This multi-center, randomized, controlled, phase III trial was conducted across eight hospitals located in Beijing, China. The trial was designed by the Peking University Cancer Hospital and supervised by the institutional review board at each study site. The study was done in accordance with the principles of the Declaration of Helsinki. All patients provided informed consent for research participation (Ethics approval number: 2009024).
Eligible patients were women aged <65 years with invasive breast cancer and positive axillary lymph nodes (diagnosed by fine needle aspiration, core needle biopsy, or sentinel lymph node biopsy). Patients with estrogen receptor-positive or progesterone receptor-positive (≥10% by immunohistochemistry) and HER2- (0 or 1+ by immunohistochemistry or HER2/chromosome enumeration probe ratio <1·8 by fluorescent in situ hybridization) were required. Patients had to complete four cycles of NCT (containing anthracycline), undergo radical surgery, and be non-responsive to NCT according to pathological assessment. An experienced pathologist employed at the Peking University cancer hospital who was blinded to the groupings used the Miller and Payne (M&P) grading system to assess the pathological response of the primary tumor, as follows: grade 1: there was no change in individual tumor cells and no decrease in overall cellularity; grade 2: <30% necrosis of tumor cells; grade 3: 30–90% necrosis of tumor cells; grade 4: >90% necrosis of tumor cells; grade 5: no tumor cells were identified in sections obtained from the tumor site; however, ductal carcinoma in situ residue alone might be present. Non-responsiveness to treatment was defined as M&P grade 1–3 disease, or any stage with residual positive lymph nodes in the surgical specimen. The other seven hospitals each had their own pathologists perform pathological grading of the specimen.
The main exclusion criteria were a history of other malignant tumors, metastatic breast cancer, and any clinically serious medical conditions.
Randomization and treatment
All intention-to-treat (ITT) patients were enrolled preoperatively, and the eligible patients were randomly assigned in a 1:1 ratio within 4 weeks postoperatively. The other ITT patients who did not meet the pathological non-responsiveness criteria were assigned to the observation group. The patients in arm A received four cycles of a non-cross-resistant regimen plus endocrine therapy (ET), while the patients in arm B patients received ET alone. The randomization sequence was created using SPSS software (version 11.0; IMB SPSS Inc., Armonk, NY, USA) and was stratified based on NCT regimens (anthracycline- or taxane-based therapy vs. concurrent anthracycline and taxane usage) and pathological response (M&P staging G1 and G2 vs. M&P staging G3) using random block sizes of four. Patients with M&P G4–5 disease and residual positive lymph nodes were assigned to the M&P G3 subgroup. For patients with limited lymph node metastasis (one or two diagnosed by sentinel lymph node biopsy) after undergoing breast-conserving surgery, we omitted axillary lymph node dissection.
Patients enrolled from Beijing Cancer Hospital started with four cycles of CEF3w (cyclophosphamide 600 mg/m2 on day 1; epirubicin 90–100 mg/m2 on day 1; 5-fluorouracil 600 mg/m2 on day 1, every 3 weeks), and after being assigned to the chemotherapy group, they received four cycles of Tq1w (paclitaxel 80 mg/m2 on days 1, 8, and 15, every 3 weeks) or TPq1w (paclitaxel 80 mg/m2 on days 1, 8, and 15 and carboplatin AUC 6 on day 1, every 3 weeks). Patients from the other hospitals started with four cycles of TAC (docetaxel 75 mg/m2, doxorubicin 50 mg/m2, and cyclophosphamide 500 mg/m2, every 3 weeks) or TE (docetaxel 75 mg/m2 and epirubicin 75 mg/m2, every 3 weeks) and received four cycles of adjuvant chemotherapy with NX (vinorelbine 25 mg/m2 on days 1 and 8 plus capecitabine 1,000 mg/m2 orally twice/day on days 1–14, every 3 weeks) or NP (vinorelbine 25 mg/m2 on days 1 and 8 and carboplatin AUC 6 on day 1, every 3 weeks) when assigned to the chemotherapy group. Patients in both arms received tamoxifen or aromatase inhibitors for 5 years and underwent whole breast irradiation after breast-conserving surgery, or chest wall and supraclavicular region irradiation after mastectomy, before 2014. After the results of the SOFT and TEXT trials were published (2014), the doctors decided on the use of gonadotropin-releasing hormone agonists for premenopausal patients with breast cancer. The adjuvant treatment regimen for patients in the observation group was also determined by doctors, with four cycles of non-cross-resistant chemotherapy plus ET or ET alone being prescribed. All enrolled patients were followed up postoperatively every 3 months over the first 2 years, every six months over the next 3–5 years, and once a year for 5 years after that.
Endpoint assessment
The primary endpoint was DDFS, defined as the interval between the date of surgery and the occurrence of the first distant event or breast cancer-specific death, whichever occurred first. The secondary endpoints were invasive disease-free survival (iDFS) and overall survival (OS). Liver metastases, loco-regional metastases, contra-lateral breast cancer, and second primary cancer were diagnosed based on pathology, while brain, bone, and lung metastases were diagnosed based on radiology.
Statistical analysis
A sample of 350 patients (ET-alone group: n = 175; non-cross-resistant chemotherapy plus ET group: n = 175) was planned due to a requirement of 74 distant disease events to provide 80% power to detect a hazard ratio (HR) of 0.75 with a two-sided significance level of 5% in the primary analysis. According to the retrospective data available at our center, we assumed that the DDFS at 5 years was 88% in the non-cross-resistant adjuvant chemotherapy group and 73% in the ET-alone group. Because the median follow-up time at the time of writing was 72.4 months, the DDFS event rates were substantially lower than originally expected, and since an additional delay of 5–8 years was considered to be unacceptably long, we deduced that “time-driven” rather than “event-driven” analyses would be more appropriate.
All cases were analyzed on an ITT basis, including patients in the observation group (Figure 1). Time-to-event endpoints were calculated using the Kaplan‒Meier method and compared between patient groups using the log-rank test. HRs and 95% confidence intervals (95% CIs) were estimated using a Cox proportional hazards regression model. A two-sided P-value <0.05 was considered statistically significant.