Marital status and survival in patients with gastric cancer

Abstract The objective of this study is to examine the impact of marital status on incidence of metastasis at diagnosis, receipt of surgery, and cause‐specific survival (CSS) in patients with gastric cancer (GC). Research data is extracted from The Surveillance, Epidemiology, and End Results (SEER) database, and 18,196 patients diagnosed with GC from 2004 to 2010 are involved. Effects of marital status on incidence of metastasis at diagnosis, receipt of surgery, and CSS are determined using multivariable logistic regression and multivariable Cox regression models, as appropriate. Single GC patients have a higher incidence of metastasis at diagnosis than married patients, while the differences between divorced/separated patients or widowed patients and married patients are not significant. Among those without distant metastasis, single patients, divorced/separated patients, and widowed patients are much less likely to accept surgery compared with married patients. Finally, in the whole group of 18,196 GC patients, single patients, divorced/separated patients, and widowed patients have shorter CSS compared with married patients, even in each of the TNM stage. Marriage had a protective effect against undertreatment and cause‐specific mortality (CSM) in GC. Spousal support may contribute to higher rate of surgery receipt and better survival in patients with GC.


Introduction
In this study, we investigated the relation between marital status and incidence of metastasis at diagnosis, receipt of surgery, and CSS in the group of 18,196 GC patients. Data are from the SEER program between 2004 and 2010.

Study population
We extracted clinical data of 18,196 cancer patients with stomach as the single primary site from SEER database. Sponsored by National Cancer Institute, the SEER program collects and publishes incidence, mortality, prevalence, survival, and lifetime risk statistics which can be used to assess the impact of cancer in the general population. The current SEER database consists of 18 population-based registries, which cover approximately 26% of the United States population [21].

Patient selection
SEER-stat software (SEER*Stat 8.1.5) was used for the data extraction and patient selection. The inclusion period was from 2004 to 2010, for the fact that several employed covariates were introduced in the SEER database in 2004 [22]. Age was limited to 18 years or older, and patients with unknown marital status were excluded. The histologic types consisted of adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma. The sixth American Joint Classification of Cancer (AJCC) TNM staging system was adopted in this study, and patients with unknown TNM stage were excluded.

Study variables
According to SEER database, marital status is described as married, single (never married), separated, divorced, and widowed. In this study, the unmarried include single, separated/divorced, and widowed patients. Race/ethnicity is classified as White (non-Hispanic), Black, Hispanic, and other (American Indian/AK Native, Asian/Pacific Islander, and unknown). The differentiation grades include well/ moderately differentiated grade, poorly differentiated/ undifferentiated grade, and unknown. Tumor location is classified as cardia and noncardia; noncardia includes fundus, body, greater curve, smaller curve, antrum, and pylorus, according to SEER database. The TNM classification system is defined by the AJCC Cancer Staging Manual (the sixth edition). Types of surgery include gastrectomy with/without regional lymph nodes removed according to the SEER database. Cause-specific survival is a net survival measure representing survival of a specified cause of death in the absence of other causes of death according to the SEER database. Estimates are calculated by specifying the cause of death. Individuals who die of causes other than the specified cause are considered to be censored. In this study, GC is the specified cause of death.

Statistical analysis
Baseline patient characteristics were analyzed with chisquared test for categorized measurements and Spearman tests for continuous measurements. Multivariable logistic regression was used to determine the association of marital status and incidence of metastasis at diagnosis; the analysis was adjusted for demographic factors (age, sex, and race/ ethnicity), tumor location, histological type, differentiated grade, and year of diagnosis. For analysis of receipt of definitive therapy, we excluded patients with metastasis at diagnosis, and 10,013 patients remained eligible. Multivariable logistic regression was used to determine the association between marital status and surgery receipt; the analysis was adjusted for demographic factors (age, sex, and race), tumor location, histological type, differentiated grade, and year of diagnosis. For CSS analysis, multivariable Cox regression analysis was adopted to assess the impact of marital status on CSS after adjustment for demographic factors, TNM stage, histological types, differentiated grades, tumor location, and year of diagnosis. The median follow-up for the cohort analyzed for CSS was 22 months (range: 1-100). All P-values were two-sided. The threshold of 0.05 was considered statistically significant. All confidence intervals (CIs) were stated at the 95% confidence level. Statistical analyses were performed using SPSS 19.0.

Patient characteristics
Among the cohort of 18,196 patients with GC, 11,114 (61.1%) were married, 2620 (14.4%) were single (never married), 201 (1.1%) were separated, 1523 (8.3%) were divorced, and 2738 (15.1%) were widowed. Eight thousand and one hundred eighty-three (8183, 44.9%) came up with distant metastasis at diagnosis, and 8580 (41.2%) people accepted surgery for GC. In the whole group, the married were 2.6 years younger than the unmarried which included the single, the separated/divorced, and the widowed (P < 0.001). The white and male had a higher percent of being married than other races (black, American Indian/AK Native, Asian/Pacific Islander, and unknown) and females, respectively (P < 0.001 for both). The rate of earlier stage (stage I/II) at diagnosis in the married group was lower than the widowed (36.2% vs. 39.7%), but higher than the single (30.3%) and the separated/divorced (35%). Details of patient demographics and pathological features were summarized in Table 1.  Spearman tests were used (r = 0.205). 2 Include American Indian/AK Native, Asian/Pacific Islander, and unknown. 3 Highly/moderately differentiated. 4 Poorly differentiated/undifferentiated. 5 Adenocarcinoma. 6 Mucinous cell adenocarcinoma. 