Survival Outcomes of Breast-Conserving Therapy versus Mastectomy in Early-Stage Breast Cancer, Including Centrally Located Breast Cancer: A SEER-Based Study

Purpose This study aims to analyze the survival outcomes of breast cancer (BC) patients, especially centrally located breast cancer (CLBC) patients undergoing breast-conserving therapy (BCT) or mastectomy. Methods Surveillance, epidemiology, and end results (SEER) data of patients with T1-T2 invasive ductal or lobular breast cancer receiving BCT or mastectomy were reviewed. We used X-tile software to convert continuous variables to categorical variables. Chi-square tests were utilized to compare baseline information. The multivariate logistic regression model was performed to evaluate the relationship between predictive variables and treatment choice. Survival outcomes were visualized by Kaplan–Meier curves and cumulative incidence function curves and compared using multivariate analyses, including the Cox proportional hazards model and competing risks model. Propensity score matching was performed to alleviate the effects of baseline differences on survival outcomes. Result A total of 180,495 patients were enrolled in this study. The breast preservation rates fluctuated around 60% from 2000 to 2015. Clinical features including invasive ductal carcinoma (IDC), lower histologic grade, smaller tumor size, fewer lymph node metastases, positive ER and PR status, and chemotherapy use were independently correlated with BCT in both BC and CLBC cohorts. In all the classic Cox models and competing risks models, BCT was an independent favorable prognostic factor for BC, including CLBC patients in most subgroups. In addition, despite the low breast-conserving rate compared with tumors located in the other areas, CLBC did not impair the prognosis of BCT patients. Conclusion BCT is optional and preferable for most early-stage BC, including CLBC patients.


Introduction
Breast-conserving therapy (BCT), which refers to breastconserving surgery plus postoperative radiotherapy, is considered a standard treatment for early-stage breast cancer. Several clinical trials, including NSABP B-06, Milan, and EORTC 10801, have proven that the survival outcomes of patients treated with BCT are equivalent to those undergoing mastectomy, despite a relatively higher risk of local recurrence [1][2][3]. In addition, BCT patients had significantly improved body image, satisfaction with treatment and sexual functioning, and there was no significant difference in fear of recurrence between patients treated with BCT and mastectomy [4,5].
Centrally located breast cancer (CLBC) usually refers to tumors located in the area within 2 cm of the nipple-areola complex (NAC) but without NAC involvement. Because of the particularity of its position, surgeons are often not inclined to perform BCT in CLBC. To date, there are only limited studies focused on the safety and prognosis of BCT compared with mastectomy in CLBC, and none of these studies are comprehensive enough [6][7][8].
To this end, we conducted a detailed retrospective study based on the SEER database to evaluate the prognosis of BC patients undergoing BCT and mastectomy, especially CLBC patients. Moreover, we used both the classic Cox proportional hazards model and competing risks model to ensure the rigor of this research and reduce statistical errors. Furthermore, we performed a series of subgroup analyses to help surgeons make the best choice according to the patient's baseline information.

Participants.
e data for this study were extracted from research plus data from 18 registries of the Surveillance, Epidemiology, and End Results (SEER) database released in November 2020. We enrolled 180,495 female patients who received mastectomy or BCT (breastconserving surgery plus postoperative radiotherapy) after

