Clinicopathologic characteristics and prognosis for male breast cancer compared to female breast cancer

Male breast cancer (MBC) is rare. Due to limited information, MBC has always been understudied. We conducted a retrospective population-based cohort study using data from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program. The clinical and biological features of female breast cancer (FBC) patients were compared with MBC patients. Cox regression models and competing risks analyses were used to identify risk factors associated with cancer-related survival in MBC and FBC groups. Results showed that MBC patients suffered from higher TNM stages, tumor grades, and a higher percentage of hormone receptor-positive tumors, compared with FBC patients (all p < 0.05). In addition, the breast tumor locations varied a lot between males and females (p < 0.05). FBC patients were associated with superior overall survival than MBC patients. Results from multivariate cox regression and competing risks analyses showed age, race, T, N, M-stages, tumor grades, estrogen receptor (ER)/progesterone receptor (PR) and human epidermal growth factor receptor-2 (HER-2) overexpression were independent prognosis factors in FBC patients (all p < 0.05). MBC patients had similar risk factors to FBC patients, but PR and HER-2 status did not independently influence survival (all p > 0.05). Tumor location was an independent prognostic factor for both gender groups.

In the current study, we attempt to compare the clinicopathologic characteristics and prognosis between MBC patients and FBC patients by drawing data from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Database from the beginning of 2010 to the end of 2014, with the aim of better understanding gender differences and specificity of MBC.

Materials and methods
This is a retrospective cohort study of breast cancer patients diagnosed in the SEER database 8.3.4 from the beginning of 2010 to the end of 2014. SEER collects cancer incidence data from population-based cancer registries covering approximately 34.6% of the U.S. population. It also records data of patients' clinicopathological characteristics, and vital status during follow-up. Breast cancer cases were identified according to the 3rd edition of the International Classification of Diseases for Oncology (ICD-O-3).
A total of 313,504 patients with breast cancer were identified in the SEER database. Patients were excluded if they had 1) survival month that was 0 or unknown; 2) T0 local disease diagnosis; 3) other malignant tumors. 4) missing information for demographic and tumor characteristics including sex, age, laterality (left, right, bilateral), tumor location, race (white, black, and other), pathological type (ductal, lobular, and other), TNM stage, histological grade (well-differentiated, moderately differentiated, poorly-differentiated and undifferentiated), surgical treatment of breast cancer, estrogen receptor (ER), progesterone receptor (PR), and HER-2. A total of 169,278 patients (1,123 males and 168,155 females) remained in the final analysis ( Fig. 1). Surgical treatment of breast cancer was defined as patients who received any type of surgical resection of the primary tumor. Histology was classified into three subtypes: ductal, lobular and others including mucinous adenocarcinoma, non-small cell carcinoma, adenocarcinoma and other rare types of cancer. All patients have given prior informed consent to being registered in SEER database. Statistical analysis. All statistical analyses were performed using SAS statistical software (version 9.4). The characteristics of the subjects with normal distribution were expressed as mean ± standard deviation and compared using t test. Categorical variables were represented as absolute value with percentage and the Chi-square test was used for comparison between male and female patients. Kaplan-Meier survival curves were generated to compare differences in survival probabilities over time between groups, and the equality of these curves was tested using a log-rank statistic.
The interval from the date of cancer diagnosis to the endpoint was calculated as survival time (in months). The endpoint was defined as one of the three events, whichever occurred first: date of breast cancer-related death, date of non-breast cancer related death, or the date used as the cutoff for the study. Cox regression models were generated to describe the relationship between clinicopathologic features and risk of breast cancer-related death among MBC patients and FBC patients.
Non-breast cancer related death may occur before the occurrence of breast cancer-related death during the follow-up period which hinders us from identifying the existence of breast cancer cases. The traditional multivariate COX regression model may markedly overestimate the risk of breast cancer 14 . To avoid overestimation and to improve accuracy, the cause-specific hazard model (CS model) and sub-distribution hazard function model (SD model) were used to calculate the absolute risk of breast cancer-related death. A two-sided p-value < 0.05 was considered statistically significant in this study.
Ethics approval and consent to participate. The study was approved by the Ethics Committee of Aerospace Center Hospital and was complied with the Declaration of Helsinki.   Among MBC patients, there were 47.91% of tumors located with the central portion, followed by 15.58% in the upper-outer quadrant, 6.14% in the nipple, 5.16% in the lower-outer quadrant, 4.27% in the upper-inner quadrant, 2.23% in the lower-inner quadrant, 0.09% in the axillary tail, and the remaining 18.61% were classified as overlapping lesion. However, in FBC patients, there were only 5.20% of tumors located in the central portion, and tumors primarily located in the upper-outer accounted for 39.16%. The rest were located in the upper-inner quadrant (14.06%), lower-inner quadrant (6.38%), lower-outer quadrant (8.44%), nipple (0.38%), axillary tail (0.53%), and overlapping lesion (25.86%).

