The Influence of Marital Status on Survival in Patients with Nasopharyngeal Carcinoma: A Surveillance, Epidemiology, and End Results (SEER) Database Analysis

Purpose: To assess the influence of marital status on survival in patients with nasopharyngeal carcinoma (NPC). Methods: We used the Surveillance, Epidemiology, and End Results (SEER) database to analyze 5477 patients who were diagnosed with NPC from 2004 to 2016. Kaplan–Meier survival analysis and Cox proportional hazard regression were used to analyze the influences of marital status on cause-specific survival (CSS) and overall survival (OS). Subgroup analyses was used to assess the influences of marital status on CSS based on different factors. Results: The 5477 patients were divided into three groups, with 61.5%, 22.4%, and 16.1% of patients being married, single/unmarried, and separated/widowed/divorced, respectively. The separated/widowed/divorced group were more likely to be female ( P <0.001); had the highest proportion of elderly subjects ( P <0.001); were mostly Caucasian ( P <0.001); had pathological grading I/II ( P <0.001); were likely to receive surgery ( P =0.032); and were registered at the northeast, north central, and south ( P < 0.001) regions. The 5-year CSS was 92.6%, 92.4%, and 85.1% in the married, single/unmarried, and separated/widowed/divorced groups, respectively ( P <0.001); and respective 5-year OS was 60.7%, 54.6%, and 40.1% ( P <0.001). Marital status was the independent prognostic factor for NPC. Compared with married patients, separated/widowed/divorced patients had a significantly increased risk of NPC-related death (hazard ratio [HR]=2.180, 95% confidence interval [CI] 1.721–2.757, P <0.001). The single/unmarried ( P =0.355) group had a similar CSS as that of the group. Conclusion: Marital status is


Introduction
Nasopharyngeal carcinoma (NPC) is a rare and malignant cancer with the highest incidences in South-Eastern Asia, Eastern Asia, Eastern Africa, and Middle Africa [1].
Radiotherapy and chemotherapy remain the most effective treatment options. However, owing to drug-induced toxicity and radiation-related sequelae, patients often experience severe side effects that significantly affect their quality of life and survival [2][3][4]. Social support plays a very important role in the lives of patients with cancer, especially marital status, which has been confirmed to affect the survival rate in different tumors [5][6][7][8].
Understandably, marital status is of significance even for patients with NPC.
A recent study analyzed the 10 leading causes of tumor-related deaths; the results showed a generally significant survival benefit for married patients as compared to unmarried patients. Moreover, the former group had lower risk of early tumor-related death than the latter [5]. Osazuwa-Peters' study of head and neck cancer (HNC) found that being married conferred a survival advantage among HNC survivors, but the study only classified patients as married or unmarried [8]. One Surveillance, Epidemiology, and End Results (SEER) database-based study on marital status found that being married had a protective effect on cause-specific survival (CSS) and overall survival (OS) compared to unmarried status. Nevertheless, this study did not compare married patients with specific subgroups of unmarried patients such as single/unmarried and separated/divorced/widowed [9]. The other SEER-based marital status study showed that widowed patients were associated with higher risk of mortality in NPC than married, divorced, or single patients, but the sample size was small. Furthermore, some essential medical data were lacking, which may have had significant effects on patient outcomes with NPC. In addition, it only recorded data from 2004 to 2013 from the SEER database [10]. However, the SEER database has been updated until 2016, and there are few studies on the impact of different marital status on NPC. Therefore, we used patient data of those diagnosed between 2004 and 2016 included in the SEER Cancer Registry to explore in detail the relationship between different marital status and patient survival of patients.

Patients
We extracted data from the SEER database released in April 2016 as the data source for this study. The SEER program is sponsored by the National Cancer Institute and consists of 17 population-based cancer registries that represents approximately 28% of the US population [11]. The SEER program provides accurate and continuous information on cancer incidence, survival, prevalence, and patient demographics. With the help of the National Cancer Institute's SEER * Stat software (Version 8.3.5; www.seer.cancer.gov/seerstat), patients diagnosed with NPC from 2004 and 2016 were included. The excluded criteria were as follows: (1) age at diagnosis was less than 18 years; (2) unknown survival month; (3) unknown marital status; (4) a prior diagnosis of malignancy; (5) more than one primary site.