Statistical Analysis.
Demographic information and clinical characteristics were compared using Chi-square tests. Continuous variables were converted to categorical variables using X-tile software (Version 3.6.1) [9]. Multivariable logistic regression was utilized to evaluate the relationship between predictive variables and treatment choice. We used the Kaplan-Meier curve to estimate survival outcomes, and the log-rank test was used to perform between-group comparisons. e Cox proportional hazards model was performed to fit demographic and clinical characteristics for overall survival (OS) and breast cancer-specific survival (BCSS). A1 :1 ratio propensity score matching (PSM) method with a caliper of 0.02 was performed to alleviate the influence of baseline differences on survival outcomes in CLBC and upper-outer breast cancer (UOBC) patients who underwent BCT. Matching variables included the year of diagnosis, age, race, histological type,  Since the Cox regression model might not accurately estimate the risk of a particular event when competing risks exist, we performed the competing risks analysis to better evaluate the relationship between therapeutic strategies and survival outcomes. We treated death from other causes as a competing event. e risk of death caused by breast cancer was estimated using the cumulative incidence function curve and compared across groups using Gray's test. e Fine-Gray model (also known as the subdistribution hazard function) and the cause-specific (CS) model were applied for multifactor competing risks analyses. A p-value < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS software (version 26.0, IBM Corporation) and SAS software (version 9.4, SAS Institute). e clinical characteristics are displayed in Table 1. Patients between 64 and 72 years old, white patients, patients with the histology of invasive ductal carcinoma (IDC), and patients with less aggressive characteristics including histologic grades I and II, T1 stage, N0 stage, and positive ER and PR status were more inclined to receive BCT. In addition, patients who underwent BCT were less likely to receive chemotherapy. Figure 1 shows the trend of BCT and mastectomy for the indicated patients from 2000 to 2015, and the breast preservation rates fluctuated around 60%.  histologic grade, smaller tumor size, fewer lymph node metastases, positive ER and PR status, and chemotherapy use were independently correlated with BCT compared with mastectomy ( Table 2).

Survival Analysis among BC Patients Treated with BCTor
Mastectomy and Subgroup Analysis. e Kaplan-Meier survival curve revealed that patients who received BCT had better overall survival (OS, p < 0.001) and breast cancerspecific survival (BCSS, p < 0.001) than those who underwent mastectomy (Figures 2(a) and 2(b)). e cumulative incidence function curve also showed that patients undergoing BCT had a lower risk of breast cancer-associated death (Figure 2(c)). en, we conducted the Cox proportional hazards model and the competing risks models for the multivariate analyses (Table 3). e results obtained from the Cox model indicated that the independent risk factors associated with the OS and BCSS of BC patients included the year of diagnosis, age, race, histological type, histologic grade, T stage, N stage, ER status, PR status, and chemotherapy status. Notably, BCT was found to be a favorable prognostic factor for OS (HR 0.764, 95% CI 0.745-0.783, p < 0.001) and BCSS (HR 0.760, 95% CI 0.734-0.787, p < 0.001). Similar results were obtained from competing risks models. BCT was still an independent risk factor in the Fine-Gray model (HR 0.807, 95% CI 0.779-0.837, p < 0.001) and the CS model (HR 0.784, 95% CI 0.757-0.812, p < 0.001). e subgroup analysis further demonstrated that patients treated with BCT had significantly better prognoses than those who received mastectomy in nearly all subgroups, except for patients of other races in the OS model and ILC patients in the BCSS, Fine-Gray, and CS models (Figure 3).     (Figure 4). Table 4. Patients between 64 and 72 years old, patients of the white race, patients with IDC, and patients with less aggressive characteristics including histologic grades I and II, T1 stage, N0 stage, ER positivity, and PR positivity tended to receive BCT. Similarly, chemotherapy was less likely to be used for patients who underwent BCT.

Predictive Factors of BCT among CLBC Patients. e clinical characteristics of CLBC patients are shown in
In the multivariate logistic regression model, features including age between 64 and 72, white race, IDC, histologic grades I and II, T1 stage, N0 stage, positive PR status, and chemotherapy use were independently associated with BCT compared with mastectomy (Table 5).

