Abstract
Purpose:
Genomic assays such as Oncotype Dx (ODX) and Mammaprint are used for risk adapted treatment decisions for early breast cancer. However, concordance between genomic assays is modest. Using real world data, we performed a comparative analysis of ODX and Mammaprint.
Methods:
A cohort of women diagnosed with early stage, hormone receptor-positive breast cancer that received ODX or Mammaprint was established using the National Cancer Database (NCDB), 2010–2016. Using the propensity score matching method, we defined two groups of patients with similar clinical and demographic characteristics; one group received ODX and another received Mammaprint. We examined the association between ODX or Mammaprint and overall survival using Cox models.
Results:
Of 451,693 eligible patients, 45.3% received ODX and 1.8% received Mammaprint testing. The use of ODX increased from 36.1% in 2010 to 49.9% in 2016, while use of Mammaprint increased from 0.5% in 2010 to 3.3% in 2016. We matched 5,042 patients who received ODX to 5,042 patients who received Mammaprint. 5-year risk of death for Mammaprint low and ODX low were 3.4% and 4.7% respectively. The prognostic value of Mammaprint was similar to ODX; the C-index was 0.614 (0.572–0.657) for Mammaprint and 0.581 (0.530–0.631) for ODX. There was a difference in the performance of ODX assay across racial/ethnic groups (p<0.001), with a slightly better performance in Whites than African Americans and Hispanics.
Conclusion:
Both ODX and Mammaprint tests are good in identifying low risk individuals who could be spared chemotherapy. Sub-optimal performance of ODX in ethnic minority deserves further investigation.
Keywords: Genomic assay, breast cancer, biomarker, endocrine receptor, racial disparity
Precis:
This study findings will inform clinicians on the prognostic comparison between the two widely used genomic assays in clinical practice. In addition, our study also emphasizes the need to understand the tumor biology underlying racial/ethnic disparity in the era of precision medicine.
INTRODUCTION
Hormone receptor positive breast cancer is the most common breast cancer affecting about 70% of women diagnosed with breast cancer in the United States.1 The heterogeneity of hormone receptor positive breast cancer led to the development of genomic biomarkers necessary to tailor therapeutic decisions and not depend solely on clinicopathologic characteristics. The most common genomic biomarker assays used in clinical practice are Oncotype Dx (ODX) and Mammaprint. The higher uptake of these two assays in clinical practice compared to other genomic biomarkers may be due to the retrospective and prospective studies conducted to validate these two assays2–8 and their recommendations in the NCCN9 and ASCO guidelines10. Although ODX was initially developed for node negative patients, in a secondary analysis of SWOG 8814 trial of women with node positive disease, RS ≥31 was predictive of chemotherapy benefit and patients with RS 18–30 did not show clear benefits of chemotherapy.11 Even though ODX and Mammaprint have been used for guiding personalized therapy in clinical practice, it is important to note that they have moderate concordance with each other12. The Optima trial showed that Kappa statistics was 0.40 between Oncotype Dx and Mammaprint.13 The lack of strong concordance amongst genomic biomarkers makes it less appropriate to interchange these biomarkers.
The use of genomic biomarkers has led to significant improvement in patient outcomes but this may not be totally translated to non-white patients particularly patients of African Ancestry. It is worthy to note that the racial disparity in the outcome of patients with breast cancer is multifactorial and not entirely limited to differences in healthcare delivery. In particular, according to Press et al14, the use of genomic biomarkers led to racial equality in treatment choices. In the recently concluded TAILORx trial which prospectively assessed the benefit of chemotherapy for women using Oncotype DX with a mid-range recurrence score (RS) of 11 to 258, analysis showed that black women (compared with white women) was associated with worse clinical outcomes and Hispanic ethnicity was generally associated with better outcomes (compared with non-Hispanic ethnicity)15. This raises the question: Do genomic biomarkers have the same ability to appropriately prognosticate clinical outcomes of breast cancer patients of racial/ethnic minorities as compared with non-Hispanic white patients? In the studies that developed or validated these genomic assays, ethnic minority patients were less representative;2, 6, 11, 16, 17 as such the prognostic and predictive value of these genomic biomarkers have not been sufficiently evaluated in minority populations. In addition, other biomarkers, such as PAM50, have shown that women of African ancestry were more likely to be luminal B subtype compared to White women18 suggesting racial difference in the spectrum of biologically distinct breast cancer.
This study aims to compare the prognostic performance of the two commonly used genomic tests, Oncotype Dx and Mammaprint, in hormone receptor positive breast cancer patients using data from National Cancer Database (NCDB). The secondary aims of the study are to compare prognostic value of Oncotype Dx across racial/ethnic groups and examine the trend of genomic test utilization from 2010 to 2016 among early stage breast cancer patients in the United States.
PATIENTS AND METHODS
The NCDB, a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society, is a nationwide facility based, oncology dataset that includes information on approximately 70% of all incident cancers diagnosed in the United States collected from more than 1400 hospitals participating in the American College of Surgeons CoC.19 Routinely collected data items include patient demographics, tumor characteristics (such as tumor site and histology), and American Joint Committee on Cancer stage. The current study used de‐identified data and was deemed exempt from human protection oversight by the Institutional Review Board of the University of Chicago.
