Breast-Conserving Therapy for Triple-Negative Breast Cancer

Importance  The aggressive triple-negative phenotype of breast cancer (negative for estrogen and progesterone receptors and v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 2 [ERBB2] [formerly human epidermal growth factor receptor 2 (HER2)]) is considered by some investigators to be a relative contraindication to breast-conserving therapy.

Objectives  To compare outcomes of breast-conserving therapy for patients with triple-negative breast cancer (TNBC) with those of patients with the luminal A, luminal B, and ERBB2 subtypes.

Design, Setting, and Participants  Prospective database review at an academic tertiary medical center with a designated breast cancer center. We included 1851 consecutive patients ages 29 to 85 years with stages I to III invasive breast cancer who underwent breast-conserving therapy at a single institution from January 1, 2000, through May 30, 2012. Of these patients, 234 (12.6%) had TNBC; 1341 (72.4%), luminal A subtype; 212 (11.5%), luminal B subtype; and 64 (3.5%), ERBB2-enriched subtype.

Exposure  Breast-conserving therapy.

Main Outcomes and Measures  The primary outcome measure was local recurrence (LR). Secondary outcome measures included regional recurrence, distant recurrence, and overall survival.

Results  Triple-negative breast cancer was associated with younger age at diagnosis (56 vs 60 years; P = .001), larger tumors (2.1 vs 1.8 cm; P < .001), more stage II vs I cancer (42.1% vs 33.6%; P = .005), and more G3 tumors (86.4% vs 28.4%; P < .001) compared with the non-TNBC subtypes. Multivariable analysis showed that TNBC did not have a significantly increased risk of LR compared with the luminal A (hazard ratio, 1.4 [95% CI, 0.6-3.3]; P = .43), luminal B (1.6 [0.5-5.2]; P = .43), and ERBB2 (1.1 [0.2-5.2]; P = .87) subtypes. Only tumor size was a significant predictor of LR (hazard ratio, 4.7 [95% CI, 1.6-14.3]; P = .006). Predictors of worse overall survival included tumor size, grade, and stage and TNBC subtype.

Conclusions and Relevance  Breast-conserving therapy for TNBC is not associated with increased LR compared with non-TNBC subtypes. However, the TNBC phenotype correlates with worse overall survival. Breast-conserving therapy is appropriate for patients with TNBC.

Breast cancers are characterized by a wide spectrum of clinical, pathologic, and molecular features. Various well-established tumor markers have been used to determine prognosis and response to therapy, including molecular biomarkers allowing for the identification of distinct subtypes of breast cancer. The most common and useful classification is based on the expression of estrogen receptors (ERs), progesterone receptors (PRs), and v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 2 (ERBB2) (formerly human epidermal growth factor receptor 2 [HER2]).^1-4

Triple-negative breast cancers (TNBCs) are those that lack or only minimally express ERs, PRs, and ERBB2. Triple-negative breast cancers account for approximately 10% to 20% of newly diagnosed breast cancers and tend to exhibit a more aggressive clinical behavior, a metastatic pattern, and poor prognosis compared with other subtypes.^5,6 Although targeted therapies have been developed for tumors that express ERs, PRs, or ERBB2, treatment for tumors that lack these markers remains challenging.

The lack of targeted therapy and the aggressive nature of TNBCs have resulted in controversy as to whether breast-conserving therapy (BCT) is appropriate treatment for these tumors. Several large randomized clinical trials demonstrate equivalent overall survival (OS) for patients undergoing mastectomy compared with those undergoing BCT; however, none included an analysis of molecular subtypes or TNBC.^7,8 For patients with TNBC who undergo BCT, many investigators suggest that local recurrence (LR), regional recurrence (RR), and distant metastasis are increased and that OS is decreased compared with the outcomes of patients with non-TNBC subtypes.^9-12

Given the aggressive nature of TNBC, one can reasonably question whether BCT is appropriate. The purpose of this study was to compare outcomes between patients with TNBC and non-TNBC who underwent BCT to determine whether LR is increased in TNBC patients.