Survival Analysis among CLBC Patients Treated with BCT and Mastectomy.
e Kaplan-Meier survival curve showed that CLBC patients treated with BCT had enhanced overall survival (OS, p < 0.001) and breast cancer-specific survival (BCSS, p < 0.001) compared with those who underwent mastectomy (Figures 5(a) and 5(b)). Besides, the cumulative incidence function curve showed that CLBC patients who received BCT were less likely to die from breast cancer ( Figure 5(c)). Moreover, the Cox proportional hazards model indicated that the year of diagnosis, age, race, histologic grade, T stage, N stage, ER status, PR status, and chemotherapy status were independent risk factors associated with the OS and BCSS of CLBC patients. BCT was also found to be a favorable prognostic factor for OS (HR 0.734, 95% CI 0.672-0.802, p < 0.001) and BCSS (HR 0.660, 95% CI 0.576-0.755, p < 0.001).
In competing risks analyses, BCT was still an independent favorable prognostic factor in the Fine-Gray model (HR 0.709, 95% CI 0.617-0.815, p < 0.001) and the CS model (HR 0.686, 95% CI 0.598-0.786, p < 0.001). However, the black race, which was proven to be a risk factor in the Cox model, was nonsignificant in the Fine-Gray model (p � 0.133) and the CS model (p � 0.109) ( Table 6). e subgroup analysis indicated that patients treated with BCT had significantly better OS in almost all subgroups, except for patients of other races. Furthermore, patients who received BCTshared improved BCSS except for patients of black or other races, with ILC, histologic grade I, and T1a stage compared with those who underwent mastectomy. In the competing risks analyses, BCT patients had better prognoses except for those diagnosed between 2000 and 2007, of black or other races, with ILC, histologic grade I, T1a stage, N3 stage, and negative ER status ( Figure 6).

Survival Analysis of BCT Patients with Differentially Located Tumors.
To further reveal the safety and prognosis of BCT in CLBC patients, we performed survival analyses among patients with tumors located in five distinct areas (Table 7). When compared to UOBC, despite a worse OS  Table 7,  Table S1). en, due to the huge differences in the patient number and clinical characteristics, we conducted propensity score matching (PSM) to reduce the influence of confounding factors. After matching, 5,864 patients in each cohort were enrolled. e results showed that patients with CLBC and UOBC had comparable prognoses in all models except the Cox-OS model ( Table 7, Table S2). Intriguingly, CLBC patients showed improved prognoses when compared to those with UIBC ( Table 7, Table S3), LOBC (Table 7,  Table S4), and LIBC (Table 7, Table S5). Subsequently, we performed subgroup analyses among patients with CLBC and those with UIBC ( Figure S1), LOBC ( Figure S2), and LIBC ( Figure S3). Patients in CLBC group showed similar prognoses compared to those with UIBC and LOBC in nearly all subgroups. When compared to LIBC, CLBC patients had better prognoses in most subgroups.
In addition, to explore whether the difference in prognosis between CLBC and LIBC is caused by internal mammary node (IMN) metastasis, we performed a survival analysis among patients without IMN metastasis in these two cohorts. e results indicated that compared to LIBC, CLBC was still an independent favorable prognostic factor among BCT patients (Table S6).

Discussion
To the best of our knowledge, this is the first populationbased retrospective study using the competing risks model to evaluate the prognosis of TI-T2 CLBC patients undergoing BCT or mastectomy.