Study Population
There were 1,573,776 breast cancer patients diagnosed between January 1, 2010 and December 31, 2016 and reported to the NCDB. To investigate the trend of multigene test utilization, we limited the analysis to women aged ≥18 years who were diagnosed with stage 1 to 3A, first primary, hormone receptor (ER or PgR) positive, HER2 negative breast cancer (according to American Society of Clinical Oncology/College of American Pathologists guidelines20) who received all or part of their first course of treatment in the reporting facility. For the analysis of the prognosis performance of multigene tests, we further excluded patients without ODX or Mammaprint, missing test results, and without data of follow-up (all patients diagnosed in 2016 had no survival data). After these selections, the analysis included 144,357 patients who received ODX and 5047 patients who received Mammaprint (Supplementary Fig. 1).
To ensure that patients receiving ODX and those receiving Mammaprint are comparable in demographic and clinical characteristics, we conducted a propensity score matching. A propensity score was the log odds of receiving Mammaprint test (rather than ODX test) estimated from the logistic regression that included age, race/ethnicity, year of diagnosis, co-morbidity index (CCI), number of positive nodes, tumor size, histology type, grade, progesterone (PgR), lymphovascular invasion, median income, and census regions. Of note, we did not match for chemotherapy because of reasonable causal pathway in which genetic tests are used to determine the use of chemotherapy and in turn chemotherapy affect survival outcome and toxicities. Matching for chemotherapy would force the two tests to be artificially concordant. Before and after 1:1 matching, we checked the standardized differences between patients who received ODX and Mammaprint to examine whether matching on the propensity score removed the observed differences21.
To allow for better comparison between ODX and Mammaprint low/high risk, ODX patients were dichotomized to two categories: low risk if RS ≤ 25 and node negative or RS < 11 and node positive; high risk if RS > 25 and node negative or RS ≥ 11 and node positive. The criteria for this dichotomization in node negative patients was based on the results reported by the TAILORx trial which showed that patients with node negative and RS ≤ 25 did not benefit from chemotherapy8. The criteria for dichotomization in node positive patients was based on the prospective West German Study Group Plan B trial which showed that patients with RS ≤ 11 did not benefit from chemotherapy,22 and our previous analysis of NCDB which showed that patients with node positive disease and RS 11–25 benefitted from chemotherapy and there was an incremental benefit with higher RS23. We continue to wait for the prospective Phase 3 RxPONDER study which should provide further insight into the RS cut-off at which a chemotherapy benefit can be detected in patients with node positive breast cancer.
Statistical Analysis
We examined the trend of multigene testing from 2010 to 2016 using generalized linear model with binary outcome and logarithmic link. We compared clinical and demographic factors between patients receiving ODX and those receiving Mammaprint using chi-square tests. Among patients in the matched groups, we examined the association between ODX or Mammaprint and overall survival using Cox proportional hazard models in the two groups separately. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated as strength of association. The prognostic values were evaluated using Harrell’s C-index. In the analysis of prognostic value of ODX across nodal status and across racial/ethnic groups, we treated ODX recurrence score as both a continuous variable and a categorical variable based on categories for TAILORx trial8.
RESULTS
Of the 451,693 eligible patients with stage 1–3A hormone receptor positive/HER2 negative breast cancer and with information on multigene testing, 48.0% received multigene testing prior to therapy. The most ordered multigene therapy was ODX (45.3%), followed by Mammaprint (1.8%) and other multigene assay tests (0.7%). The percentage of patients who received multigene testing increased from 37.6% in 2010 to 54.5% in 2016 (Figure 1). The use of ODX increased from 36.1% in 2010 to 49.9% in 2016 with an annual relative increase of 4.5% (p<0.001), while the use of Mammaprint increased from 0.5% in 2010 to 3.3% in 2016 with an annual relative increase of 30.9% (p<0.001).
Figure 1.
Utilization rate of multigene testing in early stage hormone receptor positive, Her2 negative breast cancer from year 2010 to 2016 in the U.S.
Table 1 provides the demographic and clinical characteristics of the patients who had ODX and Mammaprint. There are differences in almost all demographic and clinical characteristics between patients receiving Mammaprint and ODX, with the exception of race, progesterone receptor subtype and histology. Mammaprint was likely ordered in patients with high-risk clinicopathological features including lymph node positive, larger tumor size, higher tumor grade, and lymphovascular invasion. As summarized in the clinical risk category, which was defined according to MINDACT (“Microarray in Node-Negative and 1–3 Positive Lymph Node Disease May Avoid Chemotherapy”) modification of Adjuvant!Online7, 46.0% of the Mammaprint group was in the clinical high risk category, while 37.2% of the ODX group was in the clinical high risk category. There are geographic variations in the ordering of multigene testing, with Mammaprint testing more commonly ordered in South Atlantic and Pacific regions. The proportion of receiving chemotherapy ranged from 16.0% to 80.9% across RS categories and ranged from 11.5% to 80.0% across Mammaprint categories, suggesting that oncologists basically utilize these genetic test results to prescribe chemotherapy. In addition, the Mammaprint test placed more patients into chemotherapy than the ODX test.
Table 1.
