Triple negative breast cancer is associated with an increased risk of residual invasive carcinoma after lumpectomy†
This research was presented during the general poster session at the American Society for Radiation Oncology meeting (ASTRO 2010); October 31 to November 4, 2010; San Diego, CA.
To assess the potential mechanisms that may underlie increased local failure in triple negative (TN) breast cancers, an analysis was performed of the risk of residual carcinoma after lumpectomy with correlation to pathologic factors, including molecular phenotype.
A review of pathologic specimens was performed for women with invasive breast cancer treated with lumpectomy followed by reexcision. Data were collected on age; tumor size, grade, and nodal stage; estrogen receptor, progesterone receptor, and human endothelial growth factor receptor 2 (Her2); extensive intraductal component; lymphovascular invasion; margins; and reexcision findings. Univariate and multivariate logistic regression analyses were performed to evaluate for associations between pathologic features of the lumpectomy specimen and reexcision findings. Molecular phenotypes were defined by conventionally used immunohistochemical pattern.
Data were collected on 369 patients with breast cancer. The median age was 57 years, median tumor size was 1.5 cm, 36% had positive margins, 32% had positive lymph nodes, 73.5% had the luminal A subtype, 9.5% had the luminal B subtype, 4.5% were Her2-enriched, and 12.5% were TN. Overall, 32% of patients had invasive cancer in their reexcision specimens, and 51% of those with the TN subtype had residual invasive disease on reexcision compared with 30% to 31% for other subtypes. On univariate analysis, age, tumor size, margin status, lymphovascular invasion, nodal status, and TN subtype were associated with elevated risk of residual invasive cancer. On multivariate analysis using a forward stepwise model, TN subtype maintained significance, with an odds ratio of 3.28 (P = .002).
TN subtype has a statistically significant association with an increased risk of residual tumor. This suggests the putative increase in the risk of local failure in TN patients may be related to increased residual tumor burden. Cancer 2012. � 2012 American Cancer Society.
In 2000, Perou et al published a landmark study identifying breast cancer molecular subtypes by gene expression profiles.1 Over the past decade, the prognostic value of these subtypes has been validated in multiple datasets.2-4 The gene-expression–derived molecular subtypes can be approximated by immunohistochemical-defined molecular phenotypes using estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (Her2). The marker combinations that best matched the molecular profiles are 1) ER(+) and/or PR(+) and Her2(−) for luminal A subtype; 2) ER(+) and/or PR(+) and Her2(+) for luminal B subtype; 3) ER(−) and PR(−) and Her2(−) for the basal-like subtype; and 4) ER(−) and PR(−) and Her2(+) for the Her2-enriched subtype.5, 6
Molecular phenotype has become increasingly valuable in guiding systemic treatment decisions. Yet, the impact of molecular phenotype on local therapy has lagged, despite mounting evidence of an association between molecular phenotype and local recurrence (LR) or locoregional recurrence (LRR).7-11 Most reports find a low risk of LRR in the luminal phenotypes and increased risk of LRR in the so-called triple negative (TN; ER-, PR-, and Her2-negative) and Her2-enriched breast cancer phenotypes after breast conservation therapy, mastectomy with postmastectomy radiation therapy, and accelerated partial breast irradiation.7-14 Although trastuzumab likely mitigates excess LR risk in the Her2-enriched phenotype, the lack of targeted systemic agents for triple negative breast cancer (TNBC), makes TN status in patients a clinical challenge for surgical and radiation oncologists.
TNBCs disproportionately affect younger women, and thus the increased risk of local recurrence has been speculated to be due to smaller, more cosmetically driven surgeries, poorer imaging quality in denser breasts, and/or differences in wound-healing growth factors that are more conducive to LR. Other explanations include radioresistance12 and rapid proliferative rate of TNBC, resulting in a transiently elevated LR rate during the early posttreatment period that is overcome with longer follow-up.15
Here, we test the hypothesis that the higher risk of LR in TNBC is due to increased microscopic invasive tumor burden after lumpectomy. The objective of our study is to assess the risk of residual carcinoma after lumpectomy with correlation to multiple pathologic factors, including molecular phenotype. Prior reexcision studies that analyzed clinical and pathologic predictors for residual disease have reported correlation with age, margin status, and extensive intraductal component (EIC).16-18 However, the influence of molecular phenotype has not been previously analyzed in this context and may illuminate the mechanism of increased LR in TNBC after lumpectomy. From the analysis of reexcision data, our ultimate objective is to establish whether molecular phenotype should be factored into local management decisions, specifically whether surgical reexcisions and radiation dose should be tailored to the molecular phenotype of the tumor.
