The Impact of Post-mastectomy Radiotherapy on T1-2N0-1 Male Breast Cancer and Establishment of an Artificial Neural Network Predicting Model: Population-Based Study

Kaiyan Huang; Yushuai Yu; Xiaofen Li; Yushan Liu; Kailong Huang; Xin Wang; Jie Zhang

doi:10.4143/crt.2025.518

Articles

Page Path: HOME > J Korean Cancer Assoc > Ahead-of print articles > Article

Original Article The Impact of Post-mastectomy Radiotherapy on T1-2N0-1 Male Breast Cancer and Establishment of an Artificial Neural Network Predicting Model: Population-Based Study: Kaiyan Huang¹, Yushuai Yu², Xiaofen Li², Yushan Liu², Kailong Huang¹, Xin Wang¹, Jie Zhang²; DOI: https://doi.org/10.4143/crt.2025.518
Published online: November 5, 2025

¹Department of Breast Surgery, The Second Affiliated Hospital of Fujian Medical University, Quanzhou, China

²Department of Breast Surgery, Department of General Surgery, Fujian Medical University Union Hospital, Fuzhou, China

Correspondence: Jie Zhang, Department of Breast Surgery, Fujian Medical University Union Hospital, No. 29, Xin Quan Road, Gulou District, Fuzhou, Fujian Province 350001, China
Tel: 86-13959189637 E-mail: zjie1979@fjmu.edu.cn

*Kaiyan Huang, Yushuai Yu, Xiaofen Li, and Yushan Liu contributed equally to this work.

• Received: May 14, 2025 • Accepted: November 4, 2025

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

1,235 Views
60 Download
2 Crossref

Full Article

Download PDF

Abstract

Purpose
The objective of this study was to analyze the impact of post-mastectomy radiotherapy (PMRT) in male breast cancer (MBC) patients and develop an artificial neural network (ANN) model to identify a potential PMRT benefit population.
Materials and Methods
Data from a total of 2,247 MBC patients with T1-2N0-1M0 who underwent total mastectomy between 1998 and 2016 were enrolled from the Surveillance, Epidemiology, and End Results database. Propensity score matching was used to reduce covariate imbalances. Cox regression analysis was conducted to compare overall survival (OS) between the PMRT and no-PMRT groups. The hypothesis was that patients who had undergone PMRT and lived longer than the median OS of the no-PMRT group could benefit from PMRT. An ANN model was then developed to predict PMRT benefit population.
Results
Multivariate Cox regression analysis demonstrated better OS in the PMRT group compared to the no-PMRT group of matched patients. This survival benefit was particularly significant in patients with grade III or T2N0 and T2N1 disease, while no significant difference was observed in patients with grade I/II or T1N0 and T1N1 disease. An ANN model was established to predict PMRT benefit population based on patients with T2N0/T2N1. The optimal cut-off value for the model predicted probability was 0.51. Survival curves indicated that a score of 0.51 could accurately distinguish potential PMRT benefit population.
Conclusion
For MBC patients with T2N0, T2N1, and grade III, PMRT would improve survival. The ANN model would be used to identify patients who are likely to benefit from PMRT and aid in clinical decision-making.
Key words: Male breast neoplasms, Post mastectomy radiotherapy, Benefit population, Artificial neural network predicting model, Overall survival

Introduction

Although males account for only 1% of breast cancer patients, the incidence has been increasing in recent years, while breast cancer has become the most common malignant tumor worldwide [1-4]. Due to the lack of dedicated randomized clinical research evidence for male breast cancer (MBC), we often rely on guidelines based on studies primarily focused on female breast cancer (FBC) when treating MBC [5]. When considering the differences in clinical and pathological characteristics between MBC and FBC, we find that MBC tends to present as more aggressive, have more frequent lymph node involvement, a more advanced disease stage when initially diagnosed, higher hormone receptor–positive and higher recurrence rates, and a poorer overall survival (OS) when compared with FBC [6-9]. Therefore, it is crucial to conduct research specifically on the treatment of MBC patients.

Radiotherapy (RT) is used to enhance local control after surgery in high-risk breast cancer patients and has been demonstrated to be effective in stage T3/T4 tumor or lymph node positive FBC. Yet, the evidence for postmastectomy radiotherapy (PMRT) in MBC are not strong enough. Several studies, including two that used the Surveillance, Epidemiology, and End Results (SEER) database, focused on this topic [9-14]. Nearly all of them showed better survival in the RT group. However, there are still some controversies about who should receive RT and what benefits it brings. Yu et al. [10] found better local control in the RT group without improvement of OS. While Weir et al. [9] showed that post-mastectomy RT resulted in better OS. Madden et al. [13] and Abrams et al. [14], both using the SEER database to study their subjects, also came to different conclusions; one supported using RT in the node-positive population while the other did not. None of these studies paid adequate attention to tumor size, immunohistochemical subtype, or some other important factors. Therefore, we are still uncertain about the optimal use of RT in MBC, especially for the marginalized group such as T1-2 and N0-1 patients. Since the breast volume of MBC patients is much smaller than that of FBC patients, tumor of the same size were more likely to invade the skin or the chest wall in MBC than in FBC [15]. This raised a question of whether we should simply follow the principle of RT used in FBC or expand or narrow its indications.

Here, we conducted a study on the impact of PMRT in MBC patients with T1-2 and N0-1 disease using information from the SEER database. Our study includes the largest number of T1-2 and N0-1 MBC patients to date. We investigated the efficacy of RT in MBC in the general population and in different subgroups by using propensity score matching (PSM) analysis and established a predicted model of RT effectiveness in MBC patients by the “artificial neural network (ANN)” method. We hypothesized that our research results can be a useful reference in choosing the best treatment for MBC patients.

Materials and Methods

1. Data source and study population

We used SEER*Stat software to generate a list of cases. We enrolled 2,247 patients according to the following inclusion criteria: male; diagnosis year between 1998 and 2016; underwent total mastectomy; had breast carcinoma as the only primary malignant cancer diagnosis, and were classified as American Joint Committee on Cancer (AJCC) stage T1-T2 and AJCC stage N0-N1. Patients with only carcinoma in situ disease, a non-invasive form of breast cancer, or those with evidence of distant metastases were excluded from the study. According to whether they received PMRT, the patients were divided into two groups. We calculated follow-up durations from January 1, 1998 to December 31, 2016. Patient characteristics and treatment courses, such as age, race, marital status, PMRT status, and chemotherapy status, were identified. Additionally, tumor characteristics, including pathology, grade, AJCC stage, T category, N category, number of positive axillary lymph node, estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, and breast subtype, were also evaluated.

2. Outcome measurement

In our study, OS was used as a primary study outcome. It was defined as extending from the date of initial diagnosis to the date of death, whether the patient died of breast cancer or other causes. Patients who were alive on the date of last contact were censored.