Patient demographic and clinical characteristics by type of multigene tests, before and after propensity score matching
| Pre-matching | Post-matching | |||||
|---|---|---|---|---|---|---|
| Oncotype DX (N=144357) N (%) |
Mammaprint (N=5047) N (%) |
Oncotype DX (N=5042) N (%) |
Mammaprint (N=5042) N (%) |
Standardized difference* | ||
| p-value | ||||||
| Age at diagnosis | <0.001 | |||||
| 18–39 | 4561 (3.2) | 211 (4.2) | 201 (4.0) | 211 (4.2) | 0.011 | |
| 40–44 | 9879 (6.8) | 392 (7.8) | 396 (7.9) | 391 (7.8) | −0.002 | |
| 45–49 | 17611 (12.2) | 597 (11.8) | 567 (11.2) | 596 (11.8) | 0.018 | |
| 50–54 | 20703 (14.3) | 675 (13.4) | 668 (13.2) | 675 (13.4) | 0.011 | |
| 55–59 | 22343 (15.5) | 747 (14.8) | 729 (14.5) | 746 (14.8) | 0.015 | |
| 60–64 | 24609 (17.0) | 814 (16.1) | 834 (16.5) | 814 (16.1) | −0.015 | |
| 65–69 | 23164 (16.0) | 766 (15.2) | 776 (15.4) | 766 (15.2) | −0.008 | |
| 70–74 | 13610 (9.4) | 515 (10.2) | 532 (10.6) | 514 (10.2) | −0.020 | |
| 75–79 | 6036 (4.2) | 227 (4.5) | 222 (4.4) | 227 (4.5) | 0.003 | |
| 80+ | 1841 (1.3) | 103 (2.0) | 117 (2.3) | 102 (2.0) | −0.020 | |
| Mean (SD) | 58.4 (10.5) | 58.4 (11.1) | 0.94 | 58.6 (11.1) | 58.4 (11.1) | −0.028 |
| Race/ethnicity | 0.005 | |||||
| NH White | 121683 (84.3) | 4204 (83.3) | 4209 (83.5) | 4199 (83.3) | −0.007 | |
| NH Black | 11086 (7.7) | 432 (8.6) | 439 (8.7) | 432 (8.6) | −0.005 | |
| Hispanic | 6151 (4.3) | 246 (4.9) | 237 (4.7) | 246 (4.9) | 0.007 | |
| Asian PI | 5437 (3.8) | 165 (3.3) | 157 (3.1) | 165 (3.3) | 0.014 | |
| Year of diagnosis | <0.001 | |||||
| 2010 | 15535 (10.8) | 170 (3.4) | 170 (3.4) | 170 (3.4) | −0.001 | |
| 2011 | 20079 (13.9) | 329 (6.5) | 329 (6.5) | 329 (6.5) | 0.000 | |
| 2012 | 22716 (15.7) | 695 (13.8) | 694 (13.8) | 694 (13.8) | −0.003 | |
| 2013 | 25870 (17.9) | 1031 (20.4) | 1028 (20.4) | 1028 (20.4) | −0.011 | |
| 2014 | 28504 (19.7) | 1186 (23.5) | 1185 (23.5) | 1185 (23.5) | 0.005 | |
| 2015 | 31653 (21.9) | 1636 (32.4) | 1636 (32.4) | 1636 (32.4) | 0.007 | |
| Charlson comorbidity index | 0.025 | |||||
| 0 | 123107 (85.3) | 4354 (86.3) | 4362 (86.5) | 4350 (86.3) | −0.002 | |
| 1 | 17713 (12.3) | 597 (11.8) | 586 (11.6) | 596 (11.8) | 0.001 | |
| 2 | 3537 (2.5) | 96 (1.9) | 94 (1.9) | 96 (1.9) | 0.003 | |
| Number of positive lymph node | <0.001 | |||||
| 0 | 119328 (82.7) | 3897 (77.2) | 3894 (77.2) | 3894 (77.2) | 0.000 | |
| 1–3 | 25029 (17.3) | 1150 (22.8) | 1148 (22.8) | 1148 (22.8) | 0.000 | |
| Tumor size | <0.001 | |||||
| <=1.0 cm | 34318 (23.8) | 1051 (20.8) | 1026 (20.3) | 1050 (20.8) | 0.006 | |
| 1.1–2.0 cm | 72555 (50.3) | 2385 (47.3) | 2448 (48.6) | 2382 (47.2) | −0.027 | |
| 2.1–5.0 cm | 35457 (24.6) | 1500 (29.7) | 1464 (29.0) | 1500 (29.8) | 0.018 | |
| >5.0 cm | 2027 (1.4) | 111 (2.2) | 104 (2.1) | 110 (2.2) | 0.020 | |
| Progesterone receptor | 0.16 | |||||
| Negetive | 12879 (8.9) | 479 (9.5) | 494 (9.8) | 479 (9.5) | −0.022 | |
| Positive | 131478 (91.1) | 4568 (90.5) | 4548 (90.2) | 4563 (90.5) | 0.022 | |
| Histology | 0.045 | |||||
| Ductal | 111414 (77.2) | 3954 (78.3) | 3981 (79.0) | 3949 (78.3) | −0.007 | |
| Lobular | 17876 (12.4) | 627 (12.4) | 613 (12.2) | 627 (12.4) | −0.003 | |
| Ductal & Lobular | 10145 (7.0) | 307 (6.1) | 299 (5.9) | 307 (6.1) | 0.011 | |
| Other | 4922 (3.4) | 159 (3.2) | 149 (3.0) | 159 (3.2) | 0.007 | |
| Tumor grade | <0.001 | |||||
| 1 | 40135 (27.8) | 1254 (24.8) | 1268 (25.1) | 1252 (24.8) | −0.013 | |
