Background
Breast cancer is the most commonly diagnosed cancer worldwide, and hormone receptor (HR)–positive, node-negative disease accounts for approximately half of all cases of breast cancer in women.1 The recurrence score (RS) based on a 21-gene genomic assay or Oncotype DX was originally established to estimate the likelihood of distant recurrence in estrogen receptor–positive, node-negative breast cancer,2 which is currently an interest of stakeholders in the health care system.3 The TAILORx trial confirmed that RS was prognostic of distant recurrence4 and found no chemotherapy benefit overall in women with intermediate risk score (midrange; 11–25) in HR-positive, HER2-negative, node-negative breast cancer.5 A low risk score (low-range; 0–10) indicated endocrine therapy alone, and a high risk score (high-range; 26–100) indicated chemoendocrine therapy. In women aged ≤50 years, there was chemotherapy benefit from the higher end (16–25) of midrange scores with respect to the risk of invasive disease–free survival (DFS) and distant recurrence–free interval (dRFI). However, it remains unclear whether RS is prognostic of breast cancer–specific survival (BCSS) and overall survival (OS) after controlling for confounders, and whether it has more prognostic weight for disease-specific survival and recurrence than OS in the TAILORx trial population during a follow-up of up to 11.6 years.6
The prospectively collected data from TAILORx trial are of high quality and provide opportunities to address critical questions, such as a biomarker’s performance in predicting BCSS and OS relative to recurrence outcomes in early-stage breast cancer. The current study was designed to evaluate the association of RS with OS, BCSS, DFS, and RFI, adjusting for clinicopathologic measures. We also compared the prognostic information provided by RS to OS, BCSS, DFS, RFI, and dRFI (as reference) in the landmark TAILORx trial population.
Methods
TAILORx Clinical Trial
The TAILORx study (ClinicalTrials.gov identifier: NCT00310180) is a biomarker-driven, phase III, multicenter, randomized trial, with international participating sites, which was sponsored by the National Cancer Institute and coordinated by the ECOG-ACRIN Cancer Research Group in the United States.5 The trial was approved by the National Cancer Institute Central Institutional Review Board (IRB) and the local IRBs at participating sites in accordance with assurances filed with and approved by the Department of Health and Human Services in the United States, and patients in the trial provided written informed consent. According to the ethical regulations, IRB approval was not required for the current report because the data were anonymous.
Study Population and Endpoints
The TAILORx trial enrolled 10,273 patients aged 18 to 75 years who had HR-positive, HER2-negative, node-negative breast cancer. Based on RS measured in the primary tumor samples, patients were assigned to 1 of 4 treatment groups. Patients with an RS of ≤10 were assigned to receive endocrine therapy only, while those with a score of ≥26 were assigned to receive chemoendocrine therapy. Patients with a midrange score of 11 to 25 underwent randomization and were assigned to receive either endocrine therapy alone or chemoendocrine therapy.5
A total of 9,719 patients met NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) criteria for recommendation or consideration of adjuvant chemotherapy and received trial treatment,5 among which 8,916 had all clinicopathologic data and treatment information (see Figure S1 in the supplementary materials, available online with this article). The endpoints in this study included BCSS and OS plus DFS, RFI, and dRFI (as reference). The TAILORx trial used the Standardized Definitions for Efficacy End Points (STEEP) criteria for endpoint definitions.7 OS was defined as the length of time from registration to death due to any cause or date last known alive; BCSS was defined as the time from registration to death due to breast cancer or date last known alive; DFS was defined as the time from registration to the first event of recurrence [distant or locoregional], second primary cancer [excluding nonmelanoma skin cancers], or death without evidence of recurrence); RFI was defined as the time from registration to first event of distant or locoregional recurrence or death with recurrence if death was the first manifestation of recurrence); and dRFI was defined as the time from registration to first distant recurrence or death with distant recurrence if death was the first manifestation of distant recurrence.8–10 The prespecified objectives in this study were to evaluate the association of RS with OS, BCSS, DFS, and RFI adjusting for clinicopathologic factors by multivariable analysis and to assess whether RS had more prognostic weight for disease outcomes than for OS.
