Introduction

The 21-gene Oncotype DX Breast Recurrence Score® assay is used to guide adjuvant treatment in hormone receptor (HR) + HER2-negative early-stage breast cancer (BC) [1, 2]. The assay measures RNA expression of 16 cancer-related and 5 reference genes using quantitative real-time reverse transcriptase-polymerase chain reaction (qRT-PCR) on tumor tissue samples and calculates the Recurrence Score® (RS; range: 0–100), which is a validated prognosticator and predictor of chemotherapy benefit [1,2,3,4,5]. The cancer-related genes include 4 linked to the estrogen-signaling pathway (ESR1, PGR, BCL2, and SCUBE2), 5 to proliferation/anti-apoptosis (CCNB1, KI67, STK15, SURV, and MYBL2), 2 to the HER2 pathway (ERBB2 and GRB7), 2 to invasion (STMY3 and CTSL2), and 3 (CD68, GSTM1, and BAG1) to macrophage function, detoxification, and apoptosis, respectively. The levels of the cancer-related genes are normalized using the 5 reference genes [3]. The 21-gene assay report provides the RS result and single-gene scores for the estrogen receptor (ESR1), progesterone receptor (PGR), and ERBB2. The other 13 single-gene scores and gene group scores are not provided.

Germline pathogenic variants (PVs) in the BRCA1/2 are associated with a higher risk of developing BC, which is more likely to have aggressive disease characteristics [6, 7]. The prevalence of PVs in BRCA1/2 in BC patients varies between ethnic groups, with the highest prevalence among Ashkenazi Jews (8%) [8].

The 21-gene assay is offered to BC patients irrespective of BRCA1/2 status. In fact, often, the BRCA1/2 status comes to light after the 21-gene testing. Although the RS distribution in BC patients with PVs in BRCA1/2 was shown to be shifted toward higher RS results [9,10,11,12,13,14,15], information on the molecular basis of this shift and the association with clinicopathological characteristics and clinical outcomes is lacking.

We compared RS distribution and expression of single-gene/gene groups between a study group which consisted of estrogen receptor (ER) + HER2-negative BC patients with PVs in BRCA1/2 and the general 21-gene-tested BC patient population as reflected in a commercial use database (CDB) [16] and evaluated the associations between the RS result, single-gene/gene group expression, disease characteristics, and clinical outcomes in the study group.

Materials and methods

Study design

This retrospective cohort study included consecutive female patients with germline PVs in BRCA1/2 and N0/N1mi/N1 ER + HER2-negative BC who underwent 21-gene testing through Clalit Health Services between 2004 and 2015 and received treatment at Rabin Medical Center (RMC) or Hadassah Medical Center (HMC). No exclusion criteria were applied. RS data and single-gene/gene group expression data from the study group were compared to those from the CDB which included 799,986 BC excisional samples [16].

The study was conducted in accordance with the Declaration of Helsinki. It was approved by the institutional review boards of RMC and HMC (approval #0043-14-RMC and #0227-20-HMO) and was granted a waiver for obtaining patient consent due to its retrospective design.

Statistical considerations

Descriptive statistics were used to summarize clinicopathological characteristics and chemotherapy use in the study group. Chi-square and Wilcoxon signed-rank test were used to compare categorical and continuous parameters, respectively, between the study group and the CDB [16]. Fisher’s exact test was used to compare categorical parameters between patients with germline PVs in BRCA1 vs BRCA2.

One sample t test was used to compare the expression of each of the 16 cancer genes and the gene group scores in the study group to the CDB, for all patients, by menopausal status (for CDB, age was used as a surrogate for menopausal status), age, and BRCA-mutated gene. For the comparison to the CDB, group scores were calculated as in Paik et al. [3] without correction for the HER2 and the proliferation group scores. Independent sample t test was used to compare gene expression and gene group scores (calculated as in Paik et al. [3] with the correction as described therein) between patient categories within the study group.

Within the study group, Fisher’s exact test was used to compare patients with and without distant recurrence, with respect to categorical patient/tumor parameters and treatments received. Logistic regression was used to determine the association between the gene group scores (calculated with correction) as continuous parameters and having a distant recurrence.

JMP® Version 16 (SAS Institute Inc., Cary, NC) was used. All tests were 2-sided. p ≤ 0.05 was considered statistically significant.