| 2 | 76099 (52.7) | 2648 (52.5) | 2632 (52.2) | 2647 (52.5) | 0.016 | |
| 3 | 22164 (15.4) | 986 (19.5) | 997 (19.8) | 985 (19.5) | −0.009 | |
| Unknown | 5959 (4.1) | 159 (3.2) | 145 (2.9) | 158 (3.1) | 0.007 | |
| Lymph vascular invasion | <0.001 | |||||
| No | 110468 (76.5) | 3670 (72.7) | 3664 (72.7) | 3666 (72.7) | 0.009 | |
| Yes | 18656 (12.9) | 702 (13.9) | 712 (14.1) | 702 (13.9) | −0.015 | |
| Unknown | 15233 (10.6) | 675 (13.4) | 666 (13.2) | 674 (13.4) | 0.004 | |
| Clinical risk group† | <0.001 | |||||
| Low risk | 88255 (62.8) | 2678 (54.0) | 2722 (55.0) | 2675 (54.0) | −0.019 | |
| High risk | 52184 (37.2) | 2279 (46.0) | 2231 (45.0) | 2278 (46.0) | 0.019 | |
| Median income quartiles 2008–2012 | <0.001 | |||||
| < $38,000 | 16777 (11.6) | 673 (13.3) | 663 (13.1) | 673 (13.3) | 0.004 | |
| $38,000-$47,999 | 27897 (19.4) | 1158 (23.0) | 1174 (23.3) | 1158 (23.0) | −0.008 | |
| $48,000-$62,999 | 38784 (26.9) | 1356 (26.9) | 1353 (26.8) | 1356 (26.9) | 0.001 | |
| >=$63,000 | 60575 (42.1) | 1856 (36.8) | 1852 (36.7) | 1855 (36.8) | 0.003 | |
| Facility Location | <0.001 | |||||
| New England | 8935 (6.4) | 40 (0.8) | 40 (0.8) | 40 (0.8) | 0.000 | |
| Middle Atlantic | 26446 (18.9) | 552 (11.4) | 600 (12.4) | 550 (11.4) | −0.031 | |
| South Atlantic | 30039 (21.5) | 1690 (34.9) | 1738 (35.9) | 1689 (35.0) | −0.020 | |
| East North Central | 25563 (18.3) | 779 (16.1) | 737 (15.2) | 778 (16.1) | 0.024 | |
| East South Central | 7392 (5.3) | 279 (5.8) | 245 (5.1) | 279 (5.8) | 0.032 | |
| West North Central | 11976 (8.6) | 420 (8.7) | 434 (9.0) | 419 (8.7) | −0.010 | |
| West South Central | 7018 (5.0) | 322 (6.7) | 322 (6.7) | 322 (6.7) | 0.001 | |
| Mountain | 7893 (5.6) | 152 (3.1) | 143 (3.0) | 152 (3.1) | 0.011 | |
| Pacific | 14534 (10.4) | 602 (12.4) | 582 (12.0) | 602 (12.5) | 0.013 | |
| Propensity score, mean (SD) | −3.123 (0.613) | −3.123 (0.613) | ||||
| Receiving chemotherapy | ||||||
| Oncotype DX (RS 0–10) | 1336 (16.0) | 60 (5.8) | ||||
| Oncotype DX (RS 11–25) | 15625 (35.0) | 569 (19.9) | ||||
| Oncotype DX (RS >25) | 15513 (80.9) | 608 (73.7) | ||||
| Mammaprint low risk | 311 (11.5) | 311 (11.5) | ||||
| Mammaprint high risk | 1671 (80.0) | 1668 (80.0) | ||||
all p values for comprising the two groups were greater than 0.30 after propensity score matching.
MINDACT (“Microarray in Node-Negative and 1–3 Positive Lymph Node Disease May Avoid Chemotherapy”) modification of Adjuvant!Online assessment of clinical risk
We conducted propensity score matching to remove these clinical and demographic differences between Mammaprint and ODX. As shown in Supplementary Figure 2, the propensity score of receiving Mammaprint was almost identical between the 5,042 patients who actually received Mammaprint and the 5,042 matched patients who actually received ODX. After the matching, all the demographics and clinicopathological features were comparable between the two groups (right panel of Table 1). In the matched cohort of 5,042 patients who received Mammaprint, 2,908 had genomic low-risk score and 11.5% of them received chemotherapy, while 2,134 had genomic high-risk score and 80.0% of them received chemotherapy. Of the matched ODX cohort of 5,042 patients, 1,140 had low risk RS 0–10, 3,068 had intermediate RS 11–25, and 834 had high risk RS >25, and the percentage of receiving chemotherapy was 5.8%, 19.9%, and 73.7%, respectively, for the three risk categories (Table 1).
