Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Apr 15.
Published in final edited form as: Cancer. 2023 Dec 26;130(8):1210–1220. doi: 10.1002/cncr.35163

Validation of the RSClin Risk Calculator in the National Cancer Data Base

Augustin GL Vannier 1, Asim Dhungana 1, Fangyuan Zhao 2, Nan Chen 3, Sarah Shubeck 4, Olwen M Hahn 3, Rita Nanda 3, Nora T Jaskowiak 4, Gini F Fleming 3, Olufunmilayo I Olopade 3, Alexander T Pearson 3, Dezheng Huo 2,*, Frederick M Howard 3,*
PMCID: PMC10948297  NIHMSID: NIHMS1953850  PMID: 38146744

Abstract

Background:

Guidelines recommend the use of genomic assays such as OncotypeDx to aid decisions regarding the use of chemotherapy for hormone receptor-positive, HER2-negative (HR+/HER2-) breast cancer. The RSClin prognostic tool integrates OncotypeDx and clinicopathologic features to predict distant recurrence and chemotherapy benefit, but further validation is needed prior to broad clinical adoption.

Methods:

This study included patients from the National Cancer Database (NCDB) who were diagnosed with Stage I-III HR+/HER2− breast cancer from 2010–2020 and received adjuvant endocrine therapy with or without chemotherapy. RSClin predicted chemotherapy benefit was stratified into low (<3% reduction in distant recurrence), intermediate (3–5%), and high (>5%). Cox models were used to model mortality adjusted for age, comorbidity index, insurance, and race/ethnicity.

Results:

We identified 285,441 patients for inclusion from the NCDB, with an average age of 60 years and median follow-up of 58 months. Chemotherapy was associated with improved OS only for those predicted to have intermediate (adjusted hazard ratio [aHR] 0.68, 95% confidence interval [CI] 0.60–0.79) and high benefit per RSClin (aHR 0.66, 95% CI 0.61–0.72). Consistent benefit was seen in the subset with low OncotypeDx (<26) and intermediate (aHR 0.66, 95% CI 0.53–0.82) or high (aHR 0.71, 95% CI 0.58–0.86) RSClin predicted benefit. No survival benefit with chemotherapy was seen in patients with high OncotypeDx (≥26) and low benefit per RSClin (aHR 1.70, 95% CI 0.41–6.99).

Conclusions:

RSClin may identify high-risk patients who benefit from treatment intensification more accurately than OncotypeDx, and further prospective study is needed.

Keywords: Breast Neoplasms, Hormone Antagonists, Aromatase Inhibitors, Chemotherapy, Precision Medicine, Clinical Decision-Making, Gene Expression Profiling

Introduction:

Breast cancer is the leading cause of cancer death for women globally with an estimated 1.7 million cases diagnosed each year1. Hormone receptor positive (HR+) breast cancer is the most common subtype of breast cancer, constituting about 70% of newly diagnosed cases in the United States2. However, there is marked heterogeneity in outcome within HR+ breast cancer, and gene expression assays have been developed to more accurately characterize prognosis and/or chemotherapy benefit, including the OncotypeDx (ODX) and MammaPrint assays3,4. Although genomic assays enhance clinical decision-making, they must be used in combination with a patient’s clinical and demographic factors to individualize care. For example, subgroup analyses of the pivotal TAILORx, RxPONDER, and MINDACT trials have demonstrated that some pre-menopausal women who are not classified as ‘high-risk’ by genomic assays may still benefit from chemotherapy5–7. One recently described attempt to refine decision-making with the OncotypeDx assay is the RSClin prognostic risk tool8. RSClin incorporates age, tumor grade, tumor size, and OncotypeDx scores to provide more accurate prognostic information for patients with node-negative HR+ breast cancer. The RSClin tool provides an estimate of risk of distant metastasis at 10 years with hormonal therapy alone, as well as an estimate of chemotherapy benefit as measured by a reduction in this risk of distant metastasis when chemotherapy is added to hormonal therapy. This tool has been validated using data from a large health registry in Israel, but further real-world evaluation of the tool in larger, diverse populations is required8. In addition, whereas OncotypeDx has been prospectively validated to demonstrate that in women with scores over 26, chemotherapy should be offered, there is no clear cutoff for the RSClin tool for which chemotherapy is deemed necessary. Some authors have reported using a cutoff of greater than 5% predicted absolute chemotherapy benefit to administer chemotherapy, but additional data is needed9.

Although genomic recurrence assays have improved the care of women with HR+ breast cancer through more appropriate selection of patients for chemotherapy, accumulating evidence suggests that these assays are less applicable to non-White patients. This may in part be due to the fact that the most well-studied genomic recurrence assays were developed with samples from predominantly White patients. NSABP B-20, consisting of only 6% Black patients, was the largest cohort contributing samples to the development of OncotypeDx10. MammaPrint was developed from a single center cohort of patients under age 55 treated at the Netherlands Cancer Institute, which likely had limited representation of non-White patients4. Multiple studies have demonstrated that Black women with low recurrence scores have a higher risk of recurrence than a comparable body of White patients – although it remains unclear how much is due to differences in disease biology versus socioeconomic factors11,12. A prior analysis of the National Cancer Data Base (NCDB) found that the prognostic accuracy of OncotypeDx was nearly halved in Black and Hispanic patients13. A study of 95 African American women found that nearly two-thirds of patients classified as low or intermediate risk by OncotypeDx would be classified as high risk by MammaPrint – highlighting the difficulties of accurate risk stratification in this population14. Additionally, White women are more likely to undergo genomic testing than Black and Hispanic patients in the USA, furthering disparities15,16. RSClin utilizes the OncotypeDx score, so it is possible that this tool recapitulates the inherent disparities in OncotypeDx testing. As RSClin was developed through a patient-specific meta-analysis of the NSABP B-14 and TAILORx trials, both of which enrolled predominantly non-Hispanic White women, it may not be optimized for application in diverse patients.

