PDF
ePub
Citation
Print
INTRODUCTION
The 8th Edition of the American Joint Committee on Cancer Staging Manual (AJCC8) was implemented on January 1, 2018. The AJCC8 represented a significant change from previous editions as it incorporated biological markers into the prognostication of breast cancer [1]. Most validation studies have found that the new AJCC8 prognostic stage (AJCC8 PS) provides more accurate prognostication than the anatomic stage [2345].
Notwithstanding the advantages provided by the new AJCC8 PS, many authors have expressed concerns regarding the complexity of the new system, which includes over 170 subgroups, making it impractical for routine use without references or electronic aids [5678].
In 2017, Chavez-Macgregor et al. [9] proposed a simpler alternative to the AJCC8 PS in patients with breast cancer. The risk score system (RSS) calculated a risk score based on estrogen receptor (ER) status, human epithelial growth factor receptor 2 (HER2) receptor status, and tumor grade that was then incorporated into the anatomic stage for prognostication. Each anatomic stage would have 4 separate sub-stages with scores of 0–3. The authors verified the RSS using data from 43,938 patients in the California Cancer Registry and concluded that it provided accurate prognostic information. However, this alternative system has not yet been compared with the AJCC8 PS.
The present study was undertaken to determine the validity of the RSS, which has not been done using an external dataset at the time of this writing of this article, and explore the usefulness of the RSS compared to the AJCC8 PS.
METHODS
Study population
Patient information was obtained from the Singhealth Joint Breast Cancer Registry, a prospectively maintained database of patients from Singapore Health Services, a multi-institutional healthcare organization.
Patients diagnosed with invasive breast cancer between January 1, 2006, and December 31, 2014, were identified. This period was selected as routine HER2 testing and treatment with trastuzumab started in 2006. Patients for whom data were incomplete for ER status, HER2 receptor status, or tumor grade were excluded (n = 644). Those for whom data were missing for follow-up time, prognostic staging, or receipt of chemotherapy were also excluded (n = 114). Patients who received neoadjuvant chemotherapy were also excluded. Altogether, data for 6,656 patients were analyzed.
We defined ER, PR, and HER2 positivity in accordance with the recommendations of the 2010 and 2013 American Society of Clinical Oncology/College of American Pathologists [1011].
Cancer-specific survival (CSS) was defined as the time from diagnosis to death from breast cancer, and deaths from other causes were assessed at the last follow-up. Overall survival (OS) was defined as time from diagnosis to death from any cause and assessed at the last follow-up.
Statistical analysis
Descriptive statistics were used to evaluate the characteristics of the study population.
A Cox proportional hazards model was used to identify factors independently associated with survival. The variables included were based on those previously identified by Chavez-Macgregor et al. [9] (anatomic stage, ER receptor status, HER2 receptor status, tumor grade, age, ethnicity, and treatment received).
Patients were classified according to their anatomic stage based on the 7th edition of the AJCC [12]. Risk scores for each patient was calculated as per the method followed by Chavez-Macgregor et al. [9]: 1 point was allocated each for ER receptor-negative status, HER2 receptor-negative status, and grade 3 tumor (Table 1). These points are totaled for the calculation of a risk score of 0–3. Patients in each anatomic stage were thus subdivided into 4 ordinal categories (risk score, 0–3). Survival analysis was performed for CSS and OS using the Kaplan-Meier method based on the anatomic stage and risk score. Log-rank tests were then performed to identify differences between the categories (risk score, 0–3) at each anatomical stage.
Table 1
Chavez-Macgregor et al. [9] risk score system components
| Factors | Risk score assigned | |
|---|---|---|
| Tumour grade | ||
| 1 | 0 | |
| 2 | 0 | |
| 3 | 1 | |
| ER status | ||
| ER+ | 0 | |
| ER− | 1 | |
| HER2 receptor status | ||
| HER2+ | 0 | |
| HER2− | 1 | |
Akaike information criterion (AIC) and Harrell's concordance index (C-index) were used to compare the RSS and the AJCC8 PS. Adjustments were made for age, ethnicity, and receipt of chemotherapy since these factors affect survival outcomes [131415].
The level of significance was set at 5% for all statistical tests. All statistical analyses were performed using STATA 15.0 (StataCorp LLC, College Station, USA).
Ethics statement
All procedures performed in studies involving human participants were done so in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the Singapore Health Services Institutional Review Board (reference number 2017/2375). The institutional review board waived the need for informed consent due to the retrospective nature of this study.
RESULTS
Population characteristics
We analyzed a total of 6,656 patients from our database. Their clinicopathological characteristics are listed in Table 2. The median patient age at diagnosis was 54 (range, 19–97) years. The median follow-up was 61 (range, 0.1–43.0) months. Ductal carcinoma was the most common histology in the study cohort (87.0%). Overall, 64.0% of patients received chemotherapy and 84.3% of node-positive patients received adjuvant chemotherapy. Hormone receptor (HR) +/HER2− was the most common tumor subtype (64.5%), followed by HR+/HER2+ (15.2%), triple-negative breast cancer (TNBC; 10.8%), and HR−/HER2+ (9.4%). Of the ER+ patients, 84.6% received endocrine therapy, while 86.0% of HER2+ patients who had chemotherapy received targeted therapy. Among our patients, 14.0% (n = 932) had ER+/PR− histology. Of these patients with discordant HR, 9.8% (n = 642) were ER+/PR−/HER2− and 4.4% (n = 290) were ER+/PR−/HER2+.