In the matched cohorts, the median (interquartile range) follow-up time was 33 months (21–49 months). Figure 2 shows the Kaplan-Meier curves of overall survival by ODX and Mammaprint. 5-year risk of dying for Mammaprint high and ODX high category were 9.3% and 12.4%, respectively. Compared to patients with a low genomic risk Mammaprint score, patients with a high-risk Mammaprint score had 2.64-fold risk of dying (95% CI 1.89–3.69) with a C-index of 0.614 (95% CI 0.572–0.657). Relative to patients with a low risk ODX RS, the hazard ratio was 0.98 (95% CI 0.63–1.52) for intermediate risk RS and 2.45 (95% CI 1.54–3.91) for high-risk RS; the C-index for ODX was 0.581 (95% CI 0.530–0.631). Relative to dichotomized low-risk ODX recurrence score, the HR for high-risk ODX score was 2.19 (95% CI 1.59–3.00) with a C-index of 0.608. Multivariable Cox models gave similar results (Table 2). These results suggest that the separation between risk groups in survival was similar between the two multigene tests although the prognostic value was slightly higher for Mammaprint.
Figure 2.
Kaplan-Meier curve by Mammaprint and Oncotype DX testing results in matched samples. New Oncotype DX category: low risk if recurrence score (RS) <26 and node negative or RS <11 and node positive; high risk if RS >25 and node negative or RS >10 and node positive.
Table 2.
Prognostic value of Mammaprint and Oncotype DX in matched samples
| Multigene test | # of patients | # of death | 5-year risk (95% CI) | Hazard ratio (95% CI) | C-index (95% CI) | Adjusted hazard ratio (95% CI)* |
|---|---|---|---|---|---|---|
| Mammaprint | ||||||
| Low risk | 2908 | 52 | 3.4% (2.4%−4.7%) | 1.00 (ref.) | 0.614 (0.572–0.657) | 1.00 (ref.) |
| High risk | 2134 | 101 | 9.3% (7.4%−11.7%) | 2.64 (1.89–3.69) | 2.25 (1.56–3.25) | |
| Oncotype DX | ||||||
| Low risk (RS 0–10) | 1140 | 27 | 4.7% (3.0%−7.4%) | 1.00 (ref.) | 0.581 (0.530–0.631) | 1.00 (ref.) |
| Intermediate risk (RS 11–25) | 3068 | 74 | 5.2% (3.9%−6.8%) | 0.98 (0.63–1.52) | 1.04 (0.66–1.62) | |
| High risk (RS >25) | 834 | 52 | 12.4% (9.1%−16.8%) | 2.45 (1.54–3.91) | 1.81 (1.05–3.09) | |
| Oncotype DX + node status † | ||||||
| Low risk | 3503 | 76 | 4.7% (3.6%−6.2%) | 1.00 (ref.) | 0.608 (0.563–0.653) | 1.00 (ref.) |
| High risk | 1539 | 77 | 9.9% (7.6%−12.7%) | 2.19 (1.59–3.00) | 1.63 (1.07–2.48) |
Abbreviation: CI, confidence intervals; RS, recurrence score.
Adjusted for age, race, Charlson Comorbidity Index, nodal status, tumor size, progesterone receptor status, tumor grade and lymphovascular invasion
New Oncotype DX category: low risk if RS <26 and node negative or RS <11 and node positive; high risk if RS >25 and node negative or RS >10 and node positive
We also estimated the prognostic value of ODX among all women who received Oncotype DX test (n=144,357). The mean RS was 17.1 (SD=10.0) and it was slightly lower in patients with node positive than those with node negative diseases (Supplementary Table 1). RS was significantly higher in African Americans than in other racial/ethnic groups. Compared with patients with low-risk RS, the HR for patients with intermediate RS and high-risk RS was 0.95 (95% CI 0.88–1.03) and 2.26 (95% CI 2.06–2.47), respectively. The C-index was 0.574 (Table 3). The prognostic performance of ODX was slightly better in node positive patients than in node negative patients when analyzing RS as a categorical variable (p=0.012), but not statistically significant when analyzing RS as a continuous variable (p=0.08). There was a statistically significant difference in the performance of ODX test across racial/ethnic groups (p<0.001), with a better performance in non-Hispanic Whites (HR=1.33 for 10 unit increase of RS) than in African Americans (HR=1.18 for 10 unit increase of RS) and Hispanics (HR=1.21 for 10 unit increase of RS), but the number of patients in minority populations, especial Asian Americans and Hispanics, are limited.
Table 3.