Results

Patient characteristics

The study group included 81 female patients (all of whom were self-reported women), whereas the CDB included 799,986 BC excisional samples [16]. Baseline patient and tumor characteristics for the study group are presented in Table 1. Age at diagnosis was statistically significantly younger in the study group vs the CDB (median [IQR] 56 [47–61.5] vs 60 (51–67) years; p < 0.001). Also, in the study group, 29.6% were < 50 years at diagnosis vs 20.5% in the CDB (p = 0.030). The nodal status distribution in the study group and the CDB were similar: 80.2, 4.9, and 14.8% had N0, N1mi, and N1 disease, respectively, in the study group vs 84, 5, and 11%, respectively, in the CDB (p = 0.55). In the study group, more patients had PVs in BRCA2 than BRCA1 (59.3% vs. 39.5%). BRCA mutation information was unavailable for one patient.

Table 1 Baseline patient and tumor characteristics of the study group (N = 81)
Full size table

Age at diagnosis was not statistically significantly different between patients with PVs in BRCA1 vs BRCA2 (median [IQR]: 52 (45.5–59) vs 57 (47–63) years, respectively; p = 0.17), whereas grade distribution did. In patients with PVs in BRCA1, grade information was available for 30 patients (11 [36.7%] with grade 1–2, 19 [63.3%] with grade 3). In patients with PVs in BRCA2, grade information was available for 43 (30 [69.8%] with grade 1–2, 13 [30.2%] with grade 3) (p = 0.005).

RS results

The median RS result of the study group was statistically significantly higher than that of the CDB (25 [IQR, 18–35] vs 16 [IQR, 11–22]; p < 0.001). RS distribution also differed, with higher proportion of patients with RS 26–100 in the study group vs the CDB (49.4% vs 16.4%, p < 0.001) (Fig. 1).

Fig. 1
figure 1

Distribution of RS results in the study group vs that in the CDB [16], overall and by menopausal status (for the CDB, age at diagnosis was used as a surrogate for menopausal status; < 50 vs ≥ 50 years). ***p < 0.001 (chi-square test)

Full size image

RS results were also analyzed by BRCA-mutated gene (BRCA1 vs BRCA2). The median (IQR) RS result for the 32 patients with PVs in BRCA1 was 29 (18–37) vs 24 (16–31) for the 48 patients with PVs in BRCA2 (p = 0.18). Both were statistically significantly different than the median RS result of the CDB (16 [IQR, 11–22]) (p < 0.001 each). Among patients with PVs in BRCA1, the number of patients in the RS 0–10, 11–15, 16–20, 21–25, and 26–100 categories was 2 (6.3%), 3 (9.4%), 6 (18.8%), 3 (9.4%), and 18 (56.3%), respectively, whereas among patients with PVs in BRCA2, the respective values were 5 (10.4%), 6 (12.5%), 5 (10.4%), 11 (22.9%), and 21 (43.8%). Both these RS distributions differed significantly from the corresponding distribution in the CDB with a shift toward the high-risk RS category regardless of the BRCA-mutated gene (p < 0.001, each) (Suppl. Figure 1).

The shift toward higher risk in patients with PVs in BRCA1/2 was observed for premenopausal and postmenopausal patients separately (Fig. 1), as well as for younger and older patients separately (< 50, ≥ 50 years) (Suppl. Figure 2). Menopausal status information was not available for the CDB and a cut-off value of 50 years at diagnosis was used as a surrogate.

Single-gene expression and gene group scores overall and by BRCA-mutated gene

The expression of 12 of the 16 cancer genes differed significantly between the study group and the CDB. In all, the directionality of the difference contributed to higher RS results (lower expression in study group patients vs the CDB in PGR, SCUBE2, GSTM1, and BAG1; higher expression in CCNB1, KI67, STK15, SURV, MYBL2, GRB7, and CTSL2). Gene group scores for the ER, proliferation, and invasion gene groups, but not the HER2 gene group, differed significantly between the study group and the CDB, with the directionality of these differences contributing to higher RS results (lower ER gene group score and higher proliferation and invasion gene group scores in study group patients vs the CDB) (Table 2).

Table 2 Single-gene expression and gene group scores vs the commercial use database [16]: Overall and by BRCA mutation type
Full size table

Single-gene expression/gene group scores were compared between study group patients with PVs in BRCA1 and BRCA2 (Table 2). The only statistically significant difference between the BRCA mutation subgroups involved the ESR1 gene, where the expression was higher in patients with PVs in BRCA2 (mean [SD], 10.01 [1.22] vs 9.37 [1.54]; p = 0.041) and the MYBL2 gene, where the expression was higher in patients with PVs in BRCA1 (mean [SD], 5.83 [0.81] vs 5.37 [1.03]; p = 0.037).

