Impact of hormone receptor status and distant recurrence-free interval on survival benefits from trastuzumab in HER2-positive metastatic breast cancer

We sought to investigate the impact of hormone receptor (HR) status and distant recurrence-free interval (DRFI) on the degree of overall survival (OS) benefit from palliative trastuzumab-containing treatment in HER2-positive metastatic breast cancer (MBC). Here, we retrospectively identified 588 eligible HER2-positive patients with postoperative distant recurrence. DRFI of HR+HER2+ MBC patients (median: 30.7 months, IQR: 18.5–45.9, P < 0.001) was significant longer compared with HR−HER2+ patients. Patients were categorized into four subgroups based on HR status and palliative trastuzumab (trast+) received. The most superior outcome was observed in the HR+HER2+trast+ subgroup, with a median OS of 48.3 months. Moreover, DRFI > 24 months is an independent favourable prognostic factor for both HR−HER2+ patients (Hazard Ratio (HzR) = 0.55, 95% CI: 0.39–0.76, P < 0.001) and HR+HER2+ patients (HzR = 0.45, 95% CI: 0.32–0.64, P < 0.001). Upon further analysis of the interaction between trastuzumab and DRFI, the degree of trastuzumab benefits in HR−HER2+ MBC patients remained basically unchanged regardless of DRFI length. Unlikely, the degree in HR+HER2+ MBC patients decreased gradually along with DRFI extending, indicating that trastuzumab failed to translate into an OS benefit for late recurrent (DRFI > 5years) HR+HER2+ MBC patients.

A gene expression-based predictive model has identified a subgroup of HER2-positive primary breast cancer patients with the highest levels of ESR1 and its associated mRNA who derived no clinical benefit from adjuvant trastuzumab 9 . However, the impact of HR co-expression on the overall survival benefit from a variety of therapies for HER2-positive MBC remains poorly understood [9][10][11] . DRFI, which is simple to measure, easy to interpret and clinically meaningful, has been recognized as a predictor of trastuzumab resistance and endocrine therapy resistance. Several retrospective studies have suggested that the DRFI partially accounts for the trastuzumab benefit 12 and overall survival (OS) after metastasis 13 .
Here we report the palliative treatment pattern and clinical outcome of HER2-positive MBC patients in China. In particular, we investigate the impact of HR status on the OS benefit of trastuzumab-containing palliative therapy, and our results further elucidate the magnitude of the trastuzumab benefit interacting with DRFI as a continuous covariate.

Results
Patient characteristics. In total, 588 eligible HER2-positive MBC patients were retrospectively collected ( Fig. 1). Of these, 311 were HR-negative, and 277 were HR-positive. The median follow-up was 24.8 (IQR: 14.0-40.3) months for HR−HER2+ patients and 32.5 (IQR: 29.8-46.9) months for HR+HER2+ patients. Tables 1  and 2 give the patients' characteristics at the time of primary diagnosis and at the time of metastasis, respectively.
Obviously, HR+HER2+ MBC patients showed a longer DRFI compared with HR−HER2+ patients, with a median DRFI of 30.7 (IQR: 18.5-45.9) months versus 22.9 (IQR: 15.3~35.6) months (P < 0.001). Although HR−HER2+ MBC showed a trend towards a higher incidence of visceral metastases (66.5% vs. 59.6%; P = 0.09), the difference did not reach statistical significance. The first sites of metastases were similar between HR− HER2+ and HR+HER2+ patients, with brain (4.4% vs. 4.3%), liver (24.6% vs. 26.7%), and lung (35.7% vs. 22.4%) involvement observed. A total of 84 patients (14.3%) received a rebiopsy of metastatic lesions, of which 11 changed from HR-positive to HR-negative, 4 changed from HR-negative to HR-positive, 6 changed from HER2-positive to HER2-negative, and 2 changed from HER2-negative to HER2-positive. However, no robust statistical conclusion could be drawn due to the small number of cases. In terms of above analysis, the most notable clinical feature in HR+HER2+ patients in comparison to HR−HER2+ ones was a significant longer DRFI.

