Prognostic Performance of the Next-Generation Sequencing-Based Multigene Assay in Early Breast Cancer Patients Treated According to the 21-Gene Assay Results
Article information
Abstract
Purpose
Multigene assays guide treatment decisions in early-stage hormone receptor-positive breast cancer. OncoFREE, a next-generation sequencing assay using 179 genes, was developed for this purpose. This study aimed to evaluate the concordance between the Oncotype DX (ODX) recurrence score (RS) and the OncoFREE Decision Index (DI) and to compare their performance.
Materials and Methods
We retrospectively collected tumor blocks from patients who underwent ODX and treatment between 2012 and 2022 at four tertiary hospitals and performed OncoFREE on these samples. Distant metastasis-free survival (DMFS) was compared using RS and DI, with score cut-offs of 25 and 20, respectively.
Results
Among 838 patients, a strong correlation was observed between RS and DI (Pearson correlation coefficient 0.83). At a median follow-up of 54 months, patients with high DI had significantly worse DMFS compared to those with low-DI (log-rank p < 0.001; hazard ratio [HR], 5.73; 95% confidence interval [CI], 1.87 to 17.57; multivariable p=0.048; HR, 3.45; 95% CI, 1.01 to 11.76). In 513 patients aged ≤ 50 years, DMFS was significantly different as a function of DI (p=0.035; HR, 3.98; 95% CI, 1.00 to 15.89) but not RS (p=0.792). Among 376 patients aged ≤ 50 years who avoided chemotherapy based on low-RS, 64 with high DI had worse DMFS (p=0.015; HR, 5.91; 95% CI, 1.17 to 29.78).
Conclusion
OncoFREE showed strong concordance with ODX and effectively identified high-risk patients, particularly in younger individuals. It could be an affordable alternative to ODX for guiding treatment in hormone receptor-positive early breast cancer.
Introduction
Patients with early-stage hormone receptor (HR)–positive, human epidermal growth factor receptor 2 (HER2)–negative breast cancer have a higher chance of remaining disease-free after radical surgery [1]. One of the clinical challenges in managing these patients is identifying subgroups that can avoid unnecessary chemotherapy to prevent complications resulting from overtreatment. With the recent advancements in genomic research, numerous multigene assays have been developed and validated for use in clinical settings to aid in guiding treatment decisions [2]. Consequently, the American Joint Committee on Breast Cancer Staging (8th edition) was revised to incorporate the use of the 21-gene assay recurrence score (RS) in the pathological prognostic stage, and major guidelines recommend multigene assays to decide on adjuvant chemotherapy [3-6].
Oncotype DX (ODX) (Genomic Health [now Exact Sciences]) is a first-generation multigene assay and the most widely used, commercially available test [7-11]. This assay analyzes the expression of 21 genes in formalin-fixed, paraffin-embedded (FFPE) tissues by a quantitative reverse transcription–polymerase chain reaction based approach, providing an RS ranging from 0 to 100. Importantly, ODX is the only multigene panel that has been validated as both a prognostic and predictive tool for estimating the risk of distant recurrence and responsiveness to chemotherapy in HR-positive/HER2-negative breast cancer [12-14]. However, ODX analysis does not fully reflect the disease characteristics in younger women, and its substantial cost imposes a significant financial burden [15-16].
OncoFREE (DCGen) is a newly developed multigene assay that uses a next-generation sequencing (NGS) platform [17]. By analyzing 179 genes, this assay quantifies the prognostic risk of distant metastasis (DM) according to the Decision Index (DI) and aids in determining the need for adjuvant chemotherapy with a cut-off of 20. In the initial validation study, OncoFREE DI demonstrated an area under the receiver operating characteristic curve (AUC) of 0.76 in predicting DM, with 10-year DM-free survival (DMFS) rates of 93.2% and 64.4% for the low- and high-risk groups, respectively [17].
In this study, we aimed to investigate the concordance between ODX RS and OncoFREE DI using multicenter FFPE samples, with an ultimate goal of demonstrating the clinical utility of OncoFREE for predicting the risk of DM in HR-positive, HER2-negative breast cancer.
