Introduction

Breast cancer (BC) is a heterogeneous disease characterized by over 30 histologic types. Of these, the two most common are invasive ductal carcinoma (IDC) which represents 60–70 % of cases, and invasive lobular carcinoma (ILC) which accounts for 5–15 % [1, 2]. Several reports have described features unique to ILC that distinguish it both clinically and pathologically from IDC; these include loss of the cell–cell adhesion molecule E-cadherin [3], a tendency toward diffuse growth and multifocality in the breast, and a predominance of the luminal A biologic subtype in contrast to IDC which is distributed among other subtypes (i.e., luminal A/B/HER2, HER2, and triple-negative) [4]. Furthermore, in contrast to IDC, ILC more often requires re-excision or mastectomy following breast-conserving surgery due to challenges detecting its diffuse growth both mammographically and intra-operatively [5–9].

Despite these inherent differences, ILC is generally under-represented in clinical trials and cohort studies, and is often combined with IDC in outcome analyses. As a result, little is known about optimal approaches for the management of ILC, or about differences in modern outcomes between ILC and IDC.

Notably, recent consensus guidelines suggest that a negative resection margin of >0 mm (i.e. “no ink on tumor”) is adequate in the setting of breast-conserving therapy (BCT) for both ILC and IDC [10, 11]. These guidelines are based largely on study populations in which IDC strongly predominates [12–15], and otherwise include reports of ILC in which the distribution of sub-centimeter margins and biologic subtypes is not addressed [16–19].

Therefore, we sought to study the local recurrence (LR) patterns of ILC following BCT, with a particular focus on biologic subtype and margin status.

Methods

Patient selection

The study population included 998 consecutive patients with stage I or II invasive BC. All patients underwent BCT at the Brigham and Women’s Hospital/Dana Farber Cancer Institute from 1998–2007 and had comprehensive clinical and pathologic data available for the determination of both histologic and biologic subtype. Biologic subtypes were approximated by receptor status and tumor grade: estrogen receptor (ER) positive or progesterone receptor (PR) positive, human epidermal growth factor receptor 2 (HER-2) negative and moderately or well-differentiated = luminal A; ER + or PR +, HER-2- and poorly differentiated = luminal B; ER + or PR + and HER-2 + = luminal-HER2; ER-, PR- and HER2 + = HER2; ER-, PR- and HER-2- = triple-negative BC (TNBC).

Patients with synchronous bilateral BCs or prior malignancy (aside from non-melanoma skin cancers) were excluded, as were those who received neoadjuvant therapy or accelerated partial breast irradiation (APBI). All patient data were prospectively collected and curated by dedicated data managers at the participating institutions. This study was approved by the institutional review board of the Dana Farber/Harvard Cancer Center, and all patients signed informed consent permitting the collection of de-identified patient data at baseline and follow-up.

Treatment characteristics

All patients underwent BCT comprised breast-conserving surgery followed by whole-breast irradiation. Re-excision was performed at the discretion of the treating breast surgeon, generally based on an initial resection margin (i.e., the shortest distance from any cancer cell to the nearest inked specimen surface) of <2 mm given standard practice at the time of this study. Adjuvant systemic therapy including chemotherapy and/or hormonal therapy was administered at the discretion of the treating medical oncologist. Patients who received HER2-directed therapies were excluded from the study because this practice was not uniform throughout the study period. Whole-breast irradiation was administered in all cases using standard tangential fields followed by a boost to the region of the surgical cavity (median cumulative dose to surgical cavity = 61 Gy). The boost dose was largely uniform throughout the cohort and was not informed by histologic type, biologic subtype, or margin status. Regional nodal irradiation (i.e., axillary, supraclavicular, internal mammary) was administered at the discretion of the treating radiation oncologist. Daily fraction sizes of 1.8–2.0 Gy were used, and no patients received hypofractionation during this time frame.

