Predicting the survival benefits of postoperative radiotherapy for breast cancer with lymph node micrometastasis: A Machine Learning model

Predicting the survival benefits of postoperative radiotherapy for breast cancer with lymph node micrometastasis: A Machine Learning model | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Predicting the survival benefits of postoperative radiotherapy for breast cancer with lymph node micrometastasis: A Machine Learning model Wanwan Wang, Binjie Chen, Shuhui Yang, Qixing Tan, Changyuan Wei This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9027856/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective: This study aims to assess the impact of post-mastectomy radiotherapy (PMRT) on the survival outcomes of patients with T1-T3N1miM0 breast cancer. Additionally, we seek to develop an interpretable machine learning model to predict the individualized 5-year survival benefits for patients undergoing postoperative radiotherapy in comparison to those who do not receive such treatment. Methods: Based on data from the SEER database spanning from 2010 to 2022, a cohort of 17,994 breast cancer patients diagnosed with T1-3N1miM0 was analyzed. Propensity score matching was employed to mitigate baseline discrepancies between the radiotherapy and non-radiotherapy cohorts. The overall survival and breast cancer-specific survival between the two groups were compared using the Kaplan-Meier method. A predictive model utilizing XGBoost was developed to estimate the 5-year survival advantages associated with postoperative radiotherapy, with the model's outcomes elucidated through SHAP analysis. Results: A significant difference in survival was observed between the postoperative radiotherapy and non-radiotherapy groups (HR: 0.85, 95% CI 0.75-0.96, p < 0.01). However, no significant differences in overall survival (OS) were found among the T1 and stage I radiotherapy and non-radiotherapy groups (HR: 0.87, 95% CI 0.75-1.08, p = 0.195; HR: 0.96, 95% CI 0.77-1.19, p = 0.704). Although breast cancer-specific survival (BCSS) was assessed, the difference was not significant (HR: 0.87, 95% CI 0.74-1.02, p = 0.84). The XGBoost model we developed exhibited exceptional predictive performance and was identified as the most effective model for predicting the 5-year survival outcomes of T1-T3pN1miM0 breast cancer patients (AUC = 0.770). Conclusion: Breast cancer patients diagnosed with T1N1miM0 or Stage I（pN1miM0） may potentially forgo PMRT. An XGBoost machine learning model was created to forecast the 5-year survival advantage of radiotherapy for postoperative individuals with T1-T3N1miM0 breast cancer. Biological sciences/Cancer Health sciences/Oncology breast cancer micrometastasis Machine Learning post-mastectomy radiotherapy Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. Introduction Breast cancer,a leading cause of death from malignant tumors in women, exhibits a rising incidence, posing a significant threat to women's health [1] .The status of axillary lymph nodes plays a crucial role in prognosis assessment. Specifically, lymph node micrometastasis (pN1mi, metastatic lesion diameter > 0.2mm and ≤ 2.0mm) represents an intermediate state between no metastasis and macrometastasis, characterized by lower tumor proliferation and invasiveness compared to macrometastasis [2–4] , Previous studies have shown that axillary lymph node dissection and postoperative radiotherapy do not enhance overall survival in T1-T2pN1mi patients [5, 6] . For patients with T1-T3N1miM0 breast cancer who have undergone surgery, the necessity of post-mastectomy radiotherapy(PMRT)remains a subject of debate in clinical practice. previous research indicate that PMRT may be more appropriate for individuals with a greater lymph node tumor burden or other biologically high-risk characteristics.The primary advantage of PMRT is the reduction in the rate of local recurrence, rather than an improvement in overall survival [7, 8] .The National Comprehensive Cancer Network (NCCN) guidelines in the United States advocate for the consideration of radiotherapy in this patient population [9] , In contrast, the European Society for Medical Oncology (ESMO) guidelines stress the importance of avoiding overtreatment [10] .Other studies also suggest that PMRT should be considered for breast cancer patients with 1 to 3 axillary lymph node metastases [7, 10, 11] .However, the criteria for administering radiotherapy to the subgroup of pN1mi have not been clearly established, and there is no definitive consensus among international guidelines [7, 10, 12] . These discrepancies highlight the need for more precise tools to identify individuals who would genuinely benefit from PMRT. Previous studies have used traditional models or nomograms to predict the survival benefit of radiotherapy for NmiM0 breast cancer patients [13–15] .However, the predictive efficacy, often characterized by an Area Under the Curve (AUC) of Receiver Operating Characteristic (ROC) less than 0.7,frequently proves inadequate for supporting individualized decision-making. Machine learning technology has gained widespread application in the diagnosis and treatment of breast cancer [1, 16, 17] ,an accurate postoperative radiotherapy risk prediction model could alleviate the treatment burden for certain T1-T3pNmiM0 breast cancer patients.In this study, we developed four machine learning models: logistic regression (LR), K-Nearest Neighbor (KNN), Random Forest (RF), and XGBoost, to predict the 5-year survival benefits of radiotherapy for T1-T3NmiM0 breast cancer patients. XGBoost, as a representative of ensemble learning models, demonstrates a robust capacity to capture complex and nonlinear feature relationships, as well as high-order interactions, and excels in the domain of disease prognosis prediction [18, 19] . This study aims to utilize the Surveillance, Epidemiology, and End Results(SEER )database to first examine the survival impact of radiotherapy on patients with T1-T3NmiM0 breast cancer. Additionally, it will analyze the influence of radiotherapy on survival outcomes across various T and stage subgroups. Furthermore, the research will construct and compare multiple machine learning models, ultimately establishing a high-precision, interpretable individualized prediction model to quantitatively assess the 5-year net survival benefit derived from PMRT. This investigation will provide scientific evidence to support the exemption of low-risk patients from radiotherapy and is dedicated to offering clinicians a practical computational tool that facilitates precise clinical decision. 2. Methods 2.1 Data source and study design We analyzed data from T1-T3NmiM0 breast cancer patients collected between January 2010 and December 2022 from the SEER database of the National Cancer Institute of the United States, utilizing SEERStat software (version 9.0.4.2). The SEER database serves as an open-access resource for epidemiological and survival analyses of cancer (for details, see the official website https://seer.cancer.gov/ ). The inclusion criteria were as follows: (1) the patient had invasive breast cancer as the only malignancy; (2) there was no distant metastasis (M0); (3) the T of the tumor was T1-T3and the lymph node stage was PN1mi. The exclusion criteria included: (1) patients who received preoperative or intraoperative radiotherapy or did not undergo surgical treatment; (2) patients with two or more primary malignant tumors; (3) patients with incomplete key pathological or staging information, such as ER, PR, HER2 status, grade, or AJCC stage. Figure 1 presents the workflow of our study design and its analyses. This retrospective study comprises four common machine learning algorithms implemented in Python 3.12. The primary objective of the models is to forecast whether patients will experience 5-year survival benefits following postoperative radiotherapy. In the machine learning model, a 5-year survival state is defined as survival if the survival time is ≥ 60 months, and as death if the survival time is < 60 months. The hyperparameters for all models are predetermined based on prior research and initial experimental findings. Data preprocessing involves encoding labels for categorical variables and standardizing continuous variables. The dataset is randomly split into training and test sets at a 7:3 ratio. A 50% discount hierarchical cross-validation method is employed to assess the model's robustness on the training set, with final performance evaluation conducted on an independent test set. Receiver Operating Characteristic (ROC), calibration curves, accuracy, precision, recall rate, specificity, and F1 scores were employed to compare the performance of Random Forest (RF), extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), and Logistic Regression (LR) models [20, 21] . The clinical net benefit of the prediction model was assessed using Decision Curve Analysis (DCA) [22] . To facilitate decision-makers in the correct application of the model, it is essential to comprehend the influence of each feature on the model's predictions. Consequently, Shapley Additive exPlanations (SHAP) were utilized to analyze the impact of each feature and to rank those that significantly affect the model [23] . 2.2 Statistical analysis The patients' basic characteristics were presented as frequency (n) and percentage, and group comparisons were made using the chi-square test. To mitigate bias from confounding factors, 1:1 Propensity Score Matching (PSM) was employed to assess the impact of radiotherapy on the prognosis of T1-T3NmiM0 breast cancer patients, resulting in the matching of 8,046 patients. Subsequently, Kaplan-Meier (K-M) survival analysis and log-rank tests were conducted, stratified by T and AJCC stages among PSM-matched patients. Statistical significance was defined as P < 0.05. Univariate and multivariate COX analyses were performed to investigate the relationship between characteristics and patient survival rates within the PSM dataset. Statistical analyses were carried out using R version 4.5.1 and Python 3.12. 3. Results 3.1. Clinical characteristics of patients with T1-T3NmiM0 breast cancer We ultimately identified 17,994 eligible patients with T1-T3NmiM0 breast cancer from the SEER database (2010–2022). The baseline clinicopathological characteristics of the entire cohort prior to propensity score matching are presented in Table 1 and summarized as follows. Among the radiotherapy group (n = 11,474), a substantial majority of patients (74.3%) underwent breast-conserving surgery, whereas only 21.2% of patients in the non-radiotherapy group (n = 6,520) received this procedure. Furthermore, 48.3% of patients in the radiotherapy group received chemotherapy, compared to 41.6% in the non-radiotherapy group. The proportion of patients receiving systemic treatment in the radiotherapy group (93.8%) was significantly higher than that in the non-radiotherapy group (76.9%). Regarding tumor characteristics, the proportions of AJCC stage I, T1, ER-positive, and HER2-negative patients in the radiotherapy group (66.1%, 54.6%, 90.7%, and 88.3%, respectively) were slightly greater than those in the non-radiotherapy group (63.1%, 52.7%, 88.8%, and 86.1%, respectively). In the radiotherapy group, 46.6% of patients had a time from diagnosis to treatment of ≤ 30 days, which was higher than the 37.2% observed in the non-radiotherapy group. All of these differences were statistically significant (P < 0.001). The basic characteristics of the propensity score matched patients are shown in Table 2 . Table 1 Basic Characteristics of breast cancer patients with T1-T3NmiM0. Variable Radiation No_Radiation P Value n = 11474 n = 6520 case(%) case( %) Age < 40 791 (6.9 ) 510 (7.8 ) < 0.001 40–49 2305 (20.1 ) 1426 (21.9 ) 50–59 3120 (27.2 ) 1647 (25.3 ) 60–69 3132 (27.3 ) 1463 (22.4 ) 70–79 1670 (14.6 ) 974 (14.9 ) 80+ 456 (4 ) 500 (7.7 ) Race White 8940 (77.9 ) 5018 (77 ) 0.319 Black 1151 (10 ) 674 (10.3 ) Other 1383 (12.1 ) 828 (12.7 ) Marital status