Survival Comparisons between Breast Conservation Surgery and Mastectomy Followed by Postoperative Radiotherapy in Stage I–III Breast Cancer Patients: Analysis of the Surveillance, Epidemiology, and End Results (Seer) Program Database

Background: This study aims to evaluate the overall and breast cancer-specific survival (BCSS) after breast-conserving surgery (BCS) plus radiotherapy (RT) compared with mastectomy plus RT in resectable breast cancer. Moreover, the aim is to also identify the subgroups who benefit from BCS plus RT and establish a predictive nomogram for stage II patients. Methods: Stage I–III breast cancer patients were identified from the Surveillance, Epidemiology, and End Results (SEER) database between 1990 and 2016. Patients with available clinical information were split into two groups: BCS plus RT and mastectomy plus RT. Kaplan–Meier survival analysis, univariate and multivariate regression analysis, and propensity score matching were used in the study. Hazard ratio (HR) was calculated based on stratified Cox univariate regression analyses. A prognostic nomogram by multivariable Cox regression model was developed for stage II patients, and consistency index (C-index) and calibration curve were used to evaluate the accuracy of the nomogram in the training and validation set. Results: A total of 24,590 eligible patients were enrolled. The difference in overall survival (OS) and BCSS remained significant in stage II patients both before and after PSM (after PSM: OS: HR = 0.8536, p = 0.0115; BCSS: HR = 0.7803, p = 0.0013). In stage II patients, the survival advantage effect of BCS plus RT on OS and BCSS was observed in the following subgroups: any age, smaller tumor size (<1 cm), stage IIA (T2N0, T0–1N1), ER (+), and any PR status. Secondly, the C-indexes for BCSS prediction was 0.714 (95% CI 0.694–0.734). The calibration curves showed perfect agreement in both the training and validation sets. Conclusions: BCS plus RT significantly improved the survival rates for patients of stage IIA (T2N0, T0–1N1), ER (+). For stage II patients, the nomogram was a good predictor of 5-, 10-, and 15-year BCSS. Our study may help guide treatment decisions and prolong the survival of stage II breast cancer patients.


Background
Breast-conserving surgery (BCS) plus postoperative radiotherapy (RT) was recommended as an alternative to mastectomy for early-stage breast cancer patients based on several randomized controlled trials, demonstrating that BCS followed by postoperative RT is as effective as mastectomy [1][2][3][4]. Recently, data from several population-based studies suggest the superiority of BCS plus RT over mastectomy without RT [5][6][7][8][9][10][11]. With the advancements and standardization of surgery and postoperative adjuvant RT technology,

Data Sources
The SEER database consists of several tumor registries in different regions, collecting information about 26% to 30% of the US population. It contains detailed demographic, socioeconomic, cancer, and treatment information. The SEER data is publicly available. We obtained the clinical data of female patients diagnosed with breast cancer in the SEER database from 1990 to 2016 (November 2018 Submission, SEER 13). The data includes age at diagnosis, race, pathological type, grade, laterality, TNM stage, tumor size, lymph node infiltration, estrogen receptor (ER) status, progesterone receptor (PR) status, treatments, and follow-up data.

Cohort Selection
We used SEER *-Stat 8.3.8 software (National Cancer Institute, Bethesda, MD, USA) to extract data from the SEER database. Patients diagnosed with primary breast cancer (SEER cancer site code: 50.0) in American Joint Committee on Cancer (AJCC) stage I to III from 1990 to 2016 were included in our study. Patients who had previously undergone BCS (site-specific surgery codes 20-24) plus RT or mastectomy (site-specific surgery codes 30,40,50,60,70,76) plus RT were screened out. Patients with complete clinical and demographic information were included. Using the histological code of ICD-0-3, the patients were divided into lobular carcinoma (8570 to 8580), ductal carcinoma (8500), and other pathological types. The 7th edition of AJCC breast cancer was released in 2009, so the patients diagnosed between 1990 and 2010 were staged with 6 editions, and the patients after 2010 were staged with 7 editions. To eliminate deviations due to different RT techniques, patients only receiving external beam radiation therapy were included. The following patients were excluded: those who lack clinical and follow-up information and those whose surgical methods are unknown (site-specific surgery codes 99, site-specific surgery codes 90).

