INTRODUCTION
More than 44% of new breast cancer patients in the United States are 65 or older.1 Elderly women are more likely to have estrogen receptor (ER)-positive tumors that respond to endocrine therapy and a higher prevalence of comorbidities.2,3 Thus, the management of breast cancer in older women differs from their younger counterparts and should account for cancer biology, extent of disease, and general health status.4 The Choosing Wisely campaign5 and clinical guidelines6 recommend against surgical axillary lymph node staging for women ≥ 70 years of age with ER+, human epidermal growth factor receptor (HER-2)(–) early-stage breast cancer.
This recommendation is made partly on the basis of 2 randomized trials that demonstrated no survival advantage to performing axillary dissection in node negative elderly women.7–9 While axillary staging provides information to guide adjuvant treatment decisions, medical oncologists are less likely to prescribe chemotherapy to elderly women.10,11 The use of adjuvant chemotherapy is lowest for patients above the age of 70 and those with recognized clinical frailty.10,11 The potential prognostic information gained must be balanced against the well-described morbidity profile of axillary surgery, including lymphedema and chronic pain.12
Despite this literature and guideline recommendations, more than 80% of elderly women meeting these criteria still receive axillary staging.13 Indeed, a large analysis using the National Cancer Database in the United States found rates of axillary staging increased among elderly women from 2004 to 2016.14 There are several potential explanations. Compared to more radical surgical axillary dissections studied in randomized trials, contemporary procedures such as sentinel lymph node biopsy (SLNB) are less morbid and still accurately stage the axilla.15 Surgeons may be more liberal in utilizing what they perceive as negligibly morbid staging procedures. More recent observational studies have found survival benefits to axillary staging in elderly women, although these studies may have incompletely addressed confounding by indication and did not account for competing risks when examining cancer-specific survival.16,17 This is especially important in elderly populations where patients may be much more likely to die of comorbidities than their underlying disease, obscuring smaller treatment effects. Observational studies that have confirmed the findings of randomized trials are small or do not reflect contemporary use of SLNB over axillary dissection.18–21 Therefore, there remains active debate as to the appropriate utilization of axillary staging for elderly women and its impact on survival.
The aim of this study was to examine the impact of omitting surgical axillary staging on overall and breast cancer-specific survival (BCSS) in women ≥ 70 years of age with early-stage breast cancer, utilizing advanced methods of addressing confounding and competing risks analysis.
METHODS
Study Design and Data Source
This was a population-based cohort study using Surveillance, Epidemiology, and End Results (SEER), a cancer registry that includes 28% of the US population.22 We followed the Strengthening the Reporting of Observational Studies in Epidemiology statement in the preparation of this article (Supplementary Material 1, https://links.lww.com/AOSO/A112).23 The Research Ethics Board at the University of Toronto (Number 41545) approved this study.
Patient Population
The cohort was constructed using SEER Plus Incidence Data representing 18 registries.24 Women ≥ 70 years of age diagnosed with invasive breast cancer between 2005 and 2015 who underwent primary breast surgery were eligible. The cohort was limited to diagnoses in 2005 and later, as this marked the release of randomized evidence that omission of axillary staging in elderly patients with early disease may not impact survival.9 International Classification of Diseases for Oncology, Third Edition codes were used to select patients with invasive breast neoplasms with histology of interest (Supplementary Material 2, https://links.lww.com/AOSO/A112).
Exclusion criteria included (a) T3/T4 tumors, (b) metastatic disease, (c) atypical histology (ie, melanoma; Supplementary Material 2, https://links.lww.com/AOSO/A112), (d) neoadjuvant chemotherapy, (e) aspiration of axillary nodes as a staging strategy, (f) missing data on axillary staging, (g) zero months of survival, and (h) missing data on cause of death. Randomized controlled trial populations on this topic have included only elderly women with clinically node-negative disease; however, SEER does not report initial clinical stage; clinical stage is only used when pathologic information is not available (ie, when axillary staging is not performed).7 Therefore, clinical node status could not be used as an inclusion/exclusion criteria.