Married 7015 (61.1 ) 3735 (57.3 ) < 0.001 Single 1702 (14.8 ) 998 (15.3 ) Widow/divorced/other 2757 (24 ) 1787 (27.4 ) Primary Site Central portion of breast 620 (5.4 ) 416 (6.4 ) < 0.001 Upper-inner quadrant 1221 (10.6 ) 675 (10.4 ) Lower-inner quadrant 587 (5.1 ) 342 (5.2 ) Upper-outer quadrant 4497 (39.2 ) 2099 (32.2 ) Lower-outer quadrant 973 (8.5 ) 544 (8.3 ) Overlapping lesion 2676 (23.3 ) 1449 (22.2 ) Other 900 (7.8 ) 995 (15.3 ) Grade Well differentiated 2720 (23.7 ) 1395 (21.4 ) < 0.001 Moderate differentiated 5856 (51 ) 3359 (51.5 ) Poorly differentiated 2898 (25.3 ) 1766 (27.1 ) Laterality Left 5678 (49.5 ) 3271 (50.2 ) 0.387 Right 5796 (50.5 ) 3249 (49.8 ) Time from diagnosis to treatment < 30 days 5342 (46.6 ) 2427 (37.2 ) < 0.001 ≥ 30 days 6132 (53.4 ) 4093 (62.8 ) Histology IDC 9160 (79.8 ) 5101 (78.2 ) 0.028 ILC 1338 (11.7 ) 783 (12 ) Mixed 658 (5.7 ) 425 (6.5 ) Other 318 (2.8 ) 211 (3.2 ) AJCC Stage I 7579 (66.1 ) 4112 (63.1 ) < 0.001 II 3271 (28.5 ) 2149 (33 ) III 624 (5.4 ) 259 (4 ) T T1 6263 (54.6 ) 3437 (52.7 ) 1 1649 (14.4 ) 918 (14.1 ) ER Negative 1065 (9.3 ) 732 (11.2 ) < 0.001 Positive 10409 (90.7 ) 5788 (88.8 ) PR Negative 2024 (17.6 ) 1311 (20.1 ) < 0.001 Positive 9450 (82.4 ) 5209 (79.9 ) HER-2 Negative 10128 (88.3 ) 5615 (86.1 ) < 0.001 Positive 1346 (11.7 ) 905 (13.9 ) Breast Subtype HR+/HER2+ 1106 (9.6 ) 659 (10.1 ) < 0.001 HR+/HER2- 9377 (81.7 ) 5169 (79.3 ) HR-/HER2+ 240 (2.1 ) 246 (3.8 ) HR-/HER2- 751 (6.5 ) 446 (6.8 ) Surg Prim Site Breast-conserving surgery 8524 (74.3 ) 1380 (21.2 ) < 0.001 Simple mastectomy 2275 (19.8 ) 3927 (60.2 ) Radical mastectomy 675 (5.9 ) 1213 (18.6 ) Chemotherapy No/unknown 5928 (51.7 ) 3806 (58.4 ) < 0.001 Yes 5546 (48.3 ) 2714 (41.6 ) Systemic therapy NO 711 (6.2 ) 1504 (23.1 ) < 0.001 YES 10763 (93.8 ) 5016 (76.9 ) Table 2 Basic Characteristics of PSM-T1-T3NmiM0 Breast Cancer patients. Variable Radiation NO Radiation P value n = 4023 n = 4023 case(% ) case(% ) Age < 40 587 (14.6 ) 453 (11.3 ) < 0.001 40–49 1019 (25.3 ) 985 (24.5 ) 50–59 980 (24.4 ) 1009 (25.1 ) 60–69 790 (19.6 ) 881 (21.9 ) 70–79 439 (10.9 ) 489 (12.2 ) 80+ 208 (5.2 ) 206 (5.1 ) Race Black 437 (10.9 ) 408 (10.1 ) 0.424 Other 533 (13.2 ) 514 (12.8 ) White 3053 (75.9 ) 3101 (77.1 ) Marital status Married 2445 (60.8 ) 2449 (60.9 ) 0.214 Single 666 (16.6 ) 616 (15.3 ) Widow/divorced/other 912 (22.7 ) 958 (23.8 ) Primary Site Central portion 224 (5.6 ) 215 (5.3 ) 0.704 Lower-inner quadrant 173 (4.3 ) 200 (5.0 ) Lower-outer quadrant 304 (7.6 ) 311 (7.7 ) Other 533 (13.2 ) 522 (13.0 ) Overlapping lesion 891 (22.1 ) 926 (23.0 ) Upper-inner quadrant 419 (10.4 ) 418 (10.4 ) Upper-outer quadrant 1479 (36.8 ) 1431 (35.6 ) Grade Moderate differentiated 1994 (49.6 ) 2007 (49.9 ) 0.014 Poorly differentiated 1287 (32.0 ) 1189 (29.6 ) Well differentiated 742 (18.4 ) 827 (20.6 ) Laterality left 2009 (49.9 ) 2023 (50.3 ) 0.772 right 2014 (50.1 ) 2000 (49.7 ) Histology IDC 3024 (75.2 ) 3101 (77.1 ) 0.206 ILC 629 (15.6 ) 568 (14.1 ) Mixed 245 (6.1 ) 239 (5.9 ) Other 125 (3.1 ) 115 (2.9 ) AJCC Stage I 2113 (52.5 ) 2357 (58.6 ) < 0.001 II 1560 (38.8 ) 1453 (36.1 ) III 350 (8.7 ) 213 (5.3 ) Surg Prim Site Breast-conserving surgery 1450 (36.0 ) 1357 (33.7 ) 0.094 Radical mastectomy 585 (14.5 ) 604 (15.0 ) Simple mastectomy 1988 (49.4 ) 2062 (51.3 ) Chemotherapy No/unknown 1638 (40.7 ) 1918 (47.7 ) < 0.001 Yes 2385 (59.3 ) 2105 (52.3 ) Systemic therapy NO 706 (17.5 ) 689 (17.1 ) 0.638 YES 3317 (82.5 ) 3334 (82.9 ) Time from diagnosis to treatment < 30days 1818 (45.2 ) 1681 (41.8 ) 0.002 ≥ 30days 2205 (54.8 ) 2342 (58.2 ) T T1 1503 (37.4 ) 1853 (46.1 ) 1 931 (23.1 ) 711 (17.7 ) < 0.001 ≤ 1 3092 (76.9 ) 3312 (82.3 ) Breast Subtype HR-/HER2- 374 (9.3 ) 334 (8.3 ) 0.033 HR-/HER2+ 112 (2.8 ) 117 (2.9 ) HR+/HER2- 3044 (75.7 ) 3145 (78.2 ) HR+/HER2+ 493 (12.3 ) 427 (10.6 ) ER Negative 517 (12.9 ) 478 (11.9 ) 0.198 Positive 3506 (87.1 ) 3545 (88.1 ) PR Negative 879 (21.8 ) 831 (20.7 ) 0.2 Positive 3144 (78.2 ) 3192 (79.3 ) HER2 Negative 3418 (85.0 ) 3479 (86.5 ) 0.056 Positive 605 (15.0 ) 544 (13.5 ) 3.2. Univariate and multivariate cox regression results In PSM data, a significant difference in overall survival (OS) was observed between the radiation and no-radiation groups (HR: 0.85, 95% CI 0.75–0.96, p < 0.01). However, no significant difference was found in breast cancer-specific survival (BCSS) (HR: 0.87, 95% CI 0.74–1.02, p = 0.84). Subgroup survival analyses were conducted based on AJCC stage and T subgroups. Among AJCC stage I and T1 patients, both OS (HR: 0.87, 95% CI 0.75–1.08, p = 0.195; HR: 0.96, 95% CI 0.77–1.19, p = 0.704) and BCSS (HR: 0.96, 95% CI 0.77–1.91, p = 0.704; HR: 0.95, 95% CI 0.69–1.31, p = 0.757) did not show benefits from radiotherapy. Conversely, patients in stages II and III showed improved OS (HR: 0.83, 95% CI 0.7–0.99, p = 0.035; HR: 0.52, 95% CI 0.37–0.74, p < 0.01) with postoperative radiotherapy. Similarly, patients in T2 and T3 groups exhibited enhanced OS (HR: 0.79, 95% CI 0.67–0.93, p < 0.01; HR: 0.52, 95% CI 0.36–0.74, p < 0.01) with postoperative radiotherapy. Notably, stage II BCSS (HR: 0.88, 95% CI 0.71–1.08, p = 0.212) did not benefit from radiotherapy, and BCSS in T2 group (HR: 0.82, 95% CI 0.67-1.00, p = 0.053) did not exhibit significant survival advantages. Conversely, in AJCC stage III and T3 groups, BCSS (HR: 0.51, 95% CI 0.34–0.78, p < 0.01; HR: 0.52, 95% CI 0.34–0.79, p < 0.01) showed benefits from radiotherapy. We performed univariate and multivariate Cox analyses on the data adjusted PSM. The results of the multivariate Cox regression analysis indicated that age over 70 years, the presence of more than one positive lymph node,histological grade, radiotherapy, estrogen receptor (ER) status, progesterone receptor (PR) status, and AJCC stages II and III significantly influenced OS (Fig. 5 ) and BCSS (Fig. 6 ) in patients (p < 0.05). In contrast, age between 60 and 69 years affected OS (p < 0.05) but did not significantly impact BCSS (p = 0.995). 3.3 Establishment and evaluation of predictive models After a thorough evaluation of four machine learning models, we identified XGBoost as the optimal predictive model due to its superior predictive performance (AUC value = 0.770) (Fig. 7a). However, since the dataset exhibits an imbalanced distribution of positive and negative events, relying solely on AUC may not suffice to assess the model's effectiveness. To address this limitation, we utilized the PR (precision-recall) curve to provide a more comprehensive evaluation of the model's strengths and weaknesses. The PR curve analysis revealed that the XGBoost model outperformed other models, as indicated by its higher Average Precision (AP = 0.769) (Fig. 7b). Furthermore, the confusion matrix of the XGBoost model is presented in Fig. 7c. Overall, as illustrated in Fig. 7d, XGBoost demonstrated superior overall performance, achieving the highest Accuracy (0.691), F1 score (0.717), Specificity (0.634), and Sensitivity (0.741). The calibration curves assessed the alignment between the predicted 5-year survival benefit of radiotherapy by four models and the observed outcomes. The XGBoost model exhibited superior calibration curve performance on the test set compared to other models (Fig. 8 a). Consequently, the XGBoost model was identified as optimal. Evaluation of the clinical net benefit of the predictive model was conducted using the Decision Curve Analysis (DCA), revealing that the XGBoost model provided greater net benefit across most threshold probabilities (Fig. 8 b). The ranking of feature importance based on SHAP values indicated that T, stage, age, primary site, and grade were significant factors, with T being the most influential, followed by AJCC stage and time from diagnosis to treatment. Subsequently, the impact of variables on the model was assessed using SHAP values. Variables were ranked by mean SHAP value, highlighting tumor size and AJCC stage as the most critical factors influencing the XGBoost model (Fig. 9 ). Discussion This study utilizes extensive SEER data to elucidate the impact of PMRT on the survival outcomes of patients with T1-T3N1miM0 breast cancer. Additionally, it aims to develop an individualized prediction model to identify subgroups that may not derive significant benefit from radiotherapy. Our findings indicate that, following propensity score matching to balance confounding factors, PMRT markedly improved the OS of the cohort. However, no significant improvements in OS or BCSS were observed in patients with a mild tumor burden at T1 stage and AJCC stage I, corroborating previous studies [15, 24–26] . The EUSOMA guidelines suggest that certain low-risk breast cancer patients who have already undergone endocrine therapy may not require radiotherapy [27] . Furthermore, previous studies has demonstrated that postoperative radiotherapy does not significantly affect the OS of some elderly patients with early-stage breast cancer [28] . Our research indicates that radiotherapy has minimal impact on their overall survival for patients with a mild tumor burden (T1pmiMO and AJCC stage I with pmiMO). Given the long-term risks associated with radiotherapy, including cardiopulmonary toxicity, secondary malignancies, and lymphedema [29, 30] , the results of this study provide compelling retrospective evidence to support the clinical decision to exempt certain low-risk patients from PMRT. The subgroup analysis based on traditional clinical staging (AJCC stage) aids in identifying the general population that may be exempt from radiotherapy. Precision medicine necessitates the integration of multi-dimensional information to quantitatively predict the benefits of radiotherapy for patients. To address this limitation, the present study developed and validated a machine learning model utilizing the XGBoost model. This model exhibited excellent discriminative ability in the test set (AUC = 0.770) and outperformed LR, RF, and KNN, significantly surpassing traditional nomograms [13–15] , which often reported AUC values below 0.7 in prior studies. This investigation validated the robust predictive capacity of the XGBoost model in the specific cohort of breast cancer T1-T3N1miM0, highlighting its potential to furnish clinicians with a personalized 5-year survival benefit-risk assessment for radiotherapy. This approach facilitates tailored and precise radiotherapy decisions [18] . This study also constructed a DCA to illustrate the predictive power advantage of the XGBoost model. To further elucidate the XGBoost model, we examined the influence of each patient feature on the model using SHAP values, resulting in a clear ranking of features. Our analysis revealed that the clinical characteristics of T and AJCC stage are the most critical factors influencing the model's predictions. This finding enhances the biological and clinical rationale of the model and supports informed clinical decision-making. This study revealed that varying quantities of pN1mi lymph nodes significantly influence prognosis. Previous research examined and categorized different quantities of pN1mi lymph nodes, indicating that an escalation in micrometastatic lymph node numbers correlates with a poorer prognosis among N1mi breast cancer patients [31] . Nonetheless, this study possesses additional constraints. Initially, the absence of data such as the HER2 status in the SEER database during earlier periods could introduce data selection bias. Secondly, the SEER database lacks comprehensive information on radiotherapy dosages, chemotherapy protocols, targeted therapy, and endocrine therapy, thereby hindering the evaluation of these treatments' impact on prognosis [32] . Thirdly, Ki67 functions as a prognostic biomarker for breast cancer, while PDL-1 represents a significant therapeutic target. [33, 34] , However, the SEER database does not provide information on Ki67, PDL-1, and related factors, which constrains comprehensive research on these elements. Furthermore, the generalizability of the XGBoost model requires external validation and calibration in a multi-center cohort [18] 。