Lymphovascular invasion, race, and the 21-gene recurrence score in early estrogen receptor-positive breast cancer

Lymphovascular invasion (LVI) and Black race are associated with poorer prognosis in early breast cancer (EBC). We evaluated the association between LVI and race, and whether LVI adds prognostic benefit to the 21-gene recurrence score (RS) in EBC. Women with ER+ HER2− EBC measuring up to 5 cm, with 0–3 involved axillary nodes, diagnosed between 1 January 2010 and 1 January 2014, who underwent surgery as first treatment and had available RS, were identified in the NCDB database. Bivariate associations between two categorical variables were examined using chi-square test. Multivariate Cox proportional hazards model were used to assess the association of LVI, race, and other covariates with overall survival (OS). 77,425 women, 65,018 node-negative (N0), and 12,407 with 1–3 positive (N+) nodes, were included. LVI was present in 12.7%, and associated with poor grade, RS 26–100, and N+ (all p < 0.0001), but not Black race. In multivariate analysis, LVI was associated with worse OS in N0 [HR 1.37 (95% CI 1.27, 1.57], but not N+ EBC. LVI was associated with worse OS in N0 patients with RS 11–25 [HR 1.31 (95% CI 1.09, 1.57)] and ≥26 [HR 1.58 (95% CI 1.30, 1.93)], but not RS 0–10. No interaction between LVI and chemotherapy benefit was seen. Black race was associated with worse OS in N0 (HR 1.21, p = 0.009) and N+ (HR 1.37, p = 0.015) disease. LVI adds prognostic information in ER+, HER2−, N0 BCA with RS 11–100, but does not predict chemotherapy benefit. Black race is associated with worse OS, but not LVI.


INTRODUCTION
Lymphovascular invasion (LVI), the presence of tumor cells within the lumen of lymphatic or vascular system in the tumor primary site, is associated with a higher risk of regional lymph node metastasis at diagnosis and subsequent distant recurrence in early breast cancer (EBC) [1][2][3][4][5][6][7] . Multiparameter gene expression assays also provide prognostic information for distant recurrence in hormone receptor-positive (HR+), HER2 negative breast (HER2−) cancer 8,9 . The 21-gene recurrence score (RS) assay is both prognostic for recurrence and predictive of chemotherapy benefit in EBC [10][11][12][13] , including women with up to three positive axillary nodes 14,15 . The extent to which LVI adds prognostic information or predictive information for chemotherapy benefit to the 21-gene RS or other assays has not been adequately evaluated 16 .
Black race has also been associated with worse prognosis in EBC. Although disparities in access to care play a role in Black patients' poorer outcomes 17,18 , survival gaps are also seen for Black women receiving contemporary adjuvant therapy on NCIsponsored clinical trials [19][20][21][22] , suggesting that factors other than care disparities contribute to their poorer prognosis. For example, in the TAILORx trial, Black race was associated with a 1.4-fold higher risk of recurrence after adjustment for clinicopathologic characteristics, despite similar chemotherapy and endocrine therapy use and similar RS distribution 23 . Thus, biological factors that are not captured by the 21-gene RS, including racial differences in the tumor microenvironment 24,25 , may contribute to the worse outcomes for Black women with EBC.
The presence of LVI represents a manifestation of an interaction between the tumor and its microenvironment. The molecular mechanisms underlying LVI are complex and not fully defined, but pathways involving cytokine-receptor interactions have been implicated [26][27][28] . Studies evaluating racial differences in BCA tumor microenvironment have described the differential expression of genes associated with inflammation and angiogenesis 24,25,[29][30][31] , as well as increased microvessel density 29,31 and increased numbers of tumor-associated macrophages 29,32 , which correlate with increased lymphangiogenesis and LVI [33][34][35] , in tumors belonging to Black women. Although no studies have directly compared racial differences in LVI in BCA, a multinational study evaluating clinicopathologic characteristics of urothelial tumors in white and Asian patients showed a significant racial difference in the rate of LVI 36 . No prior studies have evaluated the impact of LVI, RS, and race in the same population. We, therefore, sought to assess whether LVI adds prognostic or predictive information to the 21gene RS, and whether there is an association between LVI and race that may contribute to racial disparities in EBC.

Patient characteristics
As shown in Fig. 1, of 2,696,734 women with breast cancer in the National Cancer Database (NCDB), 77,425 met inclusion criteria for this analysis, 65,018 of whom had no nodal involvement (N0), and 12,407 who had involvement of 1-3 axillary nodes (N+). Clinicopathologic and demographic characteristics of the study population are shown in Table 1. In all, 67,998 (87.8%) of patients were white, 5750 (7.4%) were Black, 3096 (4.0%) were members of other racial groups, and 581 (0.8%) had no racial information available. LVI was reported as present in 9856 (12.7%), absent in 59,305 (76.60%), and unknown in 8264 (10.67%).
Associations between race and other clinicopathologic features Also shown in Table 2, Black race was associated with higher rates of axillary node metastasis (17.2% for Black women vs. 16 14.5% vs. 16.3%). Although these differences were statistically significant, the absolute differences were marginal or modest. In contrast, LVI was slightly lower in Black women than in white women or women of other races (12.3% vs. 12.8% vs 13.2%, p = 0.0365).