Exposures and Covariates
The primary exposure of interest was omission of axillary staging. Women who had “No axillary nodes examined” were classified as omission of axillary staging, and those who had any results of surgical lymph node examination were classified as receipt of axillary staging (Supplementary Material 3, https://links.lww.com/AOSO/A112).25 We did not differentiate between SLNB and axillary dissection, as this distinction has been found to be unreliable in SEER.26
Covariates included age at diagnosis, year of diagnosis (2005–2015), histology, primary surgery type (lumpectomy or mastectomy), marital status (married or nonmarried), race, tumor grade (I–IV), tumor size (<2 cm or ≥2 cm), hormone receptor status (positive or negative), history of previous cancer and number of tumors, receipt of radiotherapy, and receipt of chemotherapy. Age was treated as continuous and the remaining covariates were considered categorical (Supplementary Material 3, https://links.lww.com/AOSO/A112). Race was categorized as Black, White, or other. Histology was categorized as cystic, mucinous and serous neoplasms, ductal and lobular neoplasms, or other. During the study period, HER-2 status was not consistently reported in SEER and was therefore not included in the main analysis. A sensitivity analysis was performed in patients diagnosed after 2010, for whom HER-2 status was available. Importantly, comorbidity status was not available in the datasets used for this analysis.
Outcomes
The outcomes of interest were overall survival (OS) and BCSS. OS was defined as the number of months from breast cancer diagnosis to death, or until December 2017.27 Patients were censored at the date of last contact, if this occurred prior to the study cutoff. Deaths for BCSS were classified as death due to breast cancer or death due to other causes according to SEER-provided cause of death codes.28
Statistical Analysis
Baseline characteristics stratified by axillary staging status were presented. Differences between the 2 groups were assessed using standardized mean differences. A 2-sided Cochran-Armitage test was used to determine if there was a change in the proportion of women undergoing axillary staging over time. Propensity score methods were used to reduce bias from differences in baseline patient characteristics between women that did and did not receive axillary staging. A multivariable logistic regression including baseline characteristics was generated, treating axillary staging strategy as the outcome. Receipt of chemotherapy was not included in the propensity score model as this decision is often contingent on pathological axillary nodal status. Radiotherapy was similarly not included in the propensity score model as it occurs temporally after axillary staging. However, both of these important potential confounders were included as covariates in the final survival models. The distribution of the propensity scores are presented in Supplementary Material 4 (https://links.lww.com/AOSO/A112).
Overlap propensity score weighting was used, in which a patient’s weight in the analysis is equal to the probability of being assigned to the opposing treatment group.29 This method of weighting avoids patients with extreme propensity score values (outliers) from dominating results, which may occur with inverse probability of treatment weighting (IPTW).30,31
Weighted and unweighted Kaplan-Meier curves for OS were presented for women undergoing axillary staging versus those who did not have axillary staging. The hazard ratio (HR) and 95% confidence interval (CI) of axillary staging omission was determined. A robust sandwich estimator was used to estimate the variance. For BCSS, we accounted for the competing risk of death from other causes, and the subdistribution hazard ratio (sdHR) and 95% CI was determined.32 Additional post-weighting adjustment was performed for receipt of radiotherapy or chemotherapy.
We performed several sensitivity analyses. All analyses were repeated using IPTW weights, stabilized IPTW weights, and conventional multivariable Cox regression using the same covariates in the propensity score model, including adjustment for radiotherapy and chemotherapy. Stratified analyses were additionally performed by age (<80 and 80+ years), tumor size (<2 cm and ≥2 cm), and ER status (ER+, and ER–). BCSS was also analyzed using cause-specific HRs in addition to subdistribution hazard models. We also examined only women diagnosed after 2010, when HER-2 status was available. This cohort was restricted to ER+, HER-2(–) tumors (the population discussed in the Choosing Wisely guidelines5). Finally, to explore the impact of unmeasured confounding, we calculated an E-value for the OS result. The E-value represents the strength an unmeasured confounder must have with both the exposure and outcome, independent of the measured confounders, to move the lower confidence limit to the null.33 We assumed a nonrare confounder (great than 15% prevalence).
Missing data among covariates in the propensity score model were managed using multiple imputation. Five multiply imputed datasets were created under the fully conditional specification model prior to propensity score estimation.34 Trace plots of mean and SDs were examined in each imputed dataset. Effect estimates were calculated within each imputed dataset and then combined using Rubin’s rules.35 The data analysis for this paper was generated using SAS version 9.4 (SAS Institute Inc., Cary, NC) and R (R Foundation for Statistical Computing, Vienna, Austria). All statistical tests were 2-sided, and a P ≤ 0.05 was considered statistically significant.
RESULTS
Patient Characteristics
We identified 167,635 women ≥ 70 years of age who underwent surgery for breast cancer between 2005 and 2015 in SEER. After application of exclusion criteria, 144,329 women remained (Fig. 1). This final cohort contained 121,708 (84.3%) women who underwent axillary staging and 22,621 (15.7%) women who did not undergo axillary staging.