Materials and Methods
1. Patients
The medical records of consecutive female patients who underwent breast cancer surgery and commercial ODX testing for HR-positive breast cancer between January 1, 2012, and December 31, 2022, at four tertiary hospitals in the Republic of Korea, were retrospectively reviewed. Patients with HR-negative or HER2-positive breast cancer were excluded. Moreover, male patients, as well as those with recurrent, synchronous, or metachronous breast cancer patients were also excluded. The decision to implement adjuvant chemotherapy was made after discussions with the patients based on the results of the ODX RS; patients were treated with standard therapeutic approaches.
2. Pathological assessment
Estrogen or progesterone receptor positivity was determined by staining in at least 1% of cells or an Allred score greater than 2, as identified by immunohistochemistry (IHC). HER2 status was determined according to the guidelines of the American Society of Clinical Oncology and the College of American Pathologists [18]. In case the IHC results were equivocal, fluorescence in situ hybridization or silver-enhanced in situ hybridization was performed, and the HER2 status was considered as positive if HER2/chromosome enumeration probe 17 ratio exceeded 2.0.
3. Multigene assay and risk categorization
FFPE blocks were centrally collected from all patients and tested with OncoFREE. Following the extraction of RNA from FFPE samples, the NGS-based multigene assay was conducted using the OncoFREE kit, as previously described (S1 and S2 Tables) [17]. Patients were categorized as high-and low-risk groups based on a DI of > 20 and ≤ 20, respectively. Patients whose FFPE tumor blocks were unsuitable for OncoFREE, either due to the absence of invasive tumor or insufficient number of FFPE sections for analysis, were excluded.
ODX RSs were retrospectively reviewed from the provided medical records. Clinical risk (CR) was stratified based on Adjuvant algorithms employed in the MINDACT (Microarray in Node-Negative Disease May Avoid Chemotherapy) trial [8]. The risk of integrated RS and CR was categorized based on an analysis from the TAILORx (Trial Assigning Individualized Options for Treatment) trial [11,19]. Women aged 50 years or younger, with an RS ≤ 15 or RS 16-20, with a low CR, were classified as ‘low integrated risk,’ while those having an RS ≥ 21 or RS 16-20 with a high CR were classified as ‘high integrated risk.’ For women over the age of 50, risk classification was based on an RS threshold of 25, regardless of CR.
4. Definition of recurrence
The primary endpoint of our study was the identification of DM, which was defined as any recurrence beyond the locoregional recurrence, including metastasis to bone, lung, liver, or distant lymph nodes. DMFS was calculated from the date of surgery to the date of histological confirmation of metastasis or the last follow-up. In cases where biopsy of the metastatic lesion was not feasible, DMFS was defined as the time from the date of clinical diagnosis of metastasis that led to the initiation of palliative treatments.
5. Statistical analyses
Statistical analyses were conducted at the Medical Research Collaborating Center at Seoul National University Biomedical Research Institute. Categorical variables were compared using Pearson’s chi-square test, and continuous variables were compared using Kruskal-Wallis tests. Pearson correlation coefficients and scatter plots were used to assess the correlation between ODX RS and OncoFREE DI, and the best-fit line was drawn using a simple linear regression method. The log-rank test was used to compare the survival curves derived from the Kaplan-Meier method. Cox proportional hazards regression analysis was performed to adjust for relevant clinicopathological variables and to estimate hazard ratios. Variables with a two-sided p-value of < 0.05 in the univariate analysis were included in the multivariate analysis, while those with a variance inflation factor greater than 4.0 in the multicollinearity test were ultimately excluded. Uno’s concordance index was used to quantify the performance of the prediction models. The time-dependent receiver operating characteristic curve and AUC were used to compare the ability of the two tests to predict DM 5 years after surgery. We employed a bootstrap method with 1,000 repetitions to estimate the 95% confidence intervals for the AUC. Missing data were treated using a complete case analysis approach. Statistical significance was set at p < 0.05. All analyses were conducted using SPSS ver. 29.0 (IBM Corp.), and R Statistical software ver. 3.6.3 (R Foundation for Statistical Computing). Figures were derived using GraphPad Prism ver. 10.0 (GraphPad Software).