In this study, we provide a large-scale validation of the RSClin risk tool using the NCDB, correlating tool predictions with long-term survival outcomes. Furthermore, we analyzed RSClin’s performance in multiple clinical and demographic subgroups, with particular attention to accuracy in patients of varying race and ethnicity, to determine if this novel tool can be equitably applied to all patients.

Methods:

Study Design and Data Source

This was a retrospective cohort study evaluating the RSClin prognostic risk calculator in patients with non-metastatic, node-negative HR+/HER2− breast cancer. Patients were identified using the NCDB, a large hospital-based registry maintained jointly by the American College of Surgeons and American Cancer Society, capturing approximately 70% of new cancer diagnoses in the United States17. This study was determined to be exempt from review by the University of Chicago institutional review board. No informed consent was obtained as we did not use any identifying patient information. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline18.

Study Population and Covariates

We included patients diagnosed with node-negative HR+/HER2− invasive breast carcinoma from January 1st, 2010 through December 31st, 2020, who were treated with adjuvant hormonal therapy with or without chemotherapy. We excluded patients with ductal carcinoma in situ or metastatic disease. Covariates analyzed included age, sex, race / ethnicity, Charlson/Deyo comorbidity index, insurance status, grade, tumor size, and recurrence score result – patients with unknown results for these variables were excluded (Supplemental Figure 1). For this patient cohort, each unique combination of grade, tumor size, recurrence score, and age was iteratively passed to the online RSClin calculator API to generate predictions of distant disease-free survival and absolute benefit of chemotherapy (which is listed as a reduction in distant recurrence rate compared to hormonal therapy alone) at 10 years19. As the specific hormonal therapy used is not recorded in the NCDB, the RSClin estimates of recurrence and chemotherapy benefit were calculated compared to treatment with aromatase inhibitors unless otherwise stated.

Statistical Analysis

Statistical analysis was performed using Python version 3.8.8 (Python Software Foundation, Wilmington, DE). Early studies of breast cancer chemotherapy demonstrated benefits of as low as 2–3% in postmenopausal women20, whereas clinicians may prefer to recommend chemotherapy for patients when the benefit is >5%9,21. Thus, we evaluated subgroups where RSClin predicted chemotherapy benefit was < 3% (low benefit), 3 – 5% (intermediate benefit), and > 5% (high benefit). Decisions to administer chemotherapy in the NCDB are also likely confounded by patient performance status, as well as socioeconomic factors22,23. We performed inverse probability of treatment weighting, calculating propensity scores with a logistic regression model to correct for the influence of age, race / ethnicity, comorbidity index, and insurance status on chemotherapy treatment decisions24. Additionally, as age and comorbidity index are strongly associated with non-cancer mortality in this population, adjusted hazard ratio (aHR) estimates incorporated these features as covariates to better describe the association of RSClin predictions with breast cancer specific survival.

The primary outcome was the aHR for overall survival (OS) in patients who received and did not receive chemotherapy for subgroups separated by RSClin predicted chemotherapy benefit. Secondary outcomes included assessing the prognostic accuracy of RSClin for patients treated with hormonal therapy alone, evaluating the predictive accuracy within clinical and demographic subgroups, and evaluating outcomes in patients with discordant RSClin and OncotypeDx risk categories. Prognostic accuracy of RSClin was compared to the input variables of grade, tumor size, recurrence score using Harrell’s concordance index (C-index)25. We used restricted cubic spline regression to model chemotherapy benefit as a function of RSClin predictions on continuous basis26. All statistical testing was done at the α = 0.05 significance level, and no correction for multiple hypothesis testing was performed given the exploratory nature of this analysis.

Results:

We identified 285,441 patients within the NCDB who met our inclusion criteria. The average patient age was 59.5 years, the mean recurrence score was 16.3, and 41,789 (14.6%) patients received chemotherapy (Table 1, Supplemental Table 1). All patients received endocrine therapy. The median follow-up in this cohort was 58.3 months, and 11,706 (4.6%) of patients died during this follow-up. The vast majority of patients were female (n = 238,447; 99.3%), although a few male patients met eligibility criteria (n = 1,994; 0.7%); male patients had higher grade, larger tumors with higher Oncotype scores (Supplemental Table 2). The breakdown of race / ethnicity was similar to what was seen in the TAILORx trial27: the majority (235,804; 82.6%) of patients were non-Hispanic White, 21,937 (7.7%) were non-Hispanic Black, 13,788 (4.8%) were of Hispanic ethnicity (of any race), and 11,735 (4.1%) were Asian / Pacific Islander. The average predicted risk of distant recurrence at 10 years (via RSClin) was 7.0% with aromatase inhibitors alone, and the average predicted reduction in distant recurrence with chemotherapy was 2.5%. To further characterize the distribution of RSClin predictions in the NCDB population, histograms were generated for the risk of distant recurrence (Figure 1). Notably, non-Hispanic White patients had lower predicted risk of recurrence and a lower predicted chemotherapy benefit than other racial / ethnic subgroups (Supplemental Figure 2). For example, if a threshold of 5% absolute benefit per RSClin predictions was used to select patients for chemotherapy, this would result in treatment of 11% of non-Hispanic White, 12% of Hispanic, 13% of Asian / Pacific Islander, and 16% of non-Hispanic Black patients. This is a direct reflection of heterogeneity in the input variables for the RSClin calculator. Consistent with existing studies, non-Hispanic Black patients in this NCDB cohort were younger, with higher grade, larger tumors with higher recurrence scores compared to their non-Hispanic White counterparts (Supplemental Table 3).