;)
FIGURE 1.: Development of a Surveillance, Epidemiology, and End Results database cohort of elderly breast cancer patients. Exclusions were applied sequentially in the order presented.
Baseline characteristics of the cohort are presented in Table 1. The mean age of the cohort was 77.7 years (SD, 5.8 years). The majority of women had ductal and lobular histology (93.7%), underwent lumpectomy (63.1%), were nonmarried (56.4%), and were White (86.4%). The majority of tumors were ER-positive (87.3%), progesterone receptor-positive (75.5%), and less than 2 cm in size (67.4%).
TABLE 1. -
Patient Characteristics Before and After Overlap Propensity Score Weighting
| Characteristic |
Before Propensity Score Weighting |
After Propensity Score Weighting |
| Axillary Staging (n = 121,708) |
No Axillary Staging (n = 22,621) |
SMD |
Axillary Staging (n = 16,524)*
|
No Axillary Staging (n = 16,524)*
|
SMD |
| Age, mean (SD) |
76.98 (5.34) |
81.64 (6.72) |
0.768
|
80.25 (5.88) |
80.25 (6.38) |
<0.001 |
| Year at diagnosis, n (%) |
| 2005 |
9907 (8.1) |
2129 (9.4) |
0.078 |
1517.9 (9.2) |
1517.9 (9.2) |
<0.001 |
| 2006 |
10,242 (8.4) |
2011 (8.9) |
|
1465.8 (8.9) |
1465.8 (8.9) |
|
| 2007 |
10,660 (8.8) |
1821 (8.1) |
|
1352.2 (8.2) |
1352.2 (8.2) |
|
| 2008 |
10,811 (8.9) |
1903 (8.4) |
|
1401.1 (8.5) |
1401.1 (8.5) |
|
| 2009 |
11,243 (9.2) |
1850 (8.2) |
|
1384.2 (8.4) |
1384.2 (8.4) |
|
| 2010 |
10,955 (9.0) |
1827 (8.1) |
|
1359.7 (8.2) |
1359.7 (8.2) |
|
| 2011 |
11,263 (9.3) |
2073 (9.2) |
|
1514.3 (9.2) |
1514.3 (9.2) |
|
| 2012 |
11,392 (9.4) |
2177 (9.6) |
|
1578.5 (9.6) |
1578.5 (9.6) |
|
| 2013 |
11,709 (9.6) |
2193 (9.7) |
|
1596.1 (9.7) |
1596.1 (9.7) |
|
| 2014 |
11,629 (9.6) |
2351 (10.4) |
|
1685.3 (10.2) |
1685.3 (10.2) |
|
| 2015 |
11,897 (9.8) |
2286 (10.1) |
|
1668.9 (10.1) |
1668.9 (10.1) |
|
| Marital status, n (%) |
| Married |
55,267 (45.4) |
7689 (34.0) |
0.235
|
6142.6 (37.2) |
6142.6 (37.2) |
<0.001 |
| Nonmarried |
66,441 (54.6) |
14,932 (66.0) |
|
10,381.4 (62.8) |
10,381.4 (62.8) |
|
| Race, n (%) |
| Black |
9311 (7.7) |
1622 (7.2) |
0.057 |
1204.9 (7.3) |
1204.9 (7.3) |
<0.001 |
| Other |
7565 (6.2) |
1132 (5.0) |
|
869.7 (5.3) |
869.7 (5.3) |
|
| White |
10,4832 (86.1) |
19,867 (87.8) |
|
14,449.5 (87.4) |
14,449.5 (87.4) |
|
| Histology, n (%) |
| Cystic, mucinous, and serous neoplasms |
4340 (3.6) |
1172 (5.2) |
0.111
|
780.8 (4.7) |
780.8 (4.7) |
<0.001 |
| Ductal and lobular neoplasms |
114,639 (94.2) |
20,659 (91.3) |
|
15,220.0 (92.1) |
15,220.0 (92.1) |
|
| Other |
2729 (2.2) |
790 (3.5) |
|
523.3 (3.2) |
523.3 (3.2) |
|
| Surgery type, n (%) |
| Lumpectomy |
73,871 (60.7) |
17,250 (76.3) |
0.34
|
11,976.4 (72.5) |
11,976.4 (72.5) |
<0.001 |
| Mastectomy |
47,837 (39.3) |
5371 (23.7) |
|
4547.7 (27.5) |
4547.7 (27.5) |
|
| Tumor grade, n (%) |
| I |
34,558 (28.4) |
7377 (32.6) |
0.111
|
5254.7 (31.8) |
5254.7 (31.8) |
<0.001 |
| II |
57,772 (47.5) |
10,607 (46.9) |
|
7801.2 (47.2) |
7801.2 (47.2) |
|
| III |
28,819 (23.7) |
4522 (20.0) |
|
3384.5 (20.5) |
3384.5 (20.5) |
|
| IV |
559 (0.5) |
115 (0.5) |
|
83.7 (0.5) |
83.7 (0.5) |
|
| ER positive, n (%) |
105,889 (87.0) |
20,158 (89.1) |
0.065 |
14,663.8 (88.7) |
14,663.8 (88.7) |
<0.001 |
| PR positive, n (%) |