3. Establishment and validation of predicting ANN model

We hypothesized that patients who underwent PMRT and lived longer than the median OS of the no-PMRT group had benefited from PMRT. Based on this hypothesis, participants in the PMRT group were categorized into two types: a PMRT benefit group (median OS > 108 months), and a PMRT no benefit group (median OS ≤ 108 months), measured against the median OS time of the no-PMRT group.

Multi-Layer Perceptron (MLP) is an ANN with an input layer, at least one hidden layer, and an output layer. Every node in the input layer is fully connected to all the nodes of the hidden layer. All patients in the PMRT group were randomly divided into a training group and a validation group by 7:3. The output layer was comprised of two parts: (1) the ANN model predicted probability, which indicate the predicted probability of potential PMRT benefit for MBC patients, and (2) two categories of PMRT patients (PMRT benefit population vs. PMRT no benefit population).

The ANN model was applied to the no-PMRT group to calculate the predicted probability of a potential PMRT benefit. Next, to determine the optimal cut-off value, all patients including the PMRT group and the no-PMRT group were ranked by the probability value. The area under the receiver operator characteristic curve (AUC) was employed to evaluate predictive performance of the ANN model. To validate the clinical utility of the model, Kaplan-Meier curves were established to test whether this predictive model would distinguish patients in different groups classified by cut-off values of the ANN model predicted probability.

4. Statistical analysis

The chi-square test was conducted to describe the demographic and clinical characteristics of the PMRT and no-PMRT groups. The Kaplan-Meier method was utilized to generate survival curves. The significance of the difference between the PMRT and no-PMRT groups was assessed with the log-rank test. A hazard ratio (HR) with 95% confidence intervals (CI) was calculated by using a Cox proportional hazard regression model to determine the outcome-related factors. Proportional hazard assumptions were examined by Schoenfeld residuals test. Clinical factors with a p-value of 0.05 or less in univariate analyses or those known to be important from the literature were entered into multivariate Cox regression analyses. To reduce the influence of baseline differences in demographic and clinical characteristics on differences in the results, a 1:1 PSM method was used to match patients in the PMRT group and the no-PMRT group.

These statistical analyses were conducted using R software ver. 4.3.2 (R Foundation for Statistical Computing), Python software ver. 3.13 (Python Software Foundation), and SPSS software ver. 25.0 (IBM Corp.). The X-tile software was used to determine the optimal cut-off values [16]. All statistical analyses were two-sided, and a p-value of < 0.05 was considered significant.

Results

1. Demographics and clinical characteristics of the study population

Overall, 2,247 eligible patients were enrolled in our study, including 384 patients assigned to the PMRT group and 1,863 patients to the no-PMRT group. The baseline characteristics of two groups are summarized in Table 1. There were significant differences in characteristics between the two groups, including AJCC stage, T category, N category, tumor grade, number of positive axillary lymph nodes, and chemotherapy status. The patients treated with PMRT presented with a higher proportion of AJCC stage II (87.2% vs. 55.7%, p < 0.001), T2 category (59.1% vs. 41.5%, p < 0.001), N1 category (71.4% vs. 30.1%, p < 0.001), grade I/II (43.0% vs. 32.0%, p < 0.001) and more positive axillary lymph node (three positive axillary lymph nodes, 11.5% vs. 3.4%, p < 0.001). In addition, the PMRT group was more inclined to accept chemotherapy than the no-PMRT group (63.0% vs. 30.2%, p < 0.001). Other characteristics including age, marital status, race, pathology, ER status, PR status, HER2 status, and breast subtypes, were similarly distributed between the two groups.

2. Effect of PMRT on survival before PSM

In the univariate analysis of OS, older age, unmarried, higher grade, more advanced stage, larger tumor size, more positive axillary lymph nodes, and triple-negative breast cancer subtype were identified as risk factors for poor survival (HR > 1, p < 0.05). In contrast, other races and chemotherapy were found to be protective factors for better survival (HR < 1, p < 0.05). Although the p-value was more than 0.05, the RT status was included in the multivariate analysis as it is known to be an important treatment for multiple types of cancer. All the variables mentioned above were subsequently included in the multivariate Cox analysis. After adjusting for these confounding factors above, receiving RT was associated with an improved OS (HR, 0.784; 95% CI, 0.623 to 0.986; p=0.038), compared with the no radiation group (Table 2).

3. Survival estimates in matched groups

We performed a 1:1 PSM analysis between the two groups of patients to mitigate potential biases (Table 1). We obtained a matched cohort consisting of 762 patients, with each subgroup comprising 381 patients. In the matched groups, no significant differences were observed in any of the measured variables between patients in the two groups. With a median follow-up of 87 months, the PMRT group exhibits significantly better OS compared to the no-PMRT group (HR, 0.758; 95% CI, 0.583 to 0.985; p=0.039) (Fig. 1).

A stratified analysis was used to determine the benefits of PMRT in patients with different T categories and N categories. The results are shown in Table 3. To our surprise, PMRT was found to reduce the risk of all-cause mortality in patients with T2 category breast cancer but not in those with T1 category breast cancer (T1: OS HR, 1.049; 95% CI, 0.651 to 1.691; p=0.845; T2: OS HR, 0.668; 95% CI, 0.483 to 0.924; p=0.015). Patients diagnosed with N0 category did not receive a significant benefit from PMRT (OS HR, 0.673; 95% CI, 0.416 to 1.053; p=0.106). We observed similar phenomena in the N1 category cohort (OS HR, 0.822; 95% CI, 0.600 to 1.127; p=0.224). In clinical practice, T category and N category are often considered together. Therefore, stratified analyses were conducted to investigate the different outcomes between patients receiving PMRT and those not receiving PMRT when taking into account both T category and N category. As shown in Table 3 and Fig. 2, our study determined that PMRT produces a better clinical outcome for patients with T2N0 and T2N1 (T2N0: OS HR, 0.482; 95% CI, 0.267 to 0.870; p=0.015; T2N1: OS HR, 0.681; 95% CI, 0464 to 0.998; p=0.048). However, no statistical survival differences were identified between patients with T1N0 and T1N1 who received PMRT versus those who did not receive it (T1N0: OS HR, 0.916; 95% CI, 0.409 to 2.049; p=0.831; T1N1: OS HR, 1.114; 95% CI, 0.620 to 2.001; p=0.717). Histological grade is one of the fundamental features describing breast cancer. For patients with grade I/II, no statistically significant survival differences were identified between PMRT and no-PMRT patients (OS HR, 0.972; 95% CI, 0.664 to 1.424; p=0.885). While for patients with grade III, the PMRT patients demonstrated a better prognosis than the no-PMRT patients in terms of OS (HR, 0.622; 95% CI, 0.430 to 0.899; p=0.012). We conducted an exploratory analysis of age stratification. Our findings indicate that the subgroup analysis using a 50-year cut-off did not yield statistically significant results (all p > 0.05). Further subgroup analyses based on age were performed. Patients younger than 75 years old who underwent PMRT had a better survival outcome than patients without PMRT (OS HR, 0.671; 95% CI, 0.481 to 0.936; p=0.019). However, no obvious OS benefit was found in patients equal to or older than 75 years (OS HR, 1.006; 95% CI, 0.643 to 1.574; p=0.979). Given that the patients were diagnosed over a long period, from 1998 to 2016, we stratified the year of diagnosis into three periods: 1998-2003, 2004-2010, and 2011-2016, and performed subgroup analyses. The results suggest that, across these diagnostic periods, there was a consistent trend toward survival benefit in the PMRT group compared to the no-PMRT group, although the differences did not reach statistical significance.