Table 1: Characteristics of Included Patients.

All demographic factors aside from sex were associated with receipt of chemotherapy.

Missing Overall Chemo No Chemo P-Value
n 285441 41789 243652
Age, mean (SD) 0 59.5 (10.5) 55.4 (10.7) 60.2 (10.2) <0.001
Sex, n (%) Female 0 283447 (99.3) 41501 (99.3) 241946 (99.3) 0.828
Male 1994 (0.7) 288 (0.7) 1706 (0.7)
Race / Ethnicity, n (%) Asian 0 11735 (4.1) 1918 (4.6) 9817 (4.0) <0.001
Hispanic 13788 (4.8) 2218 (5.3) 11570 (4.7)
Native American 814 (0.3) 92 (0.2) 722 (0.3)
Non-Hispanic Black 21937 (7.7) 3922 (9.4) 18015 (7.4)
Non-Hispanic White 235804 (82.6) 33437 (80.0) 202367 (83.1)
Other 1363 (0.5) 202 (0.5) 1161 (0.5)
Insurance, n (%) Government / Uninsured 0 116661 (40.9) 12776 (30.6) 103885 (42.6) <0.001
Private / Managed 168780 (59.1) 29013 (69.4) 139767 (57.4)
Charlson/Dayo Score, n (%) 0 0 239780 (84.0) 35778 (85.6) 204002 (83.7) <0.001
>= 1 45661 (16.0) 6011 (14.4) 39650 (16.3)
Grade, n (%) 1 0 84015 (29.4) 4686 (11.2) 79329 (32.6) <0.001
2 161435 (56.6) 20468 (49.0) 140967 (57.9)
3 39991 (14.0) 16635 (39.8) 23356 (9.6)
Histologic Subtype, n (%) Ductal 0 220154 (77.1) 34799 (83.3) 185355 (76.1) <0.001
Ductal and Lobular 18094 (6.3) 2116 (5.1) 15978 (6.6)
Lobular 38316 (13.4) 3873 (9.3) 34443 (14.1)
Other 8877 (3.1) 1001 (2.4) 7876 (3.2)
Tumor Size (mm), median (IQR) 0 15.0 [10.0,20.0] 17.0 [12.0,24.0] 14.0 [10.0,20.0] <0.001
Tumor Size Group, n (%) ≤ 20 mm 0 217857 (76.3) 27236 (65.2) 190621 (78.2) <0.001
21 – 50 mm 63181 (22.1) 13641 (32.6) 49540 (20.3)
> 50 mm 4403 (1.5) 912 (2.2) 3491 (1.4)
Receptor Status, n (%) ER+PR+ 6682 254280 (91.2) 31968 (79.1) 222312 (93.3) <0.001
ER+PR- 24251 (8.7) 8336 (20.6) 15915 (6.7)
ER-PR+ 228 (0.1) 108 (0.3) 120 (0.1)
ER%, mean (SD) 185652 93.2 (11.1) 90.3 (14.8) 93.4 (10.7) <0.001
PR%, mean (SD) 185582 68.4 (34.8) 51.5 (38.4) 69.8 (34.1) <0.001
KI67%, mean (SD) 237615 16.4 (14.6) 27.0 (19.9) 15.4 (13.6) <0.001
OncotypeDx Score, n (%) High (26+) 0 33468 (11.7) 24281 (58.1) 9187 (3.8) <0.001
Intermediate (11 – 25) 180769 (63.3) 16693 (39.9) 164076 (67.3)
Low (0 – 10) 71204 (24.9) 815 (2.0) 70389 (28.9)
Last Contact, mean (SD) 32554 62.5 (32.2) 73.0 (32.1) 60.6 (31.8) <0.001
Vital Status, n (%) Alive 32551 241184 (95.4) 36791 (93.9) 204393 (95.6) <0.001
Deceased 11706 (4.6) 2379 (6.1) 9327 (4.4)
RSClin Predicted Percentage Risk of Distant Recurrence at 10 Years with AI Alone, mean (SD) 0 7.0 (7.6) 16.6 (12.3) 5.4 (4.8) <0.001
RSClin Predicted Benefit of Chemotherapy (Absolute Percent Reduction in Distant Recurrence), n (%) High (> 5%) 0 31873 (11.2) 22196 (53.1) 9677 (4.0) <0.001
Intermediate (3 – 5%) 27090 (9.5) 10113 (24.2) 16977 (7.0)
Low (< 3%) 226478 (79.3) 9480 (22.7) 216998 (89.1)
Open in a new tab

Figure 1: Histogram for Distributions of RSClin Predictions.

Figure 1:

Open in a new tab

Predictions of risk of distant recurrence on aromatase inhibitor (left) and absolute chemotherapy benefit as percentage reduction in distant recurrence (right) are illustrated for the overall study population. For this study, RSClin predicted chemotherapy benefit was stratified into low (<3% reduction in distant recurrence), intermediate (3–5%), and high (>5%) – these divisions are illustrated above.