91,453 (75.1) |
17,473 (77.2) |
0.049 |
12,714.3 (76.9) |
12,714.3 (76.9) |
<0.001 |
| Tumor size, n (%) |
| <2 cm |
81,356 (66.8) |
15,884 (70.2) |
0.073 |
11,650.5 (70.5) |
11,650.5 (70.5) |
<0.001 |
| ≥2 cm |
40,352 (33.2) |
6737 (29.8) |
|
4873.5 (29.5) |
4873.5 (29.5) |
|
| History of previous cancer, n (%) |
94,550 (77.7) |
14,276 (63.1) |
0.323
|
10,879.1 (65.8) |
10,879.1 (65.8) |
<0.001 |
| Number of previous tumors, n (%) |
| 1 |
77,447 (63.6) |
11,698 (51.7) |
0.266
|
8891.2 (53.8) |
8891.2 (53.8) |
<0.001 |
| 2 |
33,741 (27.7) |
7517 (33.2) |
|
5361.8 (32.4) |
5361.8 (32.4) |
|
| 3+ |
10,520 (8.6) |
3406 (15.1) |
|
2271.0 (13.7) |
2271.0 (13.7) |
|
| Radiotherapy†, n (%) |
| Yes |
55,891 (45.9) |
5345 (23.6) |
0.481
|
7831.1 (47.4) |
4159.9 (25.2) |
0.475
|
| No/unknown |
65,817 (54.1) |
17,276 (76.4) |
|
8692.9 (52.6) |
12,364.1 (74.8) |
|
| Chemotherapy†, n (%) |
| Yes |
16,690 (13.7) |
874 (3.9) |
0.353
|
1380.1 (8.4) |
763.6 (4.6) |
0.152
|
| No/unknown |
105,018 (86.3) |
21,747 (96.1) |
|
15,144.0 (91.6) |
15,760.4 (95.4) |
|
Balance between the groups is assessed using SMDs. Larger values indicate a greater difference, with values >0.1 defined as an important difference and bolded.
*Sample sizes represent the weighted size of the cohort.
†Radiotherapy and chemotherapy were not included in the propensity score model.
PR indicates progesterone receptor; SMD, standardized mean difference.
Omission of Axillary Staging
Compared to those that underwent axillary staging, women who did not undergo axillary staging were older (81.6 vs 77.0 years; standardized mean difference, 0.768; Table 1). Advancing age was strongly associated with the decision to omit axillary staging. Among women aged 70–74, 92.1% underwent axillary staging, compared to 29.7% in women older than 95 years (Fig. 2). There were also important differences in the type of surgery used, tumor grade, marital status, and histology. After overlap propensity score weights were applied, the size of the weighted cohort was 16,524 weighted individuals in each group, and important potential confounders were well balanced (Table 1).
FIGURE 2.: Proportion of patients undergoing axillary staging by age at diagnosis. Values indicate the number of patients.
There was no significant change over time in the proportion of women not undergoing axillary staging (17.7% in 2005 vs 16.1% in 2015; Cochrane-Armitage test P = 0.50; Fig. 3).
FIGURE 3.: Proportion of patients undergoing axillary staging over the study time period. Values indicate the number of patients. Cochrane-Armitage trend test P = 0.50 (2-sided).
Receipt of Chemotherapy and Radiotherapy
After propensity score weighting, women who did not undergo axillary staging were significantly less likely to receive chemotherapy (adjusted relative risk [RR], 0.58; 95% CI, 0.54–0.62) and radiotherapy (adjusted RR, 0.53; 95% CI, 0.52–0.54) compared to women who did undergo axillary staging (Table 1). Characteristics of those who received chemotherapy and radiotherapy are presented in Supplementary Material 5 (https://links.lww.com/AOSO/A112). The most important factors associated with chemotherapy and radiotherapy receipt were tumor grade and undergoing lumpectomy, respectively.