Future research aims to improve the model's predictive accuracy and biological interpretability by incorporating multi-omics data, including radiomics and genomics. Our objective is to develop this model into an online clinical decision support tool, thereby alleviating unnecessary treatment burdens for low-risk patients and advancing precision medicine in breast cancer. Conclusion In conclusion, PMRT can enhance the overall survival of patients with T1-T3N1miM0 breast cancer. However, for patients classified as pN1miM0 at T1 Stage or AJCC Stage I, the survival benefits are limited, suggesting that PMRT may be considered for exemption. In light of this, we have developed and validated a high-precision, interpretable XGBoost machine learning model capable of quantitatively assessing the 5-year survival benefit of radiotherapy for individual patients. This study not only provides evidence supporting the exemption of PMRT for specific low-risk patients but also presents a practical predictive tool for making individualized adjuvant treatment decisions for these patients. Declarations Ethics approval and consent to participate Ethical review and approval were waived for this study due to the fact that the data are fully de-identified and no intervention on patients was performed. Consent for publication Not applicable Availability of data and material All data here are publicly available in the SEER database [https:// seer. cancer.gov/ (accessed on September 22, 2025)]. Acknowledgements We thank all staff of the SEER database for their contribution in data collection,maintenance, distribution and so on. Also we would like to thank all the developers of the R programming package and python for selflessly sharing their code. Competing interests The authors declare no competing interests. Funding This work was supported by the National Natural Science Foundation of China (Grant No.82160481, 82460522), Project funded by China Postdoctoral Science Foundation (Grant No. 2023MD744194), the Natural Science Foundation of Guangxi (Grant No. 2021GXNSFBA196015), Joint Project on Regional High-Incidence Diseases Research of Guangxi Natural Science Foundation (Grant No. 2024GXNSFAA010058), and First-class Discipline Innovation-driven Talent Program of Guangxi Medical University.Availability of data and materials Author Contributions WWW, BJC, and SHY contributed equally as co-first authors. WWW: Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft. BJC: Formal analysis, Investigation, Methodology, Supervision, Validation, Writing – original draft. SHY: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft. QXT: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing. CYW: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing. All authors read and approved the final manuscript. References XIONG X, ZHENG L W, DING Y, et al. Breast cancer: pathogenesis and treatments [J]. Signal Transduct Target Ther, 2025, 10(1): 49. MAMTANI A, PATIL S, STEMPEL M, et al. Axillary Micrometastases and Isolated Tumor Cells Are Not an Indication for Post-mastectomy Radiotherapy in Stage 1 and 2 Breast Cancer [J]. Ann Surg Oncol, 2017, 24(8): 2182-8. ZHAO D, XU M, YANG S, et al. Specific diagnosis of lymph node micrometastasis in breast cancer by targeting activatable near-infrared fluorescence imaging [J]. Biomaterials, 2022, 282: 121388. KLAUBER-DEMORE N, VAN ZEE K J, LINKOV I, et al. Biological behavior of human breast cancer micrometastases [J]. Clin Cancer Res, 2001, 7(8): 2434-9. GALIMBERTI V, COLE B F, VIALE G, et al. Axillary dissection versus no axillary dissection in patients with breast cancer and sentinel-node micrometastases (IBCSG 23-01): 10-year follow-up of a randomised, controlled phase 3 trial [J]. Lancet Oncol, 2018, 19(10): 1385-93. LIM S Z, KUSUMAWIDJAJA G, MOHD ISHAK H M, et al. Outcomes of Stage I and II Breast Cancer with Nodal Micrometastases Treated with Mastectomy without Axillary Therapy [J]. Breast Cancer Res Treat, 2021, 189(3): 837-43. WU J, FAN D, SHAO Z, et al. CACA Guidelines for Holistic Integrative Management of Breast Cancer [J]. Holist Integr Oncol, 2022, 1(1): 7. WAKS A G, WINER E P. Breast Cancer Treatment: A Review [J]. JAMA, 2019, 321(3): 288-300. GRADISHAR W J, MORAN M S, ABRAHAM J, et al. Breast Cancer, Version 3.2024, NCCN Clinical Practice Guidelines in Oncology [J]. Journal of the National Comprehensive Cancer Network, 2024, 22(5): 331-57. LOIBL S, ANDRé F, BACHELOT T, et al. Early breast cancer: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up [J]. Ann Oncol, 2024, 35(2): 159-82. RECHT A, COMEN E A, FINE R E, et al. Postmastectomy Radiotherapy: An American Society of Clinical Oncology, American Society for Radiation Oncology, and Society of Surgical Oncology Focused Guideline Update [J]. J Clin Oncol, 2016, 34(36): 4431-42. PARK K U, SOMERFIELD M R, ANNE N, et al. Sentinel Lymph Node Biopsy in Early-Stage Breast Cancer: ASCO Guideline Update [J]. J Clin Oncol, 2025, 43(14): 1720-41. WU S P, TAM M, SHAIKH F, et al. Post-mastectomy Radiation Therapy in Breast Cancer Patients with Nodal Micrometastases [J]. Ann Surg Oncol, 2018, 25(9): 2620-31. HUO D, HOU N, JASKOWIAK N, et al. Use of Postmastectomy Radiotherapy and Survival Rates for Breast Cancer Patients with T1-T2 and One to Three Positive Lymph Nodes [J]. Ann Surg Oncol, 2015, 22(13): 4295-304. TANG Y, ZHANG Y J, ZHANG N, et al. Nomogram predicting survival as a selection criterion for postmastectomy radiotherapy in patients with T1 to T2 breast cancer with 1 to 3 positive lymph nodes [J]. Cancer, 2020, 126 Suppl 16: 3857-66. QI Y J, SU G H, YOU C, et al. Radiomics in breast cancer: Current advances and future directions [J]. Cell Rep Med, 2024, 5(9): 101719. QIAN X, PEI J, HAN C, et al. A multimodal machine learning model for the stratification of breast cancer risk [J]. Nat Biomed Eng, 2025, 9(3): 356-70. CLIFT A K, DODWELL D, LORD S, et al. Development and internal-external validation of statistical and machine learning models for breast cancer prognostication: cohort study [J]. Bmj, 2023, 381: e073800. ZHU H, ZHOU Y, SHEN D, et al. An interpretable machine learning model for predicting early liver metastasis after pancreatic cancer surgery [J]. BMC Cancer, 2025, 25(1): 1117. ZHANG Y, AN W, WANG C, et al. Novel models based on machine learning to predict the prognosis of metaplastic breast cancer [J]. Breast, 2025, 79: 103858. REN J, LI Y, ZHOU J, et al. Developing machine learning models for personalized treatment strategies in early breast cancer patients undergoing neoadjuvant systemic therapy based on SEER database [J]. Sci Rep, 2024, 14(1): 22055. YU Y, HE Z, OUYANG J, et al. Magnetic resonance imaging radiomics predicts preoperative axillary lymph node metastasis to support surgical decisions and is associated with tumor microenvironment in invasive breast cancer: A machine learning, multicenter study [J]. EBioMedicine, 2021, 69: 103460. ZHONG X, LIN Y, ZHANG W, et al. Predicting diagnosis and survival of bone metastasis in breast cancer using machine learning [J]. Sci Rep, 2023, 13(1): 18301. HUGHES K S, SCHNAPER L A, BELLON J R, et al. Lumpectomy plus tamoxifen with or without irradiation in women age 70 years or older with early breast cancer: long-term follow-up of CALGB 9343 [J]. J Clin Oncol, 2013, 31(19): 2382-7. LU Z, GUO L, ZHOU J, et al. A nomogram to predict the benefit of postmastectomy radiotherapy in breast cancer with nodal micrometastases [J]. Breast cancer (Tokyo, Japan), 2025, 32(5): 935-46. LIU K, LI G Q, LI S Q, et al. Clinical treatment score Post-5 Years (CTS5) predicts the benefit of postmastectomy radiotherapy in patients with T1-2N1 luminal breast cancer [J]. Breast, 2025, 79: 103873. DARBY S, MCGALE P, CORREA C, et al. Effect of radiotherapy after breast-conserving surgery on 10-year recurrence and 15-year breast cancer death: meta-analysis of individual patient data for 10,801 women in 17 randomised trials [J]. Lancet, 2011, 378(9804): 1707-16. KUNKLER I H, WILLIAMS L J, JACK W J L, et al. Breast-Conserving Surgery with or without Irradiation in Early Breast Cancer [J]. N Engl J Med, 2023, 388(7): 585-94. LóPEZ-FERNáNDEZ T, MARCO I, AZNAR M C, et al. Breast cancer and cardiovascular health [J]. Eur Heart J, 2024, 45(41): 4366-82. JARM T, BESIC N, ARNEZ R C, et al. Breast cancer related lymphedema and shoulder mobility following radiotherapy [J]. Strahlenther Onkol, 2025. LUO S, FU W, LIN J, et al. Prognosis and local treatment strategies of breast cancer patients with different numbers of micrometastatic lymph nodes [J]. World J Surg Oncol, 2023, 21(1): 202. YANG J, LI Y, WANG S, et al. Distinct prognostic patterns of single hormone receptor-positive subtypes in HER2-negative breast cancer: a SEER-based retrospective cohort study [J]. Int J Surg, 2025. YERUSHALMI R, WOODS R, RAVDIN P M, et al. Ki67 in breast cancer: prognostic and predictive potential [J]. Lancet Oncol, 2010, 11(2): 174-83. DVIR K, GIORDANO S, LEONE J P. Immunotherapy in Breast Cancer [J]. Int J Mol Sci, 2024, 25(14). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9027856","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":605511260,"identity":"b842d0af-5573-41f5-b6bc-870b687dbc1a","order_by":0,"name":"Wanwan Wang","email":"","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wanwan","middleName":"","lastName":"Wang","suffix":""},{"id":605511261,"identity":"4b868d65-b74e-4753-ad8b-f0635f9ec192","order_by":1,"name":"Binjie Chen","email":"","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Binjie","middleName":"","lastName":"Chen","suffix":""},{"id":605511262,"identity":"c432754a-568b-4c2c-ad18-72de242e863f","order_by":2,"name":"Shuhui Yang","email":"","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shuhui","middleName":"","lastName":"Yang","suffix":""},{"id":605511263,"identity":"27e8a393-c671-4fcf-8aa6-719a75ceef0a","order_by":3,"name":"Qixing Tan","email":"","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Qixing","middleName":"","lastName":"Tan","suffix":""},{"id":605511264,"identity":"3fe9b4a6-6fc7-49b1-91ee-ece4d91548d4","order_by":4,"name":"Changyuan Wei","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/klEQVRIiWNgGAWjYDACCST2gQ9YBPFrOTiDZC3MPMRo4Z/d/Ozh17bD8uYSyQcP2/w6HG1wgPngbR4Guzycltw5Zm4s23bYcOeMtITDuX1puRsOsCVb8zAkF+PSYiCRYCYt2XaYccPtHIPDuT02QC08ZtI8DAcSG3BqSf8G0mK/4Xb+h8OWPRJALfzfCGjJMZP82HY4EWgLw2GGH2Bb2PBqkbiRUybNcC49ecP9ZwYHexvScmceZjO2nGOQjFML/4z0bZI/yqxtN5w5/PjDjz/AEDje/PDGmwo7nFpAgJmXDcpibANxwQ7Gox6k8McfGPMPPnWjYBSMglEwUgEAtslf1Y9y1RoAAAAASUVORK5CYII=","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":true,"prefix":"","firstName":"Changyuan","middleName":"","lastName":"Wei","suffix":""}],"badges":[],"createdAt":"2026-03-04 08:38:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9027856/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9027856/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104702402,"identity":"cf77c3b8-fb12-459f-b140-d79888e1d1fe","added_by":"auto","created_at":"2026-03-16 08:43:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":255344,"visible":true,"origin":"","legend":"\u003cp\u003eThe workflow described the process of conducting the study and statistical analysis. SEER the surveillance, epidemiology, and end results database; PSM propensity score matching,，COX concordance index;K–M Kaplan–Meier;XGBoost extreme Gradient Boosting; RF Random Forest; LR Logistic Regression; KNN K-Nearest Neighbor.