Human epidermal growth factor receptor 2 (HER2), a transmembrane tyrosine receptor kinase, plays an important role in the activation of signal transduction pathways that regulate cell proliferation, oppose apoptosis, and stimulate angiogenesis.1 Amplification of HER2 occurs in 15–30 % of breast cancers and is associated with poor prognosis and increased mortality.2 The successful development of trastuzumab, a humanized monoclonal antibody targeting the extracellular domain of HER2, has dramatically improved outcomes for women with HER2-positive (HER2+) breast cancer. Four randomized trials demonstrated improved disease-free and overall survival (OS) with trastuzumab added to conventional chemotherapy for women with HER2+ breast cancer.3–5

In contrast, the effect of trastuzumab on locoregional recurrence (LRR) outcomes for HER2+ breast cancer patients is not as clearly established. Although the initial reports describing a combined analysis of NSABP B-31 and NCCTG N9831 trials showed a reduction in crude LRR with the addition of trastuzumab, LRR was not a specified end point, nor was it statistically analyzed.4 A retrospective series from Memorial Sloan Kettering Cancer Center (MSKCC) evaluating LRR in early-stage patients treated with breast-conserving therapy showed a reduction in LRR with trastuzumab.6 Others have found that trastuzumab reduces LRR in hormone receptor-positive HER2+ patients but not in hormone receptor-negative HER2+ patients.7 Assessment of LRR outcomes in older studies including patients with lymph-node-positive HER2+ breast cancer have been limited by the low number of patients treated with mastectomy and trastuzumab as well as by the inclusion of patients who received neoadjuvant chemotherapy.8,9

Few studies have evaluated whether trastuzumab modulates LRR risk in HER2+ breast cancer patients who receive postmastectomy radiation (PMRT) and systemic therapy. A recent meta-analysis including nearly 5,500 mastectomy patients showed a higher risk of LRR in HER2-overexpressing tumors, the majority of whom did not receive trastuzumab (<6 %).10 Studies investigating the impact of PMRT on LRR in patients with HER2+ breast cancer who underwent contemporaneous neoadjuvant or adjuvant systemic therapy did not identify an association between LRR risk, HER2 expression status, and trastuzumab.8,11

This study aimed primarily to examine the effect of adjuvant trastuzumab on LRR in patients with HER2+ breast cancer who underwent mastectomy. Our secondary aim was to assess whether this effect varied with the use of PMRT or not.

Materials and Methods

This study was approved by the institutional review board at MSKCC. A multidisciplinary database was used to identify 556 women with stages I–III HER2+ breast cancer who underwent mastectomy at our institution from 1998 to 2007. The study defined HER2+ status as immunohistochemistry (IHC) staining of 3+ and/or ≥2.0 on fluorescence in situ hybridization (FISH) or IHC staining of 2+ and FISH greater than 2.0. Patients with inflammatory breast cancer, receipt of neoadjuvant chemotherapy or lapatinib, or unknown receipt of trastuzumab were excluded from the study, leaving 501 women in the study population.

Patient, disease, and treatment characteristics as well as LRR, distant metastasis (DM), and date of death were collected from the electronic medical records. Characteristics were assessed for association with receipt of trastuzumab or PMRT using either the χ 2 test or the Wilcoxon rank sum test.

The primary end point of the study was locoregional recurrence-free survival (LRRFS). LRR was defined as biopsy-proven ipsilateral recurrence in the chest wall or internal mammary, supraclavicular, or axillary lymph nodes. The secondary end points included DM and OS. Landmark analyses examining the effect of adjuvant treatments were performed, starting from 12 months after surgery, consistent with timing for trastuzumab, chemotherapy, and PMRT reported in the literature.12

Since 2005, trastuzumab has been routinely recommended for HER2+ patients with node-positive disease or tumors larger than 1 cm. The institutional guidelines for PMRT during the study period included patients with four or more involved lymph nodes or one to three positive nodes with other high-risk features. The number of patients and LRR events in each cohort and the patients excluded due to statistical criteria are illustrated in Fig. 1. Patients whose timing of trastuzumab initiation was not clearly documented in the medical records were excluded from the cohort used to analyze the effect of trastuzumab. Only patients alive and event free were included in the analysis of each end point. Patients were categorized as having received trastuzumab or not according to whether treatment began by the landmark time. A similar approach was applied to the analysis of PMRT.