The association of clinicopathologic characteristics with the cancer-related death risk. Dur-
ing the median follow-up of 57 (43-74) months, 116 male patients and 13,140 female patients died from breast cancer. The Kaplan-Meier method showed that the MBC patients had a worse overall survival (OS) than FBC patients. The log-rank test showed a significant difference in the OS between the two groups (log-rank, p < 0.001, Fig. 2).
During the follow up, a total of 9,255 non-breast cancer-related death cases were identified before the occurrence of breast cancer-related death. Tables 3, 4 summarizes adjusted HRs (95%CI) for the association of clinicopathological characteristics with breast cancer related death after taking competing risk events (none breast cancer-related death) into consideration. The associations of clinicopathological characteristics with breast cancer-related death were attenuated but remained significant both in the CS models and the SD models.

Discussion
This large-scale population-based study, which makes comparisons between MBC and FBC patients, provides intriguing etiologic and prognostic clues to this disease. Several significant conclusions were made. First, MBC patients have a worse prognosis than FBC patients. Second, there were differences in independent prognostic factors between MBC and FBC patients: PR and HER-2 were independent prognostic factors for FBC but not MBC patients. Finally, breast tumor locations between the two genders were different, which might have an important influence on prognostic results.
In our analysis, MBC patients had a worse overall prognosis than FBC counterparties which was in line with several previous studies. Nahleh et al. found that FBC patients had a significantly longer OS than MBC patients. The median OS for MBC patients was 7.0 years compared with 9.8 years for FBC patients (log-rank test; p < 0.05) 9 . Similarly, a study including 2,537 MBC patients also demonstrated that MBC patients had a relatively shorter 5-year survival rate than FBC patients 4 . Several explanations may help to explain this phenomenon. Better prognosis in FBC patients is partly due to the introduction of screening, public awareness, diagnosis at an earlier age with fewer complications, advances in treatment, and standardization of treatment regimens in international guidelines. However, the situation in MBC patients differs a lot compared with female counterparts. First, the breast tissue in men is sparser, and a small tumor would be able to invade the breast skin rapidly. The  10 . Third, the use of adjuvant therapy in MBC patients is not widespread as FBC patients. In a paper that included 10,173 men with HR-positive breast cancer, men were less likely to receive adjuvant endocrine therapy than women (67.3% vs 78.9%, p < 0.001) 16 .
Reliable and widespread use of adjuvant chemotherapy and radiotherapy for men is also lacking [17][18][19][20] . However, among the patients who were treated with surgery in this current study, male patients were more likely to receive mastectomy than tumorectomy (88.77% vs. 11.23%) when compared to female patients (37.84% vs. 62.16%, p < 0.001). The generally higher rate of mastectomies in men could explain why radiotherapy is less often used. There may be different risk factors between FBC and MBC especially when it is related to PR and HER-2. A population-based study indicated PR status did not appear to independently influence survival among MBC patients 4 . Matthew J's study also demonstrated PR status did not affect survival in MBC patients 21 . This may be related to the fact that PR status is not a crucial factor for endocrine therapy, and MBC patients are not as sensitive towards endocrine therapy. Little research has focused on HER-2 expression in men. The effectiveness of trastuzumab in HER-2 overexpressing MBC is unproven 22 . In addition, MBC patients with HER-2 overexpression only comprise a small portion of all MBC patients 23,24 , making it difficult to draw a reliable conclusion.