4. The performance of the ANN model

For patients with T2N0 and T2N1, the stratified analysis indicated a survival benefit from PMRT. Next, we established an ANN model to precisely predict the potential benefit population of patients with T2N0 and T2N1 disease.

The ANN model was developed based on MLP, and the abbreviated diagram of the ANN model is presented in Fig. 3. In the training group, the AUC of the ANN model was 0.905 (S1 Fig.). In the validation group, the AUC value of the ANN model was 0.868 (Fig. 4A). The baseline characteristics of the patients in the training group and validation group are shown in S2 Table, there were no significant differences in the distribution of the general clinical data between the two groups (all p > 0.05). The feature importance plot for the MLP model is shown in S3 Fig. The results indicate that HER2 status, number of lymph node metastases, and age are the three most important variables in this model.

5. Cut-off value of the ANN model predicted probability

An ANN model was used to calculate the predicted probability of all patients, regardless of whether they received PMRT or not. The model was specifically applied to the no-PMRT group, resulting in predicted probabilities ranging from 0.14 to 0.85.

The optimal cut-off value of the ANN predicted probability is 0.51, based on X-tile analysis of overall survival. A predictive benefit classification system was constructed to separate participants into two levels based on the cut-off value for 0.51: (1) a participant was classified as the PMRT benefit population if the ANN model predicted probability was more than 0.51; (2) a participant with an ANN model predicted probability of 0.51 or less was classified as part of the PMRT no benefit population.

6. Validation of the cut-off value in the ANN predicted probability

For men with an ANN model predicted probability less than or equal to 0.51, no statistically significant survival differences were identified with the use of PMRT (p=0.157) (Fig. 4B). While for patients with an ANN model predicted probability greater than 0.51, usage of PMRT was associated with greatly improved OS (p=0.015) (Fig. 4C). The survival benefits from PMRT observed in the PMRT benefit population were significantly greater than those estimated from data of the PMRT no benefit population. Therefore, our findings suggest that PMRT should be strongly recommended for T2N0/T2N1 MBC patients with the ANN model predicted probability greater than 0.51.

Discussion

While PMRT has been shown to decrease recurrence and breast cancer mortality in patients with T1-2N1M0 FBC, leading to its strong recommendation in this group by the National Comprehensive Cancer Network guidelines, it is important to consider the differences between MBC and FBC [17,18]. MBC tends to be more aggressive, and patients have a poorer prognosis. This highlights the need not to solely rely on clinical evidence from treating FBC to guide treatment decisions for MBC [7]. In addition, incorrect treatment decisions may play a significant role in inferior survival outcomes among MBC patients [9]. Therefore, our study aims to reassess the role of PMRT in T1-2N0-1M0 MBC.

As presented in the previous study, MBC patients are more likely to undergo mastectomy and less likely to receive adjuvant RT [9]. Based on the data from the SEER database we found that only 17% of T1-2N0-1M0 MBC patients received PMRT; which may have been insufficient treatment. Previous studies showed that PMRT might contribute to a better outcome, but they also pointed out that this benefit tends to be seen in patients with local advanced breast cancer [9,19]. However, there have been no studies specifically aimed at the low- and middle-risk T1-2N0-1M0 populations. In our study, both before and after PSM, multivariate analysis showed that PMRT appeared to confer a significant benefit for the overall population. This finding emphasizes the importance of PMRT in these patients and warrants further in-depth analysis.

We found that over 70% of the patients who received PMRT for MBC in our study had node-positive disease. In contrast, only 30% of patients who did not receive PMRT had lymph node involvement. This may suggest that node status is a common factor influencing the use of PMRT in MBC in current clinical practice. However, we also found that tumor size may be more important than lymph node status in making the PMRT decision. In the subgroup analysis, we found that the significant benefit of PMRT was observed only in the T2 population, but not in N1 population. After combined analysis of the tumor size and lymph node status, a significant benefit was observed in T2N0M0 and T2N1M0 disease, but not in T1N0M0 and T1N1M0. Therefore, we suggest that the traditional opinion should be changed from focusing solely on lymph node status to paying equal attention to tumor size in MBC. Consequently, PMRT is more meaningful in reducing local recurrence in this group of patients. Despite not finding a significant benefit in the entire N1 category patient group, it does not mean that PMRT should be omitted in all N1 patients. Previous research by Matthew J. Abrams demonstrated that PMRT could improve OS in the N1 population [14]. Although their study was affected by baseline characteristic imbalances between the two groups, we effectively overcame this limitation through PSM. However, further evaluation of subgroups within N1 patients that may benefit from PMRT remains meaningful. Unfortunately, due to the insufficient number of registered patients, we were unable to perform effective analyses based solely on the number of lymph nodes. This is one of the limitations in our study.

Besides tumor size and lymph nodes, there are many other factors affecting the outcome of PMRT. One is histological grading. A higher grade always signifies more aggressive tumor features and a worse prognosis that may also relate to greater benefit from PMRT. As presented in our study, the patients with histological grade III had a significant improvement in OS with PMRT. Age may be another factor. Based on the age-stratified subgroup analysis, we observed that patients under 75 years of age derived a significant survival benefit from PMRT, while no statistically significant difference was observed in patients aged 75 years or older. We therefore speculate that, in addition to being associated with an increased risk of non-cancer mortality, age itself may serve as a predictive factor for the effectiveness of PMRT.