Results
1. Patient demographics
Among 956 tumor blocks, 104 were excluded from analysis due to insufficient tumor tissue. Of the remaining samples, 14 (1.6%) did not meet the quality control criteria of OncoFREE, primarily due to insufficient RNA concentration or degradation. Consequently, 838 patients who met the inclusion criteria were analyzed in the study. The median age at the time of diagnosis was 55.0 years (interquartile range [IQR], 43.0 to 55.0). The median tumor size was 1.8 cm (IQR, 1.3 to 2.4), and 102 patients (12.2%) had axillary lymph node metastasis. Overall, 835 patients (99.6%) were administered adjuvant hormone treatment with or without ovarian function suppression, while three patients were lost to follow-up after chemotherapy. Detailed clinicopathologic features are shown in Table 1 and inter-institutional comparisons are described in S3 Table.
Demographic and clinicopathological characteristics of patients
The median ODX RS and OncoFREE DI were 17.0 (IQR, 12.0 to 23.0) and 17.9 (IQR, 14.6 to 23.2), respectively (Table 1). Based on the risk categorization, 164 (19.6%) and 322 (38.4%) patients were classified as the high-risk group according to the ODX RS and OncoFREE DI, respectively. Among the high-RS group, 152 patients (92.7%) received chemotherapy, while 616 patients (91.4%) in the low-RS group did not.
2. Concordance between ODX RS and OncoFREE DI
Among the 674 patients with an ODX RS ≤ 25, 507 (75.2%) were classified as ‘low risk’ according to the OncoFREE DI, while 155 of the 164 patients (94.5%) with a RS > 25, were identified as ‘high-risk’. The concordance rate between the two risk classifications showed moderate consistency (78.0%; 95% confidence interval [CI], 76.0 to 81.6; kappa, 0.45), and was higher when the CR was integrated into the RS classification (79.1%; 95% CI, 77.0 to 82.3; kappa, 0.57).
Next, we compared the ODX RS and OncoFREE DI. The two tests demonstrated a strongly positive correlation, with a Pearson correlation coefficient (r) of 0.833 (95% CI, 0.811 to 0.853; p < 0.001) (Fig. 1A). Subgroup analysis according to age, revealed that the concordance rate was lower in patients aged 50 years or younger (r=0.822; 95% CI, 0.792 to 0.848; p < 0.001) compared to that of individuals older than 50 years of age (r=0.849; 95% CI, 0.816 to 0.877), although the difference was not statistically significant (Fig. 1B and C).
The correlation between Oncotype DX recurrence score (RS) and OncoFREE Decision Index (DI) among all patients (A), patients aged > 50 years old (B), and patients aged ≤ 50 years old (C). The Pearson correlation coefficient was used to determine the relationship between OncoFREE and Oncotype DX RS. The x-axis provided the OncoFREE DI, while the y-axis shows the Oncotype DX RS. CI, confidence interval.
3. Survival outcomes according to the results of the multigene assays
During the median follow-up period of 54.0 months (IQR, 36.0 to 70.0), the 5- and 10-year DMFS rates were 97.5% and 95.9%, respectively. Patients with high-RS showed significantly poorer DMFS than those with low-RS (log-rank p=0.005; hazard ratio [HR], 3.63; 95% CI, 1.40 to 9.40) (Fig. 2A). Similarly, patients with a high-DI demonstrated significantly shorter DMFS than those with a low-DI (log-rank p < 0.001; HR, 5.73; 95% CI, 1.87 to 17.57) (Fig. 2B). The c-index of the OncoFREE approach (0.72; 95% CI, 0.63 to 0.82) was higher than that of the ODX assay (0.65; 95% CI, 0.52 to 0.77).
Kaplan-Meier curves representing distant metastasis-free survival according to Oncotype DX (ODX) recurrence score (RS) (A) and OncoFREE Decision Index (DI) (B). CI, confidence interval; HR, hazard ratio.