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/fdc230d17b2fa00bb0154c9c.png"},{"id":104702425,"identity":"51d82109-3b8d-4a8d-880f-02ae1ffeef18","added_by":"auto","created_at":"2026-03-16 08:43:14","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":73248,"visible":true,"origin":"","legend":"\u003cp\u003ePSM adjusted OS and BCSS of patients with radiotherapy K–M survival analysis：\u003cstrong\u003ea\u003c/strong\u003e. OS of patients with PSM;\u003cstrong\u003eb\u003c/strong\u003e. BCSS of patients with PSM.PSM propensity score matching,OS overall survival,BCSS breast cancer specific survival,HR hazard ratio, CI confidence interval\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/ebc8d4a8fc39cb120b0d9e40.png"},{"id":104702431,"identity":"12c31ddd-0d72-405a-ad3c-da1787acc748","added_by":"auto","created_at":"2026-03-16 08:43:16","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":208525,"visible":true,"origin":"","legend":"\u003cp\u003ePSM adjusted OS and BCSS of patients with radiotherapy（Stratified by AJCC stage）K–M survival analysis: \u003cstrong\u003ea\u003c/strong\u003e. OS of patients with AJCC stage I; \u003cstrong\u003eb\u003c/strong\u003e. OS of patients with AJCC stage II ;\u003cstrong\u003e c\u003c/strong\u003e. OS of patients with AJCC stage III; \u003cstrong\u003ed\u003c/strong\u003e.\u003cstrong\u003e \u003c/strong\u003eBCSS of patients with AJCC stage III; \u003cstrong\u003ee\u003c/strong\u003e. BCSS of patients with AJCC stage II; \u003cstrong\u003ef\u003c/strong\u003e. BCSS of patients with AJCC stage III\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/9debd84b9ed76d9c5063fc3e.png"},{"id":104702504,"identity":"2d04b421-38a7-485e-9e0e-c9e1d17eabd7","added_by":"auto","created_at":"2026-03-16 08:43:35","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":203187,"visible":true,"origin":"","legend":"\u003cp\u003ePSM adjusted OS and BCSS of patients with radiotherapy（Stratified by T）\u003cstrong\u003ea\u003c/strong\u003e. OS of patients with T1; \u003cstrong\u003eb\u003c/strong\u003e. OS of patients with T2; \u003cstrong\u003ec\u003c/strong\u003e. OS of patients with T3; \u003cstrong\u003ed\u003c/strong\u003e.BCSS of patients with T1; \u003cstrong\u003ee\u003c/strong\u003e. BCSS of patients with T2; \u003cstrong\u003ef\u003c/strong\u003e. BCSS of patients with T3\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/a78e731a6fcf90f342c8cdfa.png"},{"id":104702325,"identity":"a9cbf071-7627-4ea9-947d-520e2b3d4d90","added_by":"auto","created_at":"2026-03-16 08:42:48","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":326729,"visible":true,"origin":"","legend":"\u003cp\u003ePSM-OS multivariate COX regression forest plot\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/21e1b478007b2a9f39cca708.png"},{"id":104702545,"identity":"7539c26c-379a-4422-b82f-09b30738a666","added_by":"auto","created_at":"2026-03-16 08:43:45","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":323037,"visible":true,"origin":"","legend":"\u003cp\u003ePSM-BCSS multivariate COX regression forest plot\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/e842dec73d3e6245e76556d1.png"},{"id":104702481,"identity":"fabd544f-adaa-41cf-8ff9-607654c19cb8","added_by":"auto","created_at":"2026-03-16 08:43:31","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":186125,"visible":true,"origin":"","legend":"\u003cp\u003eL. Model performance. \u003cstrong\u003ea\u003c/strong\u003e. ROC curves and AUC values for KNN, LR, RF, and XGBoost derived from test data. \u003cstrong\u003eb\u003c/strong\u003e. Precision-recall curves generated from the test data of KNN, LR, RF, and XGBoost. \u003cstrong\u003ec\u003c/strong\u003e. Confusion matrix illustrating the prediction results of the XGBoost model on test data. \u003cstrong\u003ed\u003c/strong\u003e. Comparison of test data metrics for KNN, LR, RF, and XGBoost. TN True Negative; FP False Positive; FN False Negative; TP True Positive; AP Average Precision\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/a7e2b0e47965db0f20de2773.png"},{"id":104702560,"identity":"bf2af568-6553-4e6a-8f54-9811159df854","added_by":"auto","created_at":"2026-03-16 08:43:52","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":116031,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e.The calibration curves of all models; \u003cstrong\u003eb\u003c/strong\u003e. The DCA of XGBoost model\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/30e47b70d2e6caf514a8a8df.png"},{"id":104702544,"identity":"a05d6cea-da89-4b7f-8dcf-1d69f67def52","added_by":"auto","created_at":"2026-03-16 08:43:45","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":98132,"visible":true,"origin":"","legend":"\u003cp\u003eThe ranking of the mean SHAP values of features in the XGBoost model\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/7597631b89ccd309a17208f7.png"},{"id":107402052,"identity":"b2270b1b-23f0-47e5-8e0b-b520c66bedaa","added_by":"auto","created_at":"2026-04-21 07:43:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2249593,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9027856/v1/1c51b0d0-df7b-4609-a9b9-40e0b8aca366.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Predicting the survival benefits of postoperative radiotherapy for breast cancer with lymph node micrometastasis: A Machine Learning model","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eBreast cancer,a leading cause of death from malignant tumors in women, exhibits a rising incidence, posing a significant threat to women's health\u003csup\u003e[1]\u003c/sup\u003e.The status of axillary lymph nodes plays a crucial role in prognosis assessment. Specifically, lymph node micrometastasis (pN1mi, metastatic lesion diameter\u0026thinsp;\u0026gt;\u0026thinsp;0.2mm and \u0026le;\u0026thinsp;2.0mm) represents an intermediate state between no metastasis and macrometastasis, characterized by lower tumor proliferation and invasiveness compared to macrometastasis \u003csup\u003e[2\u0026ndash;4]\u003c/sup\u003e, Previous studies have shown that axillary lymph node dissection and postoperative radiotherapy do not enhance overall survival in T1-T2pN1mi patients\u003csup\u003e[5, 6]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFor patients with T1-T3N1miM0 breast cancer who have undergone surgery, the necessity of post-mastectomy radiotherapy(PMRT)remains a subject of debate in clinical practice. previous research indicate that PMRT may be more appropriate for individuals with a greater lymph node tumor burden or other biologically high-risk characteristics.The primary advantage of PMRT is the reduction in the rate of local recurrence, rather than an improvement in overall survival\u003csup\u003e[7, 8]\u003c/sup\u003e.The National Comprehensive Cancer Network (NCCN) guidelines in the United States advocate for the consideration of radiotherapy in this patient population\u003csup\u003e[9]\u003c/sup\u003e, In contrast, the European Society for Medical Oncology (ESMO) guidelines stress the importance of avoiding overtreatment \u003csup\u003e[10]\u003c/sup\u003e.Other studies also suggest that PMRT should be considered for breast cancer patients with 1 to 3 axillary lymph node metastases \u003csup\u003e[7, 10, 11]\u003c/sup\u003e.However, the criteria for administering radiotherapy to the subgroup of pN1mi have not been clearly established, and there is no definitive consensus among international guidelines\u003csup\u003e[7, 10, 12]\u003c/sup\u003e. These discrepancies highlight the need for more precise tools to identify individuals who would genuinely benefit from PMRT.\u003c/p\u003e \u003cp\u003ePrevious studies have used traditional models or nomograms to predict the survival benefit of radiotherapy for NmiM0 breast cancer patients \u003csup\u003e[13\u0026ndash;15]\u003c/sup\u003e.However, the predictive efficacy, often characterized by an Area Under the Curve (AUC) of Receiver Operating Characteristic (ROC) less than 0.7,frequently proves inadequate for supporting individualized decision-making. Machine learning technology has gained widespread application in the diagnosis and treatment of breast cancer \u003csup\u003e[1, 16, 17]\u003c/sup\u003e,an accurate postoperative radiotherapy risk prediction model could alleviate the treatment burden for certain T1-T3pNmiM0 breast cancer patients.In this study, we developed four machine learning models: logistic regression (LR), K-Nearest Neighbor (KNN), Random Forest (RF), and XGBoost, to predict the 5-year survival benefits of radiotherapy for T1-T3NmiM0 breast cancer patients. XGBoost, as a representative of ensemble learning models, demonstrates a robust capacity to capture complex and nonlinear feature relationships, as well as high-order interactions, and excels in the domain of disease prognosis prediction\u003csup\u003e[18, 19]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThis study aims to utilize the Surveillance, Epidemiology, and End Results(SEER )database to first examine the survival impact of radiotherapy on patients with T1-T3NmiM0 breast cancer. Additionally, it will analyze the influence of radiotherapy on survival outcomes across various T and stage subgroups. Furthermore, the research will construct and compare multiple machine learning models, ultimately establishing a high-precision, interpretable individualized prediction model to quantitatively assess the 5-year net survival benefit derived from PMRT. This investigation will provide scientific evidence to support the exemption of low-risk patients from radiotherapy and is dedicated to offering clinicians a practical computational tool that facilitates precise clinical decision.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Data source and study design\u003c/h2\u003e \u003cp\u003eWe analyzed data from T1-T3NmiM0 breast cancer patients collected between January 2010 and December 2022 from the SEER database of the National Cancer Institute of the United States, utilizing SEERStat software (version 9.0.4.2). The SEER database serves as an open-access resource for epidemiological and survival analyses of cancer (for details, see the official website \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://seer.cancer.gov/\u003c/span\u003e\u003cspan address=\"https://seer.cancer.gov/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The inclusion criteria were as follows: (1) the patient had invasive breast cancer as the only malignancy; (2) there was no distant metastasis (M0); (3) the T of the tumor was T1-T3and the lymph node stage was PN1mi. The exclusion criteria included: (1) patients who received preoperative or intraoperative radiotherapy or did not undergo surgical treatment; (2) patients with two or more primary malignant tumors; (3) patients with incomplete key pathological or staging information, such as ER, PR, HER2 status, grade, or AJCC stage. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the workflow of our study design and its analyses.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThis retrospective study comprises four common machine learning algorithms implemented in Python 3.12. The primary objective of the models is to forecast whether patients will experience 5-year survival benefits following postoperative radiotherapy. In the machine learning model, a 5-year survival state is defined as survival if the survival time is \u0026ge;\u0026thinsp;60 months, and as death if the survival time is \u0026lt;\u0026thinsp;60 months. The hyperparameters for all models are predetermined based on prior research and initial experimental findings. Data preprocessing involves encoding labels for categorical variables and standardizing continuous variables. The dataset is randomly split into training and test sets at a 7:3 ratio. A 50% discount hierarchical cross-validation method is employed to assess the model's robustness on the training set, with final performance evaluation conducted on an independent test set.