Next, RSClin predicted risk of distant recurrence on hormonal therapy alone was compared to observed overall survival in the NCDB cohort. The 243,652 patients treated with hormonal therapy alone were stratified by RSClin predicted distant recurrence risk in 5% increments (Supplemental Figure 3, Supplemental Table 4). Of note, survival declined monotonically with increasing predicted risk. Similar findings were seen in patients under 50, regardless of whether treatment with aromatase inhibitor or tamoxifen was selected for the RSClin calculator (Supplemental Figure 4, Supplemental Table 5 and 6). To determine if RSClin accurately incorporated clinical and demographic features to predict prognosis, we computed the C-index for a Cox model25 for overall survival with RSClin recurrence risk, and compared this to Cox models using tumor grade, size, and recurrence score (Supplemental Table 7). RSClin alone achieved a C-index of 0.574, similar to the C-index of 0.570 for a model with grade, size, and recurrence score. These models demonstrated more prognostic accuracy for overall survival than recurrence score alone (C-index 0.543). However, in absolute terms, RSClin predictions remained only a modest predictor of survival (a C-index of 0.5 is the performance that would be obtained through random chance), and age is more important factor, as models incorporating age achieve C-index of 0.708 – 0.719.

We next assessed the accuracy of RSClin, comparing survival with and without chemotherapy in patients with low (< 3%), intermediate (3 – 5%), and high (> 5%) predicted chemotherapy benefit (Figure 2, Supplemental Table 8). No survival advantage was seen with chemotherapy patients with low RSClin predicted benefit (aHR 0.99, 95% CI 0.89–1.11, p = 0.91). Significant survival benefit was seen with chemotherapy in patients with intermediate (aHR 0.68, 0.60–0.79, p < 0.01) and high (aHR 0.66, 95% CI 0.61–0.72, p < 0.01) RSClin predicted benefit. Similar findings were seen in the subset of patients under 50 (Supplemental Figure 5) when calculating RSClin with presumption of aromatase inhibitor therapy (Supplemental Table 9), but there was only a trend towards chemotherapy benefit which did not reach statistical significance when calculations were made presuming tamoxifen therapy (Supplemental Table 9). We also quantified the relationship of RSClin predicted chemotherapy benefit versus actual survival benefit using a restricted cubic spline model (Figure 3). The survival advantage of chemotherapy became significant when the predicted benefit was at least 3%.

Figure 2: Correlation of RSClin Predictions and Mortality Rate in Patients Treated with or without Chemotherapy.

Figure 2:

Open in a new tab

Mortality rate is compared between patients treated with and without chemotherapy within three subgroups of RSClin predicted chemotherapy benefit (reported as absolute reduction in 10-year distant recurrence rate with chemotherapy). A significant survival benefit is seen starting at the 3 – 5% subgroup. Inverse probability of treatment weighting is applied to mitigate effect of patient fitness on treatment decisions, and hazard ratios are listed for chemotherapy benefit in a multivariable Cox model that also incorporates age and comorbidity index.

Figure 3: Continuous Prediction of Survival Benefit with Chemotherapy as a Function of RSClin Predicted Benefit.

Figure 3:

Open in a new tab

Relationship was modeled with a cubic spline regression with four knots, with inverse probability of treatment weighting and covariates of age and comorbidity index included in the model. A significant overall survival benefit with chemotherapy was seen when RSClin predicted chemotherapy benefit was at least 3%.

As the correlation between RSClin predictions and survival may depend on several clinical and demographic factors, we repeated this analysis in additional subgroups (Figure 4). Of note, no patient subgroups experienced a significant survival benefit with chemotherapy when RSClin predicted benefit was low (< 3%). Most subgroups demonstrated a survival benefit when RSClin predicted benefit was high (> 5%). The magnitude of benefit of chemotherapy was smaller in patients with lobular histology compared to ductal histology, and did not reach significance even in the subset with high RSClin predicted benefit (aHR 0.80, 95% CI 0.60 – 1.07). In general, similar findings were seen across racial and ethnic subgroups. However, we observed no trend towards benefit in Hispanic patients regardless of RSClin predictions, even those predicted to have a high RSClin predicted benefit (aHR 1.19, 95% CI 0.66–2.14).

Figure 4: Survival Benefit of Chemotherapy, Stratified by RSClin Predicted Benefit and Clinical / Demographic Factors.

Figure 4:

Open in a new tab

Analysis is pictured with inverse probability of treatment weighting and models also incorporated age and comorbidity index as variables. No clear subgroup clearly benefits from chemotherapy when RSClin predicted benefit was < 3%, although there was a trend towards benefit in patients with grade 3 tumors. Conversely, nearly all subgroups demonstrated a clear survival advantage with chemotherapy when RSClin predicted benefit was > 5%. Notable exceptions include lobular tumors and Hispanic patients.

Finally, we evaluated treatment patterns and outcomes in patients based on RSClin and recurrence score risk categories. In our NCDB cohort, chemotherapy was administered for 73% of patients with high-risk OncotypeDx recurrence score (≥ 26, n = 33,468) and 25% of patients with high-end intermediate scores (21 – 25, n = 38,351, Supplemental Figure 6). Similarly, chemotherapy was used in 70% of patients with high (n = 31,873) and 37% of patients with intermediate (n = 27,090) predicted chemotherapy benefit per RSClin. While only 1.4% (n = 476) of patients with high-risk OncotypeDx recurrence score had low RSClin predicted benefit, 21% (n = 6,316) of patients with high and 75% (n = 13,640) of patients with intermediate RSClin had OncotypeDx recurrence scores under 26.

