Purpose: Post-mastectomy radiotherapy (PMRT) yields improvements in both locoregional control and overall survival (OS) for women with T1-2 N1 breast cancer. The value of PMRT in this population has been questioned given advances in systemic therapy. The 21-gene recurrence score (RS) assay was evaluated as a predictor of OS among women with T1-2 N1 breast cancer who received or did not receive PMRT.

Experimental Design: An observational cohort study was performed on women with T1-2 N1 estrogen receptor–positive breast cancer from the National Cancer Database (NCDB) and, as a validation cohort, from the surveillance, epidemiology, and end results (SEER) registry who underwent mastectomy and were evaluated for RS. Multivariable parametric accelerated failure time models were used to estimate associations of RS and PMRT with OS using propensity score-adjusted matched cohorts.

Results: In both the NCDB (N = 7,332) and SEER (N = 3,087) cohorts, there was a significant interaction of RS and PMRT with OS (P = 0.009 and P = 0.03, respectively). PMRT was associated with longer OS in women with a low RS [NCDB: time ratio (TR) = 1.70; 95% CI (confidence interval), 1.30–2.22; P < 0.001; SEER: TR = 1.85; 95% CI, 1.33–2.57; P < 0.001], but not in women with an intermediate RS (NCDB: TR = 0.89; 95% CI, 0.69–1.14; P = 0.35; SEER: TR = 0.84; 95% CI, 0.62–1.14; P = 0.26), or a high RS (NCDB: TR = 1.10; 95% CI, 0.91–1.34; P = 0.33; SEER: TR = 0.79; 95% CI, 0.50–1.23; P = 0.28).

Conclusions: Longer survival associated with PMRT was limited to women with a low RS. PMRT may confer the greatest OS benefit for patients at the lowest risk of distant recurrence. These results caution against omission of PMRT among women with low RS. Clin Cancer Res; 24(16); 3878–87. ©2018 AACR.

This article is featured in Highlights of This Issue, p. 3785

Translational Relevance

The American Society of Clinical Oncology focused guideline update (2017) concluded that post-mastectomy radiotherapy (PMRT) may be potentially omitted for women with T1–2 N1 breast cancer at a low risk of locoregional recurrence. The 21-gene recurrence score (RS) assay has been presented previously as a potential means of identifying patients with a low risk of locoregional recurrence for whom PMRT could be omitted. In this analysis of two large national cancer databases, patients with a low RS derived a greater survival benefit from PMRT than those with an intermediate or high RS. This may be due to a low competing risk of subclinical micrometastatic disease at diagnosis resulting in improved translation of locoregional control to a survival benefit. These results caution against omission of PMRT for women with node-positive disease on the basis of a low-risk RS, and strongly suggest the need for prospective validation prior to widespread adoption.

Post-mastectomy radiotherapy (PMRT) for women with T1–2 breast cancer and one to three positive axillary nodes (N1) remains controversial. Although PMRT has been demonstrated in multiple randomized trials to reduce the risk of locoregional recurrence (LRR) and breast cancer mortality in women with limited nodal disease, given advances in adjuvant systemic therapy it is thought that for a subset of low-risk women the potential late toxicity of PMRT may be greater than its absolute benefit on LRR (1–4). Current national guidelines recommend PMRT primarily for “high-risk” women under the assumption that those with the greatest risk of LRR will derive the greatest survival benefit from adjuvant locoregional therapy (5).

There is considerable interest in identifying prognostic factors alongside standard clinicopathologic variables that are potentially predictive of benefit for adjuvant therapy in early-stage breast cancer. The 21-gene recurrence score (RS) assay (Genomic Health) is calculated on the basis of the RNA expression levels of 21 genes and has been validated in T1-2 N1 breast cancer to be prognostic of LRR, disease-free survival, and overall survival (OS; refs. 6–10). Other gene profiling assays have demonstrated similar correlations (11–13). More recently, the 21-gene RS has been shown in retrospective analyses to be predictive of survival benefit for adjuvant chemotherapy in women with T1-2 N1 cancer with a high RS, while those with a low RS did not derive a significant benefit (6). By extension, it is hypothesized that PMRT may provide the greatest benefit for LRR, and therefore survival, in women with a high RS (10).

Potentially confounding this hypothesis, the 21-gene RS is also strongly prognostic for distant recurrence, a competing risk to LRR (8, 9). Women with a higher RS are more likely to harbor occult systemic disease and therefore potentially less likely to derive a survival benefit from locoregional treatment. To test this concept, women with T1-2 N1 estrogen receptor (ER)–positive breast cancer who underwent mastectomy were identified from the National Cancer Database (NCDB) and, as a validation cohort, from the surveillance, epidemiology, and end results (SEER) registry. Survival analyses were performed to examine the associations and interactions of the 21-gene RS and use of PMRT with OS in these cohorts.

Women with pathologic T1-2 ER-positive breast cancer with one to three positive nodes (including micrometastatic nodal disease) that underwent mastectomy and were evaluated using the RS in the 2004–2014 NCDB and the 2004–2014 SEER 18 registry were included (Fig. 1). The NCDB is a nationwide, facility-based comprehensive clinical surveillance resource oncology dataset established by the Commission on Cancer of the American College of Surgeons and the American Cancer Society in 1989 that captures 70% of all newly diagnosed malignancies (14). The American College of Surgeons has executed a Business Associate Agreement that includes a data use agreement with each of its Commission on Cancer accredited hospitals. Deidentified data from the National Cancer Institute SEER registry, including radiation and chemotherapy variables, were linked to 21-gene Oncotype DX breast RS assay results from the Genomic Health Clinical Laboratory (2004–2014) by Information Management Services. The deidentified linked dataset was provided to the investigators after SEER approval of a custom data request (15). All investigators with access to the dataset signed a Data-Use Agreement prior to receiving access. Local institutional review board approval and informed consent were not required for these analyses of deidentified data from the NCDB and SEER registry.

Figure 1.

CONSORT diagram of cohort selection within the NCDB (A) and SEER registry (B). PUF, participant user files; T1-2, tumor stage, including tumors between 0.1 cm–5 cm; N1mi-1, nodal stage, including micrometastatic nodal disease and one to three axillary lymph nodes; M, metastasis stage; RS, 21-gene RS assay risk group. ER, estrogen receptor; PMRT, post-mastectomy radiotherapy.

Figure 1.

CONSORT diagram of cohort selection within the NCDB (A) and SEER registry (B). PUF, participant user files; T1-2, tumor stage, including tumors between 0.1 cm–5 cm; N1mi-1, nodal stage, including micrometastatic nodal disease and one to three axillary lymph nodes; M, metastasis stage; RS, 21-gene RS assay risk group. ER, estrogen receptor; PMRT, post-mastectomy radiotherapy.

Close modal

Exclusion criteria were receipt of neoadjuvant chemotherapy, no adjuvant antiendocrine therapy, unknown radiation status, radiation to primary sites other than the breast/chest wall, radiation to a documented dose <45 Gy, or follow-up time <2 months. Predictor variables of interest were RS and PMRT. RS was defined as “low-risk” (score <18), “intermediate-risk”(18–30), or “high-risk” (>30; ref. 16). Covariates were included in the creation of the matched cohorts and to adjust for potential confounding during regression analyses (Supplementary Methods).

Statistical analyses

Baseline characteristics between patient groups were compared using the Fisher exact test for categorical data, the Mann–Whitney U test for nonnormally distributed numeric or ordinal data, or the t test or ANOVA for normally distributed data. Five years restricted mean survival times for OS and 5-year OS proportions were estimated using the Kaplan–Meier method and compared with the log-rank test. Survival curves were plotted as unadjusted Kaplan–Meier estimates.

Multivariable parametric accelerated failure time (AFT) models using the generalized gamma distribution were used to evaluate the association of RS and PMRT with OS. This model was chosen in place of the Cox proportional hazards (PHs) model due to the presence of significant non-PH between women within each RS subgroup based on use of PMRT when the Cox model was used for the multivariable analyses (17). When the PH assumption is found to be violated using the Cox model, the AFT model for multivariable analyses provides better goodness-of-fit to the observed data and therefore more robust statistical inference (18–21). The AFT model estimates the time ratio (TR), which describes the multiplicative factor by which the time-to-event is related between two groups. A TR >1 describes longer survival. Covariates utilized in analyses were selected a priori based on clinical knowledge and availability and are described in detail in Supplementary Methods. The prespecified statistical plan was to determine if there was a significant interaction between RS and PMRT and, if so, to perform the subset analyses of PMRT effect within each RS group. The interaction between RS and PMRT was tested using the likelihood ratio test based on the fitted model.

To reduce potential confounding, nearest neighbor propensity score-matching was performed for each RS subgroup using available variables as described in detail in the Supplementary Methods to generate well-balanced matched cohorts based on the receipt of PMRT for both the NCDB and SEER cohorts. Baseline characteristics of the generated matched cohorts are presented in Supplementary Tables S1 and S2 (22). Propensity score-matched and inverse probability-weighted cohort analyses were performed to reduce treatment-assignment biases related to measured covariates (22). Analyses of RS subsets were performed both for all patients as well as for the matched patient cohorts within each RS risk group. A composite variable was made combining RS and PMRT to calculate TR estimates relative to a single reference level (low RS/no PMRT). All statistical tests were two-tailed with an alpha of 0.05 used as the cutoff for statistical significance.

A series of separate sensitivity analyses were performed on a subgroup of patients whose receipt of chemotherapy was consistent with their RS score, as well as a subgroup of patients with Her2-negative disease. To test the robustness of the observed results, sensitivity of TR estimates to a possible unmeasured confounder was explored (23, 24). Statistical analyses were performed in Rstudio v1.0.143 using MatchIt, survival, flexsurv, Hmisc, and rms packages (19, 25–30).

