If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The risk of BCSM persists for at least 20 years from diagnosis.
•
We developed a tool, ‘ESTIMATE’, for non-metastatic, HR-positive breast cancer.
•
The tool quantifies risks of BCSM, non-BCSM and all-cause mortality.
•
Risk estimates can be quantified at any given time after diagnosis, up to 20 years.
•
The tool can be used to understand competing risks and aid clinical decision-making.
Abstract
Purpose
The risk of breast cancer-specific mortality (BCSM) persists for at least 20 years from diagnosis. Estimating the risk of BCSM over this extended period along with competing risks of death would aid clinical decision-making. We aimed to develop an interactive tool called ‘ESTIMATE’, to explore the Surveillance, Epidemiology, and End Results (SEER) registry to quantify residual risks of BCSM, non-BCSM and all-cause mortality in non-metastatic, hormone receptor (HR)-positive breast cancer patient subgroups at any given time after diagnosis, up to 20 years.
Methods
Using SEER data, we included 264,237 women with invasive, non-metastatic, HR-positive breast cancer diagnosed from 1990 to 2006. We developed a tool that provides a nonparametric estimate of the residual cumulative risk of BCSM and non-BCSM by year 20 after any specified time from initial diagnosis, among patients defined by baseline clinical and pathologic variables, using Gray’s subdistribution method.
Results
ESTIMATE allows the user to input patient and tumour characteristics and the preferred timeframe. For example, patients in the age group of 40–49 diagnosed with T1cN1, grade II breast cancer who survived 7 years, have a 14% (95% confidence interval [CI]: 11.9%–16.1%) residual cumulative risk of BCSM in the next 13 years, and a 6.4% (95% CI: 4.7%–8.1%) residual cumulative risk of non-BCSM over the same period.
Conclusions
ESTIMATE provides population-based risks of BCSM, non-BCSM and all-cause mortality through 20 years after diagnosis of HR-positive breast cancer, based on patient and tumour characteristics. ESTIMATE can inform discussions about prognosis, a balance between competing risks and aid clinical decision-making.
]. Estimating the risk of recurrence over such an extended period has become a topic of significant interest. Several tools can be considered to quantify the risk of recurrences, such as risk calculators based on traditional clinicopathologic factors and a number of gene expression assays [
Prediction of late distant recurrence in patients with oestrogen-receptor-positive breast cancer: a prospective comparison of the breast-cancer index (BCI) assay, 21-gene recurrence score, and IHC4 in the TransATAC study population.
Prediction of late distant recurrence after 5 years of endocrine treatment: a combined analysis of patients from the Austrian breast and colorectal cancer study group 8 and arimidex, tamoxifen alone or in combination randomized trials using the PAM50 risk of recurrence score.
Prognostic impact of the combination of recurrence score and quantitative estrogen receptor expression (ESR1) on predicting late distant recurrence risk in estrogen receptor-positive breast cancer after 5 Years of tamoxifen: results from NRG oncology/national surgical adjuvant breast and bowel Project B-28 and B-14.
Integration of clinical variables for the prediction of late distant recurrence in patients with estrogen receptor-positive breast cancer treated with 5 Years of endocrine therapy: CTS5.
Association of circulating tumor cells with late recurrence of estrogen receptor-positive breast cancer: a secondary analysis of a randomized clinical trial.
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.
]. However, these tools report absolute risks at fixed time points, such as 10 or 15 years, and do not consider the changes in risks over time or the impact of competing risks of death.
We previously reported the ability to estimate the 20-year risk of BC-specific mortality (BCSM), non-BCSM, and all-cause mortality using data from the Surveillance, Epidemiology, and End Results (SEER) program [
]. In contrast, among patients with stage IIIA HR-positive BC, the risks of BCSM and non-BCSM from years 0–20 were 43.2% and 21.1%, respectively, totalling a 64.4% risk of all-cause mortality by 20 years [
]. Although these studies were the first to describe population-based outcomes at such extended intervals, they were unable to estimate dynamic risks, such as the residual risk of death after having survived a given period of time or the risk of death at an earlier timepoint than 20 years post-diagnosis.
The aim of this study was to develop an interactive tool to quantify residual risks of BCSM, non-BCSM and all-cause mortality for women with stage I to III, HR-positive BC, based on specific patient and tumour characteristics and for any period of time from diagnosis up to 20 years. Such a tool would be valuable for clinicians and patients to help assess prognosis from BC, balance the risk of BCSM with competing risks of death, and aid clinical decision-making.