This study was performed with the approval of the institutional review board of Cedars-Sinai Medical Center. Informed consent was waived for this retrospective review of a prospective database. We reviewed a prospectively maintained database of patients undergoing evaluation and treatment for breast cancer from January 1, 2000, through May 30, 2012. Women ages 18 through 85 years with newly diagnosed stages I through III breast cancer were identified. Approximately 20 surgeons performed the procedures that qualified for inclusion in this study. The surgeons did not follow a standardized surgical management protocol. Men, patients undergoing neoadjuvant chemotherapy or mastectomy, and patients with fewer than 90 days of follow-up were excluded from the study. In addition, patients with positive margins were excluded. Most surgeons considered margins to be negative if the tumor-free margin was larger than 1 mm; some surgeons were content if no tumor cells were found at the ink of the surgical specimen on pathologic review. Of the remaining patients, we included only those for whom ER, PR, and ERBB2 status were available. We identified a total study sample of 1851 patients.

Patients were then categorized according to tumor phenotypic subtype using the presence or absence of tumor markers. Specifically, tumors were classified into luminal A (positive for ER or PR and negative for ERBB2), luminal B (positive for ER or PR and ERBB2), ERBB2 (negative for ER and PR and positive for ERBB2), and TNBC (negative for all 3 markers) subtypes based on immunohistochemical identification of these markers on biopsied or excised specimens. A positive ERBB2 marker was defined as immunohistochemical identification (ERBB2 receptor protein on the surface of cells in the breast cancer tissue sample) of 3+ and/or amplified (ratio, >2.0) expression of ERBB2 on fluorescence in situ hybridization. Patient and tumor characteristics examined included age at diagnosis, tumor size, histologic subtype, grade, stage, lymph node status, adjuvant systemic treatment, and adjuvant radiotherapy. As is standard practice, conventional whole-breast radiotherapy with opposing tangents was used in nearly all patients. Most patients with invasive cancer at our institution receive a boost dose of radiotherapy to the primary site. The primary outcome measure was LR. Secondary outcome measures included RR, distant recurrence (DR), and OS. As is standard practice, most of the patients were followed up with biannual mammography for 2 years and then annual mammography and physical examination thereafter. Time to recurrence was calculated from the date of diagnosis to the date of recurrence. Survival was calculated from the date of diagnosis to the date of death or the most recent follow-up for surviving patients.

Differences on normally distributed variables across the 4 cancer subtypes were assessed by analysis of variance and Tukey post hoc tests. Differences on nonnormally distributed numerical variables across the 4 subtypes were assessed by the Kruskal-Wallis test and Wilcoxon rank sum post hoc tests. Differences between TNBC and non-TNBC on normally distributed variables were assessed by the independent-samples t test. Differences between TNBC and non-TNBC on nonnormally distributed numerical variables were assessed by the Wilcoxon rank sum test. Differences on dichotomous variables were assessed by the χ² test or the Fisher exact test, as appropriate. Overall survival and freedom from recurrence were estimated by the Kaplan-Meier method and compared across groups by the log-rank test. Hazard ratios (HRs) and their 95% confidence intervals were estimated using Cox proportional hazards regression models. Multivariable Cox models were used to estimate the association of TNBC with the risk of death and recurrence, adjusting for potentially confounding variables. Time to recurrence for the Cox models was truncated at 5 years. Candidate predictors in the Cox models for recurrence were age (<50, 50 to <80, and ≥80 years), tumor size (<2, 2 to <5, and ≥5 cm), stage, and grade. Subtype (comparing TNBC with each other subtype) was the main predictor of interest. A preliminary Cox survival model revealed a very large HR for patients 80 years or older vs those younger than 50 years (HR, 7.0 [95% CI, 3.6-13.4]). Thus, separate multivariable Cox survival models were estimated in patients younger than 80 years and 80 years or older. Candidate predictors in the Cox survival model for patients younger than 80 years were age (<50 and ≥50 years), tumor size, stage, and grade; candidate predictors in the Cox survival model for the group 80 years or older were age (numerical), tumor size, stage, and grade. Subtype (comparing TNBC with each other subtype) was the main predictor of interest in both age models. We used a 2-sided .05 significance level throughout. Analyses were performed with commercially available statistical software (SAS, version 9.2; SAS Institute, Inc).