Fig. 1
figure 1

Study design demonstrating cohort selection for landmark analyses

Full size image

For all end points, time to event was calculated starting from the landmark time. Patients who did not experience the event were censored on the date of the last follow-up visit for LRRFS and OS. Kaplan–Meier methods were used to estimate 5-year LRRFS and OS, and the log-rank test was used to compare the overall difference across all time points. Hazard ratios (HRs) with 95 % confidence intervals (CIs) were estimated using Cox regression. For the analysis of DM, death was considered a competing event, and estimation, testing, and regression modeling of the cumulative incidences were analyzed by competing risk methods using the cmprsk R library (R version 2.11 http://www.r-project.org).

To reduce bias secondary to the nonrandomized assignment of trastuzumab, a propensity score analysis was performed using the twang R library. The criteria for covariates in the model included any characteristic for each dependent variable (trastuzumab or PMRT) with a p value lower than 0.10 between groups. For trastuzumab, the covariates included were tumor size, nodal status, extracapsular extension (ECE), lymphovascular invasion (LVI), hormone receptor status, age, menopausal status, and receipt of chemotherapy and PMRT. We constructed an analogous model for the probability of receiving PMRT with the same covariates, exchanging receipt of trastuzumab for PMRT and including nuclear and histologic grade, surgical margin status, and axillary surgery type. A multivariable analysis for predictors of LRR was not performed due to the low number of events, which would preclude meaningful interpretation of the results. All other statistical analyses were performed using SAS 9.2 (SAS Institute, Cary, NC, USA).

Results

The study had 501 evaluable patients, 41 % of whom received trastuzumab. The median follow-up period for the entire group was 84 months (range 0.2–177.1 months). The clinicopathologic and treatment characteristics of the trastuzumab and no trastuzumab groups are compared in Table 1. The women in the no trastuzumab group were more likely to be postmenopausal (p = 0.02) or to present with negative nodes (p = 0.007) or stage 1 disease (p = 0.001) and less likely to have LVI (p = 0.001) or ECE (p = 0.02). The majority of the study patients (84.6 %) received adjuvant chemotherapy, and 53.9 % received hormonal therapy. The patients in the trastuzumab cohort were more likely to receive adjuvant chemotherapy (99.0 vs 74.7 %; p < 0.001) and more likely to receive taxane-based regimens (87.7 vs 61.1 %; p < 0.001). The median follow-up period for survivors was longer in no trastuzumab group (100.5 vs 79.4 months).

Table 1 Patient, tumor, and treatment characteristics for the trastuzumab-treated and no trastuzumab groups
Full size table

Postmastectomy radiation was delivered to 30 % (n = 150) of the patients. Compared with the patients who did not receive PMRT, those treated with PMRT had a larger median tumor size (3.1 vs 1.8 cm; p < 0.001), a greater median number of positive lymph nodes (6 vs 0; p < 0.001), LVI, high tumor grade, and ECE (all p < 0.05; Table 2). The women who received trastuzumab were more likely to receive PMRT (37.6 vs 24.7 %; p = 0.002). The median dose to the chest wall was 5,040 cGy (range 2,250–6,640 cGy). The radiation fields included the supraclavicular lymph nodes in 83 % and an axillary boost in 21.3 % of the irradiated patients, with no difference between the trastuzumab and no trastuzumab cohorts.

Table 2 Patient, tumor, and treatment characteristics of the postmastectomy radiation (PMRT) and no PMRT groups
Full size table

Overall, the study patients experienced 24 LRR events, 82 DM events, and 83 deaths from all causes. Four LRR events that occurred before the landmark time were excluded from the analyses, all of which involved patients who did not receive trastuzumab. The majority of the recurrences occurred in the chest wall (n = 11), whereas the remainder were in the ipsilateral axillary (n = 4, all managed with sentinel lymph node biopsy alone), supraclavicular (n = 4), or internal mammary lymph nodes (n = 1). The 5-year outcomes for the trastuzumab and no trastuzumab cohorts were as follows: LRRFS (98 vs 94 %; p = 0.05; Fig. 2a), cumulative incidence of DM (8 vs 18 %; p = 0.006; Fig. 2b), OS (95 vs 85 %; p = 0.0004; Fig. 2c). Propensity score analysis to balance treatments on covariates showed that trastuzumab relatively improved LRRFS by approximately 80 % (HR 0.21; 95 % CI 0.04–0.94; p = 0.04).