\u003c/p\u003e \u003cp\u003eReceiver Operating Characteristic (ROC), calibration curves, accuracy, precision, recall rate, specificity, and F1 scores were employed to compare the performance of Random Forest (RF), extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), and Logistic Regression (LR) models\u003csup\u003e[20, 21]\u003c/sup\u003e. The clinical net benefit of the prediction model was assessed using Decision Curve Analysis (DCA)\u003csup\u003e[22]\u003c/sup\u003e. To facilitate decision-makers in the correct application of the model, it is essential to comprehend the influence of each feature on the model's predictions. Consequently, Shapley Additive exPlanations (SHAP) were utilized to analyze the impact of each feature and to rank those that significantly affect the model\u003csup\u003e[23]\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Statistical analysis\u003c/h2\u003e \u003cp\u003eThe patients' basic characteristics were presented as frequency (n) and percentage, and group comparisons were made using the chi-square test. To mitigate bias from confounding factors, 1:1 Propensity Score Matching (PSM) was employed to assess the impact of radiotherapy on the prognosis of T1-T3NmiM0 breast cancer patients, resulting in the matching of 8,046 patients. Subsequently, Kaplan-Meier (K-M) survival analysis and log-rank tests were conducted, stratified by T and AJCC stages among PSM-matched patients. Statistical significance was defined as P\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Univariate and multivariate COX analyses were performed to investigate the relationship between characteristics and patient survival rates within the PSM dataset. Statistical analyses were carried out using R version 4.5.1 and Python 3.12.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1. Clinical characteristics of patients with T1-T3NmiM0 breast cancer\u003c/h2\u003e\n \u003cp\u003eWe ultimately identified 17,994 eligible patients with T1-T3NmiM0 breast cancer from the SEER database (2010\u0026ndash;2022). The baseline clinicopathological characteristics of the entire cohort prior to propensity score matching are presented in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e and summarized as follows. Among the radiotherapy group (n\u0026thinsp;=\u0026thinsp;11,474), a substantial majority of patients (74.3%) underwent breast-conserving surgery, whereas only 21.2% of patients in the non-radiotherapy group (n\u0026thinsp;=\u0026thinsp;6,520) received this procedure. Furthermore, 48.3% of patients in the radiotherapy group received chemotherapy, compared to 41.6% in the non-radiotherapy group. The proportion of patients receiving systemic treatment in the radiotherapy group (93.8%) was significantly higher than that in the non-radiotherapy group (76.9%). Regarding tumor characteristics, the proportions of AJCC stage I, T1, ER-positive, and HER2-negative patients in the radiotherapy group (66.1%, 54.6%, 90.7%, and 88.3%, respectively) were slightly greater than those in the non-radiotherapy group (63.1%, 52.7%, 88.8%, and 86.1%, respectively). In the radiotherapy group, 46.6% of patients had a time from diagnosis to treatment of \u0026le;\u0026thinsp;30 days, which was higher than the 37.2% observed in the non-radiotherapy group. All of these differences were statistically significant (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The basic characteristics of the propensity score matched patients are shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eBasic Characteristics of breast cancer patients with T1-T3NmiM0.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRadiation\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNo_Radiation\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eP Value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003en\u0026thinsp;=\u0026thinsp;11474\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003en\u0026thinsp;=\u0026thinsp;6520\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ecase(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ecase( %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e791 (6.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e510 (7.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40\u0026ndash;49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2305 (20.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1426 (21.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50\u0026ndash;59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3120 (27.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1647 (25.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60\u0026ndash;69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3132 (27.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1463 (22.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e70\u0026ndash;79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1670 (14.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e974 (14.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e80+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e456 (4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e500 (7.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRace\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWhite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8940 (77.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5018 (77 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.319\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1151 (10 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e674 (10.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1383 (12.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e828 (12.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarital status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarried\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7015 (61.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3735 (57.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSingle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1702 (14.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e998 (15.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWidow/divorced/other\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2757 (24 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1787 (27.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrimary Site\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCentral portion of breast\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e620 (5.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e416 (6.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUpper-inner quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1221 (10.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e675 (10.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLower-inner quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e587 (5.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e342 (5.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUpper-outer quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4497 (39.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2099 (32.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLower-outer quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e973 (8.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e544 (8.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOverlapping lesion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2676 (23.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1449 (22.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e900 (7.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e995 (15.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWell differentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2720 (23.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1395 (21.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate differentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5856 (51 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3359 (51.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePoorly differentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2898 (25.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1766 (27.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLaterality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLeft\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5678 (49.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3271 (50.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.387\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5796 (50.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3249 (49.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTime from diagnosis to treatment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;30 days\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5342 (46.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2427 (37.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ge;\u0026thinsp;30 days\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6132 (53.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4093 (62.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHistology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIDC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9160 (79.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5101 (78.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.028\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eILC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1338 (11.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e783 (12 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMixed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e658 (5.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e425 (6.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e318 (2.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e211 (3.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAJCC Stage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7579 (66.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4112 (63.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3271 (28.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2149 (33 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e624 (5.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e259 (4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6263 (54.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3437 (52.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4407 (38.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2808 (43.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e804 (7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e275 (4.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRegional nodes positive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026le;\u0026thinsp;1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9825 (85.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5602 (85.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.606\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026gt;\u0026thinsp;1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1649 (14.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e918 (14.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eER\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1065 (9.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e732 (11.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10409 (90.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5788 (88.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2024 (17.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1311 (20.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9450 (82.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5209 (79.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHER-2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10128 (88.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5615 (86.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1346 (11.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e905 (13.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBreast Subtype\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR+/HER2+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1106 (9.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e659 (10.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR+/HER2-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9377 (81.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5169 (79.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR-/HER2+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e240 (2.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e246 (3.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR-/HER2-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e751 (6.