Cohorts

The NCDB and SEER began reporting 21-gene RS assay results in 2005 and 2004, respectively. A documented RS was identified in 11.1% (11,181/65, 873) and 9.5% (3,752/32, 551) of eligible women, respectively, within the NCDB and SEER registry. The majority of cases within the NCDB (98.6%) and SEER (74.7%) were coded in 2010 or after. Of the identified cases, 53.3% (N = 3,907) and 58.6% (N = 1,803) had a low-risk RS within the NCDB and SEER cohorts, respectively (Fig. 1). The median RS was 12 for the low-risk group (IQR = 8–15), 21 for the intermediate-risk group (IQR = 19–25), and 38 for the high-risk group (IQR = 33–45).

Clinicopathologic and demographic characteristics of women within each RS subgroup compared by receipt of PMRT are shown for both cohorts (Tables 1 and 2). Omission of PMRT among women in the NCDB cohort due to documented contraindication or refusal was not associated with RS (Kruskal–Wallis, P = 0.21), increased age (one-way ANOVA, P = 0.43), or higher comorbidity scores (Kruskal–Wallis, P = 0.28). Clinicopathologic and treatment characteristics of women within each RS subgroup compared by receipt of PMRT were well balanced for all matched cohorts (Supplementary Tables S1 and S2).

Table 1.

Characteristics of patients from the NCDB cohort grouped by 21-gene RS assay risk group