In order to identify the appropriate population for PMRT in T1-2N0-1M0 MBC patients more accurately, we built a predictive model based on the ANN method by analyzing the effects of ten variables including age, marital status, race, grades, pathology types, positive lymph node numbers, estrogen receptor status, progesterone receptor status, HER2 status and chemotherapy records on the prognosis of PMRT. Compared to traditional modeling methods, for example “nomograms,” “ANN” is based on artificial intelligence and can deal with multivariable data without requiring prior assumptions. Moreover, it could provide personalized measurements to better serve clinical practice [20]. Based on calculating the predicted value of benefit probability of each patient, we obtained 0.51 as the cut-off value by reliable means. After dividing the patients into two groups according to the cut-off value we determined, we found the group with a predicted value under 0.51 did not significantly benefit from PMRT. On the contrary, in the other group, a significantly better OS was confirmed after PMRT. Subsequently, validating with receiver operating characteristic curves both in the training set and validation set showed good prediction efficacy of our model. Therefore, we believe that our ANN model can effectively distinguish which population is suitable to receive PMRT, and thus serve as a valuable tool for clinical decision-making in this regard.

We analyzed one of the largest MBC samples from the SEER database ever used to evaluate the role of PMRT in T1-2N0-1M0 MBC. We used PSM to reduce selected bias as much as possible and proposed for the first time that we should pay attention to the impact of tumor size on PMRT decision-making. Next, we built a prediction model of PMRT in MBC for the first time and proved its accuracy by AUC. However, we must acknowledge that our study had several limitations. The incidence of MBC is relatively low, which resulted in small sample sizes in some subgroups, such as the ER-negative and HER2-positive groups, making it difficult to perform effective analysis. Therefore, our study’s findings may only represent the results of the ER-positive population. Similarly, the rarity of MBC poses a well-recognized challenge for obtaining a cohort of sufficient size to perform external validation and demonstrate model generalizability. We hope that with advancing international collaborations in medicine, future efforts will yield larger datasets, at which point we plan to refine and externally validate this model. Due to the limitations of the SEER database, disease-free survival and local recurrence rate were not available in this database, so we had to rely solely on OS as the endpoint. The employment of OS as the endpoint may introduce a certain degree of competing risk bias in the enrolled population of this study. Another limitation was the lack of detailed information about RT methods in the SEER database. Despite these shortcomings, we believe our research provides a new perspective on PMRT for MBC and can assist clinicians in their practice.

In our study, we found that PMRT improved survival in MBC patients with grade III, T2N0M0 and T2N1M0. Therefore, PMRT may be suitable for certain patients. To help identify the appropriate candidates for PMRT, we developed and validated an ANN model. This model can be used as a prediction tool by doctors to deliver more personalized and tailored RT for MBC patients.

Electronic Supplementary Material

Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).

CRT_25-518_S1_Fig.pdf

CRT_25-518_S2_Table.pdf

CRT_25-518_S3_Fig.pdf

NOTES

Ethical Statement

Considering SEER database is publicly available and does not require informed patient consent. So, we are exempt from Institutional Review Board approval. As this study is based on a publicly available database without identifying patient information, informed consent was not needed.

Author Contributions

Conceived and designed the analysis: Huang K (Kaiyan Huang), Yu Y, Li X, Liu Y, Zhang J.

Collected the data: Huang K (Kaiyan Huang), Yu Y, Li X, Liu Y, Zhang J.

Contributed data or analysis tools: Huang K (Kaiyan Huang), Yu Y, Li X, Liu Y, Zhang J.

Performed the analysis: Huang K (Kaiyan Huang), Yu Y, Li X, Liu Y, Huang K (Kailong Huang), Wang X, Zhang J.

Wrote the paper: Huang K (Kaiyan Huang), Yu Y.

Conflicts of Interest

Conflict of interest relevant to this article was not reported.

Funding

This research was funded by Fujian Science and Technology Innovation Joint Fund Project, Fujian Province (2019Y9054).

Fig. 1.

Comparison of overall survival between postmastectomy radiotherapy (PMRT) vs. no-PMRT in matched groups. HR, hazard ratio.

Fig. 2.

Kaplan-Meier survival curves of the effect of postmastectomy radiotherapy (PMRT) on overall survival (OS) stratified by T/N category (A-D), tumor grade (E, F), and age (G, H). CI, confidence interval; HR, hazard ratio.

Fig. 3.

Architecture of artificial neural network. ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; IDC, infiltrating duct carcinoma; PMRT, post-mastectomy radiotherapy; PR, progesterone receptor.

Fig. 4.

Receiver operating characteristic curve of the artificial neural network model (A). Kaplan-Meier curves for postmastectomy radiotherapy (PMRT) no benefit population (B) and PMRT benefit population (C) in T2N0/T2N1 male breast cancer patients.

Table 1.

Characteristics of pT1-2N0-1 male breast cancer patients before PSM and after PSM