Prognostic value of Oncotype DX in all subjects (n=144,357)
| # of patients | # of death | Hazard ratio (95% CI) | C-index (95% CI) | |
|---|---|---|---|---|
| All subjects | ||||
| Low risk (RS 0–10) | 33,731 | 803 | 1.00 (ref.) | 0.574 (0.565–0.583) |
| Intermediate risk (RS 11–25) | 89,871 | 2,166 | 0.95 (0.88–1.03) | |
| High risk (RS >25) | 20,755 | 1,201 | 2.26 (2.06–2.47) | |
| Per 10 unit increase | 1.32 (1.29–1.35) | 0.577 (0.566–0.588) | ||
| Node negative | ||||
| Low risk (RS 0–10) | 27,795 | 634 | 1.00 (ref.) | 0.578 (0.567–0.589) |
| Intermediate risk (RS 11–25) | 73,951 | 1,609 | 0.90 (0.82–0.98) | |
| High risk (RS >25) | 17,582 | 928 | 2.13 (1.93–2.36) | |
| Per 10 unit increase | 1.31 (1.28–1.35) | 0.573 (0.560–0.585) | ||
| Node positive | ||||
| Low risk (RS 0–10) | 5,936 | 169 | 1.00 (ref.) | 0.581 (0.562–0.599) |
| Intermediate risk (RS 11–25) | 15,920 | 557 | 1.15 (0.97–1.36) | |
| High risk (RS >25) | 3,173 | 273 | 2.94 (2.43–3.56) | |
| Per 10 unit increase | 1.38 (1.31–1.44) | 0.593 (0.571–0.615) | ||
| Non-Hispanic White | ||||
| Low risk (RS 0–10) | 28,534 | 679 | 1.00 (ref.) | 0.578 (0.568–0.588) |
| Intermediate risk (RS 11–25) | 76,244 | 1,844 | 0.96 (0.88–1.04) | |
| High risk (RS >25) | 16,905 | 1029 | 2.39 (2.17–2.64) | |
| Per 10 unit increase | 1.33 (1.30–1.37) | 0.581 (0.570–0.593) | ||
| African American | ||||
| Low risk (RS 0–10) | 2,420 | 92 | 1.00 (ref.) | 0.545 (0.517–0.573) |
| Intermediate risk (RS 11–25) | 6,519 | 228 | 0.89 (0.70–1.13) | |
| High risk (RS >25) | 2,147 | 127 | 1.44 (1.10–1.88) | |
| Per 10 unit increase | 1.18 (1.11–1.26) | 0.535 (0.502–0.569) | ||
| Hispanic | ||||
| Low risk (RS 0–10) | 1,475 | 18 | 1.00 (ref.) | 0.519 (0.465–0.573) |
| Intermediate risk (RS 11–25) | 3,811 | 64 | 1.23 (0.73–2.07) | |
| High risk (RS >25) | 865 | 18 | 1.41 (0.73–2.70) | |
| Per 10 unit increase | 1.21 (1.05–1.40) | 0.542 (0.477–0.608) | ||
| Asian American | ||||
| Low risk (RS 0–10) | 2,982 | 14 | 1.00 (ref.) | 0.617 (0.545–0.689) |
| Intermediate risk (RS 11–25) | 3,297 | 30 | 0.78 (0.41–1.47) | |
| High risk (RS >25) | 838 | 27 | 2.57 (1.35–4.90) | |
| Per 10 unit increase | 1.52 (1.30–1.78) | 0.586 (0.496–0.676) |
DISCUSSION
Genomic tumor profiling gives a new dimension of risk prediction, compared with Nottingham Prognostic Index24, PREDICT25, and Adjuvant Online26 which rely on patients’ clinicopathological features in prognosticating outcomes and predicting the benefit of therapy27, 28. In 2017, AJCC recognized the need to incorporate gene expression prognostic panels into the TNM staging system (eight edition).29 Of all the available multigene assays, Mammaprint and ODX were strongly recommended by ASCO/NCCN as they have been extensively validated retrospectively and prospectively. Our study shows an upward trend in the utilization of multigene assay in clinical practice over 6 years, with more than half of patients receiving multigene assay in 2016. The new AJCC 8 strongly encourages the use of multigene testing for proper staging and treatment planning, so it is expected to see a further increase in the multigene testing utilization in future.
Since multigene biomarker assays are being used to guide precise therapy decision, Mammaprint and ODX are expected to have high agreement for treatment planning to boost clinicians’ confidence in the choice of shared decision making with patients. Fan et al assessed the concordance of multiple gene assays using breast cancer tissue samples, and they showed that 77% of patients were assigned to the same risk category.12 In addition, Nunes et al showed higher discordance between Mammaprint and ODX in patients of African Ancestry, with 8 (42%) tumors with a low RS classified as high risk by MammaPrint.30 The prospective OPTIMA trial was designed to evaluate the performance of these tests31, and the preliminary results of 300 patients showed a Kappa statistic of 0.40 (95% CI: 0.30–0.49)13. The discordance between the two tests may be because (i) they have only one gene, SCUBE2, in common32, (ii) Mammaprint, a 70-gene assay, included more genes than ODX which is a 21 gene assay, (iii) ODX relied more on genes in the proliferative pathway which did not appear to be weighted in the Mammaprint assay. The lack of concordance between these tests continues to raise controversial issues and deep concerns about the precision of our tests. The PROMIS study showed that 44.5 % of patients with RS 18–30 were reclassified as low risk and 55.5% were classified as high risk according to Mammaprint and the physicians in the PROMIS study reported a higher confidence level with their therapy decision33.