MATERIALS AND METHODS
A retrospective review of pathologic records was performed for women with invasive breast cancer who were treated with lumpectomy followed by a second ipsilateral breast surgery between 2001 and 2009 at Rhode Island Hospital or Women and Infants Hospital, Providence, Rhode Island. The second surgery had to occur within 3 months of the initial surgery and included reexcision or mastectomy. The study period was begun in 2001, because that is when fluorescence in situ hybridization confirmation of Her2 intermediate score (2+) became reflexive. Data were collected on age, tumor size, grade, mitotic count, status of margins, lymph nodes, ER, PR, and Her2, as well as presence of EIC, lymphovascular invasion (LVI), multifocality, and reexcision findings.
Although data was collected as continuous variables, most of the variables were grouped into 2 or 3 categories to facilitate analysis. An effort was made to keep most variables binary. Age was dichotomized as ≤45 years old or >45 years old. Tumor size was divided into 3 groups: ≤1.0 cm (defined as the reference group), 1.1 to 2.0 cm, and >2.0 cm. Grade 1 was defined as the reference group, and grade 2 and grade 3 were analyzed separately in comparison to the reference group. High mitotic count was defined as >10 mitoses per field size of 0.152 mm2. However, mitotic count was ultimately excluded from the analysis because of the large number of missing data points (n = 78). Positive LVI was defined as extensive or when characterized as “presence of LVI not otherwise specified.” Focal, equivocal, and indeterminate LVI was classified as negative for LVI. EIC was defined as >25% intraductal component. Multifocal disease was defined as multiple pathologic foci of invasive carcinoma in the lumpectomy specimen. Positive surgical margin was defined as one or more surgical margins with invasive disease present at the margin. Close or <1 mm margins were classified as negative. Lymph node involvement was defined as any pathologic evidence of carcinoma >0.2 mm in 1 or more lymph nodes. Molecular phenotypes were defined by the following immunohistochemical patterns: Luminal A = ER(+) and/or PR(+) and Her2(−). Luminal B = ER(+) and/or PR(+) and Her2(+). Her2-enriched = ER(−), PR(−), and Her2(+). TN = ER(−), PR(−), and Her2(−). TN subtypes were analyzed in comparison to non-TN subtypes (luminal A + luminal B + Her2-enriched). Positive ER and PR status was defined as ≥5% of tumor cell nuclei showing specific staining. Specific membrane staining for c-erbB2 (HER2/neu) was reported as positive, with the intensity of the staining observed on a scale of 0 to 3. Intensity of 0 to 1+ was considered negative, whereas 3+ staining was considered positive. An intermediate score of 2+ was confirmed by fluorescence in situ hybridization amplification. Residual disease was defined as presence or absence of residual invasive disease.
Patients with residual ductal carcinoma in situ (DCIS) were excluded from the analysis, and patients treated with neoadjuvant chemotherapy were also excluded.
Statistical comparisons between groups were made using the Kruskal-Wallis test for continuous variables and the chi-square test for categorical variables. Univariate relationships between age, tumor size, grade, LVI, multifocality, margin status, nodal status, and molecular phenotype with residual invasive carcinoma were analyzed using a univariate linear regression model. A multivariate logistic regression model was developed for residual invasive carcinoma using a forward stepwise process with a p value for entry into the model of 0.05. Only patients with complete data were included in the multivariate analysis. All analyses were performed with the SAS system, version 9.2 for Windows. A probability level of alpha = 0.05 was used to determine statistical significance.
This study was performed under institutional review board approval from Rhode Island Hospital, and Women and Infants Hospital.
Pathologic records from 449 patients were reviewed. Eighty patients had residual intraductal carcinoma and were excluded, because the analysis was restricted to data from 369 patients with or without residual invasive carcinoma. Descriptive statistics for the entire cohort are shown in Table 1. The median age was 57 years, median tumor size was 1.5 cm, 29% were high grade, 18% had a high mitotic count, 21% had evidence of LVI, 36% had EIC, 36% had positive margins with respect to invasive carcinoma, 38% had close margins (≤2 mm) with respect to invasive carcinoma, 70% had close or positive margins with respect to intraductal carcinoma, 30% were multifocal, 32% had positive lymph nodes, 73.5% were luminal A, 9.5% were luminal B, 4.5% were Her2 enriched, and 12.5% were TN. Overall, 32% had evidence of invasive cancer in their reexcision specimens. Descriptive statistics by molecular phenotype are shown in Table 2. TN phenotype was more likely to have residual invasive carcinoma on reexcision: 51% compared to 30% among the luminal phenotype. TN represented a younger population of women with a median age of 50. TN phenotype was more likely to be high grade and have a high mitotic count compared to the luminal phenotypes. Margin status, size, and multifocality were similar among the TN and luminal subtypes.