Study variables and endpoints
The following variables were identified from the SEER database: age, sex, ethnicity, pathology grade, primary site, insurance, registry site, year of diagnosis, surgery, and radiotherapy. We divided patients into three categories: "married", "single/unmarried," and "separated/widowed/divorced." The pathology grade was also divided into three groups: well differentiated, grade I and moderately differentiated, grade II; poorly differentiated, grade III and undifferentiated, anaplastic, grade IV; and unknown group.
According to the classification principles of the US Census Bureau, 18 cancer registries were categorized into four regions, namely West, Northeast, North central, and South [12].
The primary sites included six groups, namely superior wall of the nasopharynx, posterior wall of the nasopharynx, lateral wall of the nasopharynx, anterior wall of the nasopharynx, overlapping lesion of the nasopharynx, and nasopharynx. Patients were also categorized based on whether they underwent surgery and radiotherapy. The endpoints in this study were the 5-year cause-specific survival (CSS) that was calculated from diagnosis until death due to NPC or last known date alive and the 5-year overall survival (OS) that was calculated as death from any cause.

Statistical analysis
Patients' baseline characteristics were compared by using the chi-squared (χ2) test. We used the Kaplan-Meier method to estimate the CSS and OS, and the survival differences were calculated by the log-rank test. The binary Cox regression model was calculated using the hazard ratio (HR) for relationships between each variable and mortality. Every variable for which P<0.05 in the univariate analyses was initially included in multivariate analyses. Notably, we still included some necessary variables for the Cox proportional hazards regression despite P>0.05 in their respective univariate analyses, because they were common confounders of NPC such as T-stage and N-stage. All confidence intervals (CIs) were stated at the 95% confidence level. Statistical analysis was conducted using SPSS software version 22.0 (IBM Corporation, Armonk, USA). P value<0.05 indicated statistical significance.

Patient baseline characteristics
Initially, 7380 patient records were extracted between 2004 and 2016 from the SEER database. Based on the inclusion and exclusion criteria, 5477 eligible patients (3886 men and 1591 women) were identified during the 12-year study period. A flow diagram of the study selection is presented in Figure 1. Among these patients, 3370 (61.5%) were married, 1226 (22.4%) were single/unmarried, and 881 (16.1%) were separated/widowed/divorced. Significant differences were noticed in all the comparisons.
With respect to sex, men were more likely to be married (74.4%), while female patients were more likely to be separated/widowed/divorced (44.0%). The separated/widowed/divorced group were more likely to be female (P< 0.001); had the highest proportion of elderly subjects (>50 years, P <0.001); were mostly Caucasian (P<0.001); had pathological grading I/II (P<0.001); were likely to receive surgery (P = 0.032); and were registered at the northeast, north central, and south (P<0.001) sites.
Compared with the other groups, married patients were more likely to be in the West registry site, have grade III/IV NPC, and the highest proportion of radiation and chemotherapy history. However, there was no significant difference in primary site cancer among the three study groups. The demographic and clinical characteristics of NPC and treatment types are summarized in Table 1.

Effect of marital status on CSS in the SEER database
The univariate log-rank test was used to evaluate the 5-year CSS, which was 92.6%, 92.4%, and 85.1% in the married, single/unmarried, and separated/widowed/divorced, respectively (P<0.001). In the multivariate analysis, the HR of the single/unmarried group was lower than that of the married group (HR = 0.811, 95% CI = 0.675-1.152, P = 0.355), and the HR of separated/widowed/divorced group was higher than that of the married group (HR = 2.180, 95% CI = 1.727-2.751, P<0.001). The CSS of NPC patients were calculated by Kaplan-Meier curve ( Figure 2). In univariate analyses, age (P<0.001), ethnicity (P<0.001), registry site (P = 0.005), grade (P = 0.001), pathologic T stage (P = 0.001), pathologic N stage (P<0.001), pathologic M stage (P = 0.003), radiotherapy (P<0.001) and chemotherapy (P<0.001) were regarded as significant risk factors for CSS in NPC patients ( Table 2). The multivariate analysis was performed by Cox regression, and variables that were validated as independent prognostic factors included age (>50 years,  Table 2).

