Research on the Cutoff Tumor Size of Omitting Radiotherapy after Breast Conserving Therapy in Women Aged 65 Years or Oder with Low-risk Invasive Breast Carcinoma: Results Based on the SEER Database

Background Radiotherapy (RT) after breast-conserving therapy (BCT) is not always necessary in older women staged T1N0M0 with low-risk invasive breast cancer, but few studies have concluded the detailed tumor size as a reference for avoiding RT. We aimed to explore and identify the cutoff tumor size for patients aged 65 or older with T1N0M0 stage and estrogen receptor positive (ER+), human epidermal growth factor receptor 2 positive (HER2-) breast cancer after BCT. that small tumors (tumor ≤ 14 mm in diameter) were associated with better overall survival (OS), BCSD and OCSD, and radiotherapy might be omitted for patients aged 65 or older with T1N0M0 stage, ER + and HER2- small breast tumors after BCT. To analyze the radiation-related survival in patients with small tumors and large tumors, a competitive risk model survival analysis was performed. A total of 38400 eligible patients were classied into small tumor group, of which 25761 (67.1%) of them received RT (small tumor& RT subgroup) and 12639 (32.9%) did not (small tumor& non-RT subgroup), while 13649 patients were classied into large tumor group, of which 9265 (67.9%) of them received RT (large tumor & RT subgroup) and 4384 (32.1%) did not (large tumor & non-RT subgroup). After competitive risk model survival analysis, as shown in Figs. 3a and 3b, patients in small tumor & non-RT subgroup had similar BCSD to those in the small tumor & RT and large tumor& RT subgroup, while patients in large tumor& non-RT subgroup tended to have higher BCSD (Gray’s test, P<0.001) and OCSD (Gray’s test, P<0.001) than the other subgroups. In addition, patients in small tumor& non-RT subgroup tended to have higher OCSD than small tumor& RT and large tumor& RT subgroups, whether before or after the application of PSM. OS, especially cancer specic survival (CSS) is an objective, reliable, and bias-free measurements for patients with FBC in long-term results. In our study, based on analysis of a large cohort of 52049 patients in SEER database from 2010 to 2016 and an integrated range of factors in competing risk model, small tumors ( ≤ 14 mm) were associated with better OS and BCSS for patients with radiotherapy as well as better BCSS for patients with no radiotherapy in comparison with large tumors ( ≥ 15 mm) through X-title analysis. Such results indicated T1N0M0

To further explore and identify the detailed tumor size which could affect the prognosis OS and BCSS for patients aged 65 years or more with negative lymph nodes and ER-positive,T1 stage FBC, we conducted a large cohort of women with FBC from 2010 to 2016 from the population-based database Surveillance, Epidemiology, and End Results (SEER) cancer registry program.

Data Source
We used data from the national cancer institute's SEER program database, which includes population-based data from 18 cancer registries and represents approximately 28% of the U.S population from 1975 to 2016 (18). SEER*Stat Software version 8.3.6 (https://seer.cancer.gov/seerstat/) (Information Management Service, Inc. Calverton, MD, USA) was used to generate the case listing. All procedures were performed in accordance with approved guidelines. This study was approved by the Ethics Committee of the First A liated Hospital of Xi'an Jiaotong University. The SEER data erases the identity information of patients, so there is no need for informed consent from the patients.

Patient Cohort
Female patients with pathologically con rmed breast cancer from 2010 to 2016 were enrolled in the study. Patients were included by After the preliminary selection, patients were excluded by following criteria: (1) unknown TNM stage; (2) the follow-up type of autopsy or death certi cate; (3) unknown tumor size; (4) unknown PR status; (5) unknown laterality; (6)unknown RT status. The selecting procedure was shown in Figure 1.
A total of 52049 elderly patients with early breast cancer (EBC) were selected. The following demographic and clinicopathological variables were sex (female), tumor grade, age at diagnosis, race, primary site, year of diagnosis, tumor size, type of surgery, radiotherapy status, chemotherapy status, ER status, HER2 status, PR status, survival months, vital status, causes of death, marital status, and Derived AJCC TNM Stage Group, 7th ed (2010-2015) and Derived SEER Cmb TNM Stg Grp (2016+).