Table 1. Descriptive Statistics
|Median age, y||367||57|
|Median tumor size, cm||368||1.5|
|High mitotic count||291||18%|
|Extensive intraductal component||357||36%|
|Positive surgical margins||365||36%|
|Positive lymph nodes||347||32%|
|Molecular phenotype||345|| |
| Luminal A and B||286||83%|
| Triple negative||43||12.5%|
|Presence of residual invasive carcinoma||367||32%|
Table 2. Descriptive Statistics by Subtype
|Median age, y||59||57||50||.0008|
|Median tumor size, cm||1.5||1.2||1.6||.39|
|High mitotic count||10%||50%||56%||<.0001|
|Extensive intraductal component||35%||63%||43%||.07|
|Positive surgical margins||38%||6%||30%||.02|
|Positive lymph nodes||32%||27%||40%||.55|
|Residual invasive carcinoma||30%||31%||51%||.02|
Univariate regression analysis results are shown in Table 3. On initial univariate analysis, after excluding patients with residual DCIS, EIC did not maintain a statistically significant correlation with residual invasive disease (P = .79) and thus was not included in the final analysis. Otherwise, univariate analysis showed statistically significant associations between all the variables assessed and presence of residual invasive carcinoma. Thus, all of the variables were included in the multivariate analysis. Multivariate regression analysis results are shown in Table 4. The analysis was limited to subjects from whom complete data was available (n = 296). Of all the variables analyzed, only nodal status (odds ratio [OR] = 3.06, P < .0001), TN status (OR = 3.28, P = .02), and tumor size (OR = 3.49, P = .001) maintained statistical significance on multivariate analysis. After nodal status, TN status, and tumor size were entered into the model, no additional variables met the 0.05 significance level for entry into the model.
Table 3. Univariate Analysis
|Age ≥45 y||365||0.45 (0.27, 0.76)||.003|
|Tumor size, cm||366|| ||<.0001|
| ≤1.0|| ||Reference|| |
| 1.1-2|| ||3.00 (1.62, 5.54)||.0005|
| >2|| ||5.70 (2.98, 10.93)||<.0001|
|Grade||346|| || |
| 1|| ||Reference|| |
| 2|| ||1.32 (0.68, 2.55)||.41|
| 3|| ||2.63 (1.31, 5.27)||.007|
|Lymphovascular invasion||367||2.56 (1.53, 4.28)||.004|
|Multifocal||367||1.78 (1.12, 2.85)||.02|
|Positive surgical margin||363||1.76 (1.12, 2.76)||.01|
|Positive lymph node||345||3.69 (2.29, 5.97)||<.0001|
|TN (vs non-TN)||343||2.48 (1.30, 4.74)||.006|
Table 4. Multivariate Analysis
|TN (vs non-TN)||3.28 (1.56-6.89)||.002|
|Positive lymph node||3.06 (1.77-5.30)||<.0001|
|Tumor size 1.1-2.0 cm vs <1.0 cm||1.89 (0.94-3.82)||.076|
|Tumor size >2.0 cm vs <1.0 cm||3.49 (1.65-7.38)||.001|
The TN phenotype is a putative surrogate for the basal-like subtype that is characterized by a gene expression profile similar to that of the basal-myoepithelial layer of normal breast.1 Although they are not synonymous terms, there is substantial overlap between TNBC and basal-like breast cancer; approximately 80% of TNBCs are also basal-like breast cancers, and 80% of basal-like breast cancers express the TN phenotype.19, 20 TN phenotype represents a heterogeneous collection of tumors that account for approximately 15% of breast cancers in the United States. TNBC disproportionately affects young women, African Americans, and BRCA1 carriers.21-24 TNBC also tends to have aggressive features such as high nuclear grade and high mitotic index. Paradoxically, TNBC is not consistently associated with higher regional lymph node involvement. Some recent analyses have suggested that TNBC has a decreased prevalence of lymph node metastasis and exhibits dissociation between tumor size and positive lymph nodes.11, 25, 26 In addition to higher LRR rates, TNBCs have a higher rate of distant metastases and are more likely to metastasize to visceral organs.27 Although relatively responsive to cytotoxic chemotherapy, overall, these patients have poorer overall survival.28 The aggressive biologic nature of these tumors in combination with lack of targeted systemic therapy makes TNBC a therapeutic challenge for oncologists.