Characteristic	Before PSM			p-value^a)	After PSM			p-value^a)
Characteristic	PMRT (n=384)	No-PMRT (n=1,863)	Total (n=2,247)	p-value^a)	PMRT (n=381)	No-PMRT (n=381)	Total (n=762)	p-value^a)
Age (yr)
< 50	42 (10.9)	210 (11.3)	252 (11.2)	0.850	42 (11.0)	38 (10.0)	80 (10.5)	0.636
≥ 50	342 (89.1)	1,653 (88.7)	1,995 (88.8)		339 (89.0)	343 (90.0)	682 (89.5)
Marital status
Married	272 (70.8)	1,383 (74.2)	1,655 (73.7)	0.168	269 (70.6)	270 (70.9)	539 (70.7)	0.937
Not married^b)	112 (29.2)	480 (25.8)	592 (26.3)		112 (29.4)	111 (29.1)	223 (29.3)
Year of diagnosis
1998-2003	83 (21.6)	330 (17.7)	413 (18.4)	0.177	77 (20.2)	83 (21.8)	160 (21.0)	0.075
2004-2010	131 (34.1)	691 (37.1)	822 (36.6)		159 (41.7)	129 (33.9)	288 (37.8)
2011-2016	170 (44.3)	842 (45.2)	1,012 (45.0)		145 (38.1)	169 (44.4)	314 (41.2)
Race
White	314 (81.7)	1,520 (81.5)	1,834 (81.6)	0.102	311 (81.6)	322 (84.5)	633 (83.1)	0.565
Black	54 (14.1)	219 (11.8)	273 (12.1)		54 (14.2)	45 (11.8)	99 (13.0)
Other^c)	16 (4.2)	124 (6.7)	140 (6.3)		16 (4.2)	14 (3.7)	30 (3.9)
Grade
I and II	165 (43.0)	597 (32.0)	762 (33.9)	< 0.001	164 (43.0)	168 (44.1)	332 (43.6)	0.770
III	219 (57.0)	1,266 (68.0)	1,485 (66.1)		217 (57.0)	213 (55.9)	430 (56.4)
Pathology
IDC	333 (86.7)	1,581 (84.9)	1,914 (85.2)	0.351	330 (86.6)	334 (87.7)	664 (87.1)	0.665
Other	51 (13.3)	282 (15.1)	333 (14.8)		51 (13.4)	47 (12.3)	98 (12.9)
AJCC stage
I	49 (12.8)	826 (44.3)	875 (38.9)	< 0.001	49 (12.9)	44 (11.5)	93 (12.2)	0.580
II	335 (87.2)	1,037 (55.7)	1,372 (61.1)		332 (87.1)	337 (88.5)	669 (87.8)
T category
T1	157 (40.9)	1,090 (58.5)	1,247 (55.5)	< 0.001	157 (41.2)	158 (41.5)	315 (41.3)	0.941
T2	227 (59.1)	773 (41.5)	1,000 (44.5)		224 (58.8)	223 (58.5)	447 (58.7)
N category
N0	110 (28.6)	1,303 (69.9)	1,413 (62.9)	< 0.001	110 (28.9)	105 (27.6)	215 (28.2)	0.687
N1	274 (71.4)	560 (30.1)	834 (37.1)		271 (71.1)	276 (72.4)	547 (71.8)
Axillary lymph node positive
0	110 (28.6)	1,304 (70.0)	1,414 (62.9)	< 0.001	110 (28.9)	106 (27.8)	216 (28.3)	0.772
1	156 (40.6)	377 (20.2)	533 (23.7)		156 (40.9)	156 (40.9)	312 (40.9)
2	74 (19.3)	120 (6.4)	194 (8.6)		73 (19.1)	68 (17.8)	141 (18.5)
3	44 (11.5)	62 (3.4)	106 (4.8)		42 (11.1)	51 (13.5)	93 (12.2)
ER status
Positive	373 (97.1)	1,813 (97.3)	2,186 (97.3)	0.843	370 (97.1)	373 (97.9)	743 (97.5)	0.486
Negative	11 (2.9)	50 (2.7)	61 (2.7)		11 (2.9)	8 (2.1)	19 (2.5)
PR status
Positive	341 (88.8)	1,652 (88.7)	1,993 (88.7)	0.943	338 (88.7)	328 (86.1)	666 (87.4)	0.275
Negative	43 (11.2)	211 (11.3)	254 (11.3)		43 (11.3)	53 (13.9)	96 (12.6)
HER2 status
Positive	19 (4.9)	112 (6.0)	131 (5.8)	0.537	19 (5.0)	14 (3.7)	33 (4.3)	0.215
Negative	159 (41.4)	801 (43.0)	960 (42.7)		159 (41.7)	141 (37.0)	300 (39.4)
Others	206 (53.7)	950 (51.0)	1,156 (51.5)		203 (53.3)	226 (59.3)	429 (56.3)
Breast subtypes
HR+/HER2–	157 (40.9)	789 (42.4)	946 (42.1)	0.694	157 (41.2)	141 (37.0)	298 (39.1)	0.104
HR+/HER2+	19 (4.9)	106 (5.7)	125 (5.6)		19 (5.0)	12 (3.1)	31 (4.1)
HR–/HER2+	0	6 (0.3)	6 (0.3)		0	2 (0.5)	2 (0.3)
TNBC	2 (0.5)	12 (0.6)	14 (0.6)		2 (0.5)	0	2 (0.3)
Unknown	206 (53.6)	950 (51.0)	1,156 (51.4)		203 (53.3)	226 (59.4)	429 (56.2)
Chemotherapy
No	142 (37.0)	1,301 (69.8)	1,443 (64.2)	< 0.001	142 (37.3)	156 (40.9)	298 (39.1)	0.299
Yes	242 (63.0)	562 (30.2)	804 (35.8)		239 (62.7)	225 (59.1)	464 (60.9)

Values are presented as number (%). AJCC, American Joint Committee on Cancer; CI, confi dence interval; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; HR, hormone receptor (ER and PR); IDC, infiltrating duct carcinoma; PMRT, postmastectomy radiotherapy; PR, progesterone receptor; PSM, propensity score matching; TNBC, triple-negative breast cancer.

^a) The p-value was calculated among all groups by the chi-square test, and p < 0.001 indicates statistical significance,

^b) “Non-married” includes divorced, separated, single (never married), unmarried or domestic partner, and widowed,

^c) “Other” includes American Indian/Alaskan native and Asian/Pacific Islander.

Table 2.

Univariate and multivariate Cox proportional hazard model of overall survival

Variable	Univariate analysis		Multivariate analysis
Variable	HR (95% CI)	p-value^a)	HR (95% CI)	p-value^a)
Age (yr)
< 50	Reference		Reference
≥ 50	2.588 (1.847-3.627)	< 0.001	2.379 (1.687-3.353)	< 0.001
Year of diagnosis
1998-2003	Reference
2004-2010	0.932 (0.773-1.125)	0.464
2011-2016	0.757 (0.567-1.011)	0.059
Marital status
Married	Reference		Reference
Not married^b)	1.588 (1.334-1.890)	< 0.001	1.503 (1.258-1.795)	< 0.001
Race
White	Reference		Reference
Black	1.192 (0.931-1.527)	0.164	1.183 (0.919-1.522)	0.192
Other^c)	0.645 (0.428-0.972)	0.036	0.641 (0.423-0.969)	0.035
Grade
I and II	Reference		Reference
III	1.425 (1.207-1.681)	< 0.001	1.287 (1.083-1.530)	0.004
Pathology
IDC	Reference
Other^c)	0.812 (0.636-1.036)	0.094
AJCC stage
I	Reference
II	1.953 (1.633-2.336)	< 0.001
T category
T1	Reference		Reference
T2	2.116 (1.796-2.493)	< 0.001	2.150 (1.655-2.794)	< 0.001
N category
N0	Reference
N1	1.535 (1.303-1.807)	< 0.001
Axillary lymph node positive
0	Reference		Reference
1	1.497 (1.240-1.808)	< 0.001	1.848 (1.443-2.368)	< 0.001
2	1.533 (1.163-2.019)	0.002	1.994 (1.434-2.773)	< 0.001
3	1.740 (1.243-2.436)	0.001	2.040 (1.400-2.972)	< 0.001
ER status
Positive	Reference
Negative	1.320 (0.877-1.988)	0.183
PR status
Positive	Reference
Negative	1.008 (0.798-1.272)	0.949
HER2 status
Positive	Reference
Negative	0.713 (0.411-1.237)	0.229
Others	0.984 (0.584-1.659)	0.953
Breast subtype
HR+/HER2–	Reference		Reference
HR+/HER2+	1.301 (0.723-2.341)	0.379	1.378 (0.763-2.489)	0.287
HR–/HER2+	5.084 (1.249-20.703)	0.023	3.042 (0.737-12.560)	0.124
TNBC	4.932 (1.555-15.648)	0.007	5.097 (1.590-16.341)	0.006
Unknown	1.419 (1.098-1.835)	0.008	1.506 (1.164-1.948)	0.002
Chemotherapy
No	Reference		Reference
Yes	0.652 (0.545-0.779)	< 0.001	0.499 (0.410-0.608)	< 0.001
PMRT status
No	Reference		Reference
Yes	0.967 (0.779-1.199)	0.759	0.784 (0.623-0.986)	0.038

AJCC, American Joint Committee on Cancer; CI, confidence interval; ER, estrogen receptor; IDC, Infiltrating duct carcinoma; HER2, human epidermal growth factor receptor 2; HR, hazard ratio; PMRT, postmastectomy radiotherapy; PR, progesterone receptor; TNBC, triple-negative breast cancer.