End Points
Patients were followed up until November 2018, and the median follow-up was 34 months (range 1-83 months). Our primary observation endpoint was OS, de ned as the time from the date of diagnosis until death caused by any reasons. The secondary outcome measurements were BCSS and BCSD, de ned as the date of diagnosis until death caused by breast cancer and the interval from the date of diagnosis to the date of death respectively.

Statistics analysis
The baseline characteristics of patients were described using summary statistics. One-to-one (1:1) propensity score matching (PSM) was used to balance baseline characteristics and potential prognostic confounders between the groups 14 . X-tile program was used to determine the cutoff points of optimal tumor size through comparing the survival between two sides of each tumor size and product a minimum p-value 15 . Fine and Gray multivariable regression model was performed to identify factors associated with risk of death from all causes, which aimed to reduce bias caused by informative censoring.
Furthermore, a competing risk analysis model was built to evaluate the impact of RT on BCSD after excluding the impact of other causespeci c death (OCSD). SPSS version 23.0 (IBM Corporation, Armonk, NY, USA) and R software (version 3.6.2, R Foundation for Statistical Computing, Vienna, Austria. http://www. R-project. org/) were used for calculations. A two-sided p value < 0.05 considered statistically signi cant.

Baseline characteristics of patients
Among the 52049 patients from our study cohort, a total of 35026 (67.3%) patients received radiotherapy (RT group), while 17023 (32.71%) had no radiotherapy (non-RT group). Of these patients with FBC, 31120 (59.8%) of them aged between 65 years and 74 years, 26113 (50.2%) had tumors located in the left breast, and 26493 (50.9%) had tumors located in the upper quadrant. In total, 45455 (87.3%) patients were White, 25941 (49.8%) of them got married. The majority of FBC were PR positive (89.5%, 46570/52049), and tumor size ranging from 1 mm to 14 mm (73.8%, 38400/52049). In total, 22026 (42.3%) cases were moderately differentiated (grade I), and 2061 (4.0%) cases underwent chemotherapy. By comparing RT and non-RT groups, signi cant differences (P<0.05) were found in age at diagnosis, year of diagnosis, marital status, primary site, histology, PR status, grade level, tumor size and chemotherapy status. After PSM, there was no signi cant difference between the groups except for the age which were grouped. The detailed methodological characteristics were shown in Table 1. Table 1 The clinical and pathological characteristics of patients before and after PSM. Abbreviations: PSM, propensity score matching; RT, radiotherapy; SD, standard deviation; PR, progesterone receptor. Small tumors were associated with better OS (P < 0.001) and BCSS (P < 0.001) in comparison with the large tumors, as shown in Figs. 2a and 2b, and better BCSS after the application of PSM (Additional le 1: Figure S1a). In addition, compared with large tumors, small tumors were also associated with better OS (P < 0.001) as well as BCSS (P < 0.001) for patients with RT, and better BCSS for patients without RT (Additional le 1: Figures S1b, 1c and 1d).

Fine And Gray Multivariable Regression Model Analysis
A total of 3846 deaths were included in the unmatched cohort, of which 11.93% (459/3846) were BCSD, and 88.07% (3387/3846) were OCSD. To further explore the independent predictive consequences of BCSD, ne and gray multivariable regression model analysis was performed ( Table 2). Result showed that patients with large tumors had poorer BCSD (HR = 2.091, 95%CI: 1.730-2.527, P < 0.001) than those with small tumors. Patients in RT group (HR = 0.499, 95%CI: 0.407-0.613, P < 0.001) had lower BCSD than those in non-RT group. After PSM, patients with large tumors (HR = 2.024, 95%CI: 1.578-2.583, P < 0.001) still had worse BCSD than those with small tumors. In addition, after BCT, patients with highly differentiated grade I tumors, getting married, receiving RT or giving up chemotherapy tended to have signi cantly better BCSD than the corresponding subgroups (P<0.05), regardless of before or after PSM.