Chemotherapy benefit
Receipt of chemotherapy was associated with improved OS in the entire cohort (p = 0.003) and in N+ patients (p = 0.001), but not in N0 patients (p = 0.134). There was no interaction between LVI and chemotherapy benefit in the 39,402 patients with N0 disease and an RS 11-25 (interaction p = 0.46) (Supplemental Table 3).
Comparison of patients who did and did not undergo RS assay  Table 4). Among women who did not have genomic testing, LVI was seen more frequently in non-white women than in white women (13.9% for white women, vs. 15.7% for Black women, vs. 17.1% for women of other races, p < 0.0001); however, the difference in frequency of LVI between white and Black women, whereas statistically significant, was numerically small in magnitude. Multivariable analysis of OS in women who did not undergo RS assay showed that LVI was associated with poorer OS in the entire cohort (HR = 1.21, p < 0.0001), and in both N0 (HR = 1.15, p = 0.008) and N+ patients (HR = 1.22, p < 0.00001). Black race was associated with poorer OS in the entire cohort (HS = 1.14, p = 0.0002) and in N+ (HR = 1.26, p < 0.0001) women only (Supplemental Table 5).

DISCUSSION
We evaluated the association between LVI, race, and the 21-gene RS in 77,425 women with ER-positive, HER2-positive EBC with 0-3 positive axillary nodes. We confirmed prior reports that LVI is associated with a worse prognosis in N0 disease 1,3,5,6 , and that Black race is associated with a worse prognosis 20,23,24 , but report for the first time no association between Black race and the presence of LVI. We also report for the first time that LVI adds prognostic information to the 21-gene RS, but not predictive information for chemotherapy benefit in those with a mid-range RS of 11-25. Three previous studies have evaluated the correlation between LVI and the 21-gene RS in ER+ N0 EBC. Mutai et al. 16 and Tan et al. 37  Prior to the publication of the randomized TAILORx data 13 , the optimal management for patients with N0 ER+ BCA and intermediate RS was unclear, and recommendations for adjuvant chemotherapy in this group of patients were often based on clinicopathologic criteria. Chen et al. 39 showed that the presence of LVI was significantly associated with receipt of chemotherapy in patients with intermediate RS diagnosed before 2013. Our study, although not randomized, did not detect a chemotherapy benefit in patients with intermediate RS and LVI who received adjuvant chemotherapy. In contrast, a retrospective evaluation of patients enrolled on two International Breast Cancer Study Group trials of chemoendocrine therapy in N0 EBC found that the benefit of ovarian suppression with goserelin in premenopausal patients with ER+ disease was exclusively confined to patients with LVI 40 . In addition, analysis of the BIG 1-98 trial showed that patients with LVI had greater than average benefit from letrozole compared with tamoxifen 41 . These findings are consistent with results of the SOFT/TEXT trials, indicating greater benefit from ovarian function suppression plus an aromatase inhibitor compared with tamoxifen in those at higher recurrence risk 42 .
Multiple factors contribute to poorer prognoses for Black breast cancer patients, including disparities in access to care 17,18 , later  stage at diagnosis 43,44 , higher incidence of triple-negative disease 45 , and decreased adherence to chemotherapy and endocrine therapy regimens 46,47 . However, four clinical trials, which collectively included 26,138 patients, 2,370 (9%) of whom were characterized as Black (either by self-reported race, or by ancestry-informative genetic markers of African ancestry) have shown poorer outcomes for Black women with HR+ EBC 20-23 , despite similar antineoplastic therapy, indicating that differences in tumor biology in HR+ EBC also play a role. This disparity is hypothesized to be at least partially due to racial differences in the tumor microenvironment 24,25 . Although Black women have higher rates of obesity, which has been associated with higher recurrence rates in HR+ HER2− disease 48,49 , evidence suggests that this racial disparity is not explained by obesity alone 22,50 . Our analysis, using real-world data from NCDB, indicates that the racial disparity in outcome of ER+ EBC is not explained by a greater incidence of LVI, or by substantially and clinically relevant differences in the 21gene RS.
Our study has several strengths and limitations, which are inherent in the use of NCDB data. Strengths include the large sample size, utilizing real-world data from CoC-accredited institutions, and representing the largest analysis evaluating the prognostic information provided by LVI, with a sufficient sample size to determine an association between LVI and Black race. Limitations include lack of information regarding cancer recurrence, both distant and locoregional, relatively short duration of follow-up, and relative under-representation of Black women in the EBC NCDB dataset 43 . An additional important limitation of our study is that the NCDB relies on local assessment of LVI, and does not distinguish between focal and extensive LVI. Prior reports indicate that up to 26% of local pathologists' assessments of LVI are altered after pathologic review by an experienced breast pathologist 51 , and that extent of LVI correlates with poorer outcomes 6 . In addition, 10.67% of our cohort had unknown LVI status.
Our study is also limited by real-world biases in the ordering of RS. Comparison between our cohort and patients in the NCDB database who did not undergo RS assay show that patients in whom RS was ordered were more likely to have intermediategrade tumors, N0 disease, and no comorbidities, confirming previously reported clinician tendencies to order RS in a state of clinical equipoise 52,53 , and in patients felt to be able to tolerate chemotherapy 53 . Lower receipt of RS in patients with LVI may reflect clinician bias in favor of administering chemotherapy to patients with LVI, a known poor prognostic factor. Black women were also somewhat under-represented in our cohort, compared with patients who did not receive RS. Previous studies evaluating racial bias in receipt of genomic testing have shown conflicting results. One study utilizing NCDB data showed that Black women were less likely than white women to undergo RS assay, and more likely to have testing ordered in guideline-discordant scenarios 54 . However, a more recent study, utilizing data from the Georgia Cancer Registry, found no racial differences in receipt of RS, but found a 1 year lag in uptake of testing in Black women 55 .
In summary, our data demonstrated that LVI adds prognostic information for OS in N0 ER+ HER2-BCA, with RS 11-100, but found no association between LVI and Black race in women undergoing RS assay. Despite the poorer prognosis associated with LVI, no chemotherapy benefit was seen in patient with RS 11-25 and LVI. Therefore the presence of LVI should not influence chemotherapy decision-making in patients with intermediate RS.