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e446 (6.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSurg Prim Site\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBreast-conserving surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8524 (74.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1380 (21.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSimple mastectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2275 (19.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3927 (60.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRadical mastectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e675 (5.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1213 (18.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChemotherapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo/unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5928 (51.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3806 (58.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5546 (48.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2714 (41.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSystemic therapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e711 (6.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1504 (23.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10763 (93.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5016 (76.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"char\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eBasic Characteristics of PSM-T1-T3NmiM0 Breast Cancer patients.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRadiation\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNO Radiation\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eP value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003en\u0026thinsp;=\u0026thinsp;4023\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003en\u0026thinsp;=\u0026thinsp;4023\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ecase(% )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ecase(% )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e587 (14.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e453 (11.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40\u0026ndash;49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1019 (25.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e985 (24.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50\u0026ndash;59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e980 (24.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1009 (25.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60\u0026ndash;69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e790 (19.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e881 (21.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e70\u0026ndash;79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e439 (10.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e489 (12.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e80+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e208 (5.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e206 (5.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRace\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e437 (10.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e408 (10.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.424\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e533 (13.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e514 (12.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWhite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3053 (75.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3101 (77.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarital status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarried\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2445 (60.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2449 (60.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.214\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSingle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e666 (16.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e616 (15.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWidow/divorced/other\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e912 (22.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e958 (23.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrimary Site\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCentral portion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e224 (5.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e215 (5.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.704\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLower-inner quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e173 (4.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e200 (5.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLower-outer quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e304 (7.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e311 (7.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e533 (13.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e522 (13.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOverlapping lesion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e891 (22.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e926 (23.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUpper-inner quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e419 (10.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e418 (10.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUpper-outer quadrant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1479 (36.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1431 (35.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate differentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1994 (49.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2007 (49.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.014\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePoorly differentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1287 (32.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1189 (29.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWell differentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e742 (18.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e827 (20.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLaterality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eleft\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2009 (49.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2023 (50.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.772\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eright\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2014 (50.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2000 (49.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHistology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIDC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3024 (75.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3101 (77.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.206\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eILC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e629 (15.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e568 (14.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMixed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e245 (6.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e239 (5.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e125 (3.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e115 (2.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAJCC Stage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2113 (52.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2357 (58.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1560 (38.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1453 (36.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e350 (8.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e213 (5.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSurg Prim Site\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBreast-conserving surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1450 (36.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1357 (33.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.094\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRadical mastectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e585 (14.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e604 (15.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSimple mastectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1988 (49.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2062 (51.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChemotherapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo/unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1638 (40.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1918 (47.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2385 (59.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2105 (52.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSystemic therapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e706 (17.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e689 (17.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.638\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3317 (82.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3334 (82.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTime from diagnosis to treatment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;30days\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1818 (45.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1681 (41.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ge;\u0026thinsp;30days\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2205 (54.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2342 (58.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1503 (37.4 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1853 (46.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2096 (52.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1933 (48.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e424 (10.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e237 (5.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRegional nodes positive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026gt;\u0026thinsp;1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e931 (23.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e711 (17.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026le;\u0026thinsp;1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3092 (76.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3312 (82.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBreast Subtype\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR-/HER2-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e374 (9.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e334 (8.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.033\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR-/HER2+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e112 (2.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e117 (2.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR+/HER2-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3044 (75.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3145 (78.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHR+/HER2+\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e493 (12.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e427 (10.6 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eER\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e517 (12.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e478 (11.9 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.198\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3506 (87.