We identified 1851 patients who received BCT and had complete data for ER, PR, and ERBB2 status. Tumors in 234 patients (12.6%) were TNBC; 1341 (72.4%), luminal A subtype; 212 (11.5%), luminal B subtype; and 64 (3.5%), ERBB2-enriched subtype. All patients were followed up for at least 90 days after diagnosis, with a median follow-up of 60 (interquartile range, 33-98) months. Patient demographic and tumor data are reported in Table 1. Most of the patients were 50 years or older (74.7%), with the mean age slightly higher in the luminal A group (60.2 years) compared with the luminal B (56.9 years), TNBC (56.1 years), and ERBB2 (56.2 years) groups (P < .001). All patients had stage I (1122 [60.6%]), stage II (641 [34.6%]), or stage III (91 [4.9%]) disease at presentation. Among the 4 subtypes, we found a substantial difference in tumor size (P < .001), histopathologic type (P < .001), pathologic grade (P < .001), and stage (P < .001).

When compared with non-TNBC patients, TNBC patients were younger at diagnosis (mean [SD] age, 56.1 [14.0] vs 59.6 [12.6] years; P = .001) and had a larger tumor size (2.1 [1.3] vs 1.8 [1.3] cm; P < .001). In addition, TNBCs were more frequently high grade (P < .001), were stage II or III vs stage I (P < .001), and had predominantly infiltrating ductal histologic type tumors (P < .001). Overall, 26.2% of patients had node-positive disease. Of the subtypes, ERBB2 had the strongest association with nodal positivity (35.0%) compared with TNBC (30.5%) or the luminal A (24.3%) or luminal B (30.4%) subtypes (P = .04). When comparing TNBC with non-TNBC patients, lymph node positivity was not significantly different (30.5% vs 25.5%; P = .10).

Patients were treated with adjuvant systemic therapy and/or radiotherapy. Overall, 49.3% of patients received chemotherapy, whereas 91.2% received radiotherapy. Patients with TNBC (85.5%) and the ERBB2 subtype (85.0%) were more likely to receive chemotherapy compared with patients with the luminal B (65.3%) and especially the luminal A (38.2%) subtypes (P < .001). A total of 143 (8.8%) patients did not receive radiotherapy. Of these, 12 patients had TNBC and 106 had luminal A, 20 had luminal B, and 5 had ERBB2 tumor subtypes. Within the TNBC group, 8 patients had comorbid conditions that precluded them from receiving radiotherapy, and 4 patients refused radiotherapy although it was recommended. Within the luminal A and B groups, patients had smaller tumors and were generally older with comorbid conditions and therefore did not receive radiotherapy. Within the ERBB2 group, 2 patients refused the treatment; 3 had significant comorbidities, and although radiotherapy was recommended, they declined. Most of the patients in all groups received adjuvant whole-breast radiotherapy or accelerated partial-breast radiotherapy (94.1% vs 3.8%).

Forty-six LRs were observed among the entire study population, including 11 (4.7%) in the TNBC group, 8 (12.5%) in the ERBB2 group, 23 (1.7%) in the luminal A group, and 4 (1.9%) in the luminal B group. Kaplan-Meier–estimated freedom from LR at 5 years was 93% for the TNBC, 96% for ERBB2, 95% for luminal A, and 96% for luminal B (P = .13) groups. Figure 1 shows Kaplan-Meier curves for LR for patients who underwent BCT stratified by subtype. On multivariable analysis for LR, TNBC did not have a significantly increased risk compared with the luminal A (HR, 1.4 [95% CI, 0.6-3.3]; P = .43), the luminal B (1.6 [0.5-5.2]; P = .43), or the ERBB2 (1.1 [0.3-5.2]; P = .87) subtype. Based on multivariable analysis, larger size T3 vs T1 classification (HR, 4.7 [95% CI, 1.6-14.3]; P = .006) was the only significant independent predictor of LR (Table 2).