Low-riskIntermediate-riskHigh-risk
VariableNo PMRT (N = 2,907)PMRT (N = 1,000)PSMDNo PMRT (N = 1,303)PMRT (N = 506)PSMDNo PMRT (N = 1,136)PMRT (N = 470)PSMD
Age (years) 60 [50–68] 58 [49–65] <0.001 0.19 59 [51–67] 56 [48–66] <0.001 0.22 61 [52–68] 58 [49–67] 0.001 0.18 
Follow-up (months) 34 [24–46] 33 [24–46] 0.80 0.01 36 [24–49] 34 [24–45] 0.05 0.10 37 [26–48] 36 [25–47] 0.50 0.03 
Tumor stage         
 pT1 1,667 (57.3) 473 (47.3) <0.001 0.20 659 (50.2) 197 (38.9) <0.001 0.23 595 (52.4) 180 (38.3) <0.001 0.27 
 pT2 1,240 (42.7) 527 (52.7)   654 (49.8) 309 (61.1)   541 (47.6) 290 (61.7)   
Nodal stage           
 pN1mi 1,079 (37.1) 160 (16.0) <0.001 0.49 428 (32.6) 103 (20.4) <0.001 0.28 363 (32.0) 95 (20.2) <0.001 0.27 
 pN1 1,828 (62.9) 840 (84.0)   885 (67.4) 403 (79.6)   773 (68.0) 375 (79.8)   
Grade         
 1 822 (28.3) 244 (24.4) 0.02  249 (19.0) 77 (15.2) 0.21  237 (20.9) 81 (17.2) 0.009  
 2 1,715 (59.0) 597 (59.7)  0.11 712 (54.2) 291 (57.5)  0.11 567 (49.9) 234 (49.8)  0.19 
 3 220 (7.6) 92 (9.2)   282 (21.5) 105 (20.8)   270 (23.8) 141 (30.0)   
 Unknown 150 (5.2) 67 (6.7)   70 (5.3) 33 (6.5)   62 (5.5) 14 (3.0)   
LVI         
 Negative 1,782 (61.3) 544 (54.4) <0.001 0.15 702 (53.5) 245 (48.4) 0.06 0.13 639 (56.2) 222 (47.2) 0.001 0.21 
 Positive 736 (25.3) 319 (31.9)   420 (32.0) 192 (37.9)   326 (28.7) 180 (38.3)   
 Unknown 389 (13.4) 137 (13.7)   191 (14.5) 69 (13.6)   171 (15.1) 68 (14.5)   
Histology         
 IDC 2,023 (69.6) 660 (66.0) 0.16  965 (73.5) 356 (70.4) 0.13  854 (75.2) 346 (73.6) 0.27  
 ILC 433 (14.9) 174 (17.4)  0.08 166 (12.6) 76 (15.0)  0.13 135 (11.9) 49 (10.4)  0.11 
 IDC/ILC 393 (13.5) 143 (14.3)   151 (11.5) 68 (13.4)   111 (9.8) 61 (13.0)   
 Other 58 (2.0) 23 (2.3)   31 (2.4) 6 (1.2)   36 (3.2) 14 (3.0)   
PR status         
 PR+ 2,838 (97.6) 970 (97.0) 0.29 0.04 1,143 (87.1) 455 (89.9) 0.11 0.09 992 (87.3) 399 (84.9) 0.20 0.07 
 PR- 69 (2.4) 30 (3.0)   170 (12.9) 51 (10.1)   144 (12.7) 71 (15.1)   
Her2 status         
 Her2- 2,841 (97.7) 978 (97.8) 0.36 0.06 1,255 (95.6) 492 (97.2) 0.30 0.09 1,039 (91.5) 435 (92.6) 0.44 0.08 
 Her2+ 36 (1.2) 8 (0.8)   31 (2.4) 7 (1.4)   53 (4.7) 23 (4.9)   
 Unknown 30 (1.0) 14 (1.4)   27 (2.1) 7 (1.4)   44 (3.9) 12 (2.6)   
Chemotherapy         
 No 2,338 (80.4) 632 (63.2) <0.001 0.40 615 (46.8) 148 (29.2) <0.001 0.37 620 (54.6) 170 (36.2) <0.001 0.38 
 Yes 518 (17.8) 347 (34.7)   689 (52.5) 353 (69.8)   487 (42.9) 288 (61.3)   
 Unknown 51 (1.8) 21 (2.1)   9 (0.7) 5 (1.0)   29 (2.6) 12 (2.6)   
Race         
 White 2,568 (88.3) 880 (88.0) 0.95  1,161 (88.4) 433 (85.6) 0.22  991 (87.2) 395 (84.0) 0.02  
 Black 200 (6.9) 74 (7.4)  0.02 93 (7.1) 50 (9.9)  0.11 100 (8.8) 39 (8.3)  0.16 
 Asian/PI 78 (2.7) 26 (2.6)   33 (2.5) 11 (2.2)   25 (2.2) 17 (3.6)   
 Other/Unknown 61 (2.1) 20 (2.0)   26 (2.0) 12 (2.4)   20 (1.8) 19 (4.0)   
Hispanic         
 No 2,680 (92.2) 930 (93.0) 0.39 0.05 1,216 (92.6) 471 (93.1) 0.84 0.03 1,043 (91.8) 428 (91.1) 0.59 0.05 
 Yes 122 (4.2) 43 (4.3)   45 (3.4) 18 (3.6)   48 (4.2) 25 (5.3)   
 Unknown 105 (3.6) 27 (2.7)   52 (4.0) 17 (3.4)   45 (4.0) 17 (3.6)   
Insurance status         
 Private 1,722 (59.2) 648 (64.8) <0.001  781 (59.5) 326 (64.4) 0.32  629 (55.4) 280 (59.6) 0.23  
 Medicare 940 (32.3) 256 (25.6)  0.17 409 (31.2) 133 (26.3)  0.11 401 (35.3) 145 (30.9)  0.14 
 Medicaid 171 (5.9) 71 (7.1)   77 (5.9) 29 (5.7)   69 (6.1) 35 (7.4)   
 None 53 (1.8) 13 (1.3)   30 (2.3) 13 (2.6)   25 (2.2) 8 (1.7)   
 Unknown 21 (0.7) 12 (1.2)   16 (1.2) 5 (1.0)   12 (1.1) 2 (0.4)   
Comorbidity score         
 0 2,376 (81.7) 828 (82.8) 0.75 0.03 1,097 (83.5) 436 (86.2) 0.37 0.08 903 (79.5) 378 (80.4) 0.89 0.03 
 1 431 (14.8) 139 (13.9)   180 (13.7) 57 (11.3)   185 (16.3) 72 (15.3)   
 2 100 (3.4) 33 (3.3)   36 (2.7) 13 (2.6)   48 (4.2) 20 (4.3)   
Prior cancer         
 No 2,100 (72.2) 760 (76.0) 0.02 0.09 979 (74.6) 393 (77.7) 0.18 0.07 833 (73.3) 376 (80.0) 0.005 0.16 
 Yes 807 (27.8) 240 (24.0)   334 (25.4) 113 (22.3)   303 (26.7) 94 (20.0)   
Facility type           
 Academic 1,017 (35.0) 326 (32.6) 0.11  467 (35.6) 180 (35.6) 0.92  386 (34.0) 138 (29.4) 0.26  
 Community 221 (7.6) 83 (8.3)  0.10 103 (7.8) 38 (7.5)  0.05 104 (9.2) 52 (11.1)  0.12 
 Comprehensive 1,279 (44.0) 480 (48.0)   594 (45.2) 229 (45.3)   520 (45.8) 217 (46.2)   
 Integrated 370 (12.7) 105 (10.5)   137 (10.4) 52 (10.3)   112 (9.9) 55 (11.7)   
 Unknown 20 (0.7) 6 (0.6)   12 (0.9) 7 (1.4)   14 (1.2) 8 (1.7)   
Income quartile         
 Top 1,187 (40.8) 401 (40.1) 0.98  498 (37.9) 190 (37.5) 0.87  412 (36.3) 171 (36.4) 0.93  
 2nd 771 (26.5) 271 (27.1)  0.02 360 (27.4) 148 (29.2)  0.04 310 (27.3) 134 (28.5)  0.04 
 3rd 601 (20.7) 207 (20.7)   279 (21.2) 104 (20.6)   224 (19.7) 92 (19.6)   
 Bottom 348 (12.0) 121 (12.1)   176 (13.4) 64 (12.6)   190 (16.7) 73 (15.5)   
Education quartile         
 Top 940 (32.3) 348 (34.8) 0.20  422 (32.1) 168 (33.2) 0.60  320 (28.2) 132 (28.1) 0.20  
 2nd 1,031 (35.5) 339 (33.9)  0.08 427 (32.5) 164 (32.4)  0.07 368 (32.4) 171 (36.4)  0.12 
 3rd 630 (21.7) 195 (19.5)   277 (21.1) 94 (18.6)   282 (24.8) 115 (24.5)   
 Bottom 306 (10.5) 118 (11.8)   187 (14.2) 80 (15.8)   166 (14.6) 52 (11.1)   
Year         
 2010 and before 536 (18.4) 145 (14.5) 0.01  301 (22.9) 74 (14.6) 0.001  237 (20.9) 86 (18.3) 0.57  
 2011 604 (20.8) 226 (22.6)  0.12 311 (23.7) 129 (25.5)  0.21 291 (25.6) 117 (24.9)  0.08 
 2012 856 (29.4) 282 (28.2)   321 (24.4) 139 (27.5)   299 (26.3) 127 (27.0)   
 2013 911 (31.3) 347 (34.7)   380 (28.9) 164 (32.4)   309 (27.2) 140 (29.8)   
Low-riskIntermediate-riskHigh-risk
VariableNo PMRT (N = 2,907)PMRT (N = 1,000)PSMDNo PMRT (N = 1,303)PMRT (N = 506)PSMDNo PMRT (N = 1,136)PMRT (N = 470)PSMD
Age (years) 60 [50–68] 58 [49–65] <0.001 0.19 59 [51–67] 56 [48–66] <0.001 0.22 61 [52–68] 58 [49–67] 0.001 0.18 
Follow-up (months) 34 [24–46] 33 [24–46] 0.80 0.01 36 [24–49] 34 [24–45] 0.05 0.10 37 [26–48] 36 [25–47] 0.50 0.03 
Tumor stage         
 pT1 1,667 (57.3) 473 (47.3) <0.001 0.20 659 (50.2) 197 (38.9) <0.001 0.23 595 (52.4) 180 (38.3) <0.001 0.27 
 pT2 1,240 (42.7) 527 (52.7)   654 (49.8) 309 (61.1)   541 (47.6) 290 (61.7)   
Nodal stage           
 pN1mi 1,079 (37.1) 160 (16.0) <0.001 0.49 428 (32.6) 103 (20.4) <0.001 0.28 363 (32.0) 95 (20.2) <0.001 0.27 
 pN1 1,828 (62.9) 840 (84.0)   885 (67.4) 403 (79.6)   773 (68.0) 375 (79.8)   
Grade         
 1 822 (28.3) 244 (24.4) 0.02  249 (19.0) 77 (15.2) 0.21  237 (20.9) 81 (17.2) 0.009  
 2 1,715 (59.0) 597 (59.7)  0.11 712 (54.2) 291 (57.5)  0.11 567 (49.9) 234 (49.8)  0.19 
 3 220 (7.6) 92 (9.2)   282 (21.5) 105 (20.8)   270 (23.8) 141 (30.0)   
 Unknown 150 (5.2) 67 (6.7)   70 (5.3) 33 (6.5)   62 (5.5) 14 (3.0)   
LVI         
 Negative 1,782 (61.3) 544 (54.4) <0.001 0.15 702 (53.5) 245 (48.4) 0.06 0.13 639 (56.2) 222 (47.2) 0.001 0.21 
 Positive 736 (25.3) 319 (31.9)   420 (32.0) 192 (37.9)   326 (28.7) 180 (38.3)   
 Unknown 389 (13.4) 137 (13.7)   191 (14.5) 69 (13.6)   171 (15.1) 68 (14.5)   
Histology         
 IDC 2,023 (69.6) 660 (66.0) 0.16  965 (73.5) 356 (70.4) 0.13  854 (75.2) 346 (73.6) 0.27  
 ILC 433 (14.9) 174 (17.4)  0.08 166 (12.6) 76 (15.0)  0.13 135 (11.9) 49 (10.4)  0.11 
 IDC/ILC 393 (13.5) 143 (14.3)   151 (11.5) 68 (13.4)   111 (9.8) 61 (13.0)   
 Other 58 (2.0) 23 (2.3)   31 (2.4) 6 (1.2)   36 (3.2) 14 (3.0)   
PR status         
 PR+ 2,838 (97.6) 970 (97.0) 0.29 0.04 1,143 (87.1) 455 (89.9) 0.11 0.09 992 (87.3) 399 (84.9) 0.20 0.07 
 PR- 69 (2.4) 30 (3.0)   170 (12.9) 51 (10.1)   144 (12.7) 71 (15.1)   
Her2 status         
 Her2- 2,841 (97.7) 978 (97.8) 0.36 0.06 1,255 (95.6) 492 (97.2) 0.30 0.09 1,039 (91.5) 435 (92.6) 0.44 0.08 
 Her2+ 36 (1.2) 8 (0.8)   31 (2.4) 7 (1.4)   53 (4.7) 23 (4.9)   
 Unknown 30 (1.0) 14 (1.4)   27 (2.1) 7 (1.4)   44 (3.9) 12 (2.6)   
Chemotherapy         
 No 2,338 (80.4) 632 (63.2) <0.001 0.40 615 (46.8) 148 (29.2) <0.001 0.37 620 (54.6) 170 (36.2) <0.001 0.38 
 Yes 518 (17.8) 347 (34.7)   689 (52.5) 353 (69.8)   487 (42.9) 288 (61.3)   
 Unknown 51 (1.8) 21 (2.1)   9 (0.7) 5 (1.0)   29 (2.6) 12 (2.6)   
Race         
 White 2,568 (88.3) 880 (88.0) 0.95  1,161 (88.4) 433 (85.6) 0.22  991 (87.2) 395 (84.0) 0.02  
 Black 200 (6.9) 74 (7.4)  0.02 93 (7.1) 50 (9.9)  0.11 100 (8.8) 39 (8.3)  0.16 
 Asian/PI 78 (2.7) 26 (2.6)   33 (2.5) 11 (2.2)   25 (2.2) 17 (3.6)   
 Other/Unknown 61 (2.1) 20 (2.0)   26 (2.0) 12 (2.4)   20 (1.8) 19 (4.0)   
Hispanic         
 No 2,680 (92.2) 930 (93.0) 0.39 0.05 1,216 (92.6) 471 (93.1) 0.84 0.03 1,043 (91.8) 428 (91.1) 0.59 0.05 
 Yes 122 (4.2) 43 (4.3)   45 (3.4) 18 (3.6)   48 (4.2) 25 (5.3)   
 Unknown 105 (3.6) 27 (2.7)   52 (4.0) 17 (3.4)   45 (4.0) 17 (3.6)   
Insurance status         
 Private 1,722 (59.2) 648 (64.8) <0.001  781 (59.5) 326 (64.4) 0.32  629 (55.4) 280 (59.6) 0.23  
 Medicare 940 (32.3) 256 (25.6)  0.17 409 (31.2) 133 (26.3)  0.11 401 (35.3) 145 (30.9)  0.14 
 Medicaid 171 (5.9) 71 (7.1)   77 (5.9) 29 (5.7)   69 (6.1) 35 (7.4)   
 None 53 (1.8) 13 (1.3)   30 (2.3) 13 (2.6)   25 (2.2) 8 (1.7)   
 Unknown 21 (0.7) 12 (1.2)   16 (1.2) 5 (1.0)   12 (1.1) 2 (0.4)   
Comorbidity score         
 0 2,376 (81.7) 828 (82.8) 0.75 0.03 1,097 (83.5) 436 (86.2) 0.37 0.08 903 (79.5) 378 (80.4) 0.89 0.03 
 1 431 (14.8) 139 (13.9)   180 (13.7) 57 (11.3)   185 (16.3) 72 (15.3)   
 2 100 (3.4) 33 (3.3)   36 (2.7) 13 (2.6)   48 (4.2) 20 (4.3)   
Prior cancer         
 No 2,100 (72.2) 760 (76.0) 0.02 0.09 979 (74.6) 393 (77.7) 0.18 0.07 833 (73.3) 376 (80.0) 0.005 0.16 
 Yes 807 (27.8) 240 (24.0)   334 (25.4) 113 (22.3)   303 (26.7) 94 (20.0)   
Facility type           
 Academic 1,017 (35.0) 326 (32.6) 0.11  467 (35.6) 180 (35.6) 0.92  386 (34.0) 138 (29.4) 0.26  
 Community 221 (7.6) 83 (8.3)  0.10 103 (7.8) 38 (7.5)  0.05 104 (9.2) 52 (11.1)  0.12 
 Comprehensive 1,279 (44.0) 480 (48.0)   594 (45.2) 229 (45.3)   520 (45.8) 217 (46.2)   
 Integrated 370 (12.7) 105 (10.5)   137 (10.4) 52 (10.3)   112 (9.9) 55 (11.7)   
 Unknown 20 (0.7) 6 (0.6)   12 (0.9) 7 (1.4)   14 (1.2) 8 (1.7)   
Income quartile         
 Top 1,187 (40.8) 401 (40.1) 0.98  498 (37.9) 190 (37.5) 0.87  412 (36.3) 171 (36.4) 0.93  
 2nd 771 (26.5) 271 (27.1)  0.02 360 (27.4) 148 (29.2)  0.04 310 (27.3) 134 (28.5)  0.04 
 3rd 601 (20.7) 207 (20.7)   279 (21.2) 104 (20.6)   224 (19.7) 92 (19.6)   
 Bottom 348 (12.0) 121 (12.1)   176 (13.4) 64 (12.6)   190 (16.7) 73 (15.5)   
Education quartile         
 Top 940 (32.3) 348 (34.8) 0.20  422 (32.1) 168 (33.2) 0.60  320 (28.2) 132 (28.1) 0.20  
 2nd 1,031 (35.5) 339 (33.9)  0.08 427 (32.5) 164 (32.4)  0.07 368 (32.4) 171 (36.4)  0.12 
 3rd 630 (21.7) 195 (19.5)   277 (21.1) 94 (18.6)   282 (24.8) 115 (24.5)   
 Bottom 306 (10.5) 118 (11.8)   187 (14.2) 80 (15.8)   166 (14.6) 52 (11.1)   
Year         
 2010 and before 536 (18.4) 145 (14.5) 0.01  301 (22.9) 74 (14.6) 0.001  237 (20.9) 86 (18.3) 0.57  
 2011 604 (20.8) 226 (22.6)  0.12 311 (23.7) 129 (25.5)  0.21 291 (25.6) 117 (24.9)  0.08 
 2012 856 (29.4) 282 (28.2)   321 (24.4) 139 (27.5)   299 (26.3) 127 (27.0)   
 2013 911 (31.3) 347 (34.7)   380 (28.9) 164 (32.4)   309 (27.2) 140 (29.8)   