Overall Survival
Over the study period, 50,032 women died (34.7%) and median survival was 120 months (95% CI, 120–121 months). After overlap propensity score weighting and adjustment for chemotherapy and radiotherapy, women who did not undergo axillary staging had significantly worse OS (adjusted HR, 1.22; 95% CI, 1.19–1.25; Fig. 4).
FIGURE 4.: Overall survival for elderly women who did and did not undergo axillary staging. A, Unweighted Kaplan-Meier curve and hazard ratio for overall survival, comparing women who underwent axillary staging and those who did not undergo axillary staging. B, Overlap propensity score-weighted Kaplan-Meier curve and hazard ratio for overall survival. Adjusted hazard ratio is additionally adjusted for receipt of radiotherapy and chemotherapy. CI indicates confidence interval; HR, hazard ratio.
Breast Cancer-Specific Survival
In our cohort of elderly breast cancer patients, 10,518 deaths (21.0%) were due to breast cancer, while the remaining 39,514 (79.0%) deaths were due to other causes. These included heart disease (10,595; 21.2%), cerebrovascular disease (2960; 5.9%), chronic obstructive pulmonary disease (2587; 5.2%), Alzheimer’s disease (2580; 5.2%), lung cancer (1852; 3.7%), and other causes.
In the competing risks analysis, patients who did not undergo axillary staging had significantly higher risk of death due to breast cancer in both the unadjusted analysis (sdHR, 1.23; 95% CI, 1.17–1.30; Fig. 5) and the overlap propensity score-weighted + chemotherapy/radiotherapy adjusted analysis (adjusted sdHR, 1.14; 95% CI, 1.08–1.21; Fig. 5). These findings were unchanged when cause-specific HRs were calculated (Supplementary Material 6, https://links.lww.com/AOSO/A112).
FIGURE 5.: Cumulative incidence function for breast cancer-specific survival, treating death from nonbreast cancer causes as a competing risk. Overlap propensity score-weighted and adjusted subdistribution hazard ratios for not undergoing axillary staging are presented. Adjustment was additionally performed for receipt of radiotherapy and chemotherapy. CI indicates confidence interval; COPD, chronic obstructive pulmonary disease; sdHR, subdistribution hazard ratio
Sensitivity Analyses
Sensitivity analyses for adjusted BCSS are presented in Figure 6. Propensity score weighting using IPTW weights and stabilized IPTW weights and multivariable Cox regression reached the same overall conclusion as overlap weighting among the entire cohort. When stratified by age <80 or ≥80 years, not undergoing axillary staging had a greater impact on BCSS among younger women compared to older (sdHR, 1.35; 95% CI, 1.23–1.47 vs 1.07; 95% CI, 1.00–1.16; Fig. 6). Less clear trends were seen for stratification by tumor size and ER status or for OS (Supplementary Material 7, https://links.lww.com/AOSO/A112). When the cohort was restricted to women who had HER-2 status available (2010 and later) with ER+, HER-2(–) tumors, 65,423 patients were included. Women who did not undergo axillary staging continued to demonstrate significantly worse BCSS (sdHR, 1.17; 95% CI, 1.05–1.31), similar to the main analysis (Fig. 6).
FIGURE 6.: Sensitivity analyses for breast cancer-specific survival using propensity score weighting and multivariable Cox regression, with additional adjustment for receipt of radiotherapy or chemotherapy. Presented are subdistribution hazard ratios and 95% CIs for no axillary staging compared to receipt of axillary staging. The ER+ and HER-2(–) subgroup only includes women diagnosed after 2010, when HER-2 status was available. MV indicates multivariable.
We calculated an E-value for the final OS result of 1.51 (on the RR scale). This indicates an unmeasured confounder would need to have an association at least this strong with both axillary staging and OS, above and beyond the included confounders, to move the lower confidence limit of the study to the null.
DISCUSSION
In this large population-based cohort of elderly breast cancer patients with T1/2 tumors, women who did not undergo axillary staging were older and had better tumor characteristics. They were also significantly less likely to have chemotherapy and radiotherapy. After adjustment for baseline characteristics using overlap propensity score weighting and chemotherapy/radiotherapy receipt, women who did not undergo axillary staging had significantly worse overall and BCSS. The adverse impact on BCSS appeared more pronounced in women <80 years. Information on comorbidity status was not available.