Comparing study group patients with PVs in BRCA1 to the CDB revealed statistically significant differences in single-gene expression in 12 of the 16 cancer genes and in 2 gene group scores, all in a direction that contributed to higher RS results in the study patients (lower expression in study group patients vs the CDB in ESR1, PGR, SCUBE2, GSTM1, BAG1, and the ER gene group score; higher expression in CCNB1, KI67, STK15, SURV, MYBL2, CTSL2, and CDK68, and the proliferation gene group score). A similar analysis for patients with PVs in BRCA2 revealed statistically significant differences in gene expression in 10 of the 16 cancer genes, and in 3 gene group scores, all in a direction that contributed to higher RS results in the study patients (lower expression in study group patients vs the CDB in PGR, GSTM1, BAG1, and the ER gene group score; higher expression in CCNB1, KI67, STK15, SURV, MYBL2, GRB7, and CTSL2, and the HER2 and the proliferation gene group scores) (Table 2).

Single-gene/gene groups analyses by menopausal status and age

In the study group, the only statistically significant difference between pre- and postmenopausal patients was the expression of PGR, which was higher in premenopausal vs postmenopausal patients (mean [SD], 6.98 [1.45] vs 5.86 [1.72], p = 0.0046). Although menopausal status information was not available for the CDB, a similar finding was noted in the CDB upon using age as a surrogate (mean [SD] for CDB patients < 50 and ≥ 50 years: 7.77 [1.47] vs 7.16 [1.83], p < 0.001) (Table 3).

Table 3 Single-gene expression and gene group scores by menopausal status, in the study group vs the commercial use database [16]
Full size table

Single-gene expression/gene group scores analysis by menopausal status (using age as a surrogate for the CDB) demonstrated that the differences between the study group and the CDB observed for the entire cohort were mostly retained (except for SCUBE2 in premenopausal patients, GRB7 in both pre- and postmenopausal patients, the ER group score in premenopausal patients, and the invasion group score in postmenopausal patients). For the proliferation gene group, the differences in each single-gene and the gene group score remained highly significant in both pre- and postmenopausal patients (Table 3).

Analysis of single-gene/gene group score by age within the study group and in comparison to the corresponding age groups in the CDB, yielded similar results to those found when the study group was stratified by menopausal status (Suppl. Table 1).

Single-gene expression by other patient/disease characteristics

Single-gene expression was similar in N0 patients vs N1mi/N1 patients, except for differences in the BAG1 gene whose expression was higher in N1mi/N1 patients (mean [SD], 8.48 [0.65] vs 8.13 [0.59], p = 0.037), and the PGR gene whose expression was also higher in N1mi/N1 patients (mean [SD], 7.10 [1.33] vs 6.04 [1.73], p = 0.025; Suppl. Table 2). In both genes, the directionality of the differences contributed to higher RS results in patients with N0 disease.

Expression of 10 genes differed significantly between patients with grade 1–2 tumors vs patients with grade 3 tumors. These included all the genes in the ER group whose expression was higher in grade 1–2 vs grade 3 tumors. Consequently, the ER gene group score was also significantly higher in grade 1–2 tumors. Additionally, the GSTM1, BAG1, and ERBB2 genes had significantly higher expression in grade 1–2 tumors and 3 genes in the proliferation group (STK15, SURV, and MYBL2) had higher expression in grade 3 tumors. Except for the ERBB2 gene, the directionality of the differences contributed to higher RS results in patients with grade 3 tumors (Suppl. Table 2).

Treatments and clinical outcomes in the study group

Overall, 37 (45.7%) patients in the study group received adjuvant chemotherapy. Of the 32 patients with PVs in BRCA1, 18 (56.3%) received chemotherapy including 1/14 (7.1%) with RS 0–25, and 17/18 (94.4%) with RS 26–100 (for one patient in the RS 26–100 group, treatment information was unavailable). Of the 48 patients with PVs in BRCA2, 19 (39.6%) received chemotherapy including 5/27 (18.5%) with RS 0–25 and 14/21 (66.7%) with RS 26–100.

With a median follow-up of 8.2 (IQR, 5.6–9.7) years from diagnosis, one patient with PVs in BRCA2 experienced BC recurrence in the same breast (3.8 years after the initial diagnosis), three had contralateral BC (1.3–6.5 years after their initial BC diagnosis), and nine experienced distant recurrence (of whom 1 had also the aforementioned recurrence in the same breast). Of these nine patients, 1 had PVs in BRCA1 and 8 in BRCA2. The median RS result of these 9 patients was 25 (range, 16–41) and 4 received adjuvant chemotherapy. The distant recurrence occurred 31–130 months from the BC diagnosis (Table 4).