NOTE: Data are presented as count (percentage) or median (interquartile range) with significance determined by Fisher exact test or Kruskal–Wallis test; 1mi: 1 (microscopic); grade 1: well differentiated, grade 2: moderately differentiated, grade 3: poorly differentiated/undifferentiated; IDC: invasive ductal carcinoma; ILC: invasive lobular carcinoma; PR: progesterone receptor; and Her2: Her2/Neu receptor.

Table 2.

Characteristics of patients from the SEER cohort grouped by 21-gene RS assay risk group

Low-riskIntermediate-riskHigh-risk
VariableNo PMRT (N = 1,469)PMRT (N = 334)PSMDNo PMRT (N = 829)PMRT (N = 216)PSMDNo PMRT (N = 176)PMRT (N = 54)PSMD
Age (years) 58 [49–66] 57 [48–66] 0.52 0.05 57 [49–66] 56 [48–65] 0.06 0.16 56 [47–66] 53 [44–62] 0.06 0.25 
Follow-up (months) 37 [23–57] 36 [23–52] 0.15 0.10 39 [24–58] 33 [20–49] 0.001 0.24 42 [27–59] 32 [19–55] 0.05 0.23 
Tumor stage         
 pT1 923 (62.8) 161 (48.2) <0.001 0.30 486 (58.6) 95 (44.0) <0.001 0.30 87 (49.4) 23 (42.6) 0.47 0.14 
 pT2 546 (37.2) 173 (51.8)   343 (41.4) 121 (56.0)   89 (50.6) 31 (57.4)   
Nodal stage             
 pN1mi 632 (43.0) 93 (27.8) <0.001 0.32 345 (41.6) 58 (26.9) <0.001 0.32 60 (34.1) 16 (29.6) 0.66 0.10 
 pN1 837 (57.0) 241 (72.2)   484 (58.4) 158 (73.1)   116 (65.9) 38 (70.4)   
Grade         
 1 488 (33.2) 87 (26.0) 0.009  162 (19.5) 25 (11.6) 0.04  8 (4.5) 1 (1.9) 0.43  
 2 825 (56.2) 201 (60.2)  0.21 464 (56.0) 129 (59.7)  0.23 67 (38.1) 25 (46.3)  0.30 
 3 111 (7.6) 39 (11.7)   188 (22.7) 56 (25.9)   97 (55.1) 28 (51.9)   
 Unknown 45 (3.1) 7 (2.1)   15 (1.8) 6 (2.8)   4 (2.3) 0 (0.0)   
Histology             
 IDC 1,090 (74.2) 228 (68.3) 0.16  653 (78.8) 154 (71.3) 0.08  159 (90.3) 44 (81.5) 0.06  
 ILC 224 (15.2) 62 (18.6)  0.14 105 (12.7) 41 (19.0)  0.19 9 (5.1) 3 (5.6)  0.34 
 IDC/ILC 125 (8.5) 34 (10.2)   61 (7.4) 17 (7.9)   3 (1.7) 5 (9.3)   
 Other 30 (2.0) 10 (3.0)   10 (1.2) 4 (1.9)   5 (2.8) 2 (3.7)   
PR status         
 PR+ 1,421 (96.7) 317 (94.9) 0.14 0.09 725 (87.5) 190 (88.0) 0.91 0.02 131 (74.4) 41 (75.9) 1.00 0.04 
 PR- 48 (3.3) 17 (5.1)   104 (12.5) 26 (12.0)   45 (25.6) 13 (24.1)   
Her2 status             
 Her2- 1,056 (71.9) 271 (81.1) <0.001 0.25 584 (70.4) 167 (77.3) 0.04 0.20 106 (60.2) 35 (64.8) 0.20 0.35 
 Her2+ 19 (1.3) 7 (2.1)   10 (1.2) 5 (2.3)   10 (5.7) 0 (0.0)   
 Unknown 394 (26.8) 56 (16.8)   235 (28.3) 44 (20.4)   60 (34.1) 19 (35.2)   
Chemotherapy         
 No/Unknown 1,136 (77.3) 212 (63.5) <0.001 0.31 447 (53.9) 72 (33.3) <0.001 0.42 49 (27.8) 3 (5.6) <0.001 0.63 
 Yes 333 (22.7) 122 (36.5)   382 (46.1) 144 (66.7)   127 (72.2) 51 (94.4)   
Race             
 White 1,232 (83.9) 279 (83.5) 0.59  699 (84.3) 178 (82.4) 0.17  146 (83.0) 45 (83.3) 0.96  
 Black 105 (7.1) 29 (8.7)  0.10 58 (7.0) 24 (11.1)  0.16 15 (8.5) 4 (7.4)  0.12 
 Asian/PI 119 (8.1) 25 (7.5)   68 (8.2) 13 (6.0)   14 (8.0) 5 (9.3)   
 Other/Unknown 13 (0.9) 1 (0.3)   4 (0.5) 1 (0.5)   1 (0.6) 0 (0.0)   
Hispanic         
 No 1,319 (89.8) 309 (92.5) 0.51 0.07 762 (91.9) 193 (89.4) 0.29 0.09 160 (90.9) 50 (92.6) 0.91 0.06 
 Yes 150 (10.2) 25 (7.5)   67 (8.1) 23 (10.6)   16 (9.1) 4 (7.4)   
Insurance status             
 Insured 1,271 (86.5) 286 (85.6) 0.61  700 (84.4)) 191 (88.4) 0.47  139 (79.0) 48 (90.6) 0.40  
 Medicaid 127 (8.6) 32 (9.6)  0.09 69 (8.3) 14 (6.5)  0.14 19 (10.8) 3 (5.7)  0.29 
 None 25 (1.7) 3 (0.9)   25 (3.0) 3 (1.4)   3 (1.7) 1 (1.9)   
 Unknown 46 (3.1) 13 (3.9)   35 (4.2) 8 (3.7)   15 (18.5) 2 (3.7)   
Prior cancer         
 No 1,355 (92.2) 307 (91.9) 0.85 0.01 762 (91.9) 202 (93.5) 0.48 0.06 165 (93.8) 50 (92.6) 0.76 0.05 
 Yes 114 (7.8) 27 (8.1)   67 (8.1) 14 (6.5)   11 (6.2) 4 (7.4)   
Socioeconomic tertile             
 Top 318 (21.6) 76 (22.8) 0.95  189 (22.8) 56 (25.9) 0.16  41 (23.3) 11 (20.4) 0.77  
 Middle 490 (33.4) 107 (32.0)  0.04 239 (28.8) 74 (34.3)  0.17 61 (34.7) 16 (29.6)  0.17 
 Bottom 637 (43.4) 146 (43.7)   386 (46.6) 83 (38.4)   72 (40.9) 26 (48.1)   
 Unknown 24 (1.6) 5 (1.5)   15 (1.8) 3 (1.4)   2 (1.1) 1 (1.9)   
Year         
 2010 and before 585 (39.8) 107 (32.0) 0.03  349 (42.1) 65 (30.1) 0.004  80 (45.5) 20 (37.0) 0.11  
 2011 235 (16.0) 71 (21.3)  0.18 138 (16.6) 37 (17.1)  0.28 32 (18.2) 5 (9.3) 0.39 
 2012 331 (22.5) 77 (23.1)   169 (20.4) 48 (22.2)   32 (18.2) 13 (24.1)   
 2013 318 (21.6) 79 (23.7)   173 (20.9) 66 (30.6)   32 (18.2) 16 (29.6)   
Low-riskIntermediate-riskHigh-risk
VariableNo PMRT (N = 1,469)PMRT (N = 334)PSMDNo PMRT (N = 829)PMRT (N = 216)PSMDNo PMRT (N = 176)PMRT (N = 54)PSMD
Age (years) 58 [49–66] 57 [48–66] 0.52 0.05 57 [49–66] 56 [48–65] 0.06 0.16 56 [47–66] 53 [44–62] 0.06 0.25 
Follow-up (months) 37 [23–57] 36 [23–52] 0.15 0.10 39 [24–58] 33 [20–49] 0.001 0.24 42 [27–59] 32 [19–55] 0.05 0.23 
Tumor stage         
 pT1 923 (62.8) 161 (48.2) <0.001 0.30 486 (58.6) 95 (44.0) <0.001 0.30 87 (49.4) 23 (42.6) 0.47 0.14 
 pT2 546 (37.2) 173 (51.8)   343 (41.4) 121 (56.0)   89 (50.6) 31 (57.4)   
Nodal stage             
 pN1mi 632 (43.0) 93 (27.8) <0.001 0.32 345 (41.6) 58 (26.9) <0.001 0.32 60 (34.1) 16 (29.6) 0.66 0.10 
 pN1 837 (57.0) 241 (72.2)   484 (58.4) 158 (73.1)   116 (65.9) 38 (70.4)   
Grade         
 1 488 (33.2) 87 (26.0) 0.009  162 (19.5) 25 (11.6) 0.04  8 (4.5) 1 (1.9) 0.43  
 2 825 (56.2) 201 (60.2)  0.21 464 (56.0) 129 (59.7)  0.23 67 (38.1) 25 (46.3)  0.30 
 3 111 (7.6) 39 (11.7)   188 (22.7) 56 (25.9)   97 (55.1) 28 (51.9)   
 Unknown 45 (3.1) 7 (2.1)   15 (1.8) 6 (2.8)   4 (2.3) 0 (0.0)   
Histology             
 IDC 1,090 (74.2) 228 (68.3) 0.16  653 (78.8) 154 (71.3) 0.08  159 (90.3) 44 (81.5) 0.06  
 ILC 224 (15.2) 62 (18.6)  0.14 105 (12.7) 41 (19.0)  0.19 9 (5.1) 3 (5.6)  0.34 
 IDC/ILC 125 (8.5) 34 (10.2)   61 (7.4) 17 (7.9)   3 (1.7) 5 (9.3)   
 Other 30 (2.0) 10 (3.0)   10 (1.2) 4 (1.9)   5 (2.8) 2 (3.7)   
PR status         
 PR+ 1,421 (96.7) 317 (94.9) 0.14 0.09 725 (87.5) 190 (88.0) 0.91 0.02 131 (74.4) 41 (75.9) 1.00 0.04 
 PR- 48 (3.3) 17 (5.1)   104 (12.5) 26 (12.0)   45 (25.6) 13 (24.1)   
Her2 status             
 Her2- 1,056 (71.9) 271 (81.1) <0.001 0.25 584 (70.4) 167 (77.3) 0.04 0.20 106 (60.2) 35 (64.8) 0.20 0.35 
 Her2+ 19 (1.3) 7 (2.1)   10 (1.2) 5 (2.3)   10 (5.7) 0 (0.0)   
 Unknown 394 (26.8) 56 (16.8)   235 (28.3) 44 (20.4)   60 (34.1) 19 (35.2)   
Chemotherapy         
 No/Unknown 1,136 (77.3) 212 (63.5) <0.001 0.31 447 (53.9) 72 (33.3) <0.001 0.42 49 (27.8) 3 (5.6) <0.001 0.63 
 Yes 333 (22.7) 122 (36.5)   382 (46.1) 144 (66.7)   127 (72.2) 51 (94.4)   
Race             
 White 1,232 (83.9) 279 (83.5) 0.59  699 (84.3) 178 (82.4) 0.17  146 (83.0) 45 (83.3) 0.96  
 Black 105 (7.1) 29 (8.7)  0.10 58 (7.0) 24 (11.1)  0.16 15 (8.5) 4 (7.4)  0.12 
 Asian/PI 119 (8.1) 25 (7.5)   68 (8.2) 13 (6.0)   14 (8.0) 5 (9.3)   
 Other/Unknown 13 (0.9) 1 (0.3)   4 (0.5) 1 (0.5)   1 (0.6) 0 (0.0)   
Hispanic         
 No 1,319 (89.8) 309 (92.5) 0.51 0.07 762 (91.9) 193 (89.4) 0.29 0.09 160 (90.9) 50 (92.6) 0.91 0.06 
 Yes 150 (10.2) 25 (7.5)   67 (8.1) 23 (10.6)   16 (9.1) 4 (7.4)   
Insurance status             
 Insured 1,271 (86.5) 286 (85.6) 0.61  700 (84.4)) 191 (88.4) 0.47  139 (79.0) 48 (90.6) 0.40  
 Medicaid 127 (8.6) 32 (9.6)  0.09 69 (8.3) 14 (6.5)  0.14 19 (10.8) 3 (5.7)  0.29 
 None 25 (1.7) 3 (0.9)   25 (3.0) 3 (1.4)   3 (1.7) 1 (1.9)   
 Unknown 46 (3.1) 13 (3.9)   35 (4.2) 8 (3.7)   15 (18.5) 2 (3.7)   
Prior cancer         
 No 1,355 (92.2) 307 (91.9) 0.85 0.01 762 (91.9) 202 (93.5) 0.48 0.06 165 (93.8) 50 (92.6) 0.76 0.05 
 Yes 114 (7.8) 27 (8.1)   67 (8.1) 14 (6.5)   11 (6.2) 4 (7.4)   
Socioeconomic tertile             
 Top 318 (21.6) 76 (22.8) 0.95  189 (22.8) 56 (25.9) 0.16  41 (23.3) 11 (20.4) 0.77  
 Middle 490 (33.4) 107 (32.0)  0.04 239 (28.8) 74 (34.3)  0.17 61 (34.7) 16 (29.6)  0.17 
 Bottom 637 (43.4) 146 (43.7)   386 (46.6) 83 (38.4)   72 (40.9) 26 (48.1)   
 Unknown 24 (1.6) 5 (1.5)   15 (1.8) 3 (1.4)   2 (1.1) 1 (1.9)   
Year         
 2010 and before 585 (39.8) 107 (32.0) 0.03  349 (42.1) 65 (30.1) 0.004  80 (45.5) 20 (37.0) 0.11  
 2011 235 (16.0) 71 (21.3)  0.18 138 (16.6) 37 (17.1)  0.28 32 (18.2) 5 (9.3) 0.39 
 2012 331 (22.5) 77 (23.1)   169 (20.4) 48 (22.2)   32 (18.2) 13 (24.1)   
 2013 318 (21.6) 79 (23.7)   173 (20.9) 66 (30.6)   32 (18.2) 16 (29.6)   