We observed a total of 21 RRs, including 3 (1.3%) in the TNBC, 4 (6.3%) in the ERBB2, 9 (0.7%) in the luminal A, and 5 (2.4%) in the luminal B groups. Kaplan-Meier–estimated freedom from RR at 5 years was 98% for the TNBC, 84% for the ERBB2, 98% for the luminal A, and 96% for the luminal B groups (P < .001) (Figure 2). Multivariable analysis revealed that stage II vs I (HR, 5.2 [95% CI, 1.7-15.8]; P = .004) and stage III vs I (8.3 [1.8-37.4]; P = .006) were significant predictors of RR (Table 3). Also, TNBC was protective for risk of RR, compared with the ERBB2 subtype (HR, 0.2 [95% CI, 0.0-0.7]; P = .02).

Distant recurrences were the most frequent recurrences seen among all patients, with a total of 66 patients presenting with metastatic disease after BCT. Of these, 21 (9.0%) were in the TNBC group, 5 (7.8%) in the ERBB2 group, 31 (2.3%) in the luminal A group, and 9 (4.2%) in the luminal B group. Kaplan-Meier–estimated freedom from DR at 5 years was 85% for the TNBC, 88% for the ERBB2, 95% for luminal A, and 92% for luminal B (P < .001) groups (Figure 3). A more advanced stage (III vs I; HR, 5.4 [95% CI, 2.2-13.4]; P < .001) and larger tumor (T3 vs T1, 5.5 [2.0-15.6]; P = .001) were associated with an increased risk of DR (Table 4). Among subtypes, TNBC patients were at increased risk for DR compared with patients with the luminal A subtype (HR, 2.5 [95% CI, 1.4-4.4]; P = .001) but not those with the luminal B or ERBB2 subtypes.

A total of 113 deaths occurred. Patients older than 80 years were excluded from the multivariable analysis of OS because of their higher risk of death (HR, 7.0 [95% CI, 3.5-13.2]). Predictors of worse OS included being 50 years or older vs younger than 50 years (HR, 2.2 [95% CI, 1.3-3.9]; P = .005), stage III vs I disease (2.8 [1.3-6.0]; P = .007), T2 vs T1 classification (1.9 [1.1-3.4]; P = .03), and T3 vs T1 classification (4.2 [1.8-9.9]; P = .001). Furthermore, TNBC was associated with worse OS compared with the luminal A (HR, 3.5 [95% CI, 02.2-5.7]; P < .001) and luminal B (3.7 [1.7-8.2]; P = .001) subtypes (Table 5). Kaplan-Meier analysis of OS is shown in Figure 4.

Breast cancer subtypes are known to have phenotypic diversity with regard to tumor aggressiveness and response to therapy.³ Triple-negative breast cancer specifically is challenging to treat because it lacks a targeted treatment approach, is associated with a poor prognosis, and, in some studies, has an increased risk of LR.^13-15 This study validates that TNBC patients are younger, present with tumors of a higher grade and larger size, and present at a more advanced stage compared with non-TNBC patients. The study also validates that OS is worse among this group of patients. These findings are consistent with the current literature, which indicates that the triple-negative subtype is generally more aggressive, with worse prognosis overall.^5,16,17

Although breast cancer patients undergoing BCT have long-term outcomes equivalent to those of patients treated with mastectomy, significant debate remains regarding BCT as an appropriate treatment of TNBC.^8,18 Several investigators^10,19,20 have reported results evaluating locoregional recurrence (LRR) related to the breast cancer subtype. Some studies^11,12,21 have included only patients who have undergone BCT, whereas others include patients who have undergone mastectomy. Rates for LR after BCT in these studies have ranged from 4% to 20% for TNBC patients compared with 3% to 14% for non-TNBC patients, suggesting higher LRR in TNBC patients.¹⁹

Data regarding rates of LRR for TNBC patients are incongruent. Arvold et al²² reviewed 1434 patients who underwent BCT and determined that 171 TNBC patients had a significantly higher risk of LR compared with patients with other subtypes (adjusted HR, 3.9; P = .001). A similar but smaller study by Nguyen et al⁹ found that, of the 793 patients who underwent BCT, the 80 TNBC patients had a dramatically increased risk for LR (adjusted HR, 7.1; P = .009). Additional studies^10-12,20,21 have demonstrated that the LRR rate after BCT was significantly greater in TNBC compared with the other subtypes.