e Breast Journal
Several clinical trials have shown that patients treated with BCT and mastectomy have equivalent prognoses. For example, the NSABP B-06 trial demonstrated no significant differences in disease-free survival, distant-disease-free survival, or overall survival between early-stage BC patients treated with BCT and mastectomy [1]. e DBCG-82TM trial also showed no significant difference in 10-year recurrence-free survival and 20-year overall survival between these two groups [10]. Several studies even showed improved BCSS and OS for BCT compared with mastectomy [11,12]. Moreover, BCT also achieved superior cosmetic outcomes than mastectomy [4]. However, our research demonstrated that nearly 40% of early-stage BC patients received mastectomy each year between 2000 and 2015, and this proportion increased to over 50% in CLBC patients. e hesitation of surgeons performing BCT for CLBC patients may be partially due to the special location or anatomic structure of tumors, including the complex lymphatic drainage [13]. Although a recent retrospective study based on the SEER database discussed the benefit of BCT in CLBC patients, only the classic Cox proportional hazards model was applied, and detailed subgroup analysis was absent [14]. Zhang's study compared the prognosis of breast-conserving surgery and mastectomy, but the postoperative radiotherapy status was not controlled [8].
In this study, we revealed a higher proportion of IDC, lower histologic grade, T stage and N stage, and positive ER and PR status to receive BCT for BC, including CLBC patients. ese factors were mostly associated with a smaller region or less malignant tumor. However, as the SEER database does not collect information on the sequence of chemotherapy and surgery, we could not clarify the influence of neoadjuvant chemotherapy on the choice of BCT or mastectomy. In addition, the status of endocrine therapy and Ki-67, which influence the survival of BC patients, was also unattainable from SEER [15,16].
Our research demonstrated significantly improved OS and BCSS for BCT in both the whole BC cohort and CLBC alone cohort, which was concordant with previous studies [14,17,18]. Importantly, we performed the competing risks models, which take into account not only deaths caused by BC but also deaths caused by other events as well as their effects. We presented the outcomes of two competing risks models: the Fine-Gray model, which is appropriate for evaluating prognostic factors [19], and the CS model, which is more suitable for etiological research [20]. In line with the Cox model, both competing risks models showed better prognoses for BCT, whether in the entire BC cohort or CLBC cohort. ese results further proved the safety and efficacy of BCT in the selected population. In addition, patients with ILC showed better survival outcomes than IDC patients, which aligned with earlier studies [21,22]. When deeply dug, most subgroups of BC could benefit from BCT. All subgroup patients of CLBC showed at least equivalent prognoses receiving BCT compared with mastectomy, and some subgroups such as white race, IDC, lower N stage, and positive ER status could benefit from BCT. Combined with previous studies, patients with these beneficial factors could be more inclined to choose BCT in future clinical decisions [23].
When comparing survival outcomes of BCT in CLBC and other areas, we found that CLBC was comparable with UOBC in the Cox-BCSS model, Fine-Gray model, and CS model after propensity score matching and better than tumors located in the other three quadrants. Some studies have shown that LIBC is an unfavorable prognostic factor for early-stage BC patients, probably due to the higher possibility of IMN metastasis [24,25]. However, our study showed that CLBC still had a better prognosis than LIBC among patients without IMN metastasis in the BCT cohort.
ere are still some limitations in our research. First, we could not evaluate the influence of neoadjuvant chemotherapy on surgical choice and survival outcome. Information on local recurrence rates was also unavailable.
us, we could not compare local recurrence rates as a secondary outcome between the BCT and mastectomy cohorts. In addition, we could not obtain data about the cosmetic results and satisfaction with body image after BCT. Finally, the status of endocrine therapy, Ki-67, and patients' income level was absent, which may introduce bias into our results.
In conclusion, utilizing the classic Cox proportional hazards model and competing risks model, our research not only revealed the superiority of BCT compared with mastectomy in most early-stage breast cancer but also proved that patients with CLBC could also obtain better prognoses from BCT.

Data Availability
e datasets analyzed during the current study are available in the SEER database. https://seer.cancer.gov/.

Ethical Approval
is study was performed in line with the principles of the Declaration of Helsinki. Institutional Review Board permission was not required because the SEER database is a deidentified national database.

Consent
Informed consent was not required because personally identifiable information was not accessed.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this article.

Authors' Contributions
TSY, FM, and BLG designed the study. WLC, TW, YQD, YD, and HYZ collected the data. TSY, ZAC, JYF, ABH, and MCL conducted the statistical analyses. TSY, FM, and YLL collectively conceptualized the manuscript. BLG edited the manuscript and provided critical comments. All authors read and approved the final manuscript.