^a) p-value was adjusted by univariate and multivariate Cox proportional hazard regression model. p < 0.05 indicates statistical significance,

^b) “Non-married” includes divorced, separated, single (never married), unmarried or domestic partner and widowed,

^c) “Other” includes American Indian/Alaskan native and Asian/Pacific Islander.

Table 3.

Multivariate Cox proportional hazard regression model of overall survival for the PMRT and no-PMRT groups, stratified according to clinical variables after PSM

Variable	HR (95% CI)^a)	p-value^b)
Tumor status
T1	1.049 (0.651-1.691)	0.845
T2	0.668 (0.483-0.924)	0.015
Lymph node status
N0	0.673(0.416-1.053)	0.106
N1	0.822 (0.600-1.127)	0.224
Tumor and lymph node status
T1N0	0.916 (0.409-2.049)	0.831
T1N1	1.114 (0.620-2.001)	0.717
T2N0	0.482 (0.267-0.870)	0.015
T2N1	0.681 (0.464-0.998)	0.048
Grade
I and II	0.972 (0.664-1.424)	0.885
III	0.622 (0.430-0.899)	0.012
Age (yr)
< 50	0.442 (0.140-1.402)	0.166
≥ 50	0.799 (0.608-1.051)	0.109
Age (yr)
< 75	0.671 (0.481-0.936)	0.019
≥ 75	1.006 (0.643-1.574)	0.979
Year of diagnosis
1998-2003	0.659 (0.431-1.006)	0.053
2004-2010	0.717 (0.486-1.058)	0.094
2011-2016	0.872 (0.389-1.954)	0.739

CI, confi dence interval; HR, hazard ratio; PMRT, postmastectomy radiotherapy; PSM, perform propensity score matching.

^a) Using No-PMRT as a reference,

^b) Statistical significance.

REFERENCES

1. Ottini L. Male breast cancer: a rare disease that might uncover underlying pathways of breast cancer. Nat Rev Cancer. 2014;14:643.Article PubMed PDF
2. Speirs V, Shaaban AM. The rising incidence of male breast cancer. Breast Cancer Res Treat. 2009;115:429–30. Article PubMed PDF
3. DeSantis CE, Ma J, Gaudet MM, Newman LA, Miller KD, Goding Sauer A, et al. Breast cancer statistics, 2019. CA Cancer J Clin. 2019;69:438–51. Article PubMed PDF
4. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2021. CA Cancer J Clin. 2021;71:7–33. Article PubMed PDF
5. Rizzolo P, Silvestri V, Tommasi S, Pinto R, Danza K, Falchetti M, et al. Male breast cancer: genetics, epigenetics, and ethical aspects. Ann Oncol. 2013;24 Suppl 8:viii75–82. Article PubMed
6. Fentiman IS, Fourquet A, Hortobagyi GN. Male breast cancer. Lancet. 2006;367:595–604. Article PubMed
7. Sarmiento S, McColl M, Musavi L, Gani F, Canner JK, Jacobs L, et al. Male breast cancer: a closer look at patient and tumor characteristics and factors that affect survival using the National Cancer Database. Breast Cancer Res Treat. 2020;180:471–9. Article PubMed PDF
8. Korde LA, Zujewski JA, Kamin L, Giordano S, Domchek S, Anderson WF, et al. Multidisciplinary meeting on male breast cancer: summary and research recommendations. J Clin Oncol. 2010;28:2114–22. Article PubMed PMC
9. Weir J, Zhao YD, Herman T, Algan O. Clinicopathologic features and radiation therapy utilization in patients with male breast cancer: a national cancer database study. Breast Cancer (Auckl). 2018;12:1178223418770687.Article PubMed PMC PDF
10. Yu E, Suzuki H, Younus J, Elfiki T, Stitt L, Yau G, et al. The impact of post-mastectomy radiation therapy on male breast cancer patients: a case series. Int J Radiat Oncol Biol Phys. 2012;82:696–700. Article PubMed
11. Jardel P, Vignot S, Cutuli B, Creisson A, Vass S, Barranger E, et al. Should adjuvant radiation therapy be systematically proposed for male breast cancer? A systematic review. Anticancer Res. 2018;38:23–31. PubMed
12. Eggemann H, Ignatov A, Stabenow R, von Minckwitz G, Rohl FW, Hass P, et al. Male breast cancer: 20-year survival data for post-mastectomy radiotherapy. Breast Care (Basel). 2013;8:270–5. Article PubMed PMC PDF
13. Madden NA, Macdonald OK, Call JA, Schomas DA, Lee CM, Patel S. Radiotherapy and male breast cancer: a population-based registry analysis. Am J Clin Oncol. 2016;39:458–62. PubMed
14. Abrams MJ, Koffer PP, Wazer DE, Hepel JT. Postmastectomy radiation therapy is associated with improved survival in node-positive male breast cancer: a population analysis. Int J Radiat Oncol Biol Phys. 2017;98:384–91. Article PubMed
15. Miao H, Verkooijen HM, Chia K, Bouchardy C, Pukkala E, Laronningen S, et al. Incidence and outcome of male breast cancer: an international population-based study. J Clin Oncol. 2011;29:4381–6. Article PubMed
16. Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res. 2004;10:7252–9. PubMed
17. McGale P, Taylor C, Correa C, Cutter D, Duane F, Ewertz M, et al. Effect of radiotherapy after mastectomy and axillary surgery on 10-year recurrence and 20-year breast cancer mortality: meta-analysis of individual patient data for 8135 women in 22 randomised trials. Lancet. 2014;383:2127–35. Article PubMed PMC
18. NCCN guideline: breast cancer, version 1.2025 [Internet]. National Comprehensive Cancer Network; 2025 [cited 2025 Jan 23]. Available from: http://www.nccn.org
19. Forster T, Kohler C, El Shafie R, Weykamp F, Konig L, Arians N, et al. Adjuvant radiation therapy for male breast cancer: a rare indication? Cancers (Basel). 2020;12:3645.Article PubMed PMC
20. Lancashire LJ, Lemetre C, Ball GR. An introduction to artificial neural networks in bioinformatics: application to complex microarray and mass spectrometry datasets in cancer studies. Brief Bioinform. 2009;10:315–29. Article PubMed