In this current study, the tumor locations between the MBC and FBC patients were markedly different. Among FBC patients, tumors were primarily located in the upper-outer and accounted for 39.16% while other sites in the breast were discovered at lower frequencies, which is in alliance with previous studies [25][26][27][28] . This basic observation of asymmetric occurrence of breast cancer has become well accepted but lacks an adequate scientific explanation. A possible explanation is that the upper-outer quadrant of the breast contains a greater proportion of the epithelial tissue, which has a greater chance to occur cancer 29 . In MBC patients, the central position (nipple and central portion) is dominant which accounted for 54.05%. The upper-outer quadrant only made up 15.58%, which was far below the central position. This discrepancy may be caused by the anatomy of the male breast, as there is a larger volume of epithelial breast tissue in the central portion in men 29 . In addition, the prognostic role of the tumor location is also underappreciated, as almost all breast cancer guidelines do not include tumor location as a prognostic factor 30,31 . Yet in our study, tumor location affected the prognosis for both MBC and FBC patients; tumors situated in the medial quadrants of the breast have a worse prognosis compared with those located in lateral quadrants. This finding was compatible with other papers. David K et. al suggested that medial tumor location adversely impacts breast cancer-specific survival and OS in breast cancer patients 32 . Similarly, the Caroline trials indicated that medial location was associated with a 50% excess risk of systemic relapse and breast cancer death compared with lateral tumors 33 .
The poor prognosis of tumors with the internal location may be associated with internal mammary nodes (IMN), which were not conventionally treated. Findings from a previous study have found occult nodal metastases in the internal mammary chain is more likely to be found in tumors with a central or medial location and female breast cancer patients with metastatic axillary nodes 34 . The tumor cell is usually clinically silent in the internal mammary chain, and it might disseminate the disease, especially in node-negative women, who did  35 . It has also been observed that 5% of breast cancer patients metastasis to IMNs alone 34 .
Competing risks are common in epidemiological research 35,36 . In the current study, cancer-unrelated death occurred before the identification of cancer-related death. In this competing risk setting, traditional Cox regression may overestimate the absolute risk of cancer-related death, because individuals with a competing (and therefore censored) event are considered to be likely to experience events of interest in the future. The CS model Table 2. Multivariate analysis for breast-cancer specific survival stratified by sex in Cox regressions. ER estrogen receptor, PR progesterone receptor, HER-2 human epidermal growth factor receptor 2, HR hazard ratio, CI confidence intervals. The definitions of T, N, M were referred to pathologic stage groups (pTNM).  14 . Future research should be conducted to explore methodological differences and expand tools to understand competitive risk methods for epidemiological data. The strength of this current study is the large quantity of data regarding MBC and FBC patients, which allows for a reliable extrapolation of the results. Furthermore, we analyzed the tumor location within the breast, which has been rarely focused on. Moreover, competing risk regressions were further used to validate our results, increasing the accuracy of the study. However, the limitations should also be acknowledged. The main limitation Table 4. Multivariate analysis for breast-cancer specific survival stratified by sex in the SD models. ER estrogen receptor, PR progesterone receptor, HER-2 human epidermal growth factor receptor 2, HR hazard ratio, CI confidence intervals, SD model sub-distribution hazard function model. The definitions of T, N, M were referred to pathologic stage groups (pTNM).