These results greatly expand upon the findings of other observational studies in this area.16,17 Using multivariable Cox regression and SEER data from 2004 to 2012, Chagpar et al16 showed elderly women who did not undergo axillary staging had significantly worse OS (adjusted HR, 1.58; 95% CI, 1.53–1.63) and BCSS (adjusted HR, 3.10; 95% CI, 2.96–3.25). Our results show a much more attenuated effect on BCSS, suggesting our approach has more effectively addressed confounding. Notably, Chagpar et al16 used cause-specific HRs to estimate BCSS. This approach censors individuals who experience the competing risk and assumes competing risks are independent, which may not be a reasonable assumption.36 In comparison, sdHRs continue to account for individuals who have died from nonbreast cancer causes and are more suited to understanding the effect of axillary staging on the probability of dying from breast cancer over time.37,38 Finally, we were able to demonstrate our results in women known to be HER-2(–), which has not been presented in previous analyses using SEER.16,18
Our use of propensity score weighting allowed for objective assessment of balance on important baseline patient characteristics, which is not possible in multivariable regression. A previous SEER analysis of 20,151 elderly women diagnosed from 1990 to 1995 found no significant association between BCSS and undergoing axillary staging after propensity score stratification (HR, 1.12; P = 0.066).18 However, this study was much smaller than our analysis and does not reflect contemporary axillary staging practices (only axillary lymph node dissection was used). Further, propensity score weighting is known to perform better than stratification,39 and the authors note there was residual imbalance among age, tumor size, and geographic area.18 Other observational studies that have suggested no impact on survival were small noncomparative studies,20 included SLNB-positive women in the incomplete axillary staging group,19 or excluded women who did not have axillary staging.21
The Choosing Wisely Guidelines recommend avoiding SLNB in HR+, HER-2(–), clinically node-negative elderly breast cancer patients.5 In survey studies, breast surgeons have expressed uncertainty regarding the strength of evidence supporting the guideline.40,41 Two randomized controlled trials8,9 that randomized elderly women to receive or not receive axillary dissection are cited in support of this recommendation. Neither showed a significant difference in survival.7 The applicability of this evidence base to contemporary axillary staging decisions has some limitations. These trials were completed between 1993 and 2002 and examined axillary dissection rather than SLNB. Both trials cited avoiding the morbidity of axillary dissection as the primary rationale for the study, and surgeons are reluctant to omit staging in part because of SLNBs lower morbidity profile.8,9,40 Most importantly, adjuvant treatments were standardized, and women in both arms received endocrine therapy and radiotherapy in equal proportions. These results do not exclude the possibility that elderly women who do not receive axillary staging outside of a randomized controlled trial may be undertreated and experience worse outcomes. Our results demonstrated even after adjustment for baseline characteristics, women who did not undergo axillary staging were approximately half as likely to receive chemotherapy and radiotherapy. This supports findings from other authors who have examined adjuvant treatment use in this setting.42 Surgeons have also cited concerns about patients receiving appropriate adjuvant treatment as a rationale for continuing to use SLNB in this population.40
This is the largest study to examine the effect of axillary staging on BCSS in elderly women,16,18 and the first to report adjusted analyses stratified by important characteristics such as age, tumor size, ER status, and ER+/HER-2(–) status. While in absolute terms, a smaller proportion of women <80 years died of breast cancer, the effect of not undergoing axillary staging on BCSS was more pronounced in this group. Further work is needed to confirm these findings.
Additional strengths of our study include rigorous handling of missing data though multiple imputations and the use of advanced methods of addressing confounding. Overlap propensity score weights are a novel weighting method with several theoretical advantages over other methods of propensity score weighting.31 Our results support the finding that overlap weights result in near-perfect balancing of included baseline characteristics and result in less extreme weights even compared to stabilized IPTW weights. However, our results were not contingent on the use of this method, as sensitivity methods utilizing more traditional adjustment did not change our conclusions. Using recently available chemotherapy and radiotherapy data, we were also able to adjust for these important confounders.
There are limitations to this analysis. Clinical nodal status is not consistently available in SEER, which prevents us from truly restricting the cohort to those who were clinically node negative. However, we excluded T3/T4 tumors, patients with metastatic disease at presentation, and those who received neoadjuvant therapy. We also excluded patients who had aspiration of axillary nodes, as this strategy may be used to confirm disease in palpable nodes. While we have included known confounders available in the SEER data, additional unmeasured confounding is likely. In particular, while comorbidity data are not available in SEER, individual health status plays an important role in surgical decision-making.40 The calculated E-value of 1.51 suggests an unmeasured confounder with a reasonably strong association with both axillary staging and OS could make the results of the study nonsignificant. There is also the possibility of misclassification in SEER when determining chemotherapy and radiotherapy receipt.43 Our competing risks analysis is potentially limited by the accuracy of SEER-reported cause of death, which is derived from death certificates. However, the variation in cancer-specific survival from SEER compared to relative survival methods has been reported to be small, particularly for breast cancer.44
There remains active debate regarding the appropriate utilization of axillary staging in elderly women with early-stage breast cancer. This study suggests that women who do not undergo axillary staging may have poorer overall and disease-specific survival. It is still unclear to what extent this disparity is due to undertreatment, more granular differences in adjuvant treatment we were not able to study, or unmeasured confounding. Further research is needed to explore these factors.