Table 4 Baseline patient/disease characteristics, treatments, and clinical outcomes of distant recurrence cases in the study group
Full size table

No statistically significant differences between the nine patients with distant recurrence and the 72 non-recurring patients were observed in terms of patient/disease characteristics (age, menopausal status, tumor grade, nodal status, RS category, and BRCA mutation type) and treatment with chemotherapy (Suppl. Table 3). A trend toward significance was observed with respect to BRCA mutation type, with recurrences among 1/32 [3.1%] patients with PVs in BRCA1 vs 8/48 [16.7%] of patients with PVs in BRCA2, p = 0.078). The proliferation and invasion gene group scores were significantly associated with the odds of having distant recurrence (proliferation group score: odds ratio [OR], 23.60 [95% CI, 1.4–396.9], p = 0.028; invasion group score: OR, 5.12 [95% CI, 1.13–23.12], p = 0.034). The ER and HER2 gene group scores were not associated with distant recurrence (Suppl. Table 4).

Discussion

This study, which compared RS results, single-gene expression of the 16 cancer genes within the 21-gene assay, and gene group scores between a cohort of ER + BC patients with PVs in BRCA1/2 (N = 81) and all 21-gene-tested BC patients (N = 799,986) demonstrated that those with PVs in BRCA1/2 had higher RS results that could not be attributed to menopausal status or age at diagnosis and stemmed from a distinct gene expression profile of the majority of these 16 cancer genes. Our findings are consistent with prior studies showing higher RS results in patients with PVs in BRCA1/2 [9,10,11,12,13,14,15]; however, this is the first study to compare the single-gene expression of the cancer genes within the assay between patients with PVs in BRCA1/2 and the general 21-gene-tested populations.

This study was also the first to explore single-gene differences between patients with PVs in BRCA1 vs BRCA2. The RS result in patients with PVs in BRCA1 was numerically higher than in BRCA2 patients, although the difference was not statistically significantly different, which is consistent with prior studies [9,10,11, 14]. ESR1 gene expression was significantly lower in patients with PVs in BRCA1 vs BRCA2, which is consistent with a study comparing 20 patients with PVs in BRCA1, 38 patients with PVs in BRCA2, and 1020 controls, where the ER index in those with PVs in BRCA1 but not BRCA2 was statistically significantly lower than that in the controls [9]. This known association between BRCA1 mutation subtype and lower ER expression, may have contributed to the higher chemotherapy use observed in those with PVs in BRCA1 vs BRCA2 (56.3% vs 39.6%).

Analysis of single-gene expression in the study patients by grade revealed significant differences in the expression of 10 genes, all except 1 (ERBB2) in a directionality contributing to higher RS results in grade 3 tumors, which is consistent with the known relationship between RS result and grade [17]. In the study group, single-gene expression was overall similar between pre- and postmenopausal patients as well as between N0 and N1mi/N1 patients. The only gene with differential expression by menopausal status was PGR, whose expression was higher in pre- compared to the postmenopausal study patients. The same result was demonstrated in the CDB (using age as a surrogate), suggesting that the differential PGR expression by age/menopausal status may be unrelated to BRCA status. Interestingly, in the study cohort, N1mi/N1 patients had higher PGR expression compared to N0 patients. Since PGR expression data by nodal status was unavailable for the study cohort, this observation warrants further investigation, particularly as positive nodes are associated with higher clinical risk in BC, whereas high PGR levels are associated with lower clinical risk [18, 19].

Our findings demonstrate that patients with PVs in BRCA1/2 are likely to have higher RS results and therefore suggest that the high-risk RS group is enriched with patients who have PVs in BRCA1/2. This observation should be considered when discussing the RS results with patients. The observed similarity in gene expression in study group patients regardless of age/menopausal status emphasizes the unique gene pattern of this population. There are no data to indicate that the 21-gene assay should be used differently in patients with PVs in BRCA1/2. Our study demonstrated a statistically significant association between the proliferation and invasion gene group scores and having a distant recurrence. There was no correlation with the ER group. Further studies with larger cohorts of patients with PVs in BRCA1/2 are warranted to better define the RS threshold and chemotherapy benefit in these patients.

The strengths of our study include its representation of real-world clinical practice, a long follow-up, and extensive clinical data on each patient in the study group. Another strength involves the robust control dataset of nearly 800,000 samples from various countries worldwide. Notably, although these countries may differ in the assay eligibility criteria, the size of the dataset mitigates against a substantial selection bias effect. Moreover, the consistency in our findings between younger and older patients further supports the absence of such a bias effect. Our study is limited by the sample size of the study group and the small number of events. Also, the CDB includes all 21-gene-tested patients regardless of BRCA1/2 status; however, the proportion of patients with PVs in BRCA1/2 in the CDB is negligible due to the very low prevalence of such mutations [7, 8, 20].

In conclusion, patients with ER + HER2-negative early BC and PVs in BRCA1/2 were characterized by higher RS results that stemmed from a distinct gene expression profile of most genes in the 21-gene assay. Further study is required to explore whether these patients should have a distinct model or RS threshold for considering chemotherapy use.