Pathologic analysis

Pathologic specimens were reviewed and histologically typed by dedicated breast pathologists at the Brigham and Women’s Hospital/Dana Farber Cancer Institute. Patients were included in this study if they exhibited IDC, ILC, or mixed histology comprising features of both IDC and ILC. Tumors exhibiting any other malignant histology were excluded (i.e., mucinous, medullary, papillary, micropapillary, and tubular, in addition to lymphomas and sarcomas). Initial margins were defined as ≥2 mm, <2 mm, or “positive” if there was ink on the tumor. An additional category termed “no additional breast tissue to take” comprised patients with final margins <2 mm in whom the treating surgeon determined that there was no further breast tissue to resect either at skin or pectoralis major. Margins were measured from the inked tissue edge to the nearest focus of invasive or in situ disease.

Follow-up and endpoints

Patients underwent follow-up 4–6 weeks after the completion of radiotherapy and 6 months after with annual mammography. Follow-up time was calculated from the date of diagnosis to the date of first event, or to the last known date of disease-free status.

Time to LR as a first event was the primary endpoint of this study. LR was defined as any ipsilateral in-breast recurrence (including invasive or in situ histology) without evidence of metastatic disease in the subsequent 4 months. Patients with evidence of distant metastases within 4 months of LR were considered to have synchronous local and distant recurrence, and were excluded from the primary endpoint.

Statistical methods

Descriptive statistics were used to illustrate patient and disease characteristics as stratified by the histologic tumor types: IDC, ILC, and mixed ILC/IDC. Since the majority of ILCs were characterized by the luminal A biologic subtype (91.1 %), we further examined outcomes by histology in luminal A tumors exclusively. Fisher’s exact test was used to evaluate the associations between histology and surgical margin status, including initial margin status, use of re-excision, and final margin status. The Kruskal–Wallis test was used to compare the re-excision tumor size and margin status among the three histologic categories. Time to LR was defined as the time from diagnosis to the date of LR or last disease assessment. The method of Kaplan–Meier was used to characterize rates of LR. The log-rank test and Cox proportional hazards models were used to assess the effect of patient characteristics and disease factors on this outcome. Variables with p-value < 0.1 on univariate analysis entered the multivariable model. All analyses were performed using SAS version 9.2 (SAS Institute, Carey, NC).

Results

Baseline distribution of prognostic factors stratified by histology

Stratification of the overall study cohort by histologic type showed that the majority of patients exhibited IDC histology (74 %), while 8 % had ILC histology and 19 % had mixed histology (Table 1). Among histologic groups, there were significant baseline differences in the distribution of biologic subtype (p < 0.001), T stage (p = 0.008), tumor grade (p < 0.001), lymphovascular invasion (LVI; p < 0.001), and systemic therapy treatment patterns (p < 0.001). Compared to patients with IDC, those with ILC were more likely to have larger tumors of lower grade, with the majority characterized by luminal A biologic subtype. ILC patients were also less likely to receive adjuvant chemotherapy, although they were more likely to receive hormonal therapy.

Table 1 Patient characteristics and disease factors by histology (all patients)
Full size table

Because major differences between histologic groups could be attributed to the luminal A predominance among ILC patients, a similar analysis was performed exclusively on those with the luminal A subtype (Table 2). Holding the luminal A subtype constant, significant baseline differences between histologic groups included tumor size (p < 0.001), lymph node involvement (p = 0.02), tumor grade (p < 0.001), and LVI (p < 0.001). Among all luminal A patients, those with ILC were more likely to present with larger tumors of lower grade and with a lower likelihood of nodal involvement or LVI. Neither hormonal- or chemotherapy, nor radiotherapy treatment patterns differed significantly by histology among luminal A patients.

Table 2 Patient characteristics and disease factors by histology (Luminal A patients only)
Full size table

Surgical management and margin assessment

The majority of patients had surgical specimens available for margin evaluation (n = 827), including tissue from both the initial resection and re-excision (Table 3). On pathologic assessment, ILC exhibited a higher likelihood of initially positive margins as compared to IDC or mixed histology (45.0 % vs 17.5 % and 18.5 %, respectively; p = 0.0002). ILC patients subsequently underwent re-excision more frequently than those with IDC or mixed histology (57.1 % vs 40.4 % vs 36.9 % respectively; p = 0.02), and final margins did not differ among the groups (p = 0.88). Of those who underwent re-excision, the residual tumor size did not significantly differ by histology (p = 0.74). Of note, residual disease was detected in only 17 % of the re-excision specimens.