2. Methods
2.1 Data source and study design
Data were obtained data from the National Cancer Institute’s SEER 18 registry (1973–2016) database using SEER∗Stat version 8.3.8. We included women with a pathologically confirmed non-metastatic, invasive BC diagnosed between 1990 and 2006 to have a minimum of 10 years of potential follow-up time. Selection of the study cohort was based on the following variables in SEER: age at diagnosis, race (White, Black, other), year of diagnosis, type of breast surgery, cancer stage, tumour size (T), nodal status (N), estrogen receptor (ER) and progesterone receptor (PR) status, number of lymph nodes examined, number of positive lymph nodes, vital status, and cause of death. Cancer stages, including T and N, were registered according to the American Joint Committee on Cancer staging system, sixth edition, using the information without modifications from the SEER∗Stat variables ‘Breast – Adjusted AJCC 6th Stage (1988–2015)’, ‘Breast – Adjusted AJCC 6th T (1988–2015)’, and ‘Breast – Adjusted AJCC 6th N (1988–2015)’. Patients with any cancer diagnosis prior to the HR-positive BC of interest were excluded (n = 22,135) because the cause of death in this SEER database is attributed to the patient’s first cancer only. Thus, we included only patients whose first lifetime cancer diagnosis was the HR-positive BC of interest, such that cause of death is more accurately attributed; however, patients may have had cancer diagnoses subsequent to the HR-positive BC. Fig. 1 describes the selection of the study cohort. eTable 1 describes the SEER query and variables of interest. Human epidermal growth factor receptor 2 (HER2) status is not included in the data, given that national reporting of HER2 did not start until 2010. Patients positive for ER or PR were considered HR-positive for analysis, and the final sample size was 264,237 patients with HR-positive disease.
Given that only deidentified registry data were used for this study, the study was considered exempt from review at Dana-Farber Cancer Institute’s Institutional Review Board.
2.2 Statistical considerations
We computed population-based results for three separate endpoints: BCSM, non-BCSM, and all-cause mortality. BCSM was the primary endpoint, defined as the interval from initial BC diagnosis to death from BC or to the date last known alive for censored patients. Non-BCSM and all-cause mortality were secondary endpoints, defined as the intervals from the initial diagnosis to death from causes other than BC, or any cause, respectively, or to the date last known alive for censored patients. Estimates of the cumulative incidence functions of BCSM and non-BCSM were obtained by computing the subdistribution hazard functions for death by each cause within subgroups, nonparametrically, as shown by Gray (1988) [
] and implemented in the cmprsk package in R. All-cause mortality was obtained by taking the complement of the Kaplan–Meier survival estimate for the composite outcome of death from any cause.
Mortality estimates were computed at the subgroup level, according to the user-input patient and tumour characteristics, including age at diagnosis, T and N stage, and tumour grade (I, II, III/IV). These characteristics were selected for inclusion based on their strong associations with mortality in HR-positive BC [
]. Age at diagnosis was categorized as <40, 40–49, 50–59, 60–69, 70–79, and ≥80 for analysis to limit the potential for small subgroups among very young and very old patients. Two additional characteristics – the number of years survived since the initial diagnosis and the number of years at which to estimate – were used to further define a given subgroup and allow for the extraction of residual mortality risks, respectively. These two inputs are synergistic such that the user can only estimate up to a maximum of 20 total years post initial diagnosis. It must be noted that the above-described inputs serve only to identify a population of interest and are not entered into a model of any kind.
3. ESTIMATE tool
The ‘ESTIMATE’ (ESTImating Mortality in breAsT cancEr) tool provides a population-based, non-parametric estimate of the cumulative risks of BCSM, non-BCSM, and all-cause mortality for HR-positive BC patients based on patient and tumour characteristics, and for any period of time between initial diagnosis and 20 years post-diagnosis. It serves as a convenient, interactive extension to subgroup analyses of mortality in HR-positive BC [
] in the context of competing risks. ESTIMATE enables the user to input a number of patient and tumour characteristics with established prognostic values, and the number of years from the initial diagnosis from which to estimate. Together, these inputs serve to subset the full SEER study cohort for the desired subgroup, after which the ‘Calculate’ button is used to generate the results. Point estimates at the user-specified time point of interest and plots of the cumulative incidence functions for BCSM, non-BCSM, and all-cause mortality are provided, each equipped with optional confidence intervals (CI). Brief text descriptions are also provided with each set of statistical output to aid interpretation. Results are displayed only when there are at least 20 total deaths within the specified estimation timeframe in the subgroup of interest in order to ensure estimates are reasonable. When there are fewer than 20 deaths, estimates are withheld, and an “Inestimable” message is shown.