Variables
The data extracted in this study included age, race, pathologic grade, laterality, stage, tumor size, lymph node infiltration, ER status, PR status, chemotherapy, and treatment strategies (BCS plus RT vs. mastectomy plus RT). The study focused on OS, BCSS, and hazard ratio (HR). OS refers to the time a patient lived from breast cancer diagnosis to death from any cause, and BCSS refers to deaths from breast cancer.

Propensity Score Matching
Propensity score matching (PSM), proposed in 1983 by Rosenbaum and Rubin, and not limited by the number of events, was performed for all relevant confounding factors to minimize selection bias [22,23]. The nearest available neighbor matching and caliper matching were used in our research in PSM [24,25]. For the patients of stage III, the ratio of PSM is 1:1. For the patients of stage II, to balance the effects of confounding, PSM with a 1:1 ratio and the caliper of 0.02 was set. Only when the propensity score of the control group (mastectomy plus RT) is within a certain distance (0.02), the control group will be matched with the case. Matched covariates include age, race, stage, tumor size, ER and PR status, and chemotherapy.

Statistical Methods
In this study, all data are counted by EXCEL, and differences in baseline characteristics between the two groups of people were assessed by the χ2 test or Fisher exact test (BCS plus RT vs. mastectomy plus RT). The potential risk factors of OS and BCSS were analyzed by the univariate and multivariate Cox regression model; 95% confidence interval and log-rank test were calculated. The 5-year, 10-year, and 15-year OS and BCSS were assessed with the Kaplan-Meier method and compared with the log-rank test. The results of the subgroup analysis were presented by the forest plot. The nomogram construction and validation were performed based on the results of the Cox proportional risk model. Receiver operating characteristic (ROC) and calibration curve were used to evaluate the accuracy of the model. The C-index value is positively correlated with the predictive performance of the model. Calibration curves were plotted at 5, 10, and 15 years by a bootstract involving 100 resamples. Ideally, the points on the calibration diagram should be close to the 45 • diagonal. The above statistical analysis was completed by R software (version 4.2.0, Vienna, Austria), using R packages such as "RMS", "Foreign", "Survival", "forestplot", "Tableone", "MatchIt", "survivalROC", "caret" etc. p < 0.05 is regarded as the statistical difference.

Patient Characteristics and Determining Independent Prognostic Factors
The data filtering process is shown in Figure 1. A total of 24,590 eligible patients were enrolled, including mastectomy plus RT (10,785, 43.9%) vs. BCS plus RT (13,805,56.1%). Our data indicated that older patients (>65), better differentiated (grade I/II), earlier staged (stage I), T1, N0, smaller size (<2 cm), and ER and PR positive patients are more likely to receive BCS plus RT, which is consistent with most clinical guidelines ( Table 1). The results of multivariate regression were consistent with the results of univariate regression. The results showed that all factors affected the Stage I-III breast cancer patients' OS and BCSS except laterality and histologic type (Supplementary Tables S1 and S2). It is worth mentioning that treatment significantly affected the OS and BCSS of Stage I-III patients. BCS plus RT can reduce the risk of death by 15% (HR = 0.8473 (CI 0.7841-0.9155) p < 0.001. Supplementary Table S1) and the risk of specific death from breast cancer by 18% (HR = 0.820 (CI 0.746-0.901) p < 0.001. Supplementary Table S2).
To explore the specific beneficiaries of BCS plus RT, the patients were stratified into two different risk groups (stage II/III) for further evaluation. The χ2 test or Fisher exact test was used to compare the clinical characteristics between BCS plus RT and mastectomy plus RT. The results showed low comparability of the multiple factors for stage II or stage III patients (Tables 2 and 3). PSM was conducted for stage II and III patients to eliminate the influence of all relevant factors, respectively (Tables 2 and 3). As shown in Figure 2C

Subgroups Benefiting from BCS plus RT in Stage II
Furthermore, 8296 stage II cases were divided into the training set and test set in a ratio of 7:3. Univariate and multivariate analyses were conducted for stage II patients based on the training set. As shown in Table 4, age, race, grade, N, intervention, tumor size, ER status, and PR status were found to be independent prognostic factors for Stage II patients. Stratified Cox univariate regression analyses were performed for stage II patients, and the results were presented by forest plot (Figure 3). The survival advantage effect of BCS plus RT on OS was observed in all age subgroups compared with mastectomy plus RT (<65 HR = 0.8385 p = 0.030; >65 HR = 0.8149 p = 0.037). However, the effect of BCS plus RT on BCSS was observed only in a subset of patients younger than 65 years (<65 HR = 0.7411 p < 0.001; > 65 HR = 0.8825 p = 0.408). The OS and BCSS benefit with BCS plus RT was observed in the following four subsets: non-blacks group, well-differentiated group, tumor size < 1 cm, and ER(+) group