Discussion
At a median follow-up of 34 months, we demonstrate that tumor size of 14 mm was the optimal cutoff for predicting OS and BCSS for patients aged 65 years or older after BCT with clinical negative lymph nodes and ER-positive FBC. To our knowledge, this was the rst and largest population-based study to assess the impact of tumor size with a cutoff as 14 mm on OS and BCSS using propensity score matching analysis, X-tile analysis, survival variables, demographic and pathological factors.
OS, especially cancer speci c survival (CSS) is an objective, reliable, and bias-free measurements for patients with FBC in long-term results. In our study, based on analysis of a large cohort of 52049 patients in SEER database from 2010 to 2016 and an integrated range of factors in competing risk model, small tumors (≤ 14 mm) were associated with better OS and BCSS for patients with radiotherapy as well as better BCSS for patients with no radiotherapy in comparison with large tumors (≥ 15 mm) through X-title analysis. Such results indicated that whether with radiotherapy or not, large ER+, HER2-, and T1N0M0 tumors was associated with poorer OS and BCSS and should be attended to for patients aged ≥ 65 years after BCT, even if estimation and selection bias were possible.
To eliminate the estimation bias and further investigate the e cacy of tumor size on BCSS or other causes of death for FBC, Fine and Gray multivariable regression model analysis was performed. In our study, the patients with small tumors had better BCSD than those with large tumors. To minimize the selection bias resulting from baseline variables inherent in retrospective studies, PSM analysis was performed since PSM could eliminate a greater proportion of baseline differences between any two treatment groups than strati cation or covariates adjustment. After PSM analysis, the tumor size 14 mm was still shown be the optimal cutoff for predicting BCSS. Patients with small tumors still had better BCSD than those with large tumors. These results suggested that small ER+, HER2-, and T1N0M0 tumors should be an independent indicator for patients aged ≥ 65 years.
To further analyze and assess the e cacy of radiotherapy in patients with small or large tumors, a subgroup was performed. In our study, after BCT, patients in small tumors but no radiotherapy subgroup had similar BCSD to the radiotherapy subgroups with small tumors or large tumors, while patients in large tumors but no radiotherapy subgroup tended to have the highest BCSD, whether before or after the application of PSM. Such results showed, for patients aged ≥ 65 years with ER+, HER2-, and T1N0M0 tumors after BCT, small tumors were associated with favorable prognosis, which suggested that radiotherapy could be abandoned with caution for that omitting radiotherapy did not decrease the BCSS rate of patients with a tumor size ≤ 14 mm signi cantly.
In addition, married patients with highly-differentiation level, and PR positive tumors tended to have better prognostic indicators with BCSD than the corresponding subgroups. These results consistent with the previous reports indicated that clinicopathological features, such as tumor differentiation level, TNM stage, PR status, marital status and income level are objective and reliable prognostic indicators in patients with breast carcinoma [16][17][18] . Moreover, patients receiving chemotherapy had worse BCSD than the corresponding subgroup.
The underlying reason maybe that the local therapeutic effect of adjuvant chemotherapy on EBC patients cannot offset the systemic damage and long-term side effects 19 .
However,there are still some possible limitations in this study. Firstly, Selection bias may have occurred due to the nature of the retrospective analysis, for studies show that randomly assigned patients into different groups by treatment method are needed. Secondly, we were unable to avoid the possibility that the observed risk reductions might exclude the in uence of potential confounders, such as family history, insurance coverage, patient anxiety, detailed regimens of endocrine therapy, frailty or co-morbid conditions known to be related to receipt of speci c treatments, and so on. These data greatly impacted the clinical decisions and even breast cancer prognosis [20][21][22] .Thirdly, media follow-up in this study was merely 34 months. Longer follow-up times may be necessary for an accurate assessment of prognostic factors for patients with T1N0M0 FBC. Finally, and most importantly, there was a lack of information about local recurrence, adverse effects of normal tissue after radiotherapy. However, we believe that the ndings of this study, which cover about 28% of the U.S. population of patients with cancer, are generalizable and will contribute to improved survival in elder patients with EBC after BCT.

Conclusions
Our study demonstrated that small tumors (tumor ≤ 14 mm in diameter) were associated with better OS, BCSD and OCSD, and radiotherapy might be omitted for patients aged 65 or older with T1N0M0 stage, ER + and HER2-small breast tumors after BCT.
Randomly controlled clinical trials are needed to provide a high-level evidence.  Eligibility, inclusion, and exclusion criteria of study population. X-tile analysis of survival data from the SEER registry. The optimal cutoff points of tumor size were obtained based on overall survival (OS) (a) and breast cancer speci c survival (BCSS) (b) of the whole population, before the application of propensity score matching (PSM). Each graph contains the X-tile plot, a histogram, the K-M curve, and the data related to optimal cut-point.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.