Interaction of the trastuzumab benefit between DRFI as an continuous variable. To further
investigate the degree of OS benefit, the interaction between degree of trastuzumab benefit and DRFI as a continuous covariate was tested using an multivariable fractional polynomial interaction (MFPI) model 14 . A plot of the log hazard ratio of trastuzumab treatment against the DRFI, together with corresponding 95% CI, is presented in Fig. 3a and b. According to the model, no significant interaction was observed between trastuzumab treatment and DRFI in HR−HER2+ MBC patients, namely, the survival advantage of trastuzumab remained fairly stable up to 6 years of DRFI with a relative hazard ratio less than −0.87 (Fig. 3a). Conversely, for HR+HER2+ MBC patients, the relative hazard ratio kept elevating with prolonged DRFI and reach above zero at DRFI beyond 5 years in trastuzumab treated patients (Fig. 3b). The early recurrent (DRFI ≤ 5 years) HR+HER2+ MBC patients substantially benefit from trastuzumab (HzR = 0.40, 95% CI: 0.27-0.58, P < 0.001), whereas the later recurrent (DRFI > 5 years) ones didn't show similar benefit (HzR = 0.84, 95% CI: 0.32-2.20, P = 0.718) (Fig. 3c). Taking into account the possible bias caused by cancer heterogeneity 15 , we further demonstrated the robustness by bootstrap resampling with 10,000 times (Supplementary Figure 1). Our finding suggests that trastuzumab-containing treatment failed to translate into OS benefit in a subgroup of late recurrent HR+HER2+ MBC patients.

Discussion
Our results confirm that trastuzumab-containing therapy substantially improves OS in both HR+HER2+ and HR−HER2+ MBC patients in China. In accordance with previous reports 16 , we found that HR+HER2+trast+ disease was associated with the most favourable outcome, with a median OS of 48.3 months. Among the prognostic factors investigated, a DRFI > 24 months was independently associated with better OS in HER2-positive MBC regardless of HR status, suggesting that DRFI was closely related to the biological characteristics of both primary tumours and their metastases. Although there is increasing recognition that HER2-positive MBC is not homogeneous, high-quality prognostic and/or predictive markers with high accuracy and robustness are remain to be identified 15 . The optimization of palliative therapies for HR+HER2+ MBC patients requires further study 8 . Previous reports suggested that high ER expression (≥30%) was associated with a reduced probability of tumour response to trastuzumab plus chemotherapy 11 . However, progression-free survival was significantly improved when maintenance endocrine therapy was added to trastuzumab in ER-positive MBC patients 17 . More importantly, the current analysis provides real-world evidence that DRFI can serve as a reliable stratification factor for OS benefit analysis from trastuzumab-containing palliative therapy in HR+HER2+ MBC, demonstrating that late recurrent (DRFI > 5 years) patients derived a less pronounced survival benefit than their early recurrent counterparts.
Over the last decade, many HER2-positive early breast cancer patients have had limited access to adjuvant trastuzumab worldwide due to reimbursement policies. From January 1, 2007 to December 31, 2011 at FUSCC, only 35.2% (258/732) of HER2-positive primary breast cancer patients received one year of adjuvant trastuzumab 18 , while 83% of such patients in the United States received this treatment. Indeed, the management of trastuzumab-naïve HER2-positive MBC patients will likely continue to be a challenge for physicians. Thus, further study should be carried out to optimize the combination and/or sequence among endocrine therapy, anti-HER2 agents and chemotherapy.
Our study has several limitations. First, we performed a single-institute retrospective database analysis rather than a prospective cohort analysis, which may have led to unaccounted biases. Second, patients who crossed over to receive lapatinib beyond progression were included. Third, adverse effects were not analysed due to unavailable medical records.   Table 3. Univariate and multivariate analysis for overall survival.
In conclusion, our findings reveal that HER2+ metastatic breast cancer can be divided into two distinct subgroups by HR status. HR+HER2+ subgroup have a significant longer DRFI compared to HR−HER2+ subgroup. The interaction between the degree of trastuzumab benefit and DRFI in HR+HER2+ subgroup differed from HR−HER2+ subgroup, demonstrating that DRFI can serve as a reliable stratification factor for the survival benefit from trastuzumab-containing palliative therapy in the HR+HER2+ subgroup.