Table 3 Distribution of surgical margin and residual disease at re-excision by histology
Full size table

Local recurrence by histology and resection margins

At a median follow-up of 119 months, 45 patients were diagnosed with an isolated LR (Table 4). The 10-year cumulative incidence of LR was 5.5 % for patients with IDC, 4.4 % for ILC, and 1.2 % for those with mixed ILC/IDC histology (Fig. 1a; p = 0.08). Among those with luminal A disease, 10-year cumulative incidence was 2.6 % for IDC, 3.4 % for ILC, and 0 % in the mixed histology group (Fig. 1b; p = 0.12). Four patients were diagnosed with concurrent LR and distant metastasis, excluding them from the primary endpoint; all four had non-luminal A IDC and margins >2 mm.

Table 4 Time to local recurrence by histology and biologic subtype
Full size table
Fig. 1
figure 1

a Kaplan–Meier estimate of local recurrence by histologic type (log-rank p = 0.08). b Kaplan–Meier estimate of local recurrence by histologic type among Luminal A tumors (log-rank p = 0.12)

Full size image

LR did not differ significantly between histologic types when stratified by width of resection margins (Table 5). For patients with margins ≥2 mm, the 10-year cumulative incidence of LR was 5.2 % for IDC, 1.8 % for ILC, and 2.2 % for mixed histology (p = 0.25). Restricting the analysis to luminal A patients, LR rates were 2.8 % for IDC, 1.9 % for ILC, and 0 % for those with mixed histology (p = 0.20). Notably, no LR events were observed among any ILC or mixed histology patients with margins <2 mm (n = 28).

Table 5 Distribution of local recurrence by histology and final margin status
Full size table

Univariate Cox regression modeling showed no significant association between the risk of LR and histologic type (Table 6). Factors associated with reduced LR on univariate analysis included increasing age at diagnosis (p = 0.002), tumor size ≤2 cm (p < 0.001), the absence of nodal involvement (dichotomous comparison of N1-3 vs N0; p < 0.001), and receipt of hormonal therapy (p < 0.001). Biologic subtype was significantly associated with LR as previously reported [20], with HER2 (HR = 7.33) and TNBC (HR = 6.14) exhibiting the highest risk in comparison to luminal A (p < 0.0001).

Table 6 Association between time to local recurrence and patient characteristics/disease factors
Full size table

Multivariate Cox regression with IDC as the reference showed that neither ILC nor mixed histology exhibited an association with the risk of LR (p = 0.52). The highest quartile of age at diagnosis (≥64 years old) compared with the lowest quartile (≤46 years old) was independently associated with a reduced risk of LR (HR = 0.18; 95 % CI 0.06–0.56), as was the use of chemotherapy (HR 0.43; 95 % CI 0.21–0.88). Tumor size > 2 cm (HR 2.05; 95 % CI 1.09–3.86) and lymph node involvement (HR 2.44; 95 % CI, 1.26–4.73) were both associated with increased risk of LR. Among luminal A patients, age at diagnosis was the only factor significantly associated with LR (HR 0.08; 95 % CI 0.01–0.65, for the highest age quartile versus the lowest).

Discussion

The results of this study show that among women with early-stage invasive BC who underwent BCT, ILC was not associated with an increased risk of LR, nor did ILC require more extensive surgical margins than IDC. To our knowledge, this is the first analysis of ILC outcomes that employs modern surgical margin definitions.

The relative minority of ILC cases among all BCs has generated a diverse body of literature for this histologic subtype. A number of studies have concluded, as we have, that despite the inherent differences between ILC and IDC, risk profiles are similar among the two [21–26]. Meanwhile, others have suggested that ILC outcomes may be superior [27–29] or inferior [2, 30–32] to their IDC counterparts, occasionally going so far as to recommend routine mastectomy for this disease [33].