Statistical Analysis
The primary analysis was to evaluate the association between RS and clinical variables with BCSS, OS, RFI, and DFS, respectively, in multivariable Cox proportional hazards regression models. Hazard ratios (HRs) were estimated using partial log-likelihood analysis of the multivariable Cox regression model, with 95% confidence intervals unsymmetric on the log-ratio scale. Baseline characteristics and median age (range) were compiled with the use of frequency distribution and descriptive statistics. To compare the prognostic information provided by RS regarding BCSS, OS, RFI, DFS, and dRFI, we performed the likelihood ratio (LR) test to quantitatively measure the relative amount of prognostic information.11 The higher the value of the likelihood, the more information the RS provides. Numbers of events and patients in each of the variable categories for all endpoints, and cause of death for OS were analyzed by frequency distribution. All statistical tests were 2-sided, and performed using lifelines version 0.27.1 (Python) and Prism version 9.4.0 (GraphPad).
Results
Patients Characteristics
The 21-gene assay, patient registration, randomization, and patient characteristics in the intention-to-treat (ITT) population according to 3 categories of RS in TAILORx trial has been described previously.2,5 Characteristics of 8,916 patients in this study are shown in Table 1, and the median age was 56 years (range, 23–75 years). Median follow-up time was 95.2 months (range, 0–138.6 months) for BCSS, 95.5 months (range, 0–138.6 months) for OS, 89.6 months (range, 0–134.8 months) for DFS, and 87.7 months (range, 0–134.8 months) for RFI. The event numbers (rate) were 153 (1.7%) for BCSS, 462 (5.2%) for OS, 1,127 (12.6%) for DFS, 502 (5.6%) for RFI, and 352 (3.9%) for dRFI (Table 2).
Baseline Characteristics According to Recurrence Score Categories
Prognostic Information Provided by RS Regarding BCSS, OS, RFI, DFS, and dRFI in the TAILORx Trial
Association of RS With Clinical Endpoints
Compared with low-range scores (RS 0–10), midrange (RS 11–25) and high-range (RS 26–100) scores were associated with higher risk of BCSS (adjusted hazard ratio [aHR], 5.12 [95% CI, 2.09–16.92; P=.002], and 8.03 [95% CI, 2.91–28.47; P=.0002], respectively), independent of tumor size, grade, endocrine therapy types, surgery type, and chemotherapy status as well as demographics (Figure 1 and Supplementary Table S1). The scores were also independently associated with an increased risk of RFI (aHR, 1.68 [95% CI, 1.23–2.36; P=.002], and 3.05 [95% CI, 2.02–4.67; P<.0001], respectively). High-range score was associated with greater risk for DFS (aHR, 1.56 [95% CI, 1.20–2.04; P=.001]) and was not significantly associated with OS (aHR, 1.44 [95% CI, 0.95–2.18; P=.09]) (Figure 2 and Supplementary Table S1). Midrange score was neither associated with OS (aHR, 1.14 [95% CI, 0.87–1.52; P=.35]) nor significantly associated with DFS (aHR, 1.15 [95% CI, 0.96–1.38; P=.13]).
Recurrence scores and clinical factors for (A) BCSS and (B) RFI prognosis. Horizontal lines represent 95% CI. Eleven cases treated with “other” were excluded from the endocrine therapy types for outcome analysis.
Abbreviations: AI, aromatase inhibitor; BCSS, breast cancer–specific survival; HR, hazard ratio; OFS, ovarian function suppression; PR, progesterone receptor; RFI, recurrence-free interval; RS, recurrence score; Tam, tamoxifen.
Citation: Journal of the National Comprehensive Cancer Network 22, 6; 10.6004/jnccn.2024.7008
Recurrence scores and clinical factors for (A) OS and (B) DFS prognosis. Horizontal lines represent 95% CI. Eleven cases treated with “other” were excluded from the endocrine therapy types for outcome analysis.