In recent decades, the risk of residual disease after local breast tumor excision has been examined in multiple studies. For example, in 1997, Wazer et al reported a positive reexcision rate (including invasive and intraductal carcinoma) of 49%.16 In the present analysis, the positive reexcision rate (including invasive and intraductal carcinoma) is 44%. Thus, despite advances in imaging, localization, and pathologic margin assessment, little progress has been made toward the preoperative identification of patients who are at high risk for residual disease. Our results suggest that factoring molecular phenotype into the decision-making regarding the need for reexcision may be useful.
Although our positive reexcision rates are similar to those of previously published reports, our findings differ. This study identified nodal status, TN status, and tumor size as independent predictors of residual invasive carcinoma. This is contrary to prior reexcision studies that showed margin status, EIC, and age to be major predictors of residual disease.16-18 However, our study was designed differently than the prior reexcision studies, because we deliberately excluded patients with residual DCIS from the analysis whereas residual DCIS was included in the prior studies as a positive outcome. We made the decision to exclude residual DCIS, because preliminary analysis of our data with the inclusion of residual DCIS as a positive outcome showed that molecular phenotype was not statistically significantly correlated with residual disease. When we excluded the patients with residual DCIS, a strong correlation between molecular phenotype and residual invasive carcinoma emerged. By excluding residual DCIS, we effectively obviated the predictive role of EIC.
Although age >45 years (OR = 0.45, P = .003) was significantly associated with residual invasive disease on univariate analysis, it did not enter into the forward stepwise multivariate model. The lack of independent predictive value of age is likely related to redundancy between TN status and young age. To further clarify the relationship between age and TN status, we also ran a multivariate analysis with age included in the multivariate model, and TN status still maintained statistical significance. Thus, the influence of TN status is maintained even independent of age, but not vice versa. Similarly, positive margin status (OR = 1.76, P = .01) was significantly correlated with residual invasive carcinoma on univariate analysis but did not enter into the multivariate model. One explanation for why margin status is not an independent predictor of residual disease in our study may be reflective of more meticulous pathologic margin assessment in the relatively recent study period. Also, margin status in our study was defined with respect to invasive carcinoma, whereas the prior studies analyzed margin status with respect to intraductal and invasive carcinoma.
Although the Her2-enriched phenotype is associated with an increased risk of LR,7 in this study it was not found to have a statistically significant correlation with the risk of residual disease (31% compared to 30% in luminal A and B). This may be related to the very small sample size of only 16 cases with the Her2-enriched phenotype. Alternatively, this could reflect selection bias, because the majority of high-risk Her2-enriched patients may have received neoadjuvant chemotherapy or mastectomy and thus would not have been included in our study.
This study is a retrospective pathologic chart review of reexcision data with inherent limitations. As with all retrospective data, selection bias is possible, because some patients with more aggressive tumors may have had up-front mastectomy and/or neoadjuvant chemotherapy and thus been excluded from this study. However, the proportion of each molecular phenotype in this study was consistent with national data on the incidence of each molecular phenotype and does not suggest an overt selection bias with regard to molecular phenoptype. The decision to reexcise was also subjective, because there was no predefined institutional policy. Rather, all cases were presented to the respective institution's prospective multidisciplinary tumor board, and recommendations were made after careful review of the clinical scenario, pathology, margin, and description of surgical procedure performed. Extensively positive margins were generally recommended for reexcision, and recommendations for close or focally positive margins were made on a case-by-case basis and largely reflect individual surgeon preferences. Although there are inherent limitations to this approach, the results are meant to help illuminate factors that should be considered when contemplating the need for reexcision. The increased risk of residual disease burden in TNBC is meant to be hypothesis-generating and provides incremental progress in understanding the mechanism as to why TNBCs have higher rates of local recurrence, and help to direct treatment accordingly.
Our study shows that TN phenotype is independently correlated with increased risk of residual disease after lumpectomy. This finding suggests that TNBCs harbor more microscopic residual disease after lumpectomy, and thus molecular phenotype should factor into decision-making regarding the extent of initial surgery (generous gross margins) and the need for reexcision. TNBCs may also benefit from dose escalation to the tumor bed region. The EORTC (European Organisation for Research and Treatment of Cancer) boost versus no boost trial revealed a statistically significant local control benefit for boost in all age groups, with the largest benefit seen in the younger patients.29 A subgroup analysis of this trial revealed that the largest proportional benefit of boost was in high-grade tumors and in patients under the age of 50 years.30 Because TNBCs occur more frequently in young patients and are overwhelmingly high-grade, one may extrapolate that TNBCs may gain a higher proportional local control benefit from the administration of boost radiation to the tumor bed.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURE
The authors made no disclosure.