ESTIMATE was developed using the shiny package in R, which provides a powerful, open-source framework for building web applications. The tool exists on a secure internet server that stores two main components: our pre-processed data set – derived from publicly available SEER data – and a set of R code that constructs the user interface (UI) and performs the statistical programming necessary to generate the results. Through our user-friendly point-and-click UI, the user bypasses the need for statistical programming savvy and can explore national mortality trends in HR-positive BC quickly and efficiently. ESTIMATE is freely available online at www.estimatetool.org/.
4. Results
4.1 Patient characteristics
The study cohort includes a total of 264,237 women diagnosed between 1990 and 2006 with a median follow-up of 14.8 years (interquartile range [IQR], 12.1 and 18.3 years). Table 1 shows additional characteristics of the study population. The median age at diagnosis was 59 years, and 26% of patients were 70 years of age or older. The predominant tumour size was T1c (40.67%, n = 107,475), and 43.79% had histologic grade II cancers (n = 115,700). Nodal status distribution was the following: N0: 64.81% (n = 171,244); N1: 23.89% (n = 63,116); N2: 7.48% (n = 19,777); N3: 3.82% (n = 10,100). Among all patients, 20.5% (n = 54,274) developed subsequent malignancies of any kind. Of all patients, 60.39% remained alive, 15.29% died from breast cancer, and 24.32% died from other causes. The distribution of causes of death for patients who died from causes other than breast cancer is listed in eTable 2.
It is important to consider that ESTIMATE is not predictive of specific outcomes for individual patients but instead prognostic for specific groups of patients with characteristics of interest. Shown below are several sets of results generated by ESTIMATE, which illustrate the tool’s utility in various clinical contexts.
5.1 Mortality after having survived any given time from initial diagnosis
ESTIMATE can be used to calculate risks in flexible timeframes after the initial diagnosis of BC. Example #1 refers to patients in the age group 40–49 years old diagnosed with a T1c, N1, grade II BC and who have survived 7 years since diagnosis. These patients have a 14% (95% CI: 11.9%–16.1%) residual cumulative risk of BCSM in the next 13 years and a 6.4% (95% CI: 4.7%–8.1%) residual cumulative risk of non-BCSM over the same period (Fig. 2A and e 3). For women in the age group 50–59 years old, diagnosed with T1c, N0, grade I BC, and who have survived for 10 years since diagnosis (like Example #2), the residual cumulative risk of BCSM over the next 10 years is 4.5% (95% CI: 3.2%–5.8%) and the residual cumulative risk of non-BCSM over the same period is 9.6% (95% CI: 7.9%–11.3%) (Fig. 2B and eTable 3).
Fig. 2ESTIMATE inputs and results for (A) patients aged 40–49 years, diagnosed with a T1c, N1, grade II BC, who have survived 7 years since diagnosis (Example #1) and (B) patients aged 50–59 years, diagnosed with T1c, N0, grade I BC, who have survived 10 years since diagnosis (Example #2).
Examples #3 and #4 represent two women in the 70–79 years old group with newly diagnosed BC: both patients have T2 primary tumours, but example #4 has nodal involvement (N1 versus N0) and higher tumour grade (grade III/IV versus grade II). The risk profiles for these two examples are outlined in eTable 4 and depicted in Fig. 3A and B. Within SEER, the tumour characteristics of example #4 confer a significantly higher average risk of BCSM over the following 10 years than do the tumour characteristics of example #3 (26.4% [95% CI: 24.0%–28.8%] versus 12.4% [95% CI: 11.2%–13.7%], respectively). The estimated risk of non-BCSM over the next 10 years for these two types of women are similar: 29.2% (95% CI: 26.7%–31.7%) for example #4 and 31.2% (95% CI: 29.5%–33.0%) for example #3.