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3545 (88.1 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e879 (21.8 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e831 (20.7 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3144 (78.2 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3192 (79.3 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHER2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3418 (85.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3479 (86.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.056\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e605 (15.0 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e544 (13.5 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2. Univariate and multivariate cox regression results\u003c/h2\u003e\n \u003cp\u003eIn PSM data, a significant difference in overall survival (OS) was observed between the radiation and no-radiation groups (HR: 0.85, 95% CI 0.75\u0026ndash;0.96, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). However, no significant difference was found in breast cancer-specific survival (BCSS) (HR: 0.87, 95% CI 0.74\u0026ndash;1.02, p\u0026thinsp;=\u0026thinsp;0.84). Subgroup survival analyses were conducted based on AJCC stage and T subgroups. Among AJCC stage I and T1 patients, both OS (HR: 0.87, 95% CI 0.75\u0026ndash;1.08, p\u0026thinsp;=\u0026thinsp;0.195; HR: 0.96, 95% CI 0.77\u0026ndash;1.19, p\u0026thinsp;=\u0026thinsp;0.704) and BCSS (HR: 0.96, 95% CI 0.77\u0026ndash;1.91, p\u0026thinsp;=\u0026thinsp;0.704; HR: 0.95, 95% CI 0.69\u0026ndash;1.31, p\u0026thinsp;=\u0026thinsp;0.757) did not show benefits from radiotherapy. Conversely, patients in stages II and III showed improved OS (HR: 0.83, 95% CI 0.7\u0026ndash;0.99, p\u0026thinsp;=\u0026thinsp;0.035; HR: 0.52, 95% CI 0.37\u0026ndash;0.74, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) with postoperative radiotherapy. Similarly, patients in T2 and T3 groups exhibited enhanced OS (HR: 0.79, 95% CI 0.67\u0026ndash;0.93, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; HR: 0.52, 95% CI 0.36\u0026ndash;0.74, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) with postoperative radiotherapy. Notably, stage II BCSS (HR: 0.88, 95% CI 0.71\u0026ndash;1.08, p\u0026thinsp;=\u0026thinsp;0.212) did not benefit from radiotherapy, and BCSS in T2 group (HR: 0.82, 95% CI 0.67-1.00, p\u0026thinsp;=\u0026thinsp;0.053) did not exhibit significant survival advantages. Conversely, in AJCC stage III and T3 groups, BCSS (HR: 0.51, 95% CI 0.34\u0026ndash;0.78, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; HR: 0.52, 95% CI 0.34\u0026ndash;0.79, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) showed benefits from radiotherapy.\u003c/p\u003e\n \u003cp\u003eWe performed univariate and multivariate Cox analyses on the data adjusted PSM. The results of the multivariate Cox regression analysis indicated that age over 70 years, the presence of more than one positive lymph node,histological grade, radiotherapy, estrogen receptor (ER) status, progesterone receptor (PR) status, and AJCC stages II and III significantly influenced OS (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e) and BCSS (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e) in patients (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In contrast, age between 60 and 69 years affected OS (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) but did not significantly impact BCSS (p\u0026thinsp;=\u0026thinsp;0.995).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3 Establishment and evaluation of predictive models\u003c/h2\u003e\n \u003cp\u003eAfter a thorough evaluation of four machine learning models, we identified XGBoost as the optimal predictive model due to its superior predictive performance (AUC value\u0026thinsp;=\u0026thinsp;0.770) (Fig. 7a). However, since the dataset exhibits an imbalanced distribution of positive and negative events, relying solely on AUC may not suffice to assess the model\u0026apos;s effectiveness. To address this limitation, we utilized the PR (precision-recall) curve to provide a more comprehensive evaluation of the model\u0026apos;s strengths and weaknesses. The PR curve analysis revealed that the XGBoost model outperformed other models, as indicated by its higher Average Precision (AP\u0026thinsp;=\u0026thinsp;0.769) (Fig. 7b). Furthermore, the confusion matrix of the XGBoost model is presented in Fig. 7c. Overall, as illustrated in Fig. 7d, XGBoost demonstrated superior overall performance, achieving the highest Accuracy (0.691), F1 score (0.717), Specificity (0.634), and Sensitivity (0.741). The calibration curves assessed the alignment between the predicted 5-year survival benefit of radiotherapy by four models and the observed outcomes. The XGBoost model exhibited superior calibration curve performance on the test set compared to other models (Fig. \u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003ea). Consequently, the XGBoost model was identified as optimal. Evaluation of the clinical net benefit of the predictive model was conducted using the Decision Curve Analysis (DCA), revealing that the XGBoost model provided greater net benefit across most threshold probabilities (Fig. \u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003eb). The ranking of feature importance based on SHAP values indicated that T, stage, age, primary site, and grade were significant factors, with T being the most influential, followed by AJCC stage and time from diagnosis to treatment. Subsequently, the impact of variables on the model was assessed using SHAP values. Variables were ranked by mean SHAP value, highlighting tumor size and AJCC stage as the most critical factors influencing the XGBoost model (Fig. \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study utilizes extensive SEER data to elucidate the impact of PMRT on the survival outcomes of patients with T1-T3N1miM0 breast cancer. Additionally, it aims to develop an individualized prediction model to identify subgroups that may not derive significant benefit from radiotherapy. Our findings indicate that, following propensity score matching to balance confounding factors, PMRT markedly improved the OS of the cohort. However, no significant improvements in OS or BCSS were observed in patients with a mild tumor burden at T1 stage and AJCC stage I, corroborating previous studies\u003csup\u003e[15, 24\u0026ndash;26]\u003c/sup\u003e. The EUSOMA guidelines suggest that certain low-risk breast cancer patients who have already undergone endocrine therapy may not require radiotherapy\u003csup\u003e[27]\u003c/sup\u003e. Furthermore, previous studies has demonstrated that postoperative radiotherapy does not significantly affect the OS of some elderly patients with early-stage breast cancer\u003csup\u003e[28]\u003c/sup\u003e. Our research indicates that radiotherapy has minimal impact on their overall survival for patients with a mild tumor burden (T1pmiMO and AJCC stage I with pmiMO). Given the long-term risks associated with radiotherapy, including cardiopulmonary toxicity, secondary malignancies, and lymphedema\u003csup\u003e[29, 30]\u003c/sup\u003e, the results of this study provide compelling retrospective evidence to support the clinical decision to exempt certain low-risk patients from PMRT.\u003c/p\u003e \u003cp\u003eThe subgroup analysis based on traditional clinical staging (AJCC stage) aids in identifying the general population that may be exempt from radiotherapy. Precision medicine necessitates the integration of multi-dimensional information to quantitatively predict the benefits of radiotherapy for patients. To address this limitation, the present study developed and validated a machine learning model utilizing the XGBoost model. This model exhibited excellent discriminative ability in the test set (AUC\u0026thinsp;=\u0026thinsp;0.770) and outperformed LR, RF, and KNN, significantly surpassing traditional nomograms\u003csup\u003e[13\u0026ndash;15]\u003c/sup\u003e, which often reported AUC values below 0.7 in prior studies. This investigation validated the robust predictive capacity of the XGBoost model in the specific cohort of breast cancer T1-T3N1miM0, highlighting its potential to furnish clinicians with a personalized 5-year survival benefit-risk assessment for radiotherapy. This approach facilitates tailored and precise radiotherapy decisions\u003csup\u003e[18]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThis study also constructed a DCA to illustrate the predictive power advantage of the XGBoost model. To further elucidate the XGBoost model, we examined the influence of each patient feature on the model using SHAP values, resulting in a clear ranking of features. Our analysis revealed that the clinical characteristics of T and AJCC stage are the most critical factors influencing the model's predictions. This finding enhances the biological and clinical rationale of the model and supports informed clinical decision-making.\u003c/p\u003e \u003cp\u003eThis study revealed that varying quantities of pN1mi lymph nodes significantly influence prognosis. Previous research examined and categorized different quantities of pN1mi lymph nodes, indicating that an escalation in micrometastatic lymph node numbers correlates with a poorer prognosis among N1mi breast cancer patients\u003csup\u003e[31]\u003c/sup\u003e. Nonetheless, this study possesses additional constraints. Initially, the absence of data such as the HER2 status in the SEER database during earlier periods could introduce data selection bias. Secondly, the SEER database lacks comprehensive information on radiotherapy dosages, chemotherapy protocols, targeted therapy, and endocrine therapy, thereby hindering the evaluation of these treatments' impact on prognosis\u003csup\u003e[32]\u003c/sup\u003e. Thirdly, Ki67 functions as a prognostic biomarker for breast cancer, while PDL-1 represents a significant therapeutic target.\u003csup\u003e[33, 34]\u003c/sup\u003e, However, the SEER database does not provide information on Ki67, PDL-1, and related factors, which constrains comprehensive research on these elements. Furthermore, the generalizability of the XGBoost model requires external validation and calibration in a multi-center cohort\u003csup\u003e[18]\u003c/sup\u003e。Future research aims to improve the model's predictive accuracy and biological interpretability by incorporating multi-omics data, including radiomics and genomics. Our objective is to develop this model into an online clinical decision support tool, thereby alleviating unnecessary treatment burdens for low-risk patients and advancing precision medicine in breast cancer.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, PMRT can enhance the overall survival of patients with T1-T3N1miM0 breast cancer. However, for patients classified as pN1miM0 at T1 Stage or AJCC Stage I, the survival benefits are limited, suggesting that PMRT may be considered for exemption. In light of this, we have developed and validated a high-precision, interpretable XGBoost machine learning model capable of quantitatively assessing the 5-year survival benefit of radiotherapy for individual patients. This study not only provides evidence supporting the exemption of PMRT for specific low-risk patients but also presents a practical predictive tool for making individualized adjuvant treatment decisions for these patients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthical review and approval were waived for this study due to the fact that the\u003c/p\u003e\n\u003cp\u003edata are fully de-identified and no intervention on patients was performed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and material\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data here are publicly available in the SEER database [https:// seer. cancer.gov/ (accessed on September 22, 2025)].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank all staff of the SEER database for their contribution in data collection,maintenance, distribution and so on. Also we would like to thank all the developers of the R programming package and python for selflessly sharing their code.