The results from our study are very important in clinical practice as it compares the prognostic value between Mammaprint and ODX using the real world data. Although the two tests are reportedly moderately concordant, our study shows that the prognostic value of these tests is comparable. After propensity score matching, the separation between risk groups in survival was slightly wider in Mammaprint compared to ODX. The C-index, which measures the ability of biomarkers to distinguish between risk groups, was similar between the two tests: 0.614 for Mammaprint, 0.581 for 3-category ODX, and 0.608 for 2-cateogry ODX in the matched cohorts. In the entire cohort of patients who received ODX, the C-index for 3-category DOX was 0.574, suggesting that matching and patient selection did not affect the result substantively.
In subgroup analysis of the prognostic value of ODX, ODX was more likely to prognosticate non-Hispanic White women compared to patients of African Ancestry and Hispanic patients. The relatively poor prognostication in patients of African Ancestry by ODX is consistent with TAILORx study, solidifying the concept of tumor biology differences being an additional notable factor to breast cancer disparity. This should be an impetus to firmly involve patients of minority populations in clinical trials and it should be cautious when generalize the management strategy of the heterogeneous endocrine breast cancer across racial groups. It is unknown why the prognostic value of ODX is lower in ethnic minority populations. It could be due to different extent of heterogeneity in the tumor biology across populations or it is due to difference in treatment following ODX test. Although we have shown that there was no racial difference in initiating chemotherapy use after ODX test14, variation between ethnic groups in other treatment incompliance may exist.
Our study was strengthened by the large, nationwide sample with sufficiently complete data on ODX for the years of analysis, and propensity score matching for ODX and Mammaprint cohorts balanced in patient demographic and clinical characteristics. Our study was subject to several limitations. First, the predictive values of these tests were not compared due to small sample size. Second, the NCDB collects data from CoC-accredited cancer programs, which may limit the generalizability of our findings. CoC-approved hospitals are larger, more frequently located in urban locations, and have more cancer-related services available than non-CoC hospitals. Hence, our findings may be limited to CoC-approved hospitals. Third, due to small sample size, we did not compare the prognostic value of Mammaprint across race/ethnicity.
In summary, the use of multigene genomic tests has been increasing in the United States and almost half of the patients with early stage hormone receptor positive breast cancer received Oncotype Dx in 2016. Use of Mammaprint test was distant second at 3.3% in 2016. The study demonstrated that Oncotype Dx and Mammaprint had similar prognostic ability in identifying low risk individuals who could be spared chemotherapy. Sub-optimal performance of Oncotype Dx in ethnic minority deserves further investigation.
Supplementary Material
Acknowledgements:
This research was partially supported by the Breast Cancer Research Foundation (BCRF) and Paul Calabresi K12 Career Development Award for Clinical Oncology (PCACO). The National Cancer Data Base (NCDB) is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. The CoC’s NCDB and the hospitals participating in the CoC NCDB are the source of the de-identified data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
Footnotes
The authors had no conflict of interest.
References
- 1.Brenton JD, Carey LA, Ahmed AA, Caldas C. Molecular classification and molecular forecasting of breast cancer: ready for clinical application? Journal of Clinical Oncology. 2005;23: 7350–7360. [DOI] [PubMed] [Google Scholar]
- 2.Paik S. Development and clinical utility of a 21-gene recurrence score prognostic assay in patients with early breast cancer treated with tamoxifen. The Oncologist. 2007;12: 631–635. [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. New England Journal of Medicine. 2004;351: 2817–2826. [DOI] [PubMed] [Google Scholar]
- 4.Ramsey SD, Barlow WE, Gonzalez-Angulo AM, et al. Integrating comparative effectiveness design elements and endpoints into a phase III, randomized clinical trial (SWOG S1007) evaluating oncotypeDX-guided management for women with breast cancer involving lymph nodes. Contemporary clinical trials. 2013;34: 1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Van’t Veer LJ, Dai H, Van De Vijver MJ, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature. 2002;415: 530. [DOI] [PubMed] [Google Scholar]
- 6.Mook S, Schmidt MK, Viale G, et al. The 70-gene prognosis-signature predicts disease outcome in breast cancer patients with 1–3 positive lymph nodes in an independent validation study. Breast Cancer Res Treat. 2009;116: 295–302. [DOI] [PubMed] [Google Scholar]
- 7.Cardoso F, van’t Veer LJ, Bogaerts J, et al. 70-gene signature as an aid to treatment decisions in early-stage breast cancer. New England Journal of Medicine. 2016;375: 717–729. [DOI] [PubMed] [Google Scholar]
- 8.Sparano JA, Gray RJ, Makower DF, et al. Adjuvant chemotherapy guided by a 21-gene expression assay in breast cancer. New England Journal of Medicine. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Telli ML, Gradishar WJ, Ward JH. NCCN Guidelines Updates: Breast Cancer. J Natl Compr Canc Netw. 2019;17: 552–555. [DOI] [PubMed] [Google Scholar]