REFERENCES
1.
Surveillance,
Epidemiology,
and End Results Program. Cancer Stat Facts: Female Breast Cancer. Available at:
https://seer.cancer.gov/statfacts/html/breast.html. Published 2019. Accessed April 12, 2019.
2. Diab SG, Elledge RM, Clark GM. Tumor characteristics and clinical outcome of elderly women with breast cancer. J Natl Cancer Inst. 2000;92:550–556.
3. Satariano WA, Ragland DR. The effect of comorbidity on 3-year survival of women with primary breast cancer. Ann Intern Med. 1994;120:104–110.
4. Extermann M, Aapro M. Assessment of the older cancer patient. Hematol Oncol Clin North Am. 2000;14:63–77, viii–ix.
5. Choosing Wisely. Society of Surgical Oncology. Available at:
https://www.choosingwisely.org/societies/society-of-surgical-oncology/. Published 2016. Accessed October 25, 2020.
6. Gradishar WJ, Anderson BO, Abraham J. NCCN guidelines. Breast Cancer. 2019;215:40.
7. Liang S, Hallet J, Simpson JS, et al. Omission of axillary staging in elderly patients with early stage breast cancer impacts regional control but not survival: a systematic review and meta-analysis. J Geriatr Oncol. 2017;8:140–147.
8. Martelli G, Boracchi P, Ardoino I, et al. Axillary dissection versus no axillary dissection in older patients with T1N0 breast cancer: 15-year results of a randomized controlled trial. Ann Surg. 2012;256:920–924.
9. Rudenstam CM, Zahrieh D, Forbes JF, et al. Randomized trial comparing axillary clearance versus no axillary clearance in older patients with breast cancer: first results of International Breast Cancer Study Group Trial 10-93. J Clin Oncol. 2006;24:337–344.
10. DeMichele A, Putt M, Zhang Y, et al. Older age predicts a decline in adjuvant chemotherapy recommendations for patients with breast carcinoma: evidence from a tertiary care cohort of chemotherapy-eligible patients. Cancer. 2003;97:2150–2159.
11. Mandelblatt JS, Cai L, Luta G, et al. Frailty and long-term mortality of older breast cancer patients: CALGB 369901 (Alliance). Breast Cancer Res Treat. 2017;164:107–117.
12. Temple LK, Baron R, Cody HS III, et al. Sensory morbidity after
sentinel lymph node biopsy and axillary dissection: a prospective study of 233 women. Ann Surg Oncol. 2002;9:654–662.
13. Wang T, Baskin AS, Dossett LA. Deimplementation of the Choosing Wisely recommendations for low-value breast cancer surgery: a systematic review. JAMA Surg. 2020;155:759–770.
14. Wang T, Bredbeck BC, Sinco B, et al. Variations in persistent use of low-value breast cancer surgery. JAMA Surg. 2021;156:353–362.
15. Schrenk P, Rieger R, Shamiyeh A, et al. Morbidity following
sentinel lymph node biopsy versus axillary lymph node dissection for patients with breast carcinoma. Cancer. 2000;88:608–614.
16. Chagpar AB, Hatzis C, Pusztai L, et al. Association of LN evaluation with survival in women
aged 70 years or older with clinically node-negative hormone receptor positive breast cancer. Ann Surg Oncol. 2017;24:3073–3081.
17. Van Leeuwen BL, Rosenkranz KM, Feng LL, et al.; Department of Surgical Oncology, MD Anderson Cancer Center. The effect of under-treatment of breast cancer in women 80 years of age and older. Crit Rev Oncol Hematol. 2011;79:315–320.
18. Aziz D, Gardner S, Pritchard K, et al. Selective application of axillary node dissection in elderly women with early breast cancer. Ann Surg Oncol. 2007;14:652–659.
19. Poodt IGM, Schipper RJ, Vugts G, et al. The rationale for and long-term outcome of incomplete axillary staging in elderly women with primary breast cancer. Eur J Surg Oncol. 2018;44:1714–1719.