NOTE: Data are presented as count (percentage) or median (interquartile range) with significance determined by Fisher exact test or Kruskal-Wallis test; 1mi: 1 (microscopic); grade 1: well differentiated, grade 2: moderately differentiated, grade 3: poorly differentiated/undifferentiated; IDC: invasive ductal carcinoma; ILC: invasive lobular carcinoma; PR: progesterone receptor; and Her2: Her2/Neu receptor.

Kaplan–Meier estimates and multivariable survival analyses

Unadjusted survival curves, estimates of 5-year survival probabilities, and restricted mean survival times are shown for both cohorts (Fig. 2, Supplementary Tables S3 and S4). In univariate analyses of the unadjusted cohorts, low-risk women who received PMRT had significantly longer 5-year OS compared with low-risk women who did not receive PMRT as well as intermediate risk and high-risk women regardless of receipt of PMRT in both the NCDB and the SEER cohorts (P < 0.001 for both cohorts; Supplementary Tables S3 and S4).

Figure 2.

Survival curves for the NCDB and SEER cohorts. Unadjusted survival curves based on Kaplan–Meier estimates of overall survival for women with low-, intermediate-, and high-risk 21-gene RS assay groups diagnosed with T1-2 N1 ER-positive breast cancer treated with (blue) or without (red) post-mastectomy radiotherapy from the NCDB (A) and SEER cohorts (B). Significance determined by Wald test of multivariable analyses of matched cohorts. Statistical results represent analyses of matched cohorts for each RS subgroup.

Figure 2.

Survival curves for the NCDB and SEER cohorts. Unadjusted survival curves based on Kaplan–Meier estimates of overall survival for women with low-, intermediate-, and high-risk 21-gene RS assay groups diagnosed with T1-2 N1 ER-positive breast cancer treated with (blue) or without (red) post-mastectomy radiotherapy from the NCDB (A) and SEER cohorts (B). Significance determined by Wald test of multivariable analyses of matched cohorts. Statistical results represent analyses of matched cohorts for each RS subgroup.

Close modal

In multivariable analysis of all patients within the NCDB cohort, receipt of PMRT was not significantly associated with OS when examined independently (TR = 1.12, 95% CI = 0.93–1.35, P = 0.21), but there was a significant interaction of RS and PMRT with OS (P = 0.009). Age, grade 3 disease, higher tumor stage (pT2), absence of PR expression, Medicaid/none insurance status, higher comorbidity score, and history of prior cancer were also significantly associated with decreased OS in this model (Supplementary Table S3). In analyses of propensity-score–matched cohorts for each RS subgroup, women with low RS who received PMRT had significantly longer OS compared with women with low RS who did not receive PMRT (5-year OS: 96.8% vs. 94.9%; TR = 1.70, 95% CI = 1.30–2.22, P < 0.001; Table 3). PMRT was not associated with longer OS in women with intermediate RS (5-year OS: 94.4% vs. 93.6%; TR = 0.89, 95% CI = 0.69–1.14, P = 0.35) or high RS (5-year OS: 92.2% versus 91.9%; TR = 1.10, 95% CI = 0.91–1.34, P = 0.33; Table 3). Unadjusted survival curves of the matched cohorts of the NCDB cohort are presented in Fig. 2.