- 10.Andre F, Ismaila N, Henry NL, et al. Use of Biomarkers to Guide Decisions on Adjuvant Systemic Therapy for Women With Early-Stage Invasive Breast Cancer: ASCO Clinical Practice Guideline Update-Integration of Results From TAILORx. J Clin Oncol. 2019: Jco1900945. [DOI] [PubMed] [Google Scholar]
- 11.Albain KS, Barlow WE, Shak S, et al. Prognostic and predictive value of the 21-gene recurrence score assay in postmenopausal women with node-positive, oestrogen-receptor-positive breast cancer on chemotherapy: a retrospective analysis of a randomised trial. The Lancet Oncology. 2010;11: 55–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression–based predictors for breast cancer. New England Journal of Medicine. 2006;355: 560–569. [DOI] [PubMed] [Google Scholar]
- 13.Bartlett J, Bayani J, Marshall A, et al. Comparing breast cancer multiparameter tests in the OPTIMA prelim trial: no test is more equal than the others. JNCI: Journal of the National Cancer Institute. 2016;108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.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: 207–220. [DOI] [PubMed] [Google Scholar]
- 15.Albain K, Gray R, Sparano J, et al. Abstract GS4–07: Race, ethnicity and clinical outcomes in hormone receptor-positive, HER2-negative, node-negative breast cancer: results from the TAILORx trial: AACR, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med. 2004;351: 2817–2826. [DOI] [PubMed] [Google Scholar]
- 17.Dowsett M, Cuzick J, Wale C, et al. Prediction of risk of distant recurrence using the 21-gene recurrence score in node-negative and node-positive postmenopausal patients with breast cancer treated with anastrozole or tamoxifen: a TransATAC study. Journal of Clinical Oncology. 2010;28: 1829–1834. [DOI] [PubMed] [Google Scholar]
- 18.Troester MA, Sun X, Allott EH, et al. Racial differences in PAM50 subtypes in the Carolina Breast Cancer Study. JNCI: Journal of the National Cancer Institute. 2017;110: 176–182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Raval MV, Bilimoria KY, Stewart AK, Bentrem DJ, Ko CY. Using the NCDB for cancer care improvement: an introduction to available quality assessment tools. Journal of Surgical Oncology. 2009;99: 488–490. [DOI] [PubMed] [Google Scholar]
- 20.Wolff AC, Hammond MEH, Hicks DG, et al. Recommendations for human epidermal growth factor receptor 2 testing in breast cancer: American Society of Clinical Oncology/College of American Pathologists clinical practice guideline update. Archives of Pathology and Laboratory Medicine. 2013;138: 241–256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity‐score matched samples. Statistics in medicine. 2009;28: 3083–3107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gluz O, Nitz UA, Christgen M, et al. West German Study Group Phase III PlanB Trial: first prospective outcome data for the 21-gene recurrence score assay and concordance of prognostic markers by central and local pathology assessment. J Clin Oncol. 2016;34: 2341–2349. [DOI] [PubMed] [Google Scholar]
- 23.Ibraheem AF, Press DJ, Olopade OI, Huo D. Community clinical practice patterns and mortality in patients with intermediate oncotype DX recurrence scores: Who benefits from chemotherapy? Cancer. 2019;125: 213–222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Galea MH, Blamey RW, Elston CE, Ellis IO. The Nottingham Prognostic Index in primary breast cancer. Breast Cancer Res Treat. 1992;22: 207–219. [DOI] [PubMed] [Google Scholar]
- 25.Wishart G, Bajdik C, Azzato E, et al. A population-based validation of the prognostic model PREDICT for early breast cancer. European Journal of Surgical Oncology (EJSO). 2011;37: 411–417. [DOI] [PubMed] [Google Scholar]
- 26.Ravdin PM, Siminoff LA, Davis GJ, et al. Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. Journal of Clinical Oncology. 2001;19: 980–991. [DOI] [PubMed] [Google Scholar]
- 27.Oakman C, Santarpia L, Di Leo A. Breast cancer assessment tools and optimizing adjuvant therapy. Nature Reviews Clinical Oncology. 2010;7: 725. [DOI] [PubMed] [Google Scholar]
- 28.Sotiriou C, Pusztai L. Gene-expression signatures in breast cancer. New England Journal of Medicine. 2009;360: 790–800. [DOI] [PubMed] [Google Scholar]
- 29.Giuliano AE, Connolly JL, Edge SB, et al. Breast Cancer—Major changes in the American Joint Committee on Cancer eighth edition cancer staging manual. CA: a cancer journal for clinicians. 2017. [DOI] [PubMed] [Google Scholar]
- 30.Nunes RA, Wray L, Mete M, et al. Genomic profiling of breast cancer in African-American women using MammaPrint. Breast Cancer Res Treat. 2016;159: 481–488. [DOI] [PubMed] [Google Scholar]
- 31.Bartlett J, Canney P, Campbell A, et al. Selecting breast cancer patients for chemotherapy: the opening of the UK OPTIMA trial. Clinical oncology. 2013;25: 109–116. [DOI] [PubMed] [Google Scholar]
- 32.Vieira AF, Schmitt F. An update on breast cancer multigene prognostic tests-emergent clinical biomarkers. Frontiers in medicine. 2018;5: 248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Tsai M, Lo S, Audeh W, et al. Association of 70-Gene Signature Assay Findings With Physicians’ Treatment Guidance for Patients With Early Breast Cancer Classified as Intermediate Risk by the 21-Gene Assay. JAMA Oncol. 2018;4: e173470. [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.