20. Chung A, Gangi A, Amersi F, et al. Not performing a sentinel node biopsy for older patients with early-stage invasive breast cancer. JAMA Surg. 2015;150:683–684.
21. Sun J, Mathias BJ, Sun W, et al. Is it wise to omit sentinel node biopsy in elderly patients with breast cancer? Ann Surg Oncol. 2021;28:320–329.
22. Doll KM, Rademaker A, Sosa JA. Practical guide to surgical data sets:
Surveillance,
Epidemiology, and End Results (SEER) database. JAMA Surg. 2018;153:588–589.
23. von Elm E, Altman DG, Egger M, et al.; STROBE Initiative. The Strengthening the Reporting of Observational Studies in
Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–1457.
24.
Surveillance,
Epidemiology,
and End Results Program. SEER Incidence Database - SEER Data & Software. Available at:
https://seer.cancer.gov/data/index.html. Published 2019. Accessed February 26, 2019.
25. American College of Surgeons. Breast - CS Site-Specific Factor 3. Available at:
http://web2.facs.org/cstage0205/breast/Breast_lac.html. Published 2013. Accessed February 28, 2019.
26.
Surveillance,
Epidemiology,
and End Results Program. Scope of Regional Lymph Nodes Surgery- SEER Documentation. Available at:
https://seer.cancer.gov/seerstat/variables/seer/regional_ln/index.html. Published 2012. Accessed February 28, 2019.
27.
Surveillance,
Epidemiology,
and End Results Program. Months Survived Based on Complete Dates. Available at:
https://seer.cancer.gov/survivaltime/index.html. Published 2013. Accessed June 29, 2020.
28.
Surveillance,
Epidemiology,
and End Results Program. SEER Cause of Death Recode 1969+ (9/17/2004) - SEER Data Reporting Tools. Available at:
https://seer.cancer.gov/codrecode/1969_d09172004/index.html. Published 2004. Accessed June 29, 2020.
29. Mehta N, Kalra A, Nowacki AS, et al. Association of use of angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers with testing positive for coronavirus disease 2019 (COVID-19). JAMA Cardiol. 2020;5:1020–1026.
30. Thomas LE, Li F, Pencina MJ. Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial. JAMA. 2020;323:2417–2418.
31. Li F, Morgan KL, Zaslavsky AM. Balancing covariates via propensity score weighting. J Am Stat Assoc. 2018;113:390–400.
32. Bolch CA, Chu H, Jarosek S, et al. Inverse probability of treatment-weighted competing risks analysis: an application on long-term risk of urinary adverse events after prostate cancer treatments. BMC Med Res Methodol. 2017;17:93.
33. VanderWeele TJ, Ding P. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med. 2017;167:268–274.
34. Leyrat C, Seaman SR, White IR, et al. Propensity score analysis with partially observed covariates: how should multiple imputation be used? Stat Methods Med Res. 2019;28:3–19.
35. Rubin DB. Multiple imputation after 18+ years. J Am Stat Assoc. 1996;91:473–489.
36. Austin PC, Lee DS, Fine JP. Introduction to the analysis of survival data in the presence of competing risks. Circulation. 2016;133:601–609.
37. Austin PC, Fine JP. Accounting for competing risks in randomized controlled trials: a review and recommendations for improvement. Stat Med. 2017;36:1203–1209.
38. Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol. 2009;170:244–256.
39. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46:399–424.
40. Smith ME, Vitous CA, Hughes TM, et al. Barriers and facilitators to de-implementation of the Choosing Wisely® guidelines for low-value breast cancer surgery. Ann Surg Oncol. 2020;27:2653–2663.
41. Minami CA, Bryan AF, Revette AC, et al. Oncologists’ perspectives on omission of
sentinel lymph node biopsy in women >70 years with early-stage hormone receptor-positive breast cancer. JCO. 2021;39:48–48.
42. Chagpar AB, Horowitz N, Sanft T, et al. Does lymph node status influence adjuvant therapy decision-making in women 70 years of age or older with clinically node negative hormone receptor positive breast cancer? Am J Surg. 2017;214:1082–1088.
43.
Surveillance,
Epidemiology,
and End Results Program. SEER Treatment Data Limitations - SEER Data & Software. Available at:
https://seer.cancer.gov/data/sample-treatment-limitations.html. Published 2020. Accessed March 29, 2021.
44. Hu CY, Xing Y, Cormier JN, et al. Assessing the utility of cancer-registry-processed cause of death in calculating cancer-specific survival. Cancer. 2013;119:1900–1907.
