Neuroendocrine Carcinoma of the Breast is More Aggressive Than Breast Invasive Ductal Carcinoma: A Propensity Score Matching Study Based on SEER Database

Background Neuroendocrine carcinoma (NEC) of the breast is a very rare malignant tumor whose clinic pathological features and prognosis are not well dened. The aim of this study was to compare the clinic pathological features and clinical outcome between breast neuroendocrine carcinoma and breast invasive ductal carcinoma (IDC) and to know the independent prognostic risk factors of NEC. Method Patients with breast NEC and breast IDC were identied through the Surveillance, Epidemiology, and End Results (SEER) database from 2010 to 2015. Then we compared the clinic pathological features and outcome between the two cohorts. Propensity score matching (PSM) was performed to balance the effects of baseline clinic pathological differences. Potential prognostic factors of breast NEC were identied by the univariate and multivariate Cox analysis. Results In total, we identied 208 patients with breast NEC and 275,878 cases with breast IDC. Compared with IDC, NEC was signicantly correlated with higher stage, larger tumor size, more distant transfers and no population of HER2 enriched subtype. There were signicant differences on the choice of treatment therapy between NEC and IDC (p=0.000). More patients chose not to undergo surgery (35.1% vs. 9.0%), and fewer patients chose breast conserving surgery than IDC patients (32.2% vs. 51.1%). Compared with IDC patients, NEC patients were more likely to choose chemotherapy (p=0.039) and more conservative in radiotherapy (p=0.000). The 66-month overall survival (OS) rate in NEC and IDC was 44.5% and 80.2% before PSM and was 49.8% and 71.4% after PSM, respectively. The 66-month breast cancer specic survival (BCSS) rate in NEC and IDC was 56.3% and 88.8% before PSM and was 54.1% and 80.5% after PSM, respectively. Finally, we identied independent prognostic factors, age, stage, pathological type, distant metastasis, chemotherapy, and molecular typing.

or had no signi cant differences between the two cohorts 6 . But some studies showed that compared with IDC, NEC was a more aggressive tumor and had worse prognosis [7][8][9][10][11] . Data regarding the prognostic signi cance of NEC are con icting in several studies [12][13] .
The treatment of NEC is consistent with that of invasive breast cancer. Clinicians knew little about NEC and usually treated them according to IDC. We took advantage of a large database of cancer cases from surveillance, Epidemiology, and end results (SEER) registries. The aim of this study was to compare the clinic pathological features and clinical outcome between breast neuroendocrine carcinoma and the most common breast invasive ductal carcinoma (IDC) and to know the independent prognostic risk factors of NEC using SEER database.

Patients
The Surveillance, Epidemiology, and End Results (SEER) database from 1973 to 2015 was used to collect patients' clinic pathological features and survival data. Since the HER2 information was available after

Clinic pathological characteristics
To investigate the clinic pathological characteristics of NEC and IDC of the breast, the following information was obtained: age, sex, race, grade, laterality, pathological type, AJCC stage, tumor size, nodal stage, metastasis status, breast molecular subtype, surgery treatment with either mastectomy or breast-conserving surgery, radiation therapy, chemotherapy, marital status at diagnosis, cause of death, and survival (months). There were only two patients with HR+/HER2+ left when we talk about independent prognostic factors after screening out unknown factors. And there were no population of type HR-/HER2+ in the collected cases, so we categorized the patients into two subtypes later: HR+ and triple negative breast cancer (TNBC).

Statistical analyses
Clinic pathological characteristics were compared between NEC and IDC by Pearson chi-square and Fisher's exact probability tests. Kaplan-Meier plots and log-rank tests were performed to compare OS and BCS between the two cohorts. Propensity score matching was conducted to calibrate the effects of the baseline of clinic pathological differences. The univariate and multivariate Cox proportional hazard model was applied for identifying prognostic factors. All the statistical analyses and graphics were performed with the SPSS statistical software, version 22.0 (IBM Corp, Armonk, NY).

Patient characteristics
In total, we identi ed 208 patients with breast NEC and 275,878 cases with breast IDC from SEER database between the years 2010 to 2015 in our study. The clinic pathological characteristics of the 208 NEC patients were compared with those of IDC, and the results are summarized in Table 1 (Table 3). Patients who are ≥60 years old, with high AJCC stage, with distant metastases, with special pathological type (large cell neuroendocrine carcinoma and small cell carcinoma), with the molecular typing of TNBC and without chemotherapy had worse clinical outcome. We made a monogram for OS in breast NEC (Figure 3, c-index=0.835).

Therapeutic Modalities
As is shown in Table 4, chemotherapy was the independent risk factor for the clinical outcome of NEC while surgery mode and radiation made no in uences. Considering the close relationship between surgery and radiotherapy, we made a comparison of interactive factors (Table 4). It seemed that patients with the therapeutic modalities of mastectomy + no radiation and breast-conserving surgery + radiation had better OS and BCSS. Considering that the choice of operation mode is closely related to the stage TNM, age, subtype and chemotherapy of patients, we adjusted hazard ratio in order to exclude the in uence of these factors. The therapeutic modalities of surgery mode and radiation had no effect on the prognosis of NEC after adjusted. been reported [14][15][16][17][18][19][20] . Finally, 208 patients with breast NEC were identi ed in the present study.

Discussion
In our study, we found NEC has distinct clinic pathological features. Compared with IDC, NEC had higher stage, larger tumor size and more distant transfers. We found that the majority of NEC patients were found at a higher stage, which may be due to the lack of speci c imaging and clinical manifestations 21 .
The mean age of NEC patients was 63 years old in our study, which was consistent with the information that most patients are postmenopausal in other studies [22][23][24][25] . There were only 4 patients with HER2 + in our study, and most of patients were HR + / HER2 -. The results of subtype were consistent with previous studies 6, [26][27] . The 66-month overall survival (OS) rate in NEC was 44.5% and 56.3%, which was worse than IDC, respectively. The poor clinic outcome may be due to its aggressive clinic pathological features. In order to eliminate the impact of the baseline, we performed a 1:4 (NEC/IDC) propensity score matching (PSM) analysis. After PSM, there were still signi cant differences on OS and BCSS of NEC and IDC.
This poor clinical outcome was similar to previous reports but also contrary to the results of some articles. Rovera et al. 5  Richter-Ehrenstein's article, mastectomy is more suitable for NEC than breast conserving surgery due to its early invasion potential. More studies should be performed to analyze the surgical procedures and radiotherapy.
The present study had several limitations. Recurrence free survival rate is an important information to evaluate the invasiveness of a tumor. Due to a lack of recurrence information in SEER database, our study didn't analyze the recurrence free survival rate. Also, as in shown in previous, Ki67 may in uence the clinic outcome and is an important clinic pathological feature, which has not been recorded in SEER database. In addition, some bias may occur due to the small sample size. For example, there were only 2 male patients included in the prognostic factor analysis. The SEER database also lacks the information of endocrine therapy and the speci c regimen of chemotherapy. According to the literature data 26 , cisplatin and etoposide are used in small cell carcinoma, which is different from the chemotherapy regimen of non-small cell carcinoma. Therefore, a further expanded study is warranted to verify our ndings.

Conclusion
We showed that breast NEC has distinct clinic pathological features and it is a rare tumor with stronger invasion and signi cantly worse clinic outcome compared with breast IDC. The independent risk factors affecting the prognosis of NEC were age, pathological type, AJCC stage, M, chemotherapy and molecular typing. Surgery mode and radiation therapy made no in uence on the clinic outcome of NEC.