In these discordant cases where survival status was available, we found that RSClin was more predictive of the impact of chemotherapy (Figure 5, Supplementary Table 11). In the small subset with high-risk OncotypeDx recurrence score patients with low RSClin, there was no discernable differences in survival associated with treatment with chemotherapy (aHR 1.70, 95% CI 0.41–6.99, p = 0.46). All these patients had grade 1 tumors, with OncotypeDx recurrence score no greater than 28, and tumor size under 1.3 cm. By contrast, patients with low OncotypeDx score and high (aHR 0.71, 95% CI 0.58–0.86, p < 0.01) or intermediate (aHR 0.66 95% CI 0.53–0.82, p<0.01) RSClin had prolonged OS with chemotherapy. Of note, the intermediate RSClin group had a median OncotypeDx score of 23 (IQR 20 – 24). Similar results were seen in patients age 50 or older (Supplemental Table 12, Supplemental Figure 7) and those younger than 50 (Supplemental Table 13), suggesting this effect is not dependent on menopausal status. Similar to the overall population, patients aged below 50 years, with low OncotypeDx score but high or intermediate RSClin-predicted benefit showed improved OS with chemotherapy. There was only a trend towards chemotherapy benefit in patients with concordant high RSClin-predicted chemotherapy benefit and OncotypeDx scores (aHR 0.78, 95% CI 0.57–1.08, p = 0.14), which may not have reached significance in part due to the imbalanced treatment decisions in this cohort (only 13%, or n = 627 received endocrine therapy alone). The results in patients 50 years or older mirrored the overall population, with an advantage to chemotherapy exclusively seen in patients with intermediate or high predicted benefit per RSClin.

Figure 5: Mortality Rate in Patients in Discordant RSClin and OncotypeDx Risk Groups.

Figure 5:

Open in a new tab

High RSClin is defined as predicted absolute chemotherapy benefit of > 5%, intermediate as between 3% and 5%, and a low risk as less than 3% benefit. High OncotypeDx risk is defined as score of 26 or higher. Inverse probability of treatment weighting is applied to survival curves.

Discussion:

In this study, we further validated the RSClin prognostic risk tool in a large cohort of patients in the NCDB. We demonstrated that increasing RSClin predicted recurrence correlates with decreased survival in patients treated with hormonal therapy alone and identifies a subset of patients who benefit from chemotherapy. Our subgroup analysis by clinical and demographic factors provides insight into the validity of RSClin in different patient populations, with notable heterogeneity in patients with lobular breast cancer, high-grade tumors, and in patients of varying race / ethnicity. Several studies have suggested that chemotherapy may be less impactful in lobular tumors, even with high risk OncotypeDx scores28,29, and our subgroup analysis similarly found that lobular tumors had a trend towards lower chemotherapy benefit (compared to ductal tumors) for each category of RSClin predicted risk.

In our study, we consistently found a survival benefit for patients with an intermediate (3 – 5%) and high (> 5%) predicted chemotherapy benefit per RSClin, although the absolute difference in mortality was more clinically meaningful in the latter group, reaching 6.8% at 8 years, and the survival advantage was more consistent across subgroups in the high predicted benefit group. In our NCDB cohort, 21% of node negative patients would have intermediate or high predicted benefit per RSClin, which is much larger than the 12% who have a high-risk OncotypeDx score (≥26). However, the TAILORx trial has also suggested chemotherapy may benefit some patients with intermediate risk OncotypeDx scores, with a trend towards lower recurrence rates in women 50 or younger with scores of 16 – 25, especially in patients with high-intermediate scores of 21 – 256. These findings in TAILORx may in part be due to chemotherapy induced menopause, and use of ovarian suppression was likely low in our study due to the timeframe. Thus, our findings should not prompt use of chemotherapy for all patients with elevated risk per RSClin. Nonetheless, our study demonstrates outcomes are suboptimal with hormonal therapy alone, and this tool may prove helpful to select patients for treatment intensification with ovarian suppression30, CDK4/6 inhibitors31, or bisphosphonates32 – with chemotherapy reserved for patients at highest risk. RSClin may also serve to de-escalate therapy - in the small group of patients with low risk via RSClin but high OncotypeDx score, no overall survival advantage was seen with chemotherapy (Figure 5, bottom left).

Importantly, we found that RSClin, like OncotypeDx, may be less prognostic in some non-White patients. There is clear heterogeneity in RSClin risk predictions in racial / ethnic subgroups, reflective of underlying differences in model inputs. It is well described that Black women are more likely to have high-risk recurrence score results12, high-grade tumors, and low ER expression33. In our study, we demonstrated that regardless of the cutoff for RSClin predictions, chemotherapy will be recommended more frequently for non-White patients – in particular, Black women would be nearly 50% more likely than White women to be recommended to undergo chemotherapy with a cutoff of 5% predicted chemotherapy benefit. However, the survival benefit of chemotherapy in our study only reached significance for the highest RSClin risk category in Black women, and was not seen for any risk category in Hispanic patients, suggesting it may not stratify patient risk in these groups. Similarly, an analysis of the TAILORx trial had found that Black patients with intermediate recurrence scores had a 60% higher distant recurrence rate than White patients, but did not experience a chemotherapy benefit27. Black women with breast cancer are 40% more likely to die of their disease than White women2 – and there are myriad contributors to this gap, including differences in disease biology, early detection, and treatment delays34,35, access to treatment22,23. It is important that diverse patient data are used to drive prognostic testing and model development to ensure that these disparities are not perpetuated given the biologic differences in breast cancer in Black and Hispanic (weaker ER expression36,37, differences in HER2 expression34,38, and higher proportion of GATA3 and TP53 mutations39–41).