Abbreviations: AI, aromatase inhibitor; DFS, invasive disease–free survival; HR, hazard ratio; OFS, ovarian function suppression; OS, overall survival; PR, progesterone receptor; RS, recurrence score; Tam, tamoxifen.
Citation: Journal of the National Comprehensive Cancer Network 22, 6; 10.6004/jnccn.2024.7008
Prognostic Values of RS to Endpoints and Cause of Death
As shown in Table 2, prognostic information provided by RS regarding RFI was greatest (LR-χ2 = 83.0) among the 5 clinical endpoints. The information provided by RS regarding OS was the least (17.5) and second to last for DFS (33.6). Notably, LR-χ2 value for BCSS was 65.1, which was much higher than that for OS. Of 462 (or 499 in the ITT population) deaths, the causes of death included 24.5% (23.6%) other cancers, 25.7% (25.3%) other diseases, and 16.7% (16.2%) unknowns in addition to 33.1% (34.9%) breast cancer (Figure 3).
Cause of death in 8,916 patients in the TAILORx trial.
Citation: Journal of the National Comprehensive Cancer Network 22, 6; 10.6004/jnccn.2024.7008
Association of Clinicopathologic Measures With Endpoints
Age ≥50 years was associated with higher risk for OS events (aHR, 1.45 [95% CI, 1.03–2.05; P=.03]) (Figure 2 and Supplementary Table S1), but a lower risk for RFI (aHR, 0.68 [95% CI, 0.52–0.89; P=.005]) (Figure 1 and Supplementary Table S1). Postmenopausal status was also associated with greater risk for OS events (aHR, 1.45 [95% CI, 1.05–2.03; P=.03]) but not significantly associated with RFI (aHR, 1.10 [95% CI, 0.84–1.45; P=.50]). Both age and postmenopausal status were not significantly associated with BCSS (Figure 1). Tumor size significantly provided unfavorable prognoses for all endpoints and so did tumor grades except grade 2 for OS (Figures 1 and 2). As shown in Figure 2, treatment with tamoxifen; aromatase inhibitors (AIs); ovarian function suppression (OFS); tamoxifen + AI; or OFS + AI was significantly associated with favorable OS and DFS. As for BCSS and dRFI, only AI therapy reached statistical significance among 5 endocrine treatment types (Figure 1 and Supplementary Figure S2). Regarding RFI, all therapy types but tamoxifen + AI were significantly associated with better outcome.
Discussion
RS derived from the 21-gene genomic assay is frequently used to estimate the risk of distant recurrence in patients with HR-positive, HER2-negative early-stage breast cancer when making decisions regarding chemotherapy. The independent role of RS in BCSS and OS prognosis has not been estimated comparatively with RFI and dRFI. Few trials and cohort studies had data for BCSS, OS, DFS, RFI, and dRFI that allowed for concurrent analysis of biomarker performance in oncology. With the landmark TAILORx breast cancer trial, we were able to evaluate the role of RS in all 5 clinical endpoints, while controlling for confounders, at up to 11.6 years of follow-up. The data demonstrated that RS was associated with BCSS and RFI independent of clinicopathologic and treatment factors, similar to the significance for distant recurrence (Supplementary Figure S2).2,4,5,10 However, midrange (11–25) and high-range (≥26) scores did not provide independent prognosis for OS in this adjuvant trial population. High-range scores only were associated with DFS. Prognostic information assessed by LR-χ2 test further supports that RS was highly informative regarding distant recurrence and breast cancer recurrence at a locoregional site or distant site, followed by BCSS and DFS, and OS. Therefore, RS provides more prognostic information about disease-specific survival and recurrence than OS and DFS that included breast cancer recurrence, second primary cancer, and death without evidence of recurrence according to the STEEP criteria.7 The constraint for not reaching OS significance, relative to RFI (and dRFI) and BCSS, is likely ascribed to, at least in part, the fact that RS is consisted of cancer-related genes such as proliferation and metastasis.2,3,12
As revealed in Figure 3, 33.1% (34.9% in the ITT population) of deaths were ascribed to breast cancer. This may provide further evidence and explanation that RS was a significant factor in predicting BCSS versus OS, given that approximately two-thirds of deaths were not related to breast cancer. The two-thirds consisted of other cancers and diseases as well as unknown reasons, and DFS included all of the causes of death, and thus was more representative of OS.13 Recently, recurrence-free survival has been shown to be an inadequate surrogate endpoint for OS in colorectal cancer by a meta-analysis of adjuvant clinical trials.14