Table 2
Patient characteristics and treatment information
All data are presented as number (%).
HR = hormone receptor; HR+ = hormone receptor positive; HR− = hormone receptor negative; HER2 = human epidermal growth factor receptor 2; HER2+ = human epidermal growth factor receptor 2 positive; HER2− = human epidermal growth factor 2 negative; LN = lymph node.
Survival analysis
Table 3 shows that on performing multivariate Cox regression analysis, anatomic stage, ER status, HER2 status, and tumor grade were independent predictors of both CSS and OS. The hazard ratio for CSS and OS increased with anatomic stage (with reference to anatomic stage I). ER− patients, compared to ER+ patients, had a hazard ratio of 1.74 and 1.62, respectively, for CSS and OS (p < 0.001 for both). HER2− patients, compared to HER2+ patients, had a hazard ratio of 1.49 and 1.50, respectively, for CSS and OS (p < 0.001 for both). Grade III patients, compared to grade I and II patients, had a hazard ratio of 1.84 and 1.59, respectively, for CSS and OS (p < 0.001 for both).
Table 3
Cox proportional hazards model evaluating determinants of CSS and OS amongst patients in this study
The CSS and OS survival curves for anatomical stages I–IV stratified by risk scores are shown in Figures 1 and 2. On performing the log-rank test, the risk score stratified well for all stages (p < 0.05) except stage IIIB (p = 0.062 for CSS, p = 0.231 for OS).
Figure 1
Five-year CSS according to stage (I–IV) stratified by risk score. (a) I, (b) IIA, (c) IIB, (d) IIIA, (e) IIIB, (f) IIIC, (g) IV.
CSS = cancer-specific survival.
Figure 2
Five-year OS according to stage (I–IV) stratified by risk score. (a) I, (b) IIA, (c) IIB, (d) IIIA, (e) IIIB, (f) IIIC, (g) IV.
OS = overall survival.
The multivariate Cox model for CSS showed that compared to risk score 0, hazard ratio was 1.10, 1.91, and 3.73 for risk scores of 1, 2, and 3 respectively (Table 4). For OS, compared to risk score 0, hazard ratio was 1.28, 1.88, and 3.55 for risk scores of 1, 2, and 3, respectively (Table 5). The RSS and AJCC8 PS demonstrated excellent discrimination of CSS and OS for our study cohort. Thus, the RSS (AIC = 10,649.45; C-index = 0.85) is not inferior to the AJCC8 PS (AIC = 10,726.654; C-index = 0.84) for CSS, nor is the RSS (AIC = 14,714.4; C-index = 0.82) inferior to the AJCC8 PS (AIC = 14,784.69; C-index = 0.81) for OS.
Table 4
Multivariable Cox models for CSS
Table 5
Multivariable Cox models for OS
Figure 3 displays the hazard ratio estimates for each combination of anatomic staging and risk score produced by a Cox proportional hazards model adjusted for age, ethnicity, and receipt of chemotherapy. This demonstrated that with more risk factors, hazard ratios generally increased within an anatomic stage; however, a patient with a lower stage but more risk factors may have a higher risk of breast cancer-related death than another patient with a higher stage of disease but fewer risk factors.
DISCUSSION
Our study of a large cohort of Asian breast cancer patients treated between 2006 and 2014 revealed that tumor grade, ER status, and HER2 receptor status were independent predictors of 5-year CSS and OS, supporting their use in the RSS. We further demonstrated that the addition of these risk factors into anatomical staging refines the prognostication of patients with breast cancer. These findings concur with those reported by Chavez-Macgregor et al. [9].
Chavez-Macgregor et al. [9] proposed a potentially simple and effective RSS, but it has not been validated by other datasets worldwide; thus, our study aimed to increase evidence of its usefulness.
To our knowledge, this is the first study comparing the RSS with the AJCC8 PS. Our analysis showed that the RSS is not inferior to the AJCC8 PS. This RSS may be a simpler and more user-friendly alternative to the AJCC8 PS with similar discrimination in the prognostication of OS and CSS in breast cancer patients.
Our study did not manage to identify a statistically significant difference between stage IIIB patients. This is likely due to the small number of patients in our cohort with stage IIIB disease (n = 90). However, a visual analysis of the group revealed separation of the survival curves over 5 years.
While we obtained statistically significant results for the stage IIIA group of patients (p < 0.05) and there was good graphical discrimination for the first 4 years, stratification appeared to be lost at year 5. This is likely due to the small numbers of patients involved at year 5 (Figure 1).
An important distinction of the RSS compared to the AJCC8 PS is that stage IV patients are stratified into different subgroups, whereas these metastatic patients under the AJCC8 PS remain in the same group. This is clinically informative and reflects the observed differences in the survival of patients with metastatic breast cancer of different subtypes. It has been shown that that patients with metastatic disease of the HR+/HER2+ subtype have longer survival than those with the HR−/HER2− subtype [161718].