Effect of marital status on OS in the SEER database
We used univariate log-rank test to evaluate the 5-year OS of NPC patients. Patients in the married group had a better 5-year OS (60.7%) than those who were single/married (54.6%) and separated/widowed/divorced (40.1%). The Kaplan-Meier curve was used to calculate the OS of NPC patients ( Figure 3). All the differences were significant independent factors for OS among NPC patients according to the univariate log-rank test (P<0.  Table 3). The variables that were significant in the univariate log-rank test were validated as independent risk factors by multivariate modeling analysis using Cox regression ( Table 3). The single/unmarried and separated/windowed/divorced patients had higher HRs (HR = 1.219, 95% CI = 1.100-1.351, P<0.001 vs. HR = 1.977, 95% CI = 1.784-2.191, P<0.001, respectively) than married patients.
Subgroup analysis by age, gender, grade, insurance, surgery, radiation, and chemotherapy We analyzed the effects of marital status on CSS according to age, sex, grade, surgery, radiation, and chemotherapy. Interestingly, univariate analysis showed that only married patients of both age groups (i.e., ≤50 years or >50 years) had better 5-year CSS than other patients (P<0.001), although the HRs among unmarried male patients were all higher than those of their female counterparts. Further, patients in the separated/windowed/divorced group had the lowest 5-year survival rate except those that had grade I/II stage NPC. Patients in the single/unmarried group had the highest number of surgery-naïve female patients, with unknown grade, and who had not undergone radiotherapy and chemotherapy. Moreover, multivariate analysis showed that the separated/windowed/divorced group had significant differences with respect to all parameters except for the grade I/II staging and being uninsured, when compared with the married group (P<0.05). The single/divorced group had no significant differences as compared to the married group, except for those that did not receive chemotherapy. The married group and single/unmarried group seemed to have a similar 5-year survival rate in with respect to sex, NPC grade, and history of radiotherapy. Here, we report several new findings. First, risk of death in the 5-year OS among uninsured patients was obviously higher than insured patients and those on Medicaid (P<0.001) and was not reported in previous studies[9, 10]. Second, the risk of death in 5year CSS and OS among primary site tumor in the anterior wall of the nasopharynx was the higher. Third, patients with pathological grading III/IV were more likely to have higher survival rate than pathological grading I/II and unknown grading in CSS and OS. Taken together, these factors may play an important role in NPC patients, especially with regard to insurance. Some studies indicated that insurance status offers protection for cancer patients [14,15], while others implied that the health insurance status is related to the mortality and presentation stage of cancer patients [16,17]. Insurance is a socioeconomic factor and understanding how populations of cancer patients are influenced by it will better clarify how to mitigate disparities in care and outcomes [15].In our subgroup multivariate analysis, single/unmarried patients had lower HRs than married patients, with respect to their insurance plans, while separated/windowed/divorced patients showed contrary results, because in addition to the obvious psychological stress of an unfavorable diagnosis, the lack of a spouse may also reduce their ability to pay for treatment [18].
Besides, cancer diagnosis can lead to psychological disorders (e.g., despair, loss, worry, fear, anxiety), and if patients are older and without a spouse, this will lead to poorer Table 4).
Cancer treatment mainly involves surgery, radiotherapy, and chemotherapy. Previous

Conclusion
Our results suggest that being separated/widowed/divorced increases the risk of NPC mortality when compared to patients who are married or single/unmarried patients. In the subgroup analysis, separated/widowed/divorced patients were regarded as a high-risk group with poor prognosis. Therefore, more social and psychosocial support should be given to cancer patients who are separated, widowed, or divorced. Furthermore, clinicians should pay more attention to patients' marital status in case of an NPC diagnosis, and provide them with an individualized care and treatment plan.

Availability of data and material
The datasets generated and/or analysed during the current study are available in the SEER database, https://seer.cancer.gov/data/.

Competing interests
The authors declare no conflicts of interest.