Patients with pT1-2N0 HR-positive HER2-negative breast cancer who received surgery in 9 medical centers between 2013 and 2016 were included in this retrospective study. Their formalin-fixed and paraffin-embedded (FFPE) tumor samples were obtained for RecurIndex assay. The inclusion criteria included: (1) women aged 18-75 years; (2) pT1-2N0 HR-positive HER2-negative breast cancer diagnosed by both immunohistochemistry (IHC) staining and RecurIndex assay; (3) breast-conserving surgery or mastectomy as the initial treatment; (4) sufficient FFPE tumor samples, complete clinicopathological and follow-up data. The patients would be excluded from the study if they met any one of the criteria below: (1) preoperative administration of neoadjuvant therapy including chemotherapy; (2) incomplete postoperative endocrine therapy and/or chemotherapy; (3) patients diagnosed as bilateral breast cancer, or with the history of other malignancies.

The study was performed according to the principles of Declaration of Helsinki and local regulations. All the patients were informed consent, and the study was approved by the Institutional Review Board of The Affiliated Hospital of Qingdao University (approval No.: QYFYKYLL801311920).

The genes included in the RecurIndex assay was composed of 18 genes (TRPV6, DDX39, BUB1B, CCR1, BLM, C16ORF7, PIM1, TPX2, PTI1, TCF3, NFATC2IP, OBSL1, MMP15, ESR1, ERBB2, CLCA2, SF3B5, and PHACTR2), along with 3 housekeeping genes (ACTB, RPLP0 and TFRC). All these genes were simultaneously measured in diverse wells. Specifically, the primer pairs of the target genes were placed into PanelChip^®, and then the total ribonucleic acid (RNA) extracted from FFPE tumor tissues was used to perform reverse-transcriptase (RT) quantitative polymerase chain reaction (qPCR) on the PanelStation platform (Quark BioSciences, Inc., Hsinchu, Taiwan). The expression of each gene in the FFPE samples was measured using qPCR. Normalization of gene expression was computed as delta CT = 25 - CT (gene of interest) + CT (mean of housekeeping genes). The RNeasy FFPE Mini Kit (Qiagen, Valencia, CA, USA) was used to extract RNA from the FFPE tissue sections with 5-10 μm in thickness, and Qubit 4.0 fluorometer (Thermo Fisher Scientific, Massachusetts, USA) and NanoDrop Microvolume Spectrophotometer (Thermo Fisher Scientific, Massachusetts, USA) were employed for determining RNA concentration. Subsequently, the extracted RNA was reserved at -80°C until use. Through the RT² First Strand and RT² SYBR Green ROX qPCR MM kits (Qiagen, Valencia, CA, USA), a total of 800 ng RNA was utilized for RT-qPCR. Briefly, 2X SYBR MasterMix, 60×ROX, 100× qPCR Control diluent, Nuclease-Free Water, and cDNA products were used for preparing the qPCR system. The sample loader attached to the PanelStation was selected, then 350 μL heat conductive oil and 80 μL qPCR mixture were loaded to the PanelChip^® and the chip was loaded to PanelStation. Finally, thermal cycling was performed at 95°C for 44 seconds and 60°C for 88 seconds, with 40 cycles in total.

As a previous study described (11), the genetic and clinical factors were combined based on the Cox proportional hazard model to determine their weighting, and concordance statistics (C index) with the receiver operating characteristic curve were used to determine the most effective model. The best cutoff values for predicting local recurrence and distant recurrence were identified by Logistic regression.

Through reviewing the electronic medical record of patients, the clinicopathological information was extracted, containing the age, tumor stage, histological grade, presence or absence of local recurrence, distant recurrence and lymphovascular invasion (LVI), as well as the conditions of adjuvant chemotherapy and adjutant endocrine therapy. All the patients were followed up after surgery through telephones and further consultations.

The RFS rate was regarded as the primary study endpoint, and distant recurrence-free survival (DRFS) and local-regional recurrence-free survival (LRFS) rates were as the secondary endpoints. RFS was defined as the duration of time from surgery until any recurrence of ipsilateral chest, breast, regional lymph node recurrence, distant metastases, or death occurred from any cause. DRFS was defined as the duration of time from surgery until distant recurrence, or death occurred from any cause. LRFS was defined as the duration of time from surgery until any recurrence of ipsilateral chest, breast, regional lymph node recurrence, or death occurred from any cause.

HR and/or HER2 status was assessed based on the guidelines (14). ER/PR was considered positive when there were at least 1% positive tumor nuclei by IHC assay. Patients with grade I-II, ER/PR-positive and HER2-negative breast cancer were classified into IHC luminal A subtype, whereas those with grade III, ER/PR-positive and HER2-negative breast cancer were categorized into IHC luminal B subtype (15, 16).

In this study, t test was used for comparison of continuous variables with normal distribution, manifesting as the mean (standard deviation, SD); Mann-Whitney U rank sum test was employed to compare continuous variables with abnormal distribution, describing as medians with interquartile range. Categorical variables expressing as frequencies and percentages were compared using Fisher’s exact test or χ² test. Survival probabilities were estimated utilizing the Kaplan-Meier method and compared using Log-rank test. The multivariate Cox proportional-hazards model was performed to analyze the influencing factors for recurrence and metastasis. All statistical tests were two-tailed, with statistically significant differences at p values <0.05. Statistical analyses were performed using SPSS 24.0 statistical software (SPSS Inc., Chicago, IL, USA).