A Predictive Nomogram for BCSS Based on Stage II Data
Based on these studies, BCS plus RT may lead to more prolonged survival, especially for stage II breast cancer patients. Age, race, grade, N, intervention, tumor size, ER status, and PR status were found to be independent prognostic factors of stage II patients. Based on the above factors, we plotted a predictive nomogram basing the results of multivariate Cox regression analysis and displayed the 5-, 10-, and 15-year BCSS probabilities ( Figure 4A). The C-index for BCSS prediction was 0.714 (95% CI 0.694-0.734). Additionally, the ROC curves ( Figure 4B) and the calibration curves of 5, 10, and 15 years all showed excellent predictive power of the model. The result of the validation set was consistent with that of the training set ( Figure 4C).

Discussion
In this study, our results indicate that stage II-III breast cancer patients who underwent BCS plus RT had better OS and BCSS than those who underwent mastectomy plus RT, with no significant difference for stage I breast cancer patients. After adjusting for confounders by PSM, the difference in OS and BCSS remained significant in patients with stage II. In a subset of stage II breast cancer, the survival advantage effect of BCS plus RT on OS and BCSS was observed regardless of PR status. The survival advantage effect of BCS plus RT on OS was observed in the subset: all age, non-blacks, well-differentiation, T2N0, 1Cm, and ERpositive. Additionally, the BCS plus RT group younger than 65, well-differentiated, T0-1N1, <1 cm, 2-3 cm, and ER-positive subtypes can significantly benefit from BCSS rather than the mastectomy plus RT group, and the difference was statistically significant (Figure 3).
For female patients with early-stage breast cancer, randomized clinical trials and recent retrospective investigations have demonstrated that BCS followed by RT is not inferior to mastectomy and may even be superior [1][2][3][4][5][6][7][8][9][10][11]. Yet, fewer studies have shown whether BCS plus RT still surpassed mastectomy in the case of RT. Our findings were similar to those of Agarwal S et al. [6], which reported that when the number of lymph nodes was less than three, BCS plus RT was preferable to mastectomy and mastectomy plus RT. However, patients with three or more positive lymph node metastasis were excluded from their study, and the number of cases of mastectomy plus RT was small, accounting for only 3% of all cases. A real-world study published in 2016 stratified the mastectomy patients by radiotherapy receipt in Supplementary Tables, demonstrating better breast cancer-specific survival with receipt of BCS plus radiotherapy compared with mastectomy plus RT among stage II-III breast cancer [10]. In 2019, Lan XW et al. reviewed 1816 stage T1-2N1M0 female patients from Sun Y at Sen Memorial Hospital, supporting the results of Agarwal S et al. [19]. Yet, Sun GY et al. demonstrated survival equivalence between BCS plus RT group and mastectomy plus RT group by analyzing 4262 T1-2N1M0 patients in 2020 [20]. There was still no conclusion for breast cancer patients in the intermediate stage, which was the best choice to result in prolonged survival. These results may hardly be representative due to a relatively small number of cases and subgroup analysis results. De Boniface J. collected 48 986 T1-2 N0-2 Swedish data on breast cancer prospectively. Compared to mastectomy with or without radiation, 5-year OS and BCSS were significantly better after BCS. Numerous variables were controlled for in his study, including tumor characteristics, treatment, demographics, comorbidity, and socioeconomic background [21]. Our study is a retrospective analysis of a large cohort. We gathered data on all stage I-III American female breast cancer cases who have received BCS or mastectomy followed by postoperative RT from the SEER database. We include T3-4 and N3 populations in addition to T0-1N1-2, and the average follow-up time for the entire population is 79 months. We made a direct survival comparison based on stage and used PSM to equilibrate confounders.
Our study was the first to identify which subgroups of patients benefit most from BCS plus RT and to develop a stage II patient prognostic nomogram.
The biological mechanism by which BCS plus RT increased survival remains unknown. A study by Onitilo AA et al. showed that BCS plus RT had a better survival rate than the mastectomy group. They concluded that both postoperative radiotherapy and the surgical method itself were likely to have contributed to such a significant survival difference [26]. The advantages of BCS plus RT in early breast cancer have been reported in the population of the United States [5,6], Canada [7], the Netherlands [10], China [27], and South Korea [28]. They tried to believe the advantages of BCS plus RT may be attributed partly to the effects of postoperative radiotherapy treatment. Likewise, a recent Chinese retrospective study corroborated the role of RT by grouping patients according to whether they had received RT, demonstrating the RT group had better 5-year DMFS, DFS, and BCSS compared with the no-RT group [19]. We believe that the improved survival noted in RT plus RT group is not only related to the surgical method but also closely related to the postoperative radiotherapy or interaction between radiotherapy and the immune microenvironment. As we all know, RT can regulate the tumor microenvironment in various ways, including regulating the release of local cytokines, chemokines, and other soluble factors, remodeling the structure of interstitial tumor cells and vascular cells, etc. [29][30][31][32][33]. Goodman CR et al. reported that postoperative radiotherapy prolonged the survival of circulating tumor cell-positivepositive patients undergoing BCS but did not prolong the survival of mastectomy [34]. This may indicate that the immune microenvironment can work with radiotherapy to improve the survival rate of BCS.
According to our results, stage II breast patients who received BCS plus RT had better survival compared with receiving mastectomy plus RT, and the results were consistent before and after matching. For patients with stage I and stage III, the survival of BCS plus RT seem to be similar to that of mastectomy plus RT. The reason for this phenomenon may be that the larger tumor burden and higher risk of metastasis in stage III patients limit the benefits provided by RT. Stage I patients are characterized by smaller tumor burden, breast-conserving surgery and mastectomy can create similar survival, and the synergistic effect of radiotherapy and immunity has not yet been exerted. In a subset of patients with stage 2A (T0-1N1, T2N0), the effect of BCS plus RT on OS and BCSS can be observed. However, the mechanism is still poorly understood, and we plan to explore the answer in the following research.
The limitation lies in that this study is a retrospective study, and the results of this study need to be proved by a robust prospective trial. Significant imbalances of certain covariates may be unavoidable in PSM. We regret that the PR status could not be perfectly matched, while other confounders presented no significant imbalance. Human epidermal growth factor receptor 2 (HER2) expression in breast cancer patients was only documented in the SEER database starting in 2010. Unfortunately, the follow-up time was insufficiently long to include HER2 in the study. In addition, the lack of data information on local recurrence and remote metastasis, the lack of Ki67, BRCA1, BRCA2 mutations, and other genetic information, and the unsubdivided chemotherapy situation are all the limitations of this study.