As noted earlier, there have been challenges in the routine use of the AJCC8 PS. The RSS has several advantages as an alternative tool. First, it is simple and easy to calculate since it uses basic clinical information. Second, it capitalizes on the extensive knowledge of the anatomic staging system that underpins the common language between medical practitioners. It is simpler to build on this basis by incorporating the risk score into the anatomic stage rather than the use of a cumbersome application to obtain the new AJCC8 PS. Third, it is easy to communicate risk and treatment to patients using this system. Clinicians can inform patients regarding the number of risk factors they have and could also associate treatment regimens such as endocrine therapy or targeted therapy with the risk factors. Finally, the RSS maintains a similar anatomic staging, allowing for comparison to previous legacy studies and registries.
Unfortunately, the RSS is not without its limitations. It does not incorporate the results of gene expression studies such as Oncotype DX™, unlike the AJCC8 PS [1]. Studies have verified the utility of gene expression studies in refining prognosis with potential implications on treatment [1920]. This may result in difficulties applying the RSS in patients with the same anatomic score and risk factors but different Oncotype DX™ recurrence scores [21]. However, due to the current limited availability of results from these assays and the absence of long-term results from clinical trials, there is little evidence to guide the use of these assays in relation with risk factors [5]. Our study was limited by the lack of Oncotype DX™ data as well.
PR status was also not included in the current RSS. In our cohort, 14.0% (n = 932) of patients were ER+/PR−. Yi et al. [22] found that the staging system that included anatomic stage, grade, and ER had the highest C-index and lowest AIC. Other prognostic nomograms, the NHS predict for example, do not consider PR status either. On the other hand, the AJCC8 PS considers PR status, and in some circumstances, ER+ but PR− upstages a patient. St Gallens 2013 consensus recommended that ER+/PR−/HER2− patients be classified as luminal B, rather than A, reflecting the expert panel's opinion that ER+/PR− is a more aggressive variant [2324]. Other studies have also shown that ER+ and PR− tumors generally have poorer outcomes. To complicate matters, Prat et al. [25] showed that more than 20% of PR-positive tumor cells were significant for predicting survival differences among luminal A patients. There is controversy regarding the clinical utility of PR testing. Studies have shown that ER−/PR+ tumors are not reproducible and that PR provides no actionable information for ER+ tumors such as reported by Hefti et al. [26]. It is beyond the scope of this study to decide whether PR should continue to be routinely tested. However, we have shown that even without PR, the RSS is not inferior to the AJCC8 PS.
The RSS is effective on the assumption that patients receive standard of care as per their biomarker status, e.g. a HER2+ patient should receive targeted therapy. HER2-positive status, which was initially recognized as a negative prognostic factor, has been reversed into a factor that prognosticates better survival with the advent of trastuzumab. Hence, this scoring system may not be applicable to cohorts of patients with HER2+ tumors who do not receive HER2-targeted therapy. This phenomenon has been observed in other studies as well, and it is a problem that the AJCC8 PS is not spared from [1727]. This scoring system may also need to be modified in the future if there are any new breakthroughs that significantly improve the survival outcomes of subtypes associated with prognosis, such as the TNBC subtype.
The RSS does not re-stage a patient into a higher or low stage; hence, it is not immediately evident how comparable 2 patients are, such as one with a low anatomic stage but a high-risk scores and another with a high anatomic stage but lower risk scores. However, by simply building upon the existing anatomical staging system familiar to most oncologists, the RSS avoids the need for other inconvenient aids such as lengthy conversion tables, making it an attractive tool for everyday use.
We included age, ethnicity, and receipt of chemotherapy in the multivariate analysis as younger patients and Malay ethnicity have been shown in separate studies of the same cohort to be independent risk factors [131415].
The strength of our study is that the database we used is prospectively maintained. Although not all patients received treatment, there was a high rate of chemotherapy (84.3% of node-positive patients received adjuvant chemotherapy), targeted agent therapy (86.0% of HER2+ breast cancer patients), and endocrine therapy (84.6% of ER+ breast cancer patients).
Our study is limited by its relatively short follow-up duration. We only included patients diagnosed after 2006 since our institution only started routine use of adjuvant trastuzumab for patients with HER2+ disease from that point onward. This is an issue since luminal cancer subtypes may relapse after a longer interval [28].
To allow for wider implementation, more centers should consider adopting the use of this RSS, which is simple yet accurate for refining the prognosis of stage I–IV breast cancers. Since this is a novel system, we recommend the following nomenclature: anatomic stage IA (RS + 1), stage IV (RS + 2), etc.
The RSS provides a simpler way to integrate biomarkers with the prevalent common language of the anatomic stage in breast cancer patients. It is comparable to the more complex AJCC8 PS in its ability to stage patients by 5-year CSS and OS. Furthermore, it provides meaningful stratification for metastatic patients. Our findings support the use of RSS as a simpler alternative to the AJCC8 PS.
XML Download