p = 0.006).
On multivariable analysis, factors associated with the choice for neoadjuvant systemic therapy instead of primary surgery included later year of diagnosis, younger age, Hispanic ethnicity, non-White race, Charlson-Deyo comorbidity score of 0, higher clinical T category (OR 5.1 cT2 versus cT1 and OR 18.6 cT3/cT4 versus cT1, each p\ 0.001), and clinical node status (OR 2.3 for cN? versus cN0, p \ 0.001, Table 2).
On multivariable analysis to evaluate the association between primary treatment strategy and BCS, there was a significant interaction between primary treatment strategy and clinical T category, such that treatment strategy increased the likelihood of BCS only in those with clinical T2-4 tumors but not in clinical T1 tumors (Table 4). In fact, for patients with clinical T1 tumors, those who received NAC or long-course NET were significantly less likely to undergo BCS than those who underwent primary surgery (OR 0.31, 95% CI 0.23-0.41 and OR 0.65, 95% CI 0.51-0.83, respectively). However, for those with clinical T2 tumors, long-course NET was associated with significantly higher odds for undergoing BCS (OR 1.51, 95% CI 1.28-1.79). Among the cases with clinical T3-4 tumors, both NAC and long-course NET were associated with increased odds of BCS (OR 1.47, 95% CI 1.21-1.79 and OR 2.54, 95% CI 1.98-3.24, respectively). Short-course NET was not associated with increased odds of BCS in any T category. Additional factors associated with BCS were older age, lower tumor grade, lower clinical N stage, lower Charlson-Deyo comorbidity score, and unifocal disease.
Considering the axilla, 87.4% of patients were clinically node negative whereas 12.6% were clinically node positive. Among cN0 patients with axillary surgery, 75.8% had 1-5 nodes examined, and 24.2% had [ 5 nodes examined, with 1-5 nodes removed among 76.6% for primary surgery, 70.3% for long-course NET, 70.0% for short-course NET, and 55.4% for NAC (p \ 0.001). Among cN? patients, 19.3% had 1-5 nodes examined, and 80.7% had[ 5 nodes examined overall; 23.8% of cN? patients treated with long-course NET had 1-5 nodes removed, compared with 19.2% for primary surgery, 18.6% for NAC, and 17.0% for short-course NET. On multivariable analysis adjusting for other clinical factors, undergoing long-course NET or NAC was associated with significantly higher odds of having less extensive axillary surgery in cN? patients (OR 1.6 and 1.4, respectively) but not cN0 patients, resulting in a significant interaction between primary treatment strategy and clinical node status (Table 5).

HER2? and TNBC Subsets
There were 4651 patients (3668 HER? and 983 TNBC) with biologic subtype other than HR?/HER2-, representing 6.0% of all stage I-III ILC cases in this database. Of these, 80.1% were treated with primary surgery and 19.1% with NAC; there were also a small number of

DISCUSSION
In this analysis of nearly 70,000 HR?/HER2-ILC cases from the NCDB, we found small but significant changes in primary treatment strategy over time, with a decrease in the use of NAC and an increase in long-course NET. For larger tumors, neoadjuvant approaches were associated with increased rates of BCS. In particular, among the clinical T2 cases, long-course NET use was associated with increased odds of BCS. Among the clinical T3/4 tumors, both long-course NET and NAC use resulted in higher odds of BCS. Interestingly, clinical T1 tumors were more likely to be treated with BCS if the primary treatment strategy was surgery. For those clinical T1 tumors receiving NAC or NET, BCS was significantly less likely. This suggests that these neoadjuvant strategies were employed for reasons other than downstaging tumors to achieve BCS, which is consistent with small tumor sizes at baseline in this group. Potential reasons for neoadjuvant therapy in this group include attempting to evaluate response to therapy, or a desire to delay surgical intervention.
While other data show that use of NAC has increased for all breast cancer subtypes, national trends show the largest increase in the triple-negative and HER2? subtypes, likely reflecting selection of patients who are most likely to have robust response. 20,21 We found this to be true in ILC as well, despite the very small proportion with triple-negative and HER2? disease. These results are consistent with publications showing that, although ILC tumors overall have low response rates to NAC, the subset of ILC with high-risk biology appears to garner similar benefit as triple-    Our analysis also found that long-course NET was associated with significantly higher rates of BCS in ILC tumors that were clinical T2 or greater, which is consistent with prior literature. 24 Although NET is associated with increased rates of BCS, the lack of survival improvement and some data showing that primary surgery with adjuvant therapy may be associated with better survival has led some investigators to question the value of neoadjuvant approaches in ILC altogether. 2,9,10 Combined with current data highlighting the potential benefit of short-course NET for tailoring subsequent treatment, we wondered whether this approach had become more common in ILC. We found no such increase but hypothesize that, until further data emerge supporting the use of change in Ki67 to determine management, this approach may remain confined to surgical window trials.
We found an association between neoadjuvant therapy and extent of nodal surgery in clinically node-positive patients. Those who received long-course NET or NAC had significantly higher odds of undergoing less extensive axillary surgery (1-5 nodes removed versus [ 5). These findings differ from a prior analysis of node-positive ILC cases in which NAC was not found to be associated with number of nodes removed. 9 This difference could be explained by our choice to evaluate number of nodes removed as a categorical variable representing typical number of nodes removed in sentinel lymph node surgery versus axillary dissection. This suggests that neoadjuvant approaches can downstage the axilla in ILC and potentially allow for omission of axillary dissection if the type and duration of neoadjuvant therapy is tailored to tumor type. However, the absence of axillary dissection does not imply nodal pathological complete response, and data confirm a decrease in performance of axillary dissection even among patients without complete nodal response to neoadjuvant therapy. 25 Since providers who utilize neoadjuvant therapy at higher rates may also be less likely to recommend axillary dissection for residual nodal disease, this is an important potential confounder to consider in retrospective analyses of neoadjuvant therapy and surgical outcomes.
Overall, these findings are consistent with the increased appreciation for tumor biology in ILC and potential for increased responsiveness to endocrine therapy instead of chemotherapy. Recent analyses have identified molecular subtypes that are specific to ILC, and separate ILC into subtypes that may have differing responses to therapy. [26][27][28] This subtyping at the molecular level may allow for further tailoring of therapy in ILC.
While this study includes a large number of pure ILC cases, the retrospective nature of this analysis limits the ability to draw conclusions about why particular treatment strategies were chosen; For example, short-course NET may have been used to evaluate change in Ki67, but these data are not available in the NCDB. Furthermore, accurate clinical staging in ILC is difficult due to lower sensitivity of imaging tests, which could impact analyses by stage.
These findings suggest that appropriate selection of ILC patients for neoadjuvant approaches can improve surgical outcomes. Additionally, these findings support the use of longer courses of NET when utilized in ILC. The low rates at which neoadjuvant approaches are currently used in this strongly HR?/HER2-tumor type suggest room for potential improvement in care. Further work in identifying predictors of response to therapy, and development of imaging tools to accurately monitor response are needed for patients with ILC.   p-Value for test of interation between primary treatment type and clinical node status. A significant interaction means that the effect of primary treatment on undergoing less extensive axillary surgery differed significantly by clinical N status; thus, treatment odds ratios specific to each clinical node category were estimated