The TAILORx trial data raise some questions. First, whether one should expect a disease-specific multigene test such as RS to be associated or strongly associated with OS, considering that death caused by second primary cancer, other disease, and unknown cause besides breast cancer are also potential factors. Second, whether there is a rationale to develop a specific multigene algorithm for OS given that it is regarded as the most relevant endpoint in clinical research and patient care.6,7 Additionally, the long-term outcome data gradually demonstrated that patients with high recurrence scores continue to have a poor prognosis despite chemotherapy.2,4,15,16 This suggests there is a need to explore new management strategies for patients with HR-positive, HER2-negative early-stage breast cancer. It has been shown that performance of clinicopathologic and molecular factors was weighted by nature of clinical end points.6,17 In other words, some measures provided more prognostic information for recurrence, such as tumor grade, while others exhibited weight at OS, such as estrogen receptor status after endocrine therapy. Indeed, grade did convey more prognostic information regarding RFI (Figure 1) than OS (Figure 2) in the TAILORx population. Grade in relation to RS for BCSS, OS, RFI, and DFS plus dRFI prognoses exhibited similar HR profiles as shown in Figures 1 and 2. All endocrine treatment regimens were significantly associated with favorable OS and DFS in contrast to BCSS, RFI, and dRFI. Notably, tumor size demonstrated independent significance for all 5 endpoints in patients with HR-positive, HER2-negative, node-negative breast cancer. Chemotherapy was not significantly associated with OS and BCSS, while demonstrating a favorable trend for DFS and RFI prognoses.
According to the American Cancer Society and European Commision,18,19 survival rates in large breast cancer populations decrease as their ages increase. However, younger or premenopausal women have been shown to have a higher risk of local recurrence and odds of contralateral breast cancer.20,21 The TAILORx trial results reported herein confirmed that older age and postmenopausal status were independently associated with higher risk of all-cause death, whereas they were associated with lower risk or insignificance for breast cancer recurrence and distant recurrence. Similar results were reported in breast cancer cohort studies.6,22 Therefore, age and menopausal status manifest a divergence for survival and recurrence outcomes in early-stage breast cancer. In addition, OFS plus AI therapy demonstrated DFS and RFI advantages23 and favorable OS for premenopausal patients. Particularly, OFS alone was significantly associated with superior DFS, RFI, and OS in premenopausal patients.
Conclusions
Results of the TAILORx trial reported herein indicate that both midrange (11–25) and high-range (26–100) RS are independent factors for poor BCSS and RFI (and dRFI) prognoses, and that high-range RS only was associated with unfavorable DFS. However, both categories of the scores were not significantly associated with OS in the multivariable Cox proportional hazards regression models, although the high-range scores showed a trend toward poor OS prognosis. RS carries more prognostic information for breast cancer–specific outcomes than OS and/or DFS. Among 462 deaths, approximately one-third were attributed to breast cancer in the TAILORx trial population. DFS is the most representative surrogate endpoint for OS among the 4 surrogate clinical endpoints of RFI, dRFI, BCSS, and DFS.
Acknowledgments
The authors would like to thank Dr. Joseph Sparano for his critical review of the manuscript and suggestion of making the study flowchart.
References
- 1.↑
DeSantis CE, Ma J, Gaudet MM, et al. Breast cancer statistics, 2019. CA Cancer J Clin 2019;69:438–451.
- 2.↑
Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 2004;351:2817–2826.
- 3.↑
Andre F, Ismaila N, Allison KH, et al. Biomarkers for adjuvant endocrine and chemotherapy in early-stage breast cancer: ASCO guideline update. J Clin Oncol 2022;40:1816–1837.