Fig. 3ESTIMATE input and results for older women (aged 70–79 years) with T2 primary tumours with (A) no nodal involvement and lower grade disease (Example #3) versus with (B) nodal involvement and higher grade disease (Example #4) 10 years from diagnosis.
During the years analyzed in this study, the extension of endocrine therapy and ovarian suppression were not routinely used. This provides an opportunity to use ESTIMATE to inform clinical discussions about these endocrine therapy options.
Example #5 refers to patients in the age group 40–49 years old with newly diagnosed T1c, N0, grade I BC. These patients have a 4.1% (95% CI: 3.0%–5.3%) cumulative risk of BCSM over the next 20 years (Fig. 4A and eTable 5), a low risk of BCSM when considering whether to incorporate ovarian suppression.
Fig. 4ESTIMATE input and results for (A) patients aged 40–49 years with newly diagnosed T1c, N0, grade I BC (Example #5) and (B, C) patients aged 50–59 years diagnosed with a T2, N1, grade II BC ([B] from year 0 and [C] from year 5; Example #6).
The tool can also demonstrate changes in patients’ risks over time to inform extended endocrine therapy. Example #6 refers to patients in the age group 50–59 years old diagnosed with a T2, N1, grade II BC. These patients have a 26.3% (95% CI: 23.8%–28.8%) cumulative risk of BCSM from year 0–20 (Fig. 4B and eTable 5). After 5 years since the initial diagnosis, the residual cumulative risk of BCSM at year 20 is 22.8% (95% CI: 20.2%–25.4%) (Fig. 4C and eTable 5), which remains considerable and presents an opportunity to discuss extended endocrine therapy.
6. Discussion
We developed the ESTIMATE tool to quantify the risks of BCSM, non-BCSM and all-cause mortality in patients with HR-positive BC. To our knowledge, this is the first tool to estimate the integrated risk of BCSM and competing risks of death in a dynamic manner, up to 20 years from initial BC diagnosis. By providing population-based estimates of mortality due to both BC and competing risks of death while incorporating patient and tumour characteristics, we believe that the tool provides important information that could inform discussions of initial therapy options for BC. Furthermore, the tool enables estimation of BCSM and non-BCSM conditional on survival to timepoints of interest and can thus, inform clinical discussions after initial diagnosis as well.
The cumulative risk of BCSM in HR-positive BC increases over time. A pivotal study showed that after 5 years of endocrine therapy, the risk of distant recurrence at 20 years was as high as 38% for patients with N2 disease [
]. Similarly, previous studies from our group using SEER data showed that, among patients with N2 and N3 disease, risks of BCSM from years 5–20 are 38% and 50%, respectively [
]. These reports showed the feasibility of calculating long-term risk estimates using both clinical trials and population-based sources of data.
As expected, our study showed that the risk of non-BCSM increases substantially with older age. The ESTIMATE tool allows one to examine the risks of BCSM and non-BCSM simultaneously, which can assist in the understanding of the balance between risks and clinical decision-making in BC patients. This becomes apparent in the reported scenarios. Examples #3 and #4 are both 72-year-old women with a similar non-BCSM risk of approximately 30%. However, due to the node-positive status and high tumour grade in Example #4, women like these have a risk of BCSM that is twice that of women in Example #3 (26.4% versus 12.4%, respectively). On the other hand, if we consider Example #3 independently, the risk of non-BCSM is more than twice the risk of BCSM, highlighting the importance of other competing causes of death, even after a diagnosis of BC. A similar observation was seen among younger women in Example #2, with a T1c N0 grade I BC, for whom the risk of non-BCSM is twice the risk of BCSM. When examining the specific causes of death for non-BCSM events, we identified diseases of the heart, cerebrovascular diseases, chronic obstructive pulmonary disease and Alzheimer’s disease as the most frequent causes (eTable 2). These are consistent with the most common causes of death in older adults.
The ability of ESTIMATE to calculate risks any time after initial diagnosis allows one to evaluate how those risks change over time, and in turn, provides an opportunity for discussion with patients. Example #6 has an estimated risk of BCSM that remains elevated beyond 5 years after initial diagnosis, which presents an opportunity to discuss the role of extended endocrine therapy. On the other hand, Example #5 are 40-year-old women with a low-risk BC who would most likely not require the addition of ovarian suppression. In these two clinical scenarios, ESTIMATE is not intended to make a treatment decision for a clinician but rather to provide valuable information on residual risks of BCSM and non-BCSM that clinicians can integrate with other elements such as treatment side effects and patient’s preference, for shared decision-making.