Among all clinicopathologic variables, tumor size, histologic grade, progesterone status, and lymphovascular invasion were determined to be relevant to DMFS (Table 2). After adjusting for said variables, OncoFREE DI remained to be significantly associated with DMFS with good prediction performance (model 2: adjusted HR, 3.45 [95% CI, 1.01 to 11.76]; p=0.048; Uno’s c-index, 0.82). Integrated ODX RS and CR also remained significantly associated with the prognosis (model 1: adjusted HR, 4.64 [95% CI, 1.45 to 14.86]; p=0.010; Uno’s c-index, 0.74).
Log-rank and Cox-regression analyses for distant metastasis
Furthermore, we calculated the time-dependent AUC to assess the clinical performance of the two tests in predicting DM. Five years post-surgery the AUC of the ODX and OncoFREE groups was 0.73 (95% CI, 0.58 to 0.90) and 0.74 (95% CI, 0.56 to 0.90), respectively (Fig. 3A). Detailed sensitivity, specificity, positive predictive value, and negative predictive value of the two tests are summarized in Table 3. The AUC of the integrated ODX RS and CR was 0.69 (95% CI, 0.55 to 0.80) (Fig. 3B).
The time-dependent receiver operating characteristic curves of the two multigene assays for predicting distant metastasis at 5 years post-surgery among all patients (A, B) and patients aged ≤ 50 years old (C, D). AUC, area under the receiver operating characteristic curve; CR, clinical risk; RS, recurrence score.
Diagnostic performance of the OncoFREE and Oncotype DX at 5 years of surgery
4. Assessments in patients aged 50 and younger
Regarding the lower concordance rate between the two tests in younger patients, we performed a subgroup analysis for patients 50 and younger versus those above 50 years of age. Patients with high-RS demonstrated comparable DMFS to those with a low-RS (log-rank p=0.792), while an integrated CR and RS risk significantly stratified the patients into two distinct survival curves (log-rank p=0.022; HR, 5.17; 95% CI, 1.07 to 24.89) (Fig. 4A and B). Notably, patients with a low-OncoFREE DI demonstrated significantly better DMFS rates, regardless of their CR (log-rank p=0.035; HR, 3.98; 95% CI, 1.00 to 15.89) (Fig. 4C). Furthermore, the time-dependent AUC at 5 years post-surgery revealed that the AUC of OncoFREE (0.71; 95% CI, 0.51 to 0.89) indicated a slightly better performance than that of ODX (0.62; 95% CI, 0.39 to 0.89) (Fig. 3C). Moreover, the AUC of integrated ODX RS and CR was higher than that of ODX alone (0.70; 95% CI, 0.51 to 0.85) (Fig. 3D). However, this difference was not statistically significant.
Kaplan-Meier curves for distant metastasis-free survival among patients aged 50 years old or younger, according to Oncotype DX (ODX) recurrence score (RS) (A), integrated RS and clinical risk (CR) (B), and OncoFREE Decision Index (DI) (C). CI, confidence interval; HR, hazard ratio.
5. Patients with intermediate risk of ODX RS 16-25
We conducted subgroup analysis for 318 patients with ODX RS between 16 and 25. Among them, 178 patients (56.0%) had an OncoFREE DI of ≤ 20, and showed a better DMFS than those with high OncoFREE DI (log-rank p=0.042; HR, 6.86; 95% CI, 0.80 to 58.80) (S4A Fig.). Furthermore, among 188 patients who are 50 years old or younger, OncoFREE DI still significantly categorized the patients with intermediate ODX RS into two distinct groups according to the DMFS (log-rank p=0.045; HR, 6.94; 95% CI, 0.77 to 62.38) (S4B Fig.). The integrated ODX RS and CR in the 318 patients categorized 200 patients into the low-risk group that showed a better DMFS than the high-risk group. (log-rank p=0.034, HR, 7.36; 95% CI, 0.85 to 63.65) (S4C Fig.).
6. Patients who avoided chemotherapy due to low-ODX RS
Among 616 patients who avoided chemotherapy for low-RS, 130 patients (21.1%) showed a high-OncoFREE DI and had a tendency of poorer DMFS than the other 486 patients with low-DI (log-rank p=0.095; HR, 3.35; 95% CI, 0.74 to 15.08) (Fig. 5A). Notably, among 376 patients (≤ 50 years old) who avoided chemotherapy based on low-RS, 64 with high DI had worse DMFS (p=0.015; HR, 5.91; 95% CI, 1.17 to 29.78). (Fig. 5B). However, among 301 young patients with a low integrated RS and CR, there was no significant difference in DMFS according to OncoFREE DI (p=0.621) (Fig. 5C).