Our study described the performance of the RSClin calculator in the largest clinical cancer registry in the world, allowing for careful assessment of performance in a variety of subgroups. However, there are several limitations to the use of the NCDB. Most notably, only OS data is collected, which limits direct validation of RSClin predictions, which are measured in terms of distant recurrence rates. OS is moderately correlated with disease-free survival in breast cancer42, and likely serves as a reasonable proxy for distant recurrence rates. The duration of follow-up also limits the correlation of OS to distant recurrence rates, as there is a several year delay between the development of metastasis to death from breast cancer. Nonetheless, the large sample size enables detection of minute differences in survival, which may compound with further follow-up. Another limitation of our study is the lack of specific treatment annotations in NCDB. We conducted much of our analysis presuming treatment with aromatase inhibitors, since the type of endocrine therapy is not specified in NCDB, but many premenopausal women included in this study may have received tamoxifen. However, in women under 50, estimating risk with tamoxifen only demonstrated a non-significant trend towards chemotherapy benefit in the RSClin risk categories defined in our analysis. Selecting tamoxifen treatment raises the RSClin predictions of chemotherapy benefit, which may assign more lower risk women to the intermediate / high benefit categories and may explain this discrepancy. The thresholds defined using aromatase inhibitor treatment as an input for RSClin still identified younger women with a significant benefit of chemotherapy, and thus may still be reasonably applied to identify younger women for treatment intensification. Additionally, NCDB does not provide annotations for menopausal status, which would be particularly important given that ovarian function suppression may be the prime driver of chemotherapy benefit in intermediate risk node-negative patients. Although we perform an age-based analysis (separating patients above and below 50 years old) and find largely similar results, the variable onset of menopause limits our confidence in this finding. Additionally, we were unable to include the specific chemotherapy regimen received or dose intensity, which may account for some of the racial disparities in chemotherapy benefit, as non-white patients are more likely to receive nonstandard regimens and are less likely to complete planned therapy43,44. Finally, limited comorbidity information is included in the NCDB. We attempted to control for the impact of comorbidities and other demographic factors in chemotherapy administration through inverse probability of treatment weighting, but in our subgroup analysis, older patients experienced a greater magnitude of chemotherapy benefit than younger patients when RSClin predicted a high benefit, which may reflect the selection of relatively healthier older patients for chemotherapy.

Conclusions:

The RSClin prognostic risk tool accurately predicts overall survival with hormonal therapy in the NCDB – performing as well as a single metric or in combination with recurrence score, grade, and tumor size. Increased mortality was seen in node-negative patients treated with hormonal therapy alone when RSClin predicted a chemotherapy benefit of at least 3%, although this mortality difference was most consistent across subgroups when predicted chemotherapy benefit was over 5%. Further study is needed to clarify optimal strategies in such patients, which may include ovarian suppression, chemotherapy, or other emerging adjuvant therapies.

Supplementary Material

Supinfo

Precis Statement:

Although developed to predict distant recurrence rates, RSClin predictions correlate well with OS for patients receiving endocrine therapy with or without chemotherapy. RSClin may better identify high risk patients who benefit from chemotherapy than OncotypeDx score alone, and further prospective study is needed.

Acknowledgements:

FMH received funding support from the NIH / NCI grant K08CA283261 and the Cancer Research Foundation. DH, ATP, and FMH. received support from the Department of Defense (BC211095P1). DH, RN, and OIO received support from the NIH/NCI (1P20-CA233307). OIO received support from Susan G Komen (SAC 210203). DH and OIO received support from Breast Cancer Research Foundation (BCRF-21-071). These funding organizations did not have a role in the design and conduct of the study; the collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; or the decision to submit the manuscript for publication.

Footnotes

Code Availability:

The code used for analysis in this paper and for generation of relevant figures and tables is available at https://github.com/fmhoward/NCDBRSClin.

Disclosures:

AV, AD, FZ, NC, SS, NJ, DH, and FMH report no competing financial or non-financial conflicts of interest. OMH reports consulting/advisory work for Pfizer Inc. RN reports contracted research with Arvinas, AstraZeneca, Celgene, Corcept Therapeutics, Genentech/Roche, Immunomedics, Merck, OBI Pharma, Odonate Therapeutics, OncoSec, Pfizer, Seattle Genetics, and Taiho and consulting fees from AstraZeneca, BeyondSpring, Cardinal Health, Fujifilm, Immunomedics/Gilead, Infinity, iTeos, Merck, OBI, Oncosec, and Seagen. OIO reports ownership interest in 54Gene, CancerIQ, and Tempus and financial interest in Color Genomics, Healthy Life for All Foundation, and Roche/Genetech.

References:

  • 1.Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians. 2018;68(6):394–424. doi: 10.3322/caac.21492 [DOI] [PubMed] [Google Scholar]
  • 2.DeSantis CE, Ma J, Gaudet MM, et al. Breast cancer statistics, 2019. CA: A Cancer Journal for Clinicians. 2019;69(6):438–451. doi: 10.3322/caac.21583@10.1002/(ISSN)2045-7634.oncology-breast-cancer [DOI] [PubMed] [Google Scholar]
  • 3.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(27):2817–2826. doi: 10.1056/NEJMoa041588 [DOI] [PubMed] [Google Scholar]
  • 4.van ‘t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature. 2002;415(6871):530–536. doi: 10.1038/415530a [DOI] [PubMed] [Google Scholar]
  • 5.Piccart M, van ‘t Veer LJ, Poncet C, et al. 70-gene signature as an aid for treatment decisions in early breast cancer: updated results of the phase 3 randomised MINDACT trial with an exploratory analysis by age. Lancet Oncol. 2021;22(4):476–488. doi: 10.1016/S1470-2045(21)00007-3 [DOI] [PubMed] [Google Scholar]
  • 6.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(25):2395–2405. doi: 10.1056/NEJMoa1904819 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Kalinsky K, Barlow WE, Meric-Bernstam F, et al. Abstract GS3–00: First results from a phase III randomized clinical trial of standard adjuvant endocrine therapy (ET) +/− chemotherapy (CT) in patients (pts) with 1–3 positive nodes, hormone receptor-positive (HR+) and HER2-negative (HER2-) breast cancer (BC) with recurrence score (RS) < 25: SWOG S1007 (RxPonder). Cancer Res. 2021;81(4 Supplement):GS3-GS3–00. doi: 10.1158/1538-7445.SABCS20-GS3-00 [DOI] [Google Scholar]
  • 8.Sparano JA, Crager MR, Tang G, Gray RJ, Stemmer SM, Shak S. 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(6):557–564. doi: 10.1200/JCO.20.03007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Zimmerman BS, Berwick S, Kessler AJ, et al. Application of RSClin to guide treatment recommendations for premenopausal patients with early-stage hormone-positive breast cancer and intermediate risk oncotype recurrence scores. JCO. 2021;39(15_suppl):e12511–e12511. doi: 10.1200/JCO.2021.39.15_suppl.e12511 [DOI] [Google Scholar]
  • 10.Fisher B, Dignam J, Emir B, et al. Tamoxifen and Chemotherapy for Lymph Node-Negative, Estrogen Receptor-Positive Breast Cancer. JNCI: Journal of the National Cancer Institute. 1997;89(22):1673–1682. doi: 10.1093/jnci/89.22.1673 [DOI] [PubMed] [Google Scholar]
  • 11.Collin LJ, Yan M, Jiang R, et al. Oncotype DX recurrence score implications for disparities in chemotherapy and breast cancer mortality in Georgia. npj Breast Cancer. 2019;5(1):1–7. doi: 10.1038/s41523-019-0129-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Hoskins KF, Danciu OC, Ko NY, Calip GS. Association of Race/Ethnicity and the 21-Gene Recurrence Score With Breast Cancer-Specific Mortality Among US Women. JAMA Oncol. 2021;7(3):370–378. doi: 10.1001/jamaoncol.2020.7320 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ibraheem A, Olopade OI, Huo D. Propensity score analysis of the prognostic value of genomic assays for breast cancer in diverse populations using the National Cancer Data Base. Cancer. 2020;126(17):4013–4022. doi: 10.1002/cncr.32956 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Robinson PA, Tsai ML, Lo SS, et al. Genomic risk classification by the 70-gene signature and 21-gene assay in African American early-stage breast cancer patients. JCO. 2021;39(15_suppl):e12568–e12568. doi: 10.1200/JCO.2021.39.15_suppl.e12568 [DOI] [Google Scholar]
  • 15.Davis BA, Aminawung JA, Abu-Khalaf MM, et al. Racial and Ethnic Disparities in Oncotype DX Test Receipt in a Statewide Population-Based Study. J Natl Compr Canc Netw. 2017;15(3):346–354. doi: 10.6004/jnccn.2017.0034 [DOI] [PubMed] [Google Scholar]
  • 16.Press DJ, Ibraheem A, Dolan ME, Goss KH, Conzen S, Huo D. Racial disparities in omission of oncotype DX but no racial disparities in chemotherapy receipt following completed oncotype DX test results. Breast Cancer Res Treat. 2018;168(1):207–220. doi: 10.1007/s10549-017-4587-8 [DOI] [PubMed] [Google Scholar]
  • 17.Bilimoria KY, Stewart AK, Winchester DP, Ko CY. The National Cancer Data Base: A Powerful Initiative to Improve Cancer Care in the United States. Ann Surg Oncol. 2008;15(3):683–690. doi: 10.1245/s10434-007-9747-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007;335(7624):806–808. doi: 10.1136/bmj.39335.541782.AD [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.RSClin™ Educational Tool. Accessed October 23, 2021. https://rsclin.genomichealth.com/#/app/home
  • 20.Polychemotherapy for early breast cancer: an overview of the randomised trials. Early Breast Cancer Trialists’ Collaborative Group. Lancet. 1998;352(9132):930–942. [PubMed] [Google Scholar]
  • 21.Stiggelbout AM, de Haes JC, van de Velde CJ. Adjuvant chemotherapy in node negative breast cancer: patterns of use and oncologists’ preferences. Ann Oncol. 2000;11(5):631–633. doi: 10.1023/a:1008379628579 [DOI] [PubMed] [Google Scholar]
  • 22.Daly B, Olopade OI. A perfect storm: How tumor biology, genomics, and health care delivery patterns collide to create a racial survival disparity in breast cancer and proposed interventions for change. CA: A Cancer Journal for Clinicians. 2015;65(3):221–238. doi: 10.3322/caac.21271 [DOI] [PubMed] [Google Scholar]
  • 23.Howard FM, Olopade OI. Epidemiology of Triple-Negative Breast Cancer: A Review. Cancer J. 2021;27(1):8–16. doi: 10.1097/PPO.0000000000000500 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med. 2015;34(28):3661–3679. doi: 10.1002/sim.6607 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Harrell FE, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. JAMA. 1982;247(18):2543–2546. [PubMed] [Google Scholar]