In contrast, Haffty et al²³ and Freedman et al²⁴ found no significant differences in LR rates for patients treated with BCT for TNBC vs non-TNBC. In the study by Haffty et al,²³ 117 patients had TNBC and 365 had non-TNBC, with no appreciable difference in 5-year breast relapse–free survival between the 2 groups (83% vs 83%; differences were not significant). When comparing 753 patients who had received BCT by subtype, Freedman et al²⁴ found no significant difference in LRR related to the subtypes (2.3% in ER- or PR-positive subtypes; 4.6% in ER- and PR-negative and ERBB2-positive subtypes, and 3.2% in TNBC; P = .36).

The present study validates that BCT is appropriate for patients with TNBC and supports the findings of Haffty et al,²³ Freedman et al,²⁴ and others⁶ who have shown no increase in LRR for patients with TNBC. Our study is particularly relevant to widespread practice. The large numbers of surgeons in the US have different practice patterns, and a variety of adjuvant systemic therapies and radiotherapy are prescribed. Our study is particularly relevant to widespread practice because of the large number of US surgeons, each with different practice patterns, who treat these patients and the variety of standard adjuvant systemic therapies and radiotherapy that are prescribed widely in the United States. On multivariable analysis, after adjusting for tumor size and grade, no significant difference was found in risk of in-breast LR among subtypes. Indeed, few LRs were seen in any group. The use of adjuvant systemic chemotherapy may contribute to the low rate of LRR seen in this study because chemotherapy was routinely used for TNBC patients with no significant comorbidities. Furthermore, the only predictive variable for statistically significant risk of LR was a T3 tumor. This finding suggests that TNBC is not associated with increased risk for LR after BCT. Similarly, an increased risk for RR was not seen in TNBC patients but in patients with the ERBB2 subtype and stages II and III cancers. However, the multivariable analysis is of limited value with so few RR.

Triple-negative breast cancer is associated with worse OS and an increased risk for DR. However, this study validates the lack of a significant risk for LR in TNBC patients who undergo BCT. Despite the aggressive nature of TNBC, the triple-negative phenotype does not lead to a significantly increased risk of LRR compared with the non-TNBC phenotype in patients undergoing BCT. Breast-conserving therapy is appropriate for TNBC patients and should be routinely offered.

Accepted for Publication: May 17, 2013.

Corresponding Author: Armando E. Giuliano, MD, Department of Surgery, Cedars-Sinai Medical Center, 310 N San Vicente Blvd, Ste 311, Los Angeles, CA 90048 (armando.giuliano@cshs.org).

Published Online: January 1, 2014. doi:10.1001/jamasurg.2013.3037.

Author Contributions: Drs Gangi and Giuliano had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Gangi, Giuliano.

Acquisition of data: Gangi, Chung, Liou, Leong, Giuliano.

Analysis and interpretation of data: Gangi, Chung, Mirocha, Giuliano.

Drafting of the manuscript: Gangi, Chung, Mirocha.

Critical revision of the manuscript for important intellectual content: Gangi, Chung, Mirocha, Liou, Giuliano.

Statistical analysis: Gangi, Mirocha.

Obtained funding: Giuliano.

Administrative, technical, and material support: Giuliano.

Study supervision: Chung, Giuliano.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the Fashion Footwear Charitable Foundation of New York, Inc, Associates for Breast and Prostate Cancer Studies, the Avon Foundation, the Margie and Robert E. Petersen Foundation, and Linda and Jim Lippman.

Role of the Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Previous Presentation: This study was presented as a poster at the 84th Annual Meeting of the Pacific Coast Surgical Association; February 17, 2013; Kauai, Hawaii, and is published after peer review and revision.