Kaplan-Meier curves for distant metastasis-free survival according to the OncoFREE Decision Index (DI) among patients who did not require chemotherapy due to low Oncotype DX risk: all ages (A), patients aged 50 years old or younger (B, C). CI, confidence interval; CR, clinical risk; HR, hazard ratio; NC, not calculable; RS, recurrence score.
Discussion
ODX is the most widely used multigene assay for predicting the prognosis of HR-positive early-stage breast cancer. In this study, OncoFREE DI and ODX RS demonstrated high concordance, presumably because OncoFREE was initially developed based on the ODX model. Moreover, OncoFREE effectively stratified patients into two distinct groups according to their risk of DM and showed a comparable but higher AUC than ODX, especially for younger women. Lastly, OncoFREE efficiently identified the patients who might benefit from chemotherapy, even among those who avoided chemotherapy due to a low-RS.
First introduced in 2004, ODX effectively identified patients with a high risk of recurrence [13]. The TAILORx trial prospectively enrolled 10,273 patients with HR-positive, node-negative breast cancer and revealed that patients aged greater than 50 with an RS of 0-25 did not benefit from chemotherapy, while patients 50 and younger with an RS of 16-25 showed a prolonged survival after chemotherapy [11,20]. Following the TAILORx trial, Sparano et al. [19] reported that integrated CR and ODX RS could more effectively predict the risk of DM for younger patients with an RS of 16-25. The RxPONDER trial, which analyzed patients with nodepositive (N1) breast cancer and a RS of 0-25, showed that the omission of chemotherapy did not significantly affect the recurrence rates in postmenopausal women, whereas premenopausal women benefited from chemotherapy [21]. Although current guidelines recommend the use of ODX in HR-positive breast cancer, several limitations exist [3-5]. The TAILORx trial included a relatively small number of younger patients and was also expensive, leading to a significant financial burden for some patients, especially those in developing countries.
The NGS-based multigene assay, OncoFREE, was recently developed in an effort to overcome commonly experienced issues [17]. RNA-sequencing technology offers higher accuracy, sensitivity, and reproducibility than conventional PCRbased tests [22-24]. RNA-sequencing drives stable gene expression through the in silico pipeline for normalization among samples and genes, enabling decentralization and widespread use. Additionally, RNA-sequencing–based tests would be more advantageous for in vitro diagnostics due to the inconsistency of housekeeping genes [25,26]. RNA-sequencing approaches are also cost-effective as they can be combined with many advanced reagents and processed in large quantities for multiple genes. A previous study demonstrated an AUC of 0.76 for predicting DM using the OncoFREE assay during a median follow-up period of 141 months [17]. Patients with an OncoFREE DI greater than 20 had a significantly poorer DMFS than those with lower scores (HR, 5.86; 95% CI, 3.62 to 9.49; p < 0.001), and this significance remained consistent for younger patients. The key strength of the OncoFREE test is that 63.0% of the patients included in the previous study were 50 years old or younger, representing a substantially larger proportion than in the TAILORx trial [11]. Moreover, the cost is approximately half that of ODX, offering a relatively affordable option for patients with low socioeconomic status.
The small number of younger patients included in the TAILORx led to the underrepresentation of this population. The benefits of chemotherapy in patients aged 50 or younger with an RS of 16-25 may have been induced by ovarian suppression treatments [11]. Given that biological features such as a higher proliferation rate and grades, more frequent genetic mutations, and greater resistance to endocrine treatment differ between younger and older individuals with breast cancer, concerns about the clinical applicability of the ODX to younger patients exist [27-29]. The greater tendency of AUC of OncoFREE compared to ODX among patients aged 50 or younger in our study suggests the potential suitability of OncoFREE for this cohort while further validation is warranted.
