Patterns of Distant Metastases in Patients With Triple-Negative Breast Cancer –– A Population-Based Study

Currently, the prognosis of triple-negative breast cancer (TNBC) patients remained poor mainly due to resistance, recurrence, metastasis and severe side effects. The study provided systematic insights into the patterns of TNBC distant metastases (DM), as well as investigating the related elements for the prognosis prediction of TNBC patients on the basis of on large sample. We screened eligible patients with triple-negative breast cancer from the Surveillance, Epidemiology, and End Results (SEER) database between 2010 and 2015. Besides, we analyzed differences in baseline characteristics among patients with diverse modes of metastasis. Meanwhile, we calculated proportional mortality ratio (PMR) and the expression of proportional trends in different patients. Subsequently, Kaplan-Meier (KM) analysis was employed to investigate the survival outcomes. Finally, the predictive and prognostic factors of DM were identied. N


Introduction
Breast cancer (BC) remained the most severe public health problem that endangers women's lives and health in the world [1,2], accounting for 10.4% of all cancers. Although there are still regional differences between different countries, breast cancer is still the leading cause of death among women aged 20-50 [3]. BC is a heterogeneous disease. In view of genetic, epigenetic and transcriptome changes, its histological and biological characteristics are different. 95% of BC is adenocarcinoma, starting as a local disease [4]. Invasive/metastatic BC can be divided into nonspeci c type (NST) cancer (60-75%) and special type (20%-25%) [5]. In the aspect of biological characteristics, three kinds of molecular biomarkers (estrogen receptor (ER), progesterone receptor (PR) and HER2) are detected by molecular biological methods for the molecular type of BC [6-8], including estrogen receptor positive (ER+) type or progesterone receptor positive (PR+) type, human epidermal receptor 2 positive (HER2+) type and triplenegative BC (TNBC) [9,10].
The incidence of TNBC in BC is about 10-15%, but it is one of the most aggressive subtypes [11]. TNBC distant metastasis (DM) refers to the metastasis of BC outside the ipsilateral breast, chest wall and regional lymph nodes, which is the main cause of death (COD) of BC. 6-10% of cases have metastasis at the time of diagnosis, and nearly 30% will relapse or metastasize [12]. About 25% of TNBC patients still have local recurrence and DM after active treatment. There are many metastases of BC such as brain, lung, liver, etc [13]. At present, there are few effective methods for the treatment of metastatic TNBC, resulting in poor prognosis of patients with metastatic TNBC. Therefore, it is urgent to further study the prediction and prognostic factors of DM in TNBC [14][15][16][17]. But, the shortcomings of previous studies are mostly single-center studies with small sample size and incomplete long-term follow-up information.
Therefore, we used SEER database to establish prognostic models for TNBC patients with different metastasis modes to further explore the risk factors affecting distant metastasis of tumor.

Methods And Materials Database
We employed the the "SEER*Stat 8.3.8" software (Version 8.3.6; NCI) to download the data from the SEER registry. The SEER database detailed information about cancer patients in the USA. In our study, we signed the data agreement and used the 10977-Nov2019 database. In addition, the Institutional Review Board allowed us to proper use of this public database.

Patient identi cation
The TNBC patients with positive pathology were retrospectively extracted from SEER 18 registry. All patients were diagnosed between 2010 and 2015 because data DM data were recorded from 2010. Inclusion criteria: (1) patients diagnosed with TNBC (C50.0-C50.6, C50.8-C50.9; AYA site recode/WHO 2008= 8.4 Carcinoma of the breast), (2) BC was the rst primary malignancy of each patient, (3) patients with complete data of long-term follow-up. Exclusion criteria: (1) patients with bilateral tumors or unknown tumor laterality, (2) patients with unknown data on marital status, insurance status, household income, the administration of surgery, T stage, N stage, race, and tumor grade, (3) unknown metastatic status, (4) data were from autopsy or death certi cate only.

Data extraction
For each patient, essential clinical characteristics and long-term survival outcomes were extracted using the "Case Listing Session". Variables including race, age, year of diagnosis, tumor grade, laterality, AJCC 7 th T stage, N stage, the administration of surgery/ chemotherapy and radiotherapy, DM status, survival months, COD, insurance status, metastatic status, vital status and, household income were identi ed. Based on DM status, all patients were categorized intoDM group and without DM group. Moreover, those in DM group were further subdivided based on the metastatic status.
For further analyses, age at diagnosis was divided into < 45, 45 -69 and ≥ 70 years old, the race was classi ed into White, Black and Other. Furthermore, the pathological grade was divided into four levels: high, moderately, poorly and undifferentiated. Median household income was calculated to de ne highand low-level household income.
Identi cation of prognostic characteristics KM curves were used to explore the overall survive of DM or the metastatic site in TNBC patients. We analyzed COD in different groups of patients who died during long-term follow-up. Uni-and multivariable cox analyses were constructed to explore the risk factors of DM in TNBC patients. Finally, COX analyses were developed to nd prognostic factors of OS and CSS.

Statistical analysis
In our study, data were mainly presented by n (%). Chi-square test was used for comparison between categorical variables. The fundamental analyses were completed on the basis of SPSS 23.0 software (SPSS Inc) and R software (Version 3.4.1). All analytical processes were two-sided, and P < 0.05 was deemed to have statistical signi cance.