In this multi-center study, totally 610 patients were diagnosed as pT1-2N0 HR-positive HER2-negative breast cancer between 2013 and 2016. When 13 cases aged <18 or >75 years, 87 cases lacking follow-up data, 61 cases failing in RecurIndex assay (the proportion of tumor cells ≤30%), 3 cases receiving neoadjuvant chemotherapy and 1 with history of other malignancies were excluded, 445 eligible cases remained for analysis. Their median follow-up duration was 6.25 years.

There were 344 low-risk patients and 101 high-risk patients according to RecurIndex assay. It can be found that statistically significant differences were all shown between low- and high-risk patients, including age (p<0.001), tumor stage (p<0.001), histological grade (p<0.001), LVI (p<0.001), local recurrence (p=0.001), distant recurrence (p=0.002), adjuvant chemotherapy (p<0.001) and endocrine therapy (p<0.049) ( Table 1 ). Through IHC assay, 347 cases were assessed as Luminal A subtype, while 98 cases as Luminal B subtype. Except for the age (p=0.082), the differences were all pronounced between luminal A- and B-type patients regarding all other characteristics (p<0.05; Table S1 ).

Figure 1 shows the 7-year RFS, DRFS and LRFS rates of the low- and high-risk patients stratified by RecurIndex assay and IHC assay. It can be observed that the low-risk patients stratified by RecurIndex assay had higher 7-year RFS rate (92.7% vs. 78.8%, p<0.0001; Figure 1A ), DRFS rate (94.4% vs. 84.5%, p=0.0011; Figure 1B ) and LRFS rate (98.0% vs. 88.0%, p<0.0001; Figure 1C ) compared with the high-risk patients. According to the results of IHC assay, the 7-year RFS rate (91.4% vs. 82.9%, p=0.013; Figure 1D ) and LRFS rate (97.7% vs. 89.2%, p=0.001; Figure 1F ) in patients with luminal A subtype were both superior to those in patients with luminal B subtype, but the difference was indistinctive in the 7-year DRFS rate (93.6% vs. 86.9%, p=0.051; Figure 1E ).

The recurrence index for distant recurrence (RI-DR) scores of the high-risk patients assessed by RecurIndex assay were significantly higher than those of the low-risk patients (p<0.0001; Figure 2A ). Similarly, patients with luminal B subtype had significantly increased RI-DR scores than those with luminal A subtype (p<0.0001; Figure 2B ).

Of 344 low-risk patients, 174 cases received adjuvant chemotherapy, while 170 didn’t. The subgroup analysis based on chemotherapy indicated no statistically significant differences between the low-risk patients with and without adjuvant chemotherapy regarding the 7-year RFS (94.0% vs. 91.7%, p=0.54; Figure 3A ), DRFS (97.0% vs. 92.3%, p=0.21; Figure 3B ), and LRFS (96.9% vs. 98.9%, p=0.59; Figure 3C ) rates. Of 101 high-risk patients, there were 97 cases with adjuvant chemotherapy and 4 without adjuvant chemotherapy. The 7-year RFS and DRFS rates of the high-risk patients with adjuvant chemotherapy were relatively higher than those without, but without significant differences (RFS rate: 78.8% vs. 75.0%, p=0.70; DRFS rate: 84.7% vs. 75.0%, p=0.43).

To further assess the benefits of adjuvant chemotherapy, we used an approximate menopausal status defined by an age cutoff at 50 years to classify the low-risk patients into the group aged over 50 years and the group aged 50 years or younger. It could be observed that there were no significant differences in survival outcomes whether in the group aged 50 years or younger (RFS rate: 90.5% vs. 90.7%, p=0.95, Figure S1A ; DRFS rate: 96.6% vs. 90.7%, p=0.34, Figure S1B ) or the group aged over 50 years (RFS rate: 95.9% vs. 93.3%, p=0.57, Figure S1C ; DRFS rate: 97.3% vs. 94.7%, p=0.63, Figure S1D ). Notably, 330 out of 344 low-risk patients received endocrine therapy. Based on presence or absence of adjuvant chemotherapy, 330 patients were further divided into two groups. The results showed that the differences were inapparent between the patients receiving endocrine therapy with/without chemotherapy regarding RFS (94.3% vs. 91.5%, p=0.41, Figure S1E ) and DRFS (97.5% vs. 92.1%, p=0.13, Figure S1F ) rates.

The risk factors for recurrence and metastasis in all patients were analyzed using a multivariate Cox proportional-hazards model. As shown in Table 2 , the risk of recurrence and metastasis was significantly higher in high-risk patients than in low-risk patients stratified by RecurIndex assay (hazard ratios [HR]: 3.106, 95% confidence interval [CI]: 1.598-6.035, p=0.001); patients with LVI had a higher risk of recurrence and metastasis compared with those without (HR: 2.730, 95%CI: 1.348-5.528, p=0.005).