Our tool has several limitations. First, SEER does not collect information on comorbidities, their severity and level of control, all of which affect the risk of non-BCSM; however, this is a common limitation of all BC risk tools available to date and can only be overcome with more comprehensive data collection. Another limitation of our tool is the lack of systemic therapy information, including treatment compliance, which affects the risk of BCSM. Users of the tool should consider that the estimates reported already incorporate treatment received because they are based on patients from the general population who received standard BC care [
]. Although the lack of systemic therapy information is a limitation, extension of endocrine therapy and ovarian suppression were not routinely used during the study period; therefore, the information provided by ESTIMATE can be used when considering the role of those therapies. Although we used population-based data, there is an underrepresentation of certain populations of interest, such as very young or very old patients and black women. As a result, estimates in the limit-cases might appear less stable or prone to overestimations or underestimations. ESTIMATE is unable to incorporate other known prognostic factors such as tumour histology, race, ethnicity and socioeconomic status due to the small subgroups that would result from the addition of those variables and the subsequent generation of inestimable results. Genomic Recurrence Score (Oncotype Dx) and information on health insurance were unavailable during the study period. Unfortunately, SEER does not provide information on cancer recurrence, which would have provided additional information for risk estimates, if available. There was a significant number of cases that were excluded due to missing HR status and missing cause of death. While these exclusions were necessary due to the tool’s characteristics and primary endpoint, they may represent a source of bias in the estimates. The SEER population used to create the tool is consistent of women exclusively; therefore, this version of the tool should not be used to quantify risks in men with breast cancer. Last, information about HER2 status and receipt of HER2-directed therapy were also not available. A prior study evaluating tumour subtypes reported that among all HR-positive BC in SEER, only 13% were HER2-positive and 87% were HER2-negative [
], a distribution that is likely to be similar in the present cohort. HER2 is unlikely to have a large role in late recurrence given the low risk of recurrence seen in years 5–10 for HR-positive, HER2-positive BC [
Incidence of late relapses in patients with HER2-positive breast cancer receiving adjuvant trastuzumab: combined analysis of NCCTG N9831 (alliance) and NRG Oncology/NSABP B-31.
]. When integrating the results from ESTIMATE into clinical decisions, it is important to remember that the tool is prognostic and not predictive. The tool does not use a predictive model to generate results; instead, the results provided are estimates from patients in SEER that have the same patient and tumour characteristics as the ones selected by the user in the tool.
Despite the limitations, ESTIMATE has several important strengths. Compared with other risk tools, our tool was developed using a very large sample size, which allows for a narrower CI in the risk estimates. The sample size included 26% of patients aged ≥70 years (n = 68,786), an important group of patients who have been historically underrepresented in clinical trials [
]. Another important strength is the tool’s ability to estimate risks of BCSM and non-BCSM for any user-defined time period within 20 years. To our knowledge, our tool is the only one that provides flexible start and stop time points, which allow for estimation of residual risks of BCSM and non-BCSM after a given amount of time the patient has already survived since the initial diagnosis. The ability to provide 20-year estimates is important not only for BCSM but also for non-BCSM, particularly in younger patients in whom the risk of non-BCSM is less relevant initially. In addition, the tool can be very valuable in regions of the world where genomic-based prognostic tools are either unavailable or difficult to access. Finally, given that this is a prognostic tool derived from population-based data, the estimates reported do not need to be independently validated. The estimates are reflective of the outcomes of the United States population. Certainly, the outcomes in another country may be different, but that would not represent an independent validation. For example, the Cancer Facts and Figures provided by the American Cancer Society do not need to be validated by an independent database because they are population-based estimates from the United States.
In summary, we developed a new interactive tool for non-metastatic, HR-positive BC that can be used in clinical practice to provide population-based estimates of the risks of BCSM, non-BCSM and all-cause mortality over a 20-year period based on patient and tumour characteristics. There are several ways in which our tool can aid in clinical decision-making, such as consideration of residual risks of mortality, quantification of risk in different age groups, understanding the balance between competing risks of death, and discussions about endocrine therapy options. Additionally, the tool can be used to inform risk stratification in the design of clinical trials evaluating late BCSM by helping to identify groups of patients at significant risk.