Figure & Data

REFERENCES

Citations

Citations to this article as recorded by

Case Report: long-term survival of a male patient with breast cancer complicated by lung adenocarcinoma treated with individualized therapy
Yadong Liu, Xuejuan Duan, Zhenru Zheng, Gang Cheng, Xianbo Zhang, Jing Zhao
Frontiers in Oncology.2026;[Epub] CrossRef
A machine learning-based prognostic model for de novo metastatic HR-positive breast cancer: SEER cohort with external validation
Sihang Lin, Wanwan Wang, Lixia Liu, Jiayu Guan, Chuanrong Cen, Huawei Yang
Frontiers in Oncology.2026;[Epub] CrossRef

PubReader
ePub Link

Cite

CITE: export Copy Download Format; Close

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

RIS — For EndNote, ProCite, RefWorks, and most other reference management software
BibTeX — For JabRef, BibDesk, and other BibTeX-specific software

Include:

Citation for the content below
Citation and abstract for the content below

The Impact of Post-mastectomy Radiotherapy on T1-2N0-1 Male Breast Cancer and Establishment of an Artificial Neural Network Predicting Model: Population-Based Study

DOI: https://doi.org/10.4143/crt.2025.518

XML Download

Abstract
Introduction
Materials and Methods
Results
Discussion
Electronic Supplementary Material
NOTES
REFERENCES

The Impact of Post-mastectomy Radiotherapy on T1-2N0-1 Male Breast Cancer and Establishment of an Artificial Neural Network Predicting Model: Population-Based Study

Fig. 1. Comparison of overall survival between postmastectomy radiotherapy (PMRT) vs. no-PMRT in matched groups. HR, hazard ratio.

Fig. 2. Kaplan-Meier survival curves of the effect of postmastectomy radiotherapy (PMRT) on overall survival (OS) stratified by T/N category (A-D), tumor grade (E, F), and age (G, H). CI, confidence interval; HR, hazard ratio.

Fig. 3. Architecture of artificial neural network. ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; IDC, infiltrating duct carcinoma; PMRT, post-mastectomy radiotherapy; PR, progesterone receptor.

Fig. 4. Receiver operating characteristic curve of the artificial neural network model (A). Kaplan-Meier curves for postmastectomy radiotherapy (PMRT) no benefit population (B) and PMRT benefit population (C) in T2N0/T2N1 male breast cancer patients.

Fig. 1.

Fig. 2.

Fig. 3.

Fig. 4.

The Impact of Post-mastectomy Radiotherapy on T1-2N0-1 Male Breast Cancer and Establishment of an Artificial Neural Network Predicting Model: Population-Based Study

Characteristic	Before PSM			p-value^a)	After PSM			p-value^a)
Characteristic	PMRT (n=384)	No-PMRT (n=1,863)	Total (n=2,247)	p-value^a)	PMRT (n=381)	No-PMRT (n=381)	Total (n=762)	p-value^a)
Age (yr)
< 50	42 (10.9)	210 (11.3)	252 (11.2)	0.850	42 (11.0)	38 (10.0)	80 (10.5)	0.636
≥ 50	342 (89.1)	1,653 (88.7)	1,995 (88.8)		339 (89.0)	343 (90.0)	682 (89.5)
Marital status
Married	272 (70.8)	1,383 (74.2)	1,655 (73.7)	0.168	269 (70.6)	270 (70.9)	539 (70.7)	0.937
Not married^b)	112 (29.2)	480 (25.8)	592 (26.3)		112 (29.4)	111 (29.1)	223 (29.3)
Year of diagnosis
1998-2003	83 (21.6)	330 (17.7)	413 (18.4)	0.177	77 (20.2)	83 (21.8)	160 (21.0)	0.075
2004-2010	131 (34.1)	691 (37.1)	822 (36.6)		159 (41.7)	129 (33.9)	288 (37.8)
2011-2016	170 (44.3)	842 (45.2)	1,012 (45.0)		145 (38.1)	169 (44.4)	314 (41.2)
Race
White	314 (81.7)	1,520 (81.5)	1,834 (81.6)	0.102	311 (81.6)	322 (84.5)	633 (83.1)	0.565
Black	54 (14.1)	219 (11.8)	273 (12.1)		54 (14.2)	45 (11.8)	99 (13.0)
Other^c)	16 (4.2)	124 (6.7)	140 (6.3)		16 (4.2)	14 (3.7)	30 (3.9)
Grade
I and II	165 (43.0)	597 (32.0)	762 (33.9)	< 0.001	164 (43.0)	168 (44.1)	332 (43.6)	0.770
III	219 (57.0)	1,266 (68.0)	1,485 (66.1)		217 (57.0)	213 (55.9)	430 (56.4)
Pathology
IDC	333 (86.7)	1,581 (84.9)	1,914 (85.2)	0.351	330 (86.6)	334 (87.7)	664 (87.1)	0.665
Other	51 (13.3)	282 (15.1)	333 (14.8)		51 (13.4)	47 (12.3)	98 (12.9)
AJCC stage
I	49 (12.8)	826 (44.3)	875 (38.9)	< 0.001	49 (12.9)	44 (11.5)	93 (12.2)	0.580
II	335 (87.2)	1,037 (55.7)	1,372 (61.1)		332 (87.1)	337 (88.5)	669 (87.8)
T category
T1	157 (40.9)	1,090 (58.5)	1,247 (55.5)	< 0.001	157 (41.2)	158 (41.5)	315 (41.3)	0.941
T2	227 (59.1)	773 (41.5)	1,000 (44.5)		224 (58.8)	223 (58.5)	447 (58.7)
N category
N0	110 (28.6)	1,303 (69.9)	1,413 (62.9)	< 0.001	110 (28.9)	105 (27.6)	215 (28.2)	0.687
N1	274 (71.4)	560 (30.1)	834 (37.1)		271 (71.1)	276 (72.4)	547 (71.8)
Axillary lymph node positive
0	110 (28.6)	1,304 (70.0)	1,414 (62.9)	< 0.001	110 (28.9)	106 (27.8)	216 (28.3)	0.772
1	156 (40.6)	377 (20.2)	533 (23.7)		156 (40.9)	156 (40.9)	312 (40.9)
2	74 (19.3)	120 (6.4)	194 (8.6)		73 (19.1)	68 (17.8)	141 (18.5)
3	44 (11.5)	62 (3.4)	106 (4.8)		42 (11.1)	51 (13.5)	93 (12.2)
ER status
Positive	373 (97.1)	1,813 (97.3)	2,186 (97.3)	0.843	370 (97.1)	373 (97.9)	743 (97.5)	0.486
Negative	11 (2.9)	50 (2.7)	61 (2.7)		11 (2.9)	8 (2.1)	19 (2.5)
PR status
Positive	341 (88.8)	1,652 (88.7)	1,993 (88.7)	0.943	338 (88.7)	328 (86.1)	666 (87.4)	0.275
Negative	43 (11.2)	211 (11.3)	254 (11.3)		43 (11.3)	53 (13.9)	96 (12.6)
HER2 status
Positive	19 (4.9)	112 (6.0)	131 (5.8)	0.537	19 (5.0)	14 (3.7)	33 (4.3)	0.215
Negative	159 (41.4)	801 (43.0)	960 (42.7)		159 (41.7)	141 (37.0)	300 (39.4)
Others	206 (53.7)	950 (51.0)	1,156 (51.5)		203 (53.3)	226 (59.3)	429 (56.3)
Breast subtypes
HR+/HER2–	157 (40.9)	789 (42.4)	946 (42.1)	0.694	157 (41.2)	141 (37.0)	298 (39.1)	0.104
HR+/HER2+	19 (4.9)	106 (5.7)	125 (5.6)		19 (5.0)	12 (3.1)	31 (4.1)
HR–/HER2+	0	6 (0.3)	6 (0.3)		0	2 (0.5)	2 (0.3)
TNBC	2 (0.5)	12 (0.6)	14 (0.6)		2 (0.5)	0	2 (0.3)
Unknown	206 (53.6)	950 (51.0)	1,156 (51.4)		203 (53.3)	226 (59.4)	429 (56.2)
Chemotherapy
No	142 (37.0)	1,301 (69.8)	1,443 (64.2)	< 0.001	142 (37.3)	156 (40.9)	298 (39.1)	0.299
Yes	242 (63.0)	562 (30.2)	804 (35.8)		239 (62.7)	225 (59.1)	464 (60.9)