  • 26.Desquilbet L, Mariotti F. Dose-response analyses using restricted cubic spline functions in public health research. Stat Med. 2010;29(9):1037–1057. doi: 10.1002/sim.3841 [DOI] [PubMed] [Google Scholar]
  • 27.Albain KS, Gray RJ, Makower DF, et al. Race, Ethnicity, and Clinical Outcomes in Hormone Receptor-Positive, HER2-Negative, Node-Negative Breast Cancer in the Randomized TAILORx Trial. JNCI: Journal of the National Cancer Institute. 2021;113(4):390–399. doi: 10.1093/jnci/djaa148 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Kizy S, Huang JL, Marmor S, Tuttle TM, Hui JYC. Impact of the 21-gene recurrence score on outcome in patients with invasive lobular carcinoma of the breast. Breast Cancer Res Treat. 2017;165(3):757–763. doi: 10.1007/s10549-017-4355-9 [DOI] [PubMed] [Google Scholar]
  • 29.Chen XH, Zhang WW, Wang J, et al. 21-gene recurrence score and adjuvant chemotherapy decisions in patients with invasive lobular breast cancer. Biomark Med. 2019;13(2):83–93. doi: 10.2217/bmm-2018-0396 [DOI] [PubMed] [Google Scholar]
  • 30.Francis PA, Pagani O, Fleming GF, et al. Tailoring Adjuvant Endocrine Therapy for Premenopausal Breast Cancer. N Engl J Med. 2018;379(2):122–137. doi: 10.1056/NEJMoa1803164 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Johnston SRD, Harbeck N, Hegg R, et al. Abemaciclib Combined With Endocrine Therapy for the Adjuvant Treatment of HR+, HER2-, Node-Positive, High-Risk, Early Breast Cancer (monarchE). J Clin Oncol. 2020;38(34):3987–3998. doi: 10.1200/JCO.20.02514 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Eisen A, Somerfield MR, Accordino MK, et al. Use of Adjuvant Bisphosphonates and Other Bone-Modifying Agents in Breast Cancer: ASCO-OH (CCO) Guideline Update. JCO. 2022;40(7):787–800. doi: 10.1200/JCO.21.02647 [DOI] [PubMed] [Google Scholar]
  • 33.Warner ET, Tamimi RM, Hughes ME, et al. Racial and Ethnic Differences in Breast Cancer Survival: Mediating Effect of Tumor Characteristics and Sociodemographic and Treatment Factors. J Clin Oncol. 2015;33(20):2254–2261. doi: 10.1200/JCO.2014.57.1349 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Shubeck S, Zhao F, Howard FM, Olopade OI, Huo D. Response to Treatment, Racial and Ethnic Disparity, and Survival in Patients With Breast Cancer Undergoing Neoadjuvant Chemotherapy in the US. JAMA Network Open. 2023;6(3):e235834. doi: 10.1001/jamanetworkopen.2023.5834 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Zhao F, Miyashita M, Hattori M, et al. Racial Disparities in Pathological Complete Response Among Patients Receiving Neoadjuvant Chemotherapy for Early-Stage Breast Cancer. JAMA Network Open. 2023;6(3):e233329. doi: 10.1001/jamanetworkopen.2023.3329 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Purrington KS, Gorski D, Simon MS, et al. Racial differences in estrogen receptor staining levels and implications for treatment and survival among estrogen receptor positive, HER2-negative invasive breast cancers. Breast Cancer Res Treat. 2020;181(1):145–154. doi: 10.1007/s10549-020-05607-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Patel TA, Colon-Otero G, Bueno Hume C, Copland JA, Perez EA. Breast Cancer in Latinas: Gene Expression, Differential Response to Treatments, and Differential Toxicities in Latinas Compared with Other Population Groups. Oncologist. 2010;15(5):466–475. doi: 10.1634/theoncologist.2010-0004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Peiffer DS, Zhao F, Chen N, et al. Clinicopathologic Characteristics and Prognosis of ERBB2-Low Breast Cancer Among Patients in the National Cancer Database. JAMA Oncology. Published online February 23, 2023. doi: 10.1001/jamaoncol.2022.7476 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Ademuyiwa FO, Tao Y, Luo J, Weilbaecher K, Ma CX. Differences in the mutational landscape of triple negative breast cancer in African Americans and Caucasians. Breast Cancer Res Treat. 2017;161(3):491–499. doi: 10.1007/s10549-016-4062-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Pitt JJ, Riester M, Zheng Y, et al. Characterization of Nigerian breast cancer reveals prevalent homologous recombination deficiency and aggressive molecular features. Nat Commun. 2018;9(1):4181. doi: 10.1038/s41467-018-06616-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Miyashita M, Bell JSK, Wenric S, et al. Molecular profiling of a real-world breast cancer cohort with genetically inferred ancestries reveals actionable tumor biology differences between European ancestry and African ancestry patient populations. Breast Cancer Research. 2023;25(1):58. doi: 10.1186/s13058-023-01627-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Ng R, Pond GR, Tang PA, MacIntosh PW, Siu LL, Chen EX. Correlation of changes between 2-year disease-free survival and 5-year overall survival in adjuvant breast cancer trials from 1966 to 2006. Ann Oncol. 2008;19(3):481–486. doi: 10.1093/annonc/mdm486 [DOI] [PubMed] [Google Scholar]
  • 43.Griggs JJ, Culakova E, Sorbero MES, et al. Social and racial differences in selection of breast cancer adjuvant chemotherapy regimens. J Clin Oncol. 2007;25(18):2522–2527. doi: 10.1200/JCO.2006.10.2749 [DOI] [PubMed] [Google Scholar]
  • 44.Green AK, Aviki EM, Matsoukas K, Patil S, Korenstein D, Blinder V. Racial Disparities in Chemotherapy Administration for Early Stage Breast Cancer: A Systematic Review and Meta-Analysis. Breast Cancer Res Treat. 2018;172(2):247–263. doi: 10.1007/s10549-018-4909-5 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supinfo
NIHMS1953850-supplement-Supinfo.docx (1.1MB, docx)

RESOURCES