Methods
Ethics statement. Our study was approved by the independent ethical committee/institutional review board of FUSCC. All patients provided written informed consent before inclusion in this study. We obtained permission of FUSCC to collect data from the follow-up database. Our methods were performed in accordance with the relevant guidelines and regulations. Patients. This was a retrospective single-institution study conducted at Fudan University Shanghai Cancer Center (FUSCC). Among women (≥18 years old) who were initially diagnosed consecutively with stage I to III breast cancer and received curative surgery and/or radiotherapy at FUSCC from January 1, 2000 to December 31, 2012, 390 subsequently developed distant metastases before December 31, 2013. In addition, HER2-positive MBC patients, who were confirmed by pathology and received palliative therapy in FUSCC from January 1, 2006 to December 31, 2012, were also included (Fig. 1). The following cases were excluded: (1) de novo metastatic disease, (2) bilateral primary breast cancer, (3) inflammatory breast cancer, (4) second primary invasive cancer, and (4) prior lapatinib exposure before trastuzumab.
The following data were collected: age at initial diagnosis, menopausal status, date of diagnosis of primary breast cancer, histological type, size of primary tumour, number of positive lymph nodes, oestrogen receptor (ER) and progesterone receptor (PgR) status, HER2 status (including immunohistochemistry (IHC) results and fluorescent in situ hybridization (FISH) results if available), Ki67 index, date of metastasis, sites of metastases, receptor status of metastatic lesion (if reassessed), and date of death. The treatment history was also collected, including prior (neo)adjuvant chemotherapy, (neo)adjuvant endocrine therapy, and (neo)adjuvant anti-HER2 therapy. The dates of the first and last courses of trastuzumab were also recorded.
Positivity of ER and PgR status was defined as immunostaining ≥1% of invasive tumour cells. HER2 positivity was defined as neu staining of 3+ by immunohistochemistry or 2+ with gene amplification by FISH. Patients were considered to have visceral disease if they had one of the following organs involved: liver, lung, pleura, adrenal gland, pericardial/peritoneal cavity, spleen, thyroid or brain. For patients who developed multi-organ metastasis, the first site of metastasis was ranked according to the prognosis: brain, liver, lung, bone, and other (e.g., lymph node, soft tissue, skin, pleura, and other rarer sites). Statistical analysis. DRFI was defined as the interval between the diagnosis of primary breast cancer and the diagnosis of the first distant relapse 19 . OS was defined from the diagnosis of distant metastatic disease to the date of death from any cause. Follow-up was conducted by specific staff members until March, 31, 2016. If patients were lost during the follow-up period, the data were censored at the last date recorded.
The patient demographics and tumour characteristics by HR status were compared using the using Chi-square and the t test. Kaplan-Meier curves for overall survival rates were created using GraphPad Prism (Version 4.03). Univariate regression between variables and overall survival were evaluated. All variables associated with P < 0.10 on univariate analysis were included in multivariate Cox proportional hazards model. Hazard ratios (HzR) with 95% confidence intervals were tested for all entered variables with backward method (SPSS Version 20, IBM Corporation, Armonk, NY, USA). The interaction between trastuzumab benefit and DRFI was tested with multivariable fractional polynomial interaction (MFPI) using Stata (Version 13) 14 . All statistical tests were two-sided, and a P-value < 0.05 was considered significant.