Variable	Univariate analysis		Multivariate analysis
Variable	HR (95% CI)	p-value^a)	HR (95% CI)	p-value^a)
Age (yr)
< 50	Reference		Reference
≥ 50	2.588 (1.847-3.627)	< 0.001	2.379 (1.687-3.353)	< 0.001
Year of diagnosis
1998-2003	Reference
2004-2010	0.932 (0.773-1.125)	0.464
2011-2016	0.757 (0.567-1.011)	0.059
Marital status
Married	Reference		Reference
Not married^b)	1.588 (1.334-1.890)	< 0.001	1.503 (1.258-1.795)	< 0.001
Race
White	Reference		Reference
Black	1.192 (0.931-1.527)	0.164	1.183 (0.919-1.522)	0.192
Other^c)	0.645 (0.428-0.972)	0.036	0.641 (0.423-0.969)	0.035
Grade
I and II	Reference		Reference
III	1.425 (1.207-1.681)	< 0.001	1.287 (1.083-1.530)	0.004
Pathology
IDC	Reference
Other^c)	0.812 (0.636-1.036)	0.094
AJCC stage
I	Reference
II	1.953 (1.633-2.336)	< 0.001
T category
T1	Reference		Reference
T2	2.116 (1.796-2.493)	< 0.001	2.150 (1.655-2.794)	< 0.001
N category
N0	Reference
N1	1.535 (1.303-1.807)	< 0.001
Axillary lymph node positive
0	Reference		Reference
1	1.497 (1.240-1.808)	< 0.001	1.848 (1.443-2.368)	< 0.001
2	1.533 (1.163-2.019)	0.002	1.994 (1.434-2.773)	< 0.001
3	1.740 (1.243-2.436)	0.001	2.040 (1.400-2.972)	< 0.001
ER status
Positive	Reference
Negative	1.320 (0.877-1.988)	0.183
PR status
Positive	Reference
Negative	1.008 (0.798-1.272)	0.949
HER2 status
Positive	Reference
Negative	0.713 (0.411-1.237)	0.229
Others	0.984 (0.584-1.659)	0.953
Breast subtype
HR+/HER2–	Reference		Reference
HR+/HER2+	1.301 (0.723-2.341)	0.379	1.378 (0.763-2.489)	0.287
HR–/HER2+	5.084 (1.249-20.703)	0.023	3.042 (0.737-12.560)	0.124
TNBC	4.932 (1.555-15.648)	0.007	5.097 (1.590-16.341)	0.006
Unknown	1.419 (1.098-1.835)	0.008	1.506 (1.164-1.948)	0.002
Chemotherapy
No	Reference		Reference
Yes	0.652 (0.545-0.779)	< 0.001	0.499 (0.410-0.608)	< 0.001
PMRT status
No	Reference		Reference
Yes	0.967 (0.779-1.199)	0.759	0.784 (0.623-0.986)	0.038

Variable	HR (95% CI)^a)	p-value^b)
Tumor status
T1	1.049 (0.651-1.691)	0.845
T2	0.668 (0.483-0.924)	0.015
Lymph node status
N0	0.673(0.416-1.053)	0.106
N1	0.822 (0.600-1.127)	0.224
Tumor and lymph node status
T1N0	0.916 (0.409-2.049)	0.831
T1N1	1.114 (0.620-2.001)	0.717
T2N0	0.482 (0.267-0.870)	0.015
T2N1	0.681 (0.464-0.998)	0.048
Grade
I and II	0.972 (0.664-1.424)	0.885
III	0.622 (0.430-0.899)	0.012
Age (yr)
< 50	0.442 (0.140-1.402)	0.166
≥ 50	0.799 (0.608-1.051)	0.109
Age (yr)
< 75	0.671 (0.481-0.936)	0.019
≥ 75	1.006 (0.643-1.574)	0.979
Year of diagnosis
1998-2003	0.659 (0.431-1.006)	0.053
2004-2010	0.717 (0.486-1.058)	0.094
2011-2016	0.872 (0.389-1.954)	0.739

Table 1. Characteristics of pT1-2N0-1 male breast cancer patients before PSM and after PSM

The p-value was calculated among all groups by the chi-square test, and p < 0.001 indicates statistical significance,

“Non-married” includes divorced, separated, single (never married), unmarried or domestic partner, and widowed,

“Other” includes American Indian/Alaskan native and Asian/Pacific Islander.

Table 2. Univariate and multivariate Cox proportional hazard model of overall survival

p-value was adjusted by univariate and multivariate Cox proportional hazard regression model. p < 0.05 indicates statistical significance,

“Non-married” includes divorced, separated, single (never married), unmarried or domestic partner and widowed,

“Other” includes American Indian/Alaskan native and Asian/Pacific Islander.

Table 3. Multivariate Cox proportional hazard regression model of overall survival for the PMRT and no-PMRT groups, stratified according to clinical variables after PSM

CI, confi dence interval; HR, hazard ratio; PMRT, postmastectomy radiotherapy; PSM, perform propensity score matching.