Fig. 2
figure 2

a Locoregional recurrence-free survival curves stratified by receipt of trastuzumab. b Cumulative incidence of distant metastases stratified by receipt of trastuzumab. c Overall survival curves stratified by receipt of trastuzumab

Full size image

Multivariate analyses (MVA) with adjustment for tumor size, nodal status, ECE, LVI, hormone receptor status, age, menopausal status, and receipt of chemotherapy and PMRT showed that trastuzumab significantly decreased the risk of DM (HR 0.34; 95 % CI 0.18–0.64; p = 0.0009) and improved OS (HR 0.36; 95 % CI 0.18–0.75; p = 0.006). The 5-year LRRFS rate was 98 % with PMRT and 95 % without PMRT (p = 0.27; Fig. 3a). Propensity score analysis, with adjustment for differences in tumor size, nodal status, ECE, LVI, nuclear and histologic grade, hormone receptor status, age, menopausal status, surgical margin status, axillary surgery type, and receipt of chemotherapy and trastuzumab use between groups showed that PMRT did not significantly decrease the rates of LRR (HR 0.27; 95 % CI 0.07–1.08; p = 0.06).

Fig. 3
figure 3

a Locoregional recurrence-free survival curves for patients stratified by receipt of postmastectomy radiation (PMRT). b Locoregional recurrence-free survival curves of PMRT patients treated with trastuzumab or no trastuzumab. c Locoregional recurrence-free survival curves of patients treated without PMRT stratified by receipt of trastuzumab

Full size image

Subset analyses were performed to examine the effect of trastuzumab on LRR according to the use of PMRT (Fig. 3b, c). Among 129 women who received PMRT, the 5-year LRRFS was 100 % for the trastuzumab group (n = 62) and 95 % for the no trastuzumab group (n = 67), trending toward significance (p = 0.07). Among 318 patients who did not receive PMRT, the 5-year LRRFS was 97 % for the trastuzumab group (n = 107) and 94 % for the no trastuzumab group (n = 211). The difference was not significant (p = 0.26). The limited number of events precluded a formal statistical comparison of the trastuzumab effect between the PMRT and no PMRT groups.

Discussion

Transcriptional profiling of breast cancers has led to identification of distinct intrinsic subtypes with varying prognoses, one of which is the HER2+ subtype.13 Overexpression of HER2 has been associated with higher risks of both local recurrence and mortality in studies that predate the use of trastuzumab.14–16 Improvements in survival with trastuzumab for HER2+ breast cancer are well established, but the evidence is limited regarding its impact on LRR in patients undergoing mastectomy, particularly those receiving PMRT.

In our study of 501 women undergoing mastectomy, adjuvant trastuzumab lowered the relative rate of LRR by 80 % over 5 years. This reduction was marginally significant after adjustment for tumor size, nodal status, ECE, LVI, hormone receptor status, age, menopausal status, and receipt of adjuvant chemotherapy and PMRT. This difference is consistent with our findings of the impact that adjuvant trastuzumab has on LRR in HER2+ patients treated with breast-conserving surgery and radiation.6 As anticipated from the findings in large randomized trials, a 10 % absolute improvement in OS was observed with trastuzumab.5,17–20

These results contrast with those reported by others who have examined the effect of trastuzumab on LRR in HER2+ breast cancer patients. In a retrospective study of 1,185 HER2+ breast cancer patients treated with lumpectomy or mastectomy at the MD Anderson Cancer Center, trastuzumab improved local control rates in hormone receptor-positive patients (5-year LRR rate, 3 vs 6 % in ±trastuzumab) but not in hormone receptor-negative breast cancer patients (5-year LRR rate, 6 vs 6 % in ±trastuzumab).7 In our cohort of patients, the small number events (n = 10) and the absence of statistically significant interaction between ER status and trastuzumab (p = 0.38) made it impossible to explore this subpopulation further. In a multi-institutional study of T1-2N0 HER2+ breast cancer, high local control rates were observed in the mastectomy-treated subset of 346 patients regardless of trastuzumab use, with a 5-year LRRFS of 99.4 % among the mastectomy patients who received trastuzumab compared with 97.4 % among the patients who did not (p = 0.37).21