Notably, OncoFREE itself identified the patients who could benefit from chemotherapy without regarding the clinical risk, even among those with intermediate ODX RS of 16-25. The CR was first integrated with RS for patients aged 50 or younger in the TAILORx trial due to the absolute chemotherapy benefit shown in younger patients with intermediate ODX RS [19]. Instead, the substantial representation of younger patients during OncoFREE development may have contributed to the assay achieving a more accurate risk prediction without regarding CR. This would explain the concordant cut-off value of 20 for OncoFREE DI and the RS regardless of CR in the integrated RS and CR classification.
An inherent limitation of this study was that all patients were already being treated according to ODX results. Additionally, because 92.7% of high-RS patients received chemotherapy, the DM rate for these patients might have been reduced, resulting in a lower hazard ratio between high- and low-RS patients. It might also affect the results regarding OncoFREE high- vs. low-DI. Considering this limitation, all results in this study should be cautiously interpreted.
Other limitations of the study are as follows: (1) This was a retrospective study, and patients could have been selected with bias. Additionally, patients who chose not to use ODX due to high costs or personal preference were excluded. (2) We did not evaluate metrics of cost-effectiveness for the two tests. (3) The relatively short follow-up period requires cautious interpretation of our results, considering that HR-positive breast cancer shows a higher incidence of late recurrence [30]. However, since the benefits of chemotherapy for HR-positive breast cancer are mainly observed during the early follow-up period, the likelihood that this significantly affected our results is low [1].
In conclusion, ODX RS and OncoFREE DI demonstrated a high concordance rate. OncoFREE effectively identified patients at high risk of developing DM, especially among younger women. OncoFREE would be an affordable alternative to ODX in HR-positive early-stage breast cancer.
Electronic Supplementary Material
Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).
Notes
Ethical Statement
All procedures were conducted following the standards set by the Declaration of Helsinki, and informed consent was obtained from all patients. All experiments were approved by the institutional review board (IRB) of each institution (IRB numbers: Seoul National University Hospital, H-2105-129-1220; Samsung Medical Center, 2023-10-034; Asan Medical Center, 2023-0311; Kyungpook National University Hospital: 2023-09-027).
Author Contributions
Conceived and designed the analysis: Chung W, Han W, Lee HB, Lee SB, Ryu JM.
Collected the data: Kang E, Lee J, Koh J, Lee HJ, Park WK, Lee HB, Lee SB, Ryu JM.
Contributed data or analysis tools: Kang E, Lee J, Koh J, Lee H, Park JY, Lee HJ, Kang B, Park WK, Lee SB, Ryu JM.
Performed the analysis: Kang E, Cheun JH, Lee H, Park JY, Kang B, Son J, Kim B, Chung W, Lee SB, Ryu JM.
Wrote the paper: Kang E, Cheun JH,
Supervise: Lee SB, Ryu JM.
Conflicts of Interest
H.J. Lee is listed as a co-inventor of patents for the NGS-based assay used in this study, which is owned by DCGen Co. Ltd., from which royalties are paid. J. Son and B.Kim are employees of DCGen, Co., Ltd. W. Chung, W. Han and H.-B. Lee report being members of the board of directors and holding stock and ownership interests in DCGen Co. Ltd. They are also listed as co-inventors on patents for the NGS-based assay in this study, owned by DCGen Co. Ltd., from which royalties are paid. S.B. Lee is listed as a co-inventor of patents for the NGS-based assay in this study, owned by DCGen Co. Ltd., from which royalties are paid. The other authors have declared that no conflict of interest exists.
Funding
This work was supported by the National Research Foundation of Korea (NRF) grant supported funded by the Korea government (MIST) (NRF-2022K1A3A1A39090208) to HBL and EK.
Acknowledgments
Yoon-Hee Choi, PhD, Seoul National University Hospital, and the Medical Research Collaborating Center at Seoul National University Hospital provided statistical assistance. Fees were paid to the Medical Research Collaborating Center.
EK has full access to the data reported in this study and takes full responsibility for the integrity of the data and the accuracy of the analysis. Furthermore, summarized statistical data will be available from the EK (rkd4327@naver.com) upon reasonable request after approval of a proposal.