Results
Characteristics of baseline and prognosis Figure 1 presented the selection ow chart of this study. 42.00%, N0: 63.79%), and tumor-directed surgery (92.56%). Compared with none-DM patients, those DM patients had older age (P = 0.003), higher probability of black (P < 0.001), later stage of diseases (P < 0.001), lower median household income (P = 0.002), lower married rate (P < 0.001) and lower insurance rate (P < 0.001). In addition, the rate of surgery and radiotherapy was signi cantly higher (P < 0.001) in patients without DM. But, there was no statistical difference in diagnosis year, laterality and chemotherapy The most common sites of metastasis in DM patients were bone (24.46%) and lung (23.78%), while the least common organ of metastasis was brain (3.61%). Besides, we found that nearly 36.94% patients had two or more metastases ( Table 2). Compared with patients with multiple organ metastases, patients with single organ metastases were more likely to undergo surgery and less likely to receive radiotherapy. (52.86% vs. 28.50%, P < 0.001; 30.6% vs. 36.9%, P = 0.037). But, no statistically signi cant differences were found in comparison with other variables. Eventually, the patients were chopped up into four groups (simple brain, simple bone, simple liver, and simple lung), and the comparison between groups was shown in Table S1. Compared with other sites, patients with lung metastasis later stage of diseases while patients with brain metastasis underwent radiotherapy more frequently (P = 0.021, P < 0.001).

Proportional Death Rate
The results of subgroup analyses are shown in Figure 3. Compared with the non-DM group, the mortality rate from BC in DM group was signi cantly increased (77.57-94.75%), while the mortality rate from other causes (21.15%-4.68%) and the mortality rate from unknown causes (1.28%-0.57%) were signi cantly decreased ( Figure 2A). In other words, once BC patients had DM, They have a greater chance of dying from their disease. This trend became more pronounced with the increase of metastatic sites ( Figure 2B).
It is worth noting that DM in brain causes all the deaths ( Figure 2C).

Survival Results
As shown in Figure 3A, B, Non-dm patients had a better survival advantage (OS and CSS) than DM patients. Besides, patients with a single metastasis had a better survival advantage (OS and CSS) than patients with multiple metastases.patient ( Figure 3C, D) In patients with multiple metastases, patients with two metastases had better OS and CSS than patients with more than two metastases ( Figure 3E, F). What is more, the greater the number of metastases in DM patients, the worse the survival advantage (OS and CSS) ( Figure 3G, H). Patients with brain metastasis or liver metastases had the worst OS and CSS of all metastatic sites ( Figure 4A, B). Figure 4C-4F indicated the survival analyses of patients with two and three metastatic sites. However, in patients with two or three metastatic sites, there were no statistically signi cant differences in OS and CSS between different metastatic types. Table 3 showed that older TNBC patients, higher T, and higher N were at higher risk for DM, stage T, insurance status, number of managed surgery, chemotherapy, government metastatic style, and important factors affecting the operating system of DM patients (Table 4), and stage T, insurance status, surgical management, number of metastatic lesions, and marital status were signi cantly correlated with CSS (Table 5). Besides, the multivariate COX analysis indicated that T stage, insurance status, chemotherapy management, surgical method and metastatic site were obvious factors affecting OS in patients with a single metastatic site.
Page 11/30 (Table 6). T stage, insurance status, surgical method, chemotherapy method, and site of metastasis were correlated with CSS (Table 7).     Moreover, only TNBC was included in this study, which is more convincing to think that this data is the only representative of TNBC. Last but not least, with the rapid improvement in the understanding and diagnosis of the disease, earlier TNBC or smaller breast lumps will be detected in clinical practice.
There have also been previous studies on distant metastasis in BC patients. A retrospective study collected and analyzed information from 2033 BC patients from 2012 to 2014 and showed that high tumor grade, T, and N were signi cant risk factors for DM [21]. Another study identi ed 1173 BC liver metastases from the SEER database. Classi cation, marital status, surgery, radiotherapy, chemotherapy, tumor size and tumor subtypes were identi ed as risk factors for liver metastases from BC [22]. These conclusions are similar to our results. But only a few studies have focused on the triple negative subtype and combined metastatic pattern based on a larger sample size. In our study, 36.94% (379/1,026) of metastatic patients had multiple metastases. In addition, compared with patients with a single metastasis, patients with multiple metastases had a poorer survival advantage, and the more metastases there were, the worse the prognosis was (Table 4, 5). Furthermore, diverse combination of metastatic sites represented different prognosis.
Our results indicated that patients with DM have a signi cantly poor prognosis. Besides, the greater the number of metastases, the worse the prognosis. Hence, we further to identify the risk factors of DM and metastasis in patients with prognostic factors is very essential. In primary bladder cancer, high pathological grade, N and T were positively correlated with bone metastasis [20]. Another meta-analysis showed that poor tumor differentiation was related to the risk of metastasis in cutaneous squamous cell carcinoma [23]. In DM patients, the results of multi-COX analysis indicated that tumor grade, age at rst diagnosis, T, N, marital status and surgical treatment were the signi cant in uence affecting the OS. This is in line with what many previous studies have con rmed. In a BC study based on Asian female patients, age, grade, TNM stage, and chemotherapy have been shown to be associated with BC long-term survival [24]. Another SEER database study found that race, age, grade, molecular subtype, surgery, brain and liver metastases were independently associated with BC speci c survival [25].
However, there remain several limitations in our study that should not be ignored. First, we failed to get more information from SEER database, including lymphatic or vascular invasion, multifocality, the sequence and speci c arrangement of multiple metastases and even molecular biomarkers. Secondly, the database lacked several important clinical information, including LDH, hemoglobin, neutrophil count, platelet count, etc. If we include these, we can improve the comprehensiveness of analysis and conclusion. Furthermore, limitations include a lack of information on rare subtypes of TNBC that may alter treatment, such as metaplasia, adenoid cystic, and acrosine subtypes. Last but not least, the main population for this study is Americans, and whether the results applied to other populations was questionable.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableS1.doc