Differences in patient populations may explain the discrepancies between our results and those of the aforementioned studies. The latter studies included HER2+ patients treated with both breast-conserving therapy and mastectomy as well as receipt of neoadjuvant chemotherapy, and lower proportions of patients with advanced-stage disease. In contrast, our study population was composed exclusively of HER2+ patients who underwent mastectomy, with node-positive patients constituting 57 % of the overall cohort and all treatments delivered in the adjuvant setting. The proportional reduction in LRR of 80 % with trastuzumab observed in our study parallels that reported in the combined NSABP B-31 and NCCTG N9831 trials, which consisted of HER2+ patients treated with both mastectomy and breast-conserving therapy and excluded patients with small node-negative HER2+ breast cancer. In these trials, the addition of trastuzumab decreased the crude rates of LRR as a first event by approximately 50 % (from 4 to 1.7 % in the NSABP B-31 trial and from 2.7 to 1.5 % in the NCCTG trial).19 Our results also are consistent with the LRR relative reduction of 60 % demonstrated in our prior study of HER2+ patients with early-stage disease who received breast-conserving therapy.6

The relative contribution of PMRT to local control in HER2+ patients receiving trastuzumab has not been well quantified. In the randomized trials of PMRT from the 1990s, the HER2 status was unknown. In a post hoc analysis of the Danish 82 b/c trials, modest reductions in LRR were seen with the addition of PMRT in hormone receptor-negative/HER+ tumors.22 Although these findings, supported by preclinical studies, have led to the postulation that HER2+ breast cancer is radioresistant, other conflicting reports show that HER2 receptor blockade enhances DNA damage due to radiation.23,24 Neither of these theories has been supported by findings in clinical studies.

In our study, women who received both PMRT and trastuzumab experienced no LRR events, whereas the 5-year rate of LRR was 5 % for PMRT/no trastuzumab patients, which trended toward significance (p = 0.07). Among the nonirradiated patients, many of whom had low-risk features (Table 2), the effect of trastuzumab on LRR did not reach significance. Cautiously taken together, these results suggest that trastuzumab may be more effective in reducing LRR when delivered together with adjuvant radiotherapy. However, given the low number of events, the relatively short follow-up period, and the limited patient numbers comprising the subsets, these results are hypothesis generating and require validation in larger, well-designed studies.

Other limitations to this study warrant discussion. There was significant heterogeneity in patient, tumor, and treatment characteristics between cohorts, particularly in receipt of PMRT and taxane-based chemotherapy regimens. We attempted to mitigate these differences with statistical methods that balanced prognostic factors for LRR between groups. The use of trastuzumab was evolving during the study period, resulting in variable receipt of trastuzumab in the cohort. By using a landmark analysis, we attempted to minimize potential bias due to differential timing of adjuvant treatments. Additionally, the low number of LRR events prevented meaningful analysis of other predictors of LRR and likely contributed to the finding that the reduction in LRR with PMRT did not reach statistical significance. Finally, these results must be considered in the context of the growing use of additional targeted agents for HER2+ breast cancer, such as pertuzumab. Although administration of pertuzumab is not yet considered a standard adjuvant anti-HER2 therapy as we await the results of the large phase 3 APHINITY (BIG 4–11/BO25126/TOC 4939G) study, some clinicians are administering pertuzumab in the adjuvant setting with endorsement from the National Comprehensive Cancer Network (www.nccn.org) based on the significant benefits observed in the metastatic and neoadjuvant settings.25,26 Whether dual blockade confers additional local control benefits or not remains to be seen.

In summary, high rates of LRRFS (5-year rate, 98 %) were observed among the patients who received trastuzumab and mastectomy, a select proportion of whom received PMRT. These results complement the findings from our prior study, which demonstrated the benefit of trastuzumab for breast-conserved patients,6 further underscoring the effectiveness of trastuzumab for locoregional control in patients with HER2+ breast cancer.