The consideration of biologic subtypes has revealed limitations among these retrospective analyses, particularly since ILC is predominantly characterized by the favorable luminal A subtype, whereas IDC is more broadly distributed among other aggressive classifications. As a result, without accounting for biologic subtype, an ILC cohort might demonstrate superior outcomes due to a disproportionately favorable luminal A distribution. Conversely, studies affected by selection bias may suffer from an enrichment of unfavorable subtypes in the ILC cohort, consequently yielding worse outcomes. In light of these considerations, we compared the local outcomes of ILC and IDC among luminal A patients and found no significant differences in the rates of LR. This observation suggests that within the luminal A subtype, ILC does not exhibit a higher risk of LR than IDC, and may not require more aggressive local management. Further investigation is needed to determine if these findings can be extrapolated to other ILC subtypes, including those that do not express the ER.

An additional departure from past studies includes the widespread use of technologies that have improved pathologic and mammographic evaluation; these, along with recent advances in systemic therapy, have resulted in substantial reductions of LR, particularly among patients with luminal cancers [20]. Indeed, a high percentage of patients in our series received adjuvant systemic therapy in contrast to prior reports.

Consensus guidelines issued jointly by the American Society for Radiation Oncology (ASTRO) and the Society of Surgical Oncology (SSO) have established the use of no ink on tumor (i.e., no malignant cells adjacent to any inked specimen surface) as an adequate margin for the majority of patients with invasive BC treated with BCT. These recommendations focus largely on a body of literature that is predominated by IDC. Efforts to address surgical margins for ILC are necessarily constrained by a paucity of definitive studies, although limited informative reports do exist.

Galimberti et al. analyzed 382 patients with pure ILC who underwent BCT from 1996 to 2002 [19]. Margin status in this study was dichotomized to <10 mm versus ≥10 mm, and systemic therapy did not differ by surgical margins (p = 0.47). With regard to adjuvant treatment, 5 % of patients received no systemic therapy, 58 % received endocrine therapy alone, 8 % received chemotherapy alone, and 29 % received both. At a median follow-up of 8.4 years, LR in the originally involved quadrant was observed in 3.7 % of patients with ≥10 mm margins and in 4.6 % of those with <10 mm margins (log-rank, p = 0.70). Analyses of other outcomes including distant metastasis, disease-free survival, and overall survival, similarly showed no significant association with margin status. In another report, Biglia et al. retrospectively studied a cohort of 1,650 patients, 243 of whom had ILC [16]. Consistent with our analysis, they found that patients with ILC more frequently required a second surgery due to inadequate (“involved/close”) resection margins. This observation did not translate into a difference in DFS between ILC and IDC (89 % vs 85 %, respectively; p = NS), although the distributions of final margin width and biologic subtypes were not reported. A similar assessment of ILC outcomes was published in 1998 using the National Cancer Database [18]. This study of 291,273 BC patients demonstrated no significant differences in 5-year OS or DFS between ILC and IDC, although “breast preservation” was less frequent for patients with ILC. These findings are particularly informative in light of the relatively sparse use of hormonal therapy in prior decades when these studies were conducted [34, 35]. Since the majority of ILCs express ER, the increasing prevalence of hormonal therapy today would be expected to further reduce LR in the modern era.

Given minimal prior data regarding margins in the 0–2 mm range for ILC, we specifically analyzed margin width by histologic type to elucidate the implications of current guidelines on LR. Stratifying by IDC, ILC, and mixed histology, we found no significant differences in local control between margins of <2 mm and those ≥2 mm. Notably, no LR events were observed among any ILC patients with margins of <2 mm, possibly owing to the relatively small number of patients in this group, and/or to the indolent biology of this disease.

These data must be interpreted in the context of the study design. The use of ER, PR, HER2 status and grade are approximations of biologic subtypes that were originally defined by transcriptional profiling [36–38]. As a result, our findings may not necessarily apply to molecularly characterized subtypes. Another possible limitation is that our analysis of margin width yielded relatively few patients with margins of <2 mm, thereby reducing our power to conclusively analyze this group. This distribution likely arose from the longstanding practice of re-excising any margin <2 mm, established in an era with now-outdated imaging techniques and systemic therapy. Further studies are needed to confirm our findings in patients with sub-millimeter margins given that the collective experience with this approach has been limited thus far.

In summary, this study demonstrates that the rate of LR among patients with ILC is similar to that among patients with IDC or mixed histology in the context of modern BCT. In addition, we did not find that ILC required more extensive surgical margins than IDC.