- 4.↑
Sparano JA, Crager MR, Tang G, et al. Development and validation of a tool integrating the 21-gene recurrence score and clinical-pathological features to individualize prognosis and prediction of chemotherapy benefit in early breast cancer. J Clin Oncol 2021;39:557–564.
- 5.↑
Sparano JA, Gray RJ, Makower DF, et al. Adjuvant chemotherapy guided by a 21-gene expression assay in breast cancer. N Engl J Med 2018;379:111–121.
- 6.↑
Nguyen D, Yu J, Reinhold WC, et al. Association of independent prognostic factors and treatment modality with survival and recurrence outcomes in breast cancer. JAMA Netw Open 2020;3:e207213.
- 7.↑
Hudis CA, Barlow WE, Costantino JP, et al. Proposal for standardized definitions for efficacy end points in adjuvant breast cancer trials: the STEEP system. J Clin Oncol 2007;25:2127–2132.
- 8.↑
Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996;15:361–387.
- 9.↑
Sparano JA, Wang M, Martino S, et al. Weekly paclitaxel in the adjuvant treatment of breast cancer. N Engl J Med 2008;358:1663–1671.
- 10.↑
Sparano JA, Gray RJ, Makower DF, et al. Prospective validation of a 21-gene expression assay in breast cancer. N Engl J Med 2015;373:2005–2014.
- 11.↑
Dowsett M, Sestak I, Lopez-Knowles E, et al. Comparison of PAM50 risk of recurrence score with oncotype DX and IHC4 for predicting risk of distant recurrence after endocrine therapy. J Clin Oncol 2013;31:2783–2790.
- 12.↑
Gradishar WJ, Moran MS, Abraham J, et al. NCCN Clinical Practice Guidelines in Oncology: Breast Cancer. Version 2.2024. Accessed March 20, 2024. To view the most recent version, visit https://www.nccn.org
- 13.↑
Untch M, Pérol D, Mayer EL, et al. Disease-free survival (DFS) as a surrogate for overall survival (OS) in patients (pts) with HR+/HER2− early breast cancer (EBC): a correlation analysis. J Clin Oncol 2023;41(Suppl):Abstract 535.
- 14.↑
Ecker BL, Lee J, Saadat LV, et al. Recurrence-free survival versus overall survival as a primary endpoint for studies of resected colorectal liver metastasis: a retrospective study and meta-analysis. Lancet Oncol 2022;23:1332–1342.
- 15.↑
Sparano JA, Gray RJ, Makower DF, et al. Trial assigning individualized options for treatment (TAILORx): an update including 12-year event rates. Cancer Res 2023;83(Suppl 5):Abstract GS1-05.
- 16.↑
Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol 2006;24:3726–3734.
- 17.↑
Lafourcade A, His M, Baglietto L, et al. Factors associated with breast cancer recurrences or mortality and dynamic prediction of death using history of cancer recurrences: the French E3N cohort. BMC Cancer 2018;18:171.
- 18.↑
American Cancer Society. Breast cancer: facts & figures 2022–2024. Accessed October 28, 2023. Available at: https://www.cancer.org/content/dam/cancer-org/research/cancer-facts-and-statistics/breast-cancer-facts-and-figures/2022-2024-breast-cancer-fact-figures-acs.pdf
- 19.↑
European Commission. Long term estimates of cancer incidence and mortality, for all countries. Accessed March 8, 2023. Available at: https://ecis.jrc.ec.europa.eu/explorer.php
- 20.↑
Sparano JA, Gray RJ, Ravdin PM, et al. Clinical and genomic risk to guide the use of adjuvant therapy for breast cancer. N Engl J Med 2019;380:2395–2405.
- 22.↑
Brandt J, Garne JP, Tengrup I, et al. Age at diagnosis in relation to survival following breast cancer: a cohort study. World J Surg Oncol 2015;13:33.
- 23.↑
Francis PA, Pagani O, Fleming GF, et al. Tailoring adjuvant endocrine therapy for premenopausal breast cancer. N Engl J Med 2018;379:122–137.