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National Natural Science Foundation of China (Grant No.82160481, 82460522), Project funded by China Postdoctoral Science Foundation (Grant No. 2023MD744194), the Natural Science Foundation of Guangxi (Grant No. 2021GXNSFBA196015), Joint Project on Regional High-Incidence Diseases Research of Guangxi Natural Science Foundation (Grant No. 2024GXNSFAA010058), and First-class Discipline Innovation-driven Talent Program of Guangxi Medical University.Availability of data and materials\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWWW, BJC, and SHY contributed equally as co-first authors.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWWW:\u003c/strong\u003e Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft.\u003cbr\u003e\u003cstrong\u003eBJC:\u003c/strong\u003e Formal analysis, Investigation, Methodology, Supervision, Validation, Writing – original draft.\u003cbr\u003e\u003cstrong\u003eSHY:\u003c/strong\u003e Data curation, Formal analysis, Investigation, Methodology, Writing – original draft.\u003cbr\u003e\u003cstrong\u003eQXT:\u003c/strong\u003e Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review \u0026amp; editing.\u003cbr\u003e\u003cstrong\u003eCYW:\u003c/strong\u003e Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003eAll authors read and approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eXIONG X, ZHENG L W, DING Y, et al. Breast cancer: pathogenesis and treatments [J]. Signal Transduct Target Ther, 2025, 10(1): 49.\u003c/li\u003e\n\u003cli\u003eMAMTANI A, PATIL S, STEMPEL M, et al. Axillary Micrometastases and Isolated Tumor Cells Are Not an Indication for Post-mastectomy Radiotherapy in Stage 1 and 2 Breast Cancer [J]. Ann Surg Oncol, 2017, 24(8): 2182-8.\u003c/li\u003e\n\u003cli\u003eZHAO D, XU M, YANG S, et al. Specific diagnosis of lymph node micrometastasis in breast cancer by targeting activatable near-infrared fluorescence imaging [J]. Biomaterials, 2022, 282: 121388.\u003c/li\u003e\n\u003cli\u003eKLAUBER-DEMORE N, VAN ZEE K J, LINKOV I, et al. Biological behavior of human breast cancer micrometastases [J]. Clin Cancer Res, 2001, 7(8): 2434-9.\u003c/li\u003e\n\u003cli\u003eGALIMBERTI V, COLE B F, VIALE G, et al. Axillary dissection versus no axillary dissection in patients with breast cancer and sentinel-node micrometastases (IBCSG 23-01): 10-year follow-up of a randomised, controlled phase 3 trial [J]. Lancet Oncol, 2018, 19(10): 1385-93.\u003c/li\u003e\n\u003cli\u003eLIM S Z, KUSUMAWIDJAJA G, MOHD ISHAK H M, et al. Outcomes of Stage I and II Breast Cancer with Nodal Micrometastases Treated with Mastectomy without Axillary Therapy [J]. Breast Cancer Res Treat, 2021, 189(3): 837-43.\u003c/li\u003e\n\u003cli\u003eWU J, FAN D, SHAO Z, et al. CACA Guidelines for Holistic Integrative Management of Breast Cancer [J]. Holist Integr Oncol, 2022, 1(1): 7.\u003c/li\u003e\n\u003cli\u003eWAKS A G, WINER E P. Breast Cancer Treatment: A Review [J]. JAMA, 2019, 321(3): 288-300.\u003c/li\u003e\n\u003cli\u003eGRADISHAR W J, MORAN M S, ABRAHAM J, et al. Breast Cancer, Version 3.2024, NCCN Clinical Practice Guidelines in Oncology [J]. Journal of the National Comprehensive Cancer Network, 2024, 22(5): 331-57.\u003c/li\u003e\n\u003cli\u003eLOIBL S, ANDR\u0026eacute; F, BACHELOT T, et al. Early breast cancer: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up [J]. Ann Oncol, 2024, 35(2): 159-82.\u003c/li\u003e\n\u003cli\u003eRECHT A, COMEN E A, FINE R E, et al. Postmastectomy Radiotherapy: An American Society of Clinical Oncology, American Society for Radiation Oncology, and Society of Surgical Oncology Focused Guideline Update [J]. J Clin Oncol, 2016, 34(36): 4431-42.\u003c/li\u003e\n\u003cli\u003ePARK K U, SOMERFIELD M R, ANNE N, et al. Sentinel Lymph Node Biopsy in Early-Stage Breast Cancer: ASCO Guideline Update [J]. J Clin Oncol, 2025, 43(14): 1720-41.\u003c/li\u003e\n\u003cli\u003eWU S P, TAM M, SHAIKH F, et al. Post-mastectomy Radiation Therapy in Breast Cancer Patients with Nodal Micrometastases [J]. Ann Surg Oncol, 2018, 25(9): 2620-31.\u003c/li\u003e\n\u003cli\u003eHUO D, HOU N, JASKOWIAK N, et al. Use of Postmastectomy Radiotherapy and Survival Rates for Breast Cancer Patients with T1-T2 and One to Three Positive Lymph Nodes [J]. Ann Surg Oncol, 2015, 22(13): 4295-304.\u003c/li\u003e\n\u003cli\u003eTANG Y, ZHANG Y J, ZHANG N, et al. Nomogram predicting survival as a selection criterion for postmastectomy radiotherapy in patients with T1 to T2 breast cancer with 1 to 3 positive lymph nodes [J]. Cancer, 2020, 126 Suppl 16: 3857-66.\u003c/li\u003e\n\u003cli\u003eQI Y J, SU G H, YOU C, et al. Radiomics in breast cancer: Current advances and future directions [J]. Cell Rep Med, 2024, 5(9): 101719.\u003c/li\u003e\n\u003cli\u003eQIAN X, PEI J, HAN C, et al. A multimodal machine learning model for the stratification of breast cancer risk [J]. Nat Biomed Eng, 2025, 9(3): 356-70.\u003c/li\u003e\n\u003cli\u003eCLIFT A K, DODWELL D, LORD S, et al. Development and internal-external validation of statistical and machine learning models for breast cancer prognostication: cohort study [J]. Bmj, 2023, 381: e073800.\u003c/li\u003e\n\u003cli\u003eZHU H, ZHOU Y, SHEN D, et al. An interpretable machine learning model for predicting early liver metastasis after pancreatic cancer surgery [J]. BMC Cancer, 2025, 25(1): 1117.\u003c/li\u003e\n\u003cli\u003eZHANG Y, AN W, WANG C, et al. Novel models based on machine learning to predict the prognosis of metaplastic breast cancer [J]. Breast, 2025, 79: 103858.\u003c/li\u003e\n\u003cli\u003eREN J, LI Y, ZHOU J, et al. Developing machine learning models for personalized treatment strategies in early breast cancer patients undergoing neoadjuvant systemic therapy based on SEER database [J]. Sci Rep, 2024, 14(1): 22055.\u003c/li\u003e\n\u003cli\u003eYU Y, HE Z, OUYANG J, et al. Magnetic resonance imaging radiomics predicts preoperative axillary lymph node metastasis to support surgical decisions and is associated with tumor microenvironment in invasive breast cancer: A machine learning, multicenter study [J]. EBioMedicine, 2021, 69: 103460.\u003c/li\u003e\n\u003cli\u003eZHONG X, LIN Y, ZHANG W, et al. Predicting diagnosis and survival of bone metastasis in breast cancer using machine learning [J]. Sci Rep, 2023, 13(1): 18301.\u003c/li\u003e\n\u003cli\u003eHUGHES K S, SCHNAPER L A, BELLON J R, et al. Lumpectomy plus tamoxifen with or without irradiation in women age 70 years or older with early breast cancer: long-term follow-up of CALGB 9343 [J]. J Clin Oncol, 2013, 31(19): 2382-7.\u003c/li\u003e\n\u003cli\u003eLU Z, GUO L, ZHOU J, et al. A nomogram to predict the benefit of postmastectomy radiotherapy in breast cancer with nodal micrometastases [J]. Breast cancer (Tokyo, Japan), 2025, 32(5): 935-46.\u003c/li\u003e\n\u003cli\u003eLIU K, LI G Q, LI S Q, et al. Clinical treatment score Post-5 Years (CTS5) predicts the benefit of postmastectomy radiotherapy in patients with T1-2N1 luminal breast cancer [J]. Breast, 2025, 79: 103873.\u003c/li\u003e\n\u003cli\u003eDARBY S, MCGALE P, CORREA C, et al. Effect of radiotherapy after breast-conserving surgery on 10-year recurrence and 15-year breast cancer death: meta-analysis of individual patient data for 10,801 women in 17 randomised trials [J]. Lancet, 2011, 378(9804): 1707-16.\u003c/li\u003e\n\u003cli\u003eKUNKLER I H, WILLIAMS L J, JACK W J L, et al. Breast-Conserving Surgery with or without Irradiation in Early Breast Cancer [J]. N Engl J Med, 2023, 388(7): 585-94.\u003c/li\u003e\n\u003cli\u003eL\u0026oacute;PEZ-FERN\u0026aacute;NDEZ T, MARCO I, AZNAR M C, et al. Breast cancer and cardiovascular health [J]. Eur Heart J, 2024, 45(41): 4366-82.\u003c/li\u003e\n\u003cli\u003eJARM T, BESIC N, ARNEZ R C, et al. Breast cancer related lymphedema and shoulder mobility following radiotherapy [J]. Strahlenther Onkol, 2025.\u003c/li\u003e\n\u003cli\u003eLUO S, FU W, LIN J, et al. Prognosis and local treatment strategies of breast cancer patients with different numbers of micrometastatic lymph nodes [J]. World J Surg Oncol, 2023, 21(1): 202.\u003c/li\u003e\n\u003cli\u003eYANG J, LI Y, WANG S, et al. Distinct prognostic patterns of single hormone receptor-positive subtypes in HER2-negative breast cancer: a SEER-based retrospective cohort study [J]. Int J Surg, 2025.\u003c/li\u003e\n\u003cli\u003eYERUSHALMI R, WOODS R, RAVDIN P M, et al. Ki67 in breast cancer: prognostic and predictive potential [J]. Lancet Oncol, 2010, 11(2): 174-83.\u003c/li\u003e\n\u003cli\u003eDVIR K, GIORDANO S, LEONE J P. Immunotherapy in Breast Cancer [J]. Int J Mol Sci, 2024, 25(14).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"breast cancer, micrometastasis, Machine Learning, post-mastectomy radiotherapy","lastPublishedDoi":"10.21203/rs.3.rs-9027856/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9027856/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective:\u003c/strong\u003e This study aims to assess the impact of post-mastectomy radiotherapy (PMRT) on the survival outcomes of patients with T1-T3N1miM0 breast cancer. Additionally, we seek to develop an interpretable machine learning model to predict the individualized 5-year survival benefits for patients undergoing postoperative radiotherapy in comparison to those who do not receive such treatment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e Based on data from the SEER database spanning from 2010 to 2022, a cohort of 17,994 breast cancer patients diagnosed with T1-3N1miM0 was analyzed. Propensity score matching was employed to mitigate baseline discrepancies between the radiotherapy and non-radiotherapy cohorts. The overall survival and breast cancer-specific survival between the two groups were compared using the Kaplan-Meier method. A predictive model utilizing XGBoost was developed to estimate the 5-year survival advantages associated with postoperative radiotherapy, with the model's outcomes elucidated through SHAP analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eA significant difference in survival was observed between the postoperative radiotherapy and non-radiotherapy groups (HR: 0.85, 95% CI 0.75-0.96, p \u0026lt; 0.01). However, no significant differences in overall survival (OS) were found among the T1 and stage I radiotherapy and non-radiotherapy groups (HR: 0.87, 95% CI 0.75-1.08, p = 0.195; HR: 0.96, 95% CI 0.77-1.19, p = 0.704). Although breast cancer-specific survival (BCSS) was assessed, the difference was not significant (HR: 0.87, 95% CI 0.74-1.02, p = 0.84). The XGBoost model we developed exhibited exceptional predictive performance and was identified as the most effective model for predicting the 5-year survival outcomes of T1-T3pN1miM0 breast cancer patients (AUC = 0.770).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e Breast cancer patients diagnosed with T1N1miM0 or Stage I（pN1miM0） may potentially forgo PMRT. An XGBoost machine learning model was created to forecast the 5-year survival advantage of radiotherapy for postoperative individuals with T1-T3N1miM0 breast cancer.\u003c/p\u003e","manuscriptTitle":"Predicting the survival benefits of postoperative radiotherapy for breast cancer with lymph node micrometastasis: A Machine Learning model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-16 08:40:11","doi":"10.21203/rs.3.rs-9027856/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cd39ddb0-e404-48eb-a39a-df93532e6fee","owner":[],"postedDate":"March 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64441200,"name":"Biological sciences/Cancer"},{"id":64441201,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2026-04-21T07:41:56+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-16 08:40:11","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9027856","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9027856","identity":"rs-9027856","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}