Authors’ contributions
Conceptualization JPL, BAL, NT; Data curation JPL, NG; Formal Analysis NG, NT; Funding acquisition not applicable investigation JPL, NG, SMT, BAL, MJH, JL, CTV, EPW, NUL, NT; Methodology JPL, NG, BAL, RAF, JL, CTV, EPW, NUL, NT; Project administration JPL, NG Resources JPL, NG Software JPL, NG, NT; Supervision BAL, EPW, NUL, NT, Validation not applicable, visualization not applicable Writing – original draft JPL, NG; Writing – review and editing JPL, NG, SMT, RAF, MJH, JL, CTV, EPW, NUL, NT.
Data availability
The data underlying this article will be shared on reasonable request to the corresponding author.
Funding
None received.
Conflict of interest statement
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: JPL received research funding from Kazia Therapeutics and consulting honoraria from Minerva Biotechnologies. RAF has institutional research funding from Puma Biotechnology. SMT receives institutional research funding from AstraZeneca, Lilly, Merck, Nektar, Novartis, Pfizer, Genentech/Roche, Immunomedics/Gilead, Exelixis, Bristol-Myers Squibb, Eisai, Nanostring, Cyclacel, Odonate, and Seattle Genetics; has served as an advisor/consultant to AstraZeneca, Eli Lilly, Merck, Nektar, Novartis, Pfizer, Genentech/Roche, Immunomedics/Gilead, Bristol-Myers Squibb, Eisai, Nanostring, Puma, Sanofi, Silverback Therapeutics, G1 Therapeutics, Athenex, OncoPep, Kyowa Kirin Pharmaceuticals, Daiichi-Sankyo, Ellipsis, Infinity, 4D Pharma, Samsung Bioepsis Inc., Chugai Pharmaceuticals, BeyondSpring Pharmaceuticals, OncXerna, OncoSec Medical Incorporated, Certara, Mersana Therapeutics, CytomX, Seattle Genetics. EPW reports institutional research funding from Genentech/Roche; serving as a consultant for Athenex, Carrick Therapeutics, G1 Therapeutics, Genentech/Roche, Genomic Health, Gilead, GlaxoSmithKline, GSK, Jounce, Lilly, St. Lucia, Syros, and Zymeworks; a non-paid scientific advisory board membership at Leap Therapeutics; and serving as President-Elect of the American Society of Clinical Oncology (ASCO). NUL reports institutional research funding from Genentech, Merck, Pfizer, Seattle Genetics, AstraZeneca, and Zion Pharmaceuticals; consultant/advisory board work for Pfizer, Puma, Seattle Genetics, Daiichi Sankyo, AstraZeneca, Prelude Therapeutics, Denali Therapeutics, Olema Pharmaceuticals, Aleta BioPharma, and Affinia Therapeutics; and royalties from UpToDate. All other authors report no conflicts.
Appendix ASupplementary data
The following is the supplementary data to this article:
Prediction of late distant recurrence in patients with oestrogen-receptor-positive breast cancer: a prospective comparison of the breast-cancer index (BCI) assay, 21-gene recurrence score, and IHC4 in the TransATAC study population.
Prediction of late distant recurrence after 5 years of endocrine treatment: a combined analysis of patients from the Austrian breast and colorectal cancer study group 8 and arimidex, tamoxifen alone or in combination randomized trials using the PAM50 risk of recurrence score.
Prognostic impact of the combination of recurrence score and quantitative estrogen receptor expression (ESR1) on predicting late distant recurrence risk in estrogen receptor-positive breast cancer after 5 Years of tamoxifen: results from NRG oncology/national surgical adjuvant breast and bowel Project B-28 and B-14.
Integration of clinical variables for the prediction of late distant recurrence in patients with estrogen receptor-positive breast cancer treated with 5 Years of endocrine therapy: CTS5.
Association of circulating tumor cells with late recurrence of estrogen receptor-positive breast cancer: a secondary analysis of a randomized clinical trial.
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.
Incidence of late relapses in patients with HER2-positive breast cancer receiving adjuvant trastuzumab: combined analysis of NCCTG N9831 (alliance) and NRG Oncology/NSABP B-31.
00380-X&id=gr1.jpg){kind=link}
00380-X&id=gr2.jpg){kind=link}
00380-X&id=gr3.jpg){kind=link}
00380-X&id=gr4.jpg){kind=link}