Case selection
We utilized a data set derived from the 2005-2016 National Cancer Database (NCDB). NCDB is a nationwide, facility-based database jointly sponsored by the American Cancer Society and the American College of Surgeons Commission on Cancer (CoC). The NCDB contains data collected on over 34,000,000 cancer cases from over 1500 CoC-accredited hospitals, representing over 70% of newly diagnosed cancers 56 , and 80% of newly diagnosed breast cancers in the United States 57 . The NCDB Participant User File (PUF) is a HIPAA-compliant data file, which is made available to investigators from CoC-accredited cancer programs who complete an application process. The data used in the study are derived from a deidentified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigator. The NCDB PUF contains de-identified patient data, including demographic information, tumor site, and pathology data, first course of treatment, and mortality. The NCDB PUF does not contain information on recurrence. The NCDB began collecting information on both LVI and on results of gene expression assays in 2010.
The study population included women age 75 and under-diagnosed between 1 January 2010 and 1 January 2014 with estrogen receptorpositive (ER+) HER2− BCA, measuring up to 5 cm, with 0-3 pathologically involved axillary nodes, treated with definitive surgery as first treatment, and with numeric RS available. Demographic information obtained included age, race, estimated annual household income and educational attainment. Race was classified as white, Black, other (including all other racial codes captured by the NCDB) or unknown. Clinical characteristics included tumor size in mm, histologic grade, axillary node involvement, LVI, and numeric RS. Micrometastatic nodal involvement (pN1mi) was classified as node positive. LVI was characterized as present, absent, or unknown. RS was characterized as low, intermediate, or high using TAILORx cutpoints, where 0-10 was defined as low, 11-25 as intermediate, and 26-100 as high.

Statistical analysis
Bivariate associations between categorical variables were examined using the chi-square test. Multivariate Cox proportional hazards model were used to assess the association of LVI, chemotherapy, and RS on OS, while adjusting age, race, tumor size, grade, LN status, Charlson-Deyo comorbidity score, median income, and education level. The estimated HR for each variable in the model, along with its 95% CI, was reported. All tests are two-sided with a significance level ≤5%. Proportionality assumption was examined, and no violation was detected. All analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

Ethics
NCDB PUF data are de-identified, and compliant with HIPAA. Hospitals, health-care providers, and patients are not identified. Patient-informed consent is not obtained prior to institutional data submission to NCDB. As this study utilizes de-identified patient data, with no attempt made to contact or re-identify the subjects, it is deemed exempt from oversight by the Institutional Review Board of Albert Einstein College of Medicine.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

DATA AVAILABILITY
The data generated and analyzed during this study are described in the following data record: https://doi.org/10.6084/m9.figshare.13517210 58 . The data are stored in the National Cancer Database (NCDB) Participant User File (PUF) "NCDBPUF_-Breast.0.2016.0.dat". This is a HIPAA-compliant data file. The data use agreement between the researchers and the NCDB prohibits sharing of data, but the file can be made available to investigators from CoC-accredited cancer programs who complete an application process (CoC is the American College of Surgeons Commission on Cancer). Information regarding the application process is available at https://www. facs.org/quality-programs/cancer/ncdb/puf. CODE AVAILABILITY SAS 9.4 was used, and the code is available upon request. D. Makower et al.