Conclusions
In summary, BCS plus RT had a superior treatment effect as that of mastectomy plus RT for stage II patients and also provided an equivalent survival for stage I or stage III patients. Moreover, the nomogram was a good predictor of 5-, 10-, and 15-year BCSS for stage II patients. Our results may help guide treatment decisions and prolong the survival of breast cancer patients.  Figure S1: Subgroup analysis for OS and BCSS for stage III patients.
Author Contributions: W.X. and C.W. are co-first authors. W.X. and C.W. contributed to the acquisition of the data and data analysis, and W.X. drafted the manuscript. H.Y. contributed to the major concept of this study and provided funding support. S.F., N.L. and H.W. provided professional advice on analyzing and interpreting the data. All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement: Ethical review and approval were waived for this study due to this study using de-identified data from the National Cancer Institute's Surveillance, Epidemiology, and End Results program.

Informed Consent Statement:
This study used de-identified data from the National Cancer Institute's Surveillance, Epidemiology, and End Results program, and informed consent was not required.

Data Availability Statement:
The datasets analyzed during the current study are available from the National Cancer Institute's Surveillance, Epidemiology, and End Results program database: https://seer.cancer.gov/data/access.html (accessed on 25 June 2020).

Conflicts of Interest:
The authors declare no conflict of interest.

BCSS
Breast cancer-specific survival BCS Breast-conserving surgery RT Radiotherapy OS Overall survival PSM Propensity score matching ER Estrogen receptor PR Progesterone receptor HER2 Human epidermal growth factor receptor 2