Table 3.

Kaplan–Meier estimates and parametric accelerated failure time models for overall survival from the NCDB and SEER cohorts

Multivariable AFT analysis
Kaplan–Meier estimatesAll patientsMatched cohorts
VariableN at risk (N events)Restricted mean OS (mo; 95% CI)5y OS (95% CI)Log-rank PTR (95% CI)PTR (95% CI)P
NCDB cohort 
Low-risk 
 No PMRT 2,907 (79) 91.3 (88.4–94.1) 94.9 (93.6–96.2)  1.00 (reference)  1.00 (reference)  
 PMRT 1,000 (9) 94.7 (93.5–96.0) 96.8 (93.9–99.7)  1.72 (1.14–2.58) 0.009 1.70 (1.30–2.22) <0.001 
Intermediate-risk    
 No PMRT 1,313 (45) 91.3 (88.8–93.8) 93.6 (91.4–95.8) <0.001 1.00 (reference)  1.00 (reference)  
 PMRT 506 (17) 91.0 (87.3–94.8) 94.4 (91.3–97.6)  0.85 (0.60–1.20) 0.35 0.89 (0.69–1.14) 0.35 
High-risk    
 No PMRT 1,136 (51) 90.8 (89.1–92.6) 91.9 (89.5–94.4)  1.00 (reference)  1.00 (reference)  
 PMRT 470 (19) 89.3 (83.8–94.4) 92.2 (87.9–96.7)  1.09 (0.83–1.43) 0.54 1.10 (0.91–1.34) 0.33 
SEER cohort 
Low-risk    
 No PMRT 1,469 (56) 111.3 (108.8–113.9) 94.2 (92.4–96.0)  1.00 (reference)  1.00 (reference)  
 PMRT 334 (6) 117.1 (115.1–119.0) 97.4 (95.3–99.5)  1.65 (1.03–2.64) 0.04 1.85 (1.33–2.57) <0.001 
Intermediate-risk   
 No PMRT 829 (53) 106.8 (103.6–110.0) 90.4 (87.5–93.4) <0.001 1.00 (reference)  1.00 (reference)  
 PMRT 216 (11) 107.7 (101.3–114.1) 91.7 (86.3–97.4)  0.84 (0.52–1.37) 0.49 0.84 (0.62–1.14) 0.26 
High-risk   
 No PMRT 176 (16) 100.4 (92.1–108.8) 87.8 (81.0–95.3)  1.00 (reference)  1.00 (reference)  
 PMRT 54 (6) 99.5 (86.4–112.5) 77.0 (61.6–96.2)  0.49 (0.23–1.03) 0.06 0.79 (0.50–1.23) 0.28 
Multivariable AFT analysis
Kaplan–Meier estimatesAll patientsMatched cohorts
VariableN at risk (N events)Restricted mean OS (mo; 95% CI)5y OS (95% CI)Log-rank PTR (95% CI)PTR (95% CI)P
NCDB cohort 
Low-risk 
 No PMRT 2,907 (79) 91.3 (88.4–94.1) 94.9 (93.6–96.2)  1.00 (reference)  1.00 (reference)  
 PMRT 1,000 (9) 94.7 (93.5–96.0) 96.8 (93.9–99.7)  1.72 (1.14–2.58) 0.009 1.70 (1.30–2.22) <0.001 
Intermediate-risk    
 No PMRT 1,313 (45) 91.3 (88.8–93.8) 93.6 (91.4–95.8) <0.001 1.00 (reference)  1.00 (reference)  
 PMRT 506 (17) 91.0 (87.3–94.8) 94.4 (91.3–97.6)  0.85 (0.60–1.20) 0.35 0.89 (0.69–1.14) 0.35 
High-risk    
 No PMRT 1,136 (51) 90.8 (89.1–92.6) 91.9 (89.5–94.4)  1.00 (reference)  1.00 (reference)  
 PMRT 470 (19) 89.3 (83.8–94.4) 92.2 (87.9–96.7)  1.09 (0.83–1.43) 0.54 1.10 (0.91–1.34) 0.33 
SEER cohort 
Low-risk    
 No PMRT 1,469 (56) 111.3 (108.8–113.9) 94.2 (92.4–96.0)  1.00 (reference)  1.00 (reference)  
 PMRT 334 (6) 117.1 (115.1–119.0) 97.4 (95.3–99.5)  1.65 (1.03–2.64) 0.04 1.85 (1.33–2.57) <0.001 
Intermediate-risk   
 No PMRT 829 (53) 106.8 (103.6–110.0) 90.4 (87.5–93.4) <0.001 1.00 (reference)  1.00 (reference)  
 PMRT 216 (11) 107.7 (101.3–114.1) 91.7 (86.3–97.4)  0.84 (0.52–1.37) 0.49 0.84 (0.62–1.14) 0.26 
High-risk   
 No PMRT 176 (16) 100.4 (92.1–108.8) 87.8 (81.0–95.3)  1.00 (reference)  1.00 (reference)  
 PMRT 54 (6) 99.5 (86.4–112.5) 77.0 (61.6–96.2)  0.49 (0.23–1.03) 0.06 0.79 (0.50–1.23) 0.28 

NOTE: AFT models for the matched cohorts were covariate-adjusted and inverse probability-weighted using propensity-weighted matched cohorts. Significance determined by log-rank test or Wald test. AFT, accelerated failure time.

In multivariable analysis of all patients within the SEER cohort, receipt of PMRT did not independently impact OS (TR = 1.11, 95% CI = 0.83–1.47, P = 0.48), but there was a significant interaction of RS and PMRT with OS (P = 0.03). Age, Medicaid/none insurance status, and history of prior cancer were also significantly associated with decreased OS in this model (Supplementary Table S4). In subset analyses, women with low RS who received PMRT had significantly longer OS compared with women with low RS who did not receive PMRT (5-year OS: 97.4% vs. 94.2%; TR = 1.85, 95% CI = 1.33–2.57, P < 0.001; Table 3). PMRT was not associated with longer OS in women with intermediate RS (5-year OS: 91.7% vs. 90.4%; TR = 0.84, 95% CI = 0.62–1.14, P = 0.26), or high RS (5-year OS: 77.0% vs. 87.8%; TR = 0.79, 95% CI = 0.50–1.23, P = 0.28; Table 3). Unadjusted survival curves of the matched cohorts of the SEER cohort are presented in Fig. 2.

Sensitivity analyses were performed in a subgroup of patients within the NCDB cohort whose receipt of chemotherapy was consistent with their RS. In this subgroup, women with low RS did not receive chemotherapy, while women with high RS did receive chemotherapy; women with intermediate RS who both received or did not receive chemotherapy were included. Results of this analysis were consistent with those of the overall cohort: women with low RS who received PMRT had significantly longer OS compared with women with low RS who did not receive PMRT (TR = 1.84; 95% CI = 1.15–2.97, P = 0.01), while PMRT was not associated with longer OS in women with intermediate RS (TR = 0.88, 95% CI = 0.62–1.14, P = 0.44) or high RS (TR = 1.64, 95% CI = 0.82–3.30, P = 0.16). Results were similarly consistent in a subgroup of analysis of only women within the NCDB cohort with Her2-negative disease: women with low RS who received PMRT had significantly longer OS compared to women with low RS who did not receive PMRT (TR = 1.80; 95% CI = 1.18–2.74, P = 0.007), while PMRT was not associated with longer OS in women with intermediate RS (TR = 0.86, 95% CI = 0.60–1.21, P = 0.38) or high RS (TR = 1.04, 95% CI = 0.81–1.35, P = 0.74). Due to the small number of women with Her2-positive disease, it was not possible to perform a separate analysis for women with Her2-positive disease.

TR estimates for decreased OS associated with low RS and omission of PMRT within the NCDB cohort were very robust to reasonable assumptions about possible unmeasured confounding. For example, in order for the observed effect of PMRT to be rendered nonsignificant, there would need to be a moderately imbalanced (prevalence of 70% vs. 30%) unmeasured confounder that demonstrated an association with OS with a TR ≥3 (Supplementary Table S3).

In this analysis of a large cohort of women from the NCDB with T1–2 N1 ER-positive breast cancer who underwent mastectomy, the 21-gene RS assay predicted benefit of PMRT for OS. This finding was validated in a separate cohort using the SEER registry. In both cohorts, women with low RS who received PMRT had significantly longer OS compared with women with low RS who did not receive PMRT. The survival advantage seen among women with low RS who received PMRT was approximately 2% to 3% at 5 years. In the matched cohort analysis of the low RS subgroup of the NCDB cohort, 29 of 1,000 women who did not receive PMRT had died, while only nine of 1,000 women died who did receive PMRT. Receipt of PMRT was not associated with longer OS among women with intermediate or high RS. The use of robust statistical analyses and the consistency across two datasets and sensitivity analyses adds to the strength of this finding. The 21-gene RS assay may therefore be useful as a predictive marker for potential OS benefit from PMRT in women with T1–2 N1 ER-positive breast cancer.

The value of PMRT in patients with T1–2 N1 breast cancer remains a topic of great clinical relevance and controversy. Until recently, it had been assumed that the absolute reduction in LRR achieved by PMRT would be proportional to its benefit in OS. The 2005 Early Breast Cancer Trialists Collaborative Group (EBCTCG) meta-analysis posited that for every four LRRs prevented at 5 years, one death from breast cancer would be prevented at 15 years (1). More recently, the 2014 EBCTCG meta-analysis demonstrated that PMRT reduced the 10-year risk of any recurrence by 10.6% and the 20-year risk of breast cancer–specific death by 8.1% (3). Consensus statements that incorporate more modern data suggest that there has been a reduction in absolute benefit of PMRT attributable to improvements in other modalities, namely systemic therapy; more recent series have reported rates of 10-year LRR following mastectomy and without PMRT to range between 4% and 10% (5). The steadily improving rates of LRR have led many to question whether the long-term risks associated with PMRT may outweigh its potential absolute benefits in women with one to three positive nodes (4, 5, 10).

The American Society of Clinical Oncology (ASCO) focused guideline update (2017) concluded that PMRT may be potentially omitted for patients with a low risk of LRR (5). Running counter to this line of thinking, however, is the understanding that adjuvant locoregional therapy will necessarily confer the greatest survival benefit to patients at the lowest risk of harboring subclinical micrometastatic disease. Patients with a low RS are at the lost risk of future distant metastasis. Sterilization of potential residual disease may therefore be of the greatest clinical utility to patients within this subgroup. This point is most elegantly demonstrated by the subgroup analysis of the DBCG 82 b&c trials in which patients within the low-risk subgroup experienced the smallest absolute reduction in LRR following PMRT, but the largest benefit in OS (31). In contrast, patients within the high-risk group achieved a large absolute reduction in LRR following PMRT, but demonstrated no OS benefit. It is hypothesized that this may be potentially due to the competing risk of distant disease recurrence ultimately negating any survival benefit from a locoregional treatment. Another potential contributor to this effect may be that radiosensitivity differs between biologic subtypes; tumors with lower risk of LRR may derive a larger relative benefit in OS due to greater radiosensitivity. The potential translation of a small absolute benefit in LRR to a surprisingly large downstream benefit is similarly exhibited in both the MA-20 and EORTC 22922 trials, in which the improvement in distant disease-free survival exceeded the reduction in LRR (32, 33).

The 21-gene RS assay has been validated as an independent prognosticator of LRR, distant recurrence, and OS in women with node-positive ER-positive breast cancer (6–10). Furthermore, RS has been shown to predict benefit to adjuvant systemic chemotherapy for women with node-negative cancer, and is endorsed by the ASCO 2016 Focused Guideline Update to be used in decision making regarding systemic treatment (34–36). Consistent with this, a retrospective analysis of the Southwest Oncology Group (SWOG)-8814 demonstrated that women with one to three positive nodes and low RS may not derive a significant benefit to adjuvant chemotherapy; the RxPONDER trial is currently accruing patients for prospective validation of this finding (NCT01272037; ref. 6). Similarly, it has been hypothesized that the 21-gene RS assay may be prognostic for LRR and therefore predicts the value of PMRT. Mamounas and colleagues (10) recently demonstrated that RS has independent prognostic value for assessing risk of LRR in node-positive patients and may be useful in identifying patients with a low risk of LRR for whom PMRT could be omitted.

The argument posited in the analysis by Mamounas and colleagues is logical. It is also reasonable, however, to surmise the opposite: patients with high RS who are at the highest risk for subclinical micrometastatic disease may not derive a survival benefit from a locoregional treatment due to a competing risk of distant failure. Our results are consistent with the latter hypothesis: women with a low RS had longer OS associated with receipt of PMRT, while women with intermediate or high RS did not. Survival benefit may therefore not be proportional to the absolute reduction of LRR in this patient population. Instead, the translation from LRR benefit to survival benefit appears to be heterogeneous and varies between subpopulations on the basis of distant recurrence risk and/or intrinsic radiosensitivity.

The current analyses are most critically limited by cohort size, particularly for the high-risk subset within the SEER cohort, which had very few high-risk patients who received PMRT. Although the size of cohorts used is large compared with many other studies evaluating RS, a relatively small proportion of eligible patients underwent testing for RS. Given that the 21-gene RS is not routinely ordered for women with node-positive disease, the majority of patients who were otherwise eligible for this analysis were not evaluated for RS. This may be due to a range of factors, including type of treatment facility, physician specialty, and other physician-related characteristics (37). Survival estimates of the selected cohorts, however, were reflective of those calculated for all eligible T1–2 N1 women within the NCDB and SEER registry, and were also consistent with recently published studies of this patient population (38). Cohort size precluded subset analyses by race, which may be relevant given that among women with node-negative disease, non-Hispanic black women have higher RS compared with non-Hispanic white women (39).

Interestingly, a larger proportion than expected of women with high RS within the NCDB cohort did not receive chemotherapy (49.2%; N = 790/1,606), while 22.1% (N = 865/3,907) of patients with low RS did receive chemotherapy. To address this, a sensitivity analysis was performed on the subgroup of patients within the NCDB cohort whose chemotherapy status was consistent with their reported RS; these results were very similar to those of the overall cohort. In addition, although the sensitivity of the SEER registry for recording receipt of chemotherapy and radiotherapy is moderate to high, there may be patients who are incorrectly categorized as not having received either or both treatments (40). Lastly, the short median follow-up times, necessarily limited by the availability of RS in the databases, may have precluded finding survival advantages that may appear later in time. Indeed, it was surprising to find such an obvious survival advantage so early in the low RS subsets. It should also be noted that although the NCDB and SEER registry are created independently and sampled for inclusion differently, there are likely shared patients between these two datasets that would make the two cohorts not entirely unique. The NCDB reports approximately 70% to 75% of cases within the United States, while the SEER registry reports approximately 28% of cases (41). As a result, the majority of patients within the NCDB cohort are likely not represented in the SEER cohort (42).

Given that the NCDB and SEER do not provide data regarding LRR and are subject to notable limitations, the current findings require prospective validation. The Canadian Tailor RT trial (MA.39), to begin accrual in mid-2018, will randomize women with T1–2 N1 disease and documented low RS who underwent BCS or mastectomy to either regional radiation or no regional radiation, with breast cancer recurrence-free interval as the primary objective. This trial should provide clarity regarding the value of RS as a predictor of the value of radiotherapy in women with T1–2 N1 breast cancer.

In conclusion, the current analyses provide initial evidence that survival benefit of PMRT in women with T1–2 N1 breast ER-positive breast cancer may be more pronounced in, or even limited to, women with low-risk RS. These data are remarkable for consistent validation with robust statistical analysis across two large cohorts. These results support and extend the findings from subgroup analyses of seminal trials which demonstrated that PMRT conferred the greatest improvement in survival to patients within the most favorable prognostic group (31). Although women with high RS may experience the greatest absolute reduction of LRR, women with low RS may derive the greatest survival benefit from PMRT due to a low competing risk of subclinical micrometastatic disease at diagnosis. These results caution against omission of PMRT for women with node-positive disease on the basis of a low-risk 21-gene RS alone and strongly suggest the need for prospective validation of this strategy prior to widespread adoption.

M. Kocherginsky is listed as a co-inventor on a patent issued “Methods and compositions related to glucocorticoid receptor (GR) antagonists and breast cancer,” which has been licensed to Corcept Therapeutics by The University of Chicago. No potential conflicts of interest were disclosed by the other authors.

The NCDB and SEER have not verified and are not responsible for the statistical validity of the data analysis or conclusions.

Conception and design: C.R. Goodman, B.-L.L. Seagle, S. Shahabi, J.B. Strauss

Development of methodology: C.R. Goodman, B.-L.L. Seagle, J.B. Strauss

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.R. Goodman, B.-L.L. Seagle

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.R. Goodman, B.-L.L. Seagle, M. Kocherginsky, E.D. Donnelly

Writing, review, and/or revision of the manuscript: C.R. Goodman, B.-L.L. Seagle, M. Kocherginsky, E.D. Donnelly, J.B. Strauss

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.R. Goodman

Study supervision: S. Shahabi

We acknowledge the NCDB and SEER for collecting the data used in these analyses. The NCDB and SEER and their participating hospitals are the source of the data.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Clarke
M
,
Collins
R
,
Darby
S
,
Davies
C
,
Elphinstone
P
,
Evans
V
, et al
Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials
.
Lancet
2005
;
366
:
2087
106
.
2.
Headon
H
,
Kasem
A
,
Almukbel
R
,
Mokbel
K
. 
Improvement of survival with postmastectomy radiotherapy in patients with 1-3 positive axillary lymph nodes: a systematic review and meta-analysis of the current literature
.
Mol Clin Oncol
2016
;
5
:
429
36
.
3.
McGale
P
,
Taylor
C
,
Correa
C
,
Cutter
D
,
Duane
F
,
Ewertz
M
, et al
Effect of radiotherapy after mastectomy and axillary surgery on 10-year recurrence and 20-year breast cancer mortality: meta-analysis of individual patient data for 8135 women in 22 randomised trials
.
Lancet
2014
;
383
:
2127
35
.
4.
McBride
A
,
Allen
P
,
Woodward
W
,
Kim
M
,
Kuerer
HM
,
Drinka
EK
, et al
Locoregional recurrence risk for patients with T1,2 breast cancer with 1–3 positive lymph nodes treated with mastectomy and systemic treatment
.
Int J Radiat Oncol Biol Phys
2014
;
89
:
392
8
.
5.
Recht
A
,
Comen
EA
,
Fine
RE
,
Fleming
GF
,
Hardenbergh
PH
,
Ho
AY
, et al
Postmastectomy radiotherapy: an american society of clinical oncology, american society for radiation oncology, and society of surgical oncology focused guideline update
.
Ann Surg Oncol
2017
;
24
:
38
51
.
6.
Albain
KS
,
Barlow
WE
,
Shak
S
,
Hortobagyi
GN
,
Livingston
RB
,
Yeh
IT
, 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
.
Lancet Oncol
2010
;
11
:
55
65
.
7.
Mamounas
EP
,
Tang
G
,
Fisher
B
,
Paik
S
,
Shak
S
,
Costantino
JP
, et al
Association between the 21-gene recurrence score assay and risk of locoregional recurrence in node-negative, estrogen receptor-positive breast cancer: results from NSABP B-14 and NSABP B-20
.
J Clin Oncol
2010
;
28
:
1677
83
.
8.
Dowsett
M
,
Cuzick
J
,
Wale
C
,
Forbes
J
,
Mallon
EA
,
Salter
J
, 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
.
J Clin Oncol
2010
;
28
:
1829
34
.
9.
Wolmark
N
,
Mamounas
EP
,
Baehner
FL
,
Butler
SM
,
Tang
G
,
Jamshidian
F
, et al
Prognostic impact of the combination of recurrence score and quantitative estrogen receptor expression (ESR1) on predicting late distant recurrence risk in estrogen receptor-positive breast cancer after 5 years of tamoxifen: results from NRG oncology/national surgical adjuvant breast and bowel project B-28 and B-14
.
J Clin Oncol
2016
;
34
:
2350
8
.
10.
Mamounas
EP
,
Liu
Q
,
Paik
S
,
Baehner
FL
,
Tang
G
,
Jeong
JH
, et al
21-gene recurrence score and locoregional recurrence in node-positive/ER-positive breast cancer treated with chemo-endocrine therapy
.
J Natl Cancer Inst
2017
;
109
:
pii: djw259
.
11.
Nimeus-Malmstrom
E
,
Krogh
M
,
Malmstrom
P
,
Strand
C
,
Fredriksson
I
,
Karlsson
P
, et al
Gene expression profiling in primary breast cancer distinguishes patients developing local recurrence after breast-conservation surgery, with or without postoperative radiotherapy
.
Breast Cancer Res
2008
;
10
:
R34
.
12.
Kreike
B
,
Halfwerk
H
,
Armstrong
N
,
Bult
P
,
Foekens
JA
,
Veltkamp
SC
, et al
Local recurrence after breast-conserving therapy in relation to gene expression patterns in a large series of patients
.
Clin Cancer Res
2009
;
15
:
4181
90
.
13.
Nuyten
DS
,
Kreike
B
,
Hart
AA
,
Chi
JT
,
Sneddon
JB
,
Wessels
LF
, et al
Predicting a local recurrence after breast-conserving therapy by gene expression profiling
.
Breast Cancer Res
2006
;
8
:
R62
.
14.
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
:
683
90
.
15.
Surveillance, Epidemiology, and End Results (SEER) Program
SEER*Stat Database: Incidence SEER 18 Regs (Excl AK) Custom Data Malignant Breast (with Oncotype DX and Additional Treatment Fields and Census Tract SES), Nov 2016 Sub (2004–2013)–Linked To Census Tract Attributes–Time Dependent (2000–2014)–SEER 18 (excl AK) Census 2000/2010 Geographies with Index Field Quantiles. Released June 2017, based on the November 2016 submission
. Available from: http://www.seer.cancer.gov/.
16.
Wong
WB
,
Ramsey
SD
,
Barlow
WE
,
Garrison
LP
 Jr.
,
Veenstra
DL
. 
The value of comparative effectiveness research: projected return on investment of the RxPONDER trial (SWOG S1007)
.
Contemp Clin Trials
2012
;
33
:
1117
23
.
17.
Grambsch
PM
,
Therneau
T
. 
Proportional hazards tests and diagnostics based on weighted residuals
.
Biometrika
1994
;
81
:
515
26
.
18.
Box
GEP
,
Hunter
JS
,
Hunter
WG
.
Statistics for experimenters: design, innovation and discovery
. Second edition.
Hoboken, NJ
:
John Wiley & Sons
; 
2005
.
19.
Jackson
CH
. 
Flexsurv: a platform for parametric survival modeling in R
.
J Stat Softw
2016
;
70
:
1
33
.
20.
Kwong
GPS
,
Hutton
JL
. 
Choice of parametric models in survival analysis: applications to monotherapy for epilepsy and cerebral palsy
.
Roy Stat Soc
2003
;
52
:
153
68
.
21.
Collett
D
.
Modelling survival data in medical research
. Third edition.
Boca Raton, FL
:
Chapman and Hall/CRC
; 
2014
.
22.
Austin
PC
. 
An introduction to propensity score methods for reducing the effects of confounding in observational studies
.
Multivariate Behav Res
2011
;
46
:
399
424
.
23.
Mitra
N
,
Heitjan
DF
. 
Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study
.
Stat Med
2007
;
26
:
1398
414
.
24.
Austin
PC
. 
The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments
.
Stat Med
2014
;
33
:
1242
58
.
25.
Ho
DE
,
Imai
K
,
King
G
,
Stuart
EA
. 
MatchIt: nonparametric preprocessing for parametric causal inference
.
J Stat Softw
2011
;
42
:
1
28
.
26.
Harrell
FE
 Jr
,
Dupont
C
. 
Hmisc: harrell miscellaneous. R package version 4.0–2
; 
2016
. Available from: https://CRAN.R-project.org/package=Hmisc.
27.
Therneau
TM
. 
A package for survival analysis in S_. version 2.382015
. Available from: https://CRAN.R-project.org/package=survival.
28.
Yoshida
K
,
Bohn
J
. 
tableone: Create "Table 1" to describe baseline characteristics. R package version 0.7.3
; 
2015
. Available from: https://CRAN.R-project.org/package=tableone.
29.
R Development Core Team
. 
R: A Language and Environment for Statistical Computing
.
Vienna, Austria
:
R Foundation for Statistical Computing
; 
2016
.
30.
van Buuren
S
,
Groothuis-Oudshoorn
K
. 
mice: multivariate imputation by chained equations in R
.
J Stat Softw
2011
;
45
:
1
67
.
31.
Kyndi
M
,
Overgaard
M
,
Nielsen
HM
,
Sorensen
FB
,
Knudsen
H
,
Overgaard
J
. 
High local recurrence risk is not associated with large survival reduction after postmastectomy radiotherapy in high-risk breast cancer: a subgroup analysis of DBCG 82 b&c
.
Radiother Oncol
2009
;
90
:
74
9
.
32.
Poortmans
PM
,
Collette
S
,
Kirkove
C
,
Van Limbergen
E
,
Budach
V
,
Struikmans
H
, et al
Internal mammary and medial supraclavicular irradiation in breast cancer
.
N Engl J Med
2015
;
373
:
317
27
.
33.
Whelan
TJ
,
Olivotto
IA
,
Levine
MN
. 
Regional nodal irradiation in early-stage breast cancer
.
N Engl J Med
2015
;
373
:
1878
9
.
34.
Paik
S
,
Tang
G
,
Shak
S
,
Kim
C
,
Baker
J
,
Kim
W
, et al
Gene expression and benefit of chemotherapy in women with node-negative, estrogen receptor–positive breast cancer
.
J Clin Oncol
2006
;
24
:
3726
34
.
35.
Harris
LN
,
Ismaila
N
,
McShane
LM
,
Andre
F
,
Collyar
DE
,
Gonzalez-Angulo
AM
, et al
Use of biomarkers to guide decisions on adjuvant systemic therapy for women with early-stage invasive breast cancer: American Society of Clinical Oncology Clinical Practice Guideline
.
J Clin Oncol
2016
;
34
:
1134
50
.
36.
Sparano
JA
,
Gray
RJ
,
Makower
DF
,
Pritchard
KI
,
Albain
KS
,
Hayes
DF
, et al
Prospective validation of a 21-gene expression assay in breast cancer
.
N Engl J Med
2015
;
373
:
2005
14
.
37.
Wilson
LE
,
Pollack
CE
,
Greiner
MA
,
Dinan
MA
. 
Association between physician characteristics and the use of 21-gene recurrence score genomic testing among Medicare beneficiaries with early-stage breast cancer, 2008–2011
.
Breast Cancer Res Treat
2018 Mar 13
.
[Epub ahead of print]
.
38.
Roberts
MC
,
Miller
DP
,
Shak
S
,
Petkov
VI
. 
Breast cancer-specific survival in patients with lymph node-positive hormone receptor-positive invasive breast cancer and Oncotype DX Recurrence Score results in the SEER database
.
Breast Cancer Res Treat
2017
;
163
:
303
10
.
39.
Holowatyj
AN
,
Cote
ML
,
Ruterbusch
JJ
,
Ghanem
K
,
Schwartz
AG
,
Vigneau
FD
, et al
Racial differences in 21-gene recurrence scores among patients with hormone receptor-positive, node-negative breast cancer
.
J Clin Oncol
2018
;
36
:
652
8
.
40.
Noone
AM
,
Lund
JL
,
Mariotto
A
,
Cronin
K
,
McNeel
T
,
Deapen
D
, et al
Comparison of SEER treatment data with medicare claims
.
Med Care
2016
;
54
:
e55
64
.
41.
Mohanty
S
,
Bilimoria
KY
. 
Comparing national cancer registries: the national cancer data base (NCDB) and the surveillance, epidemiology, and end results (SEER) program
.
J Surg Oncol
2014
;
109
:
629
30
.
42.
Murphy
M
,
Alavi
K
,
Maykel
J
. 
Working with existing databases
.
Clin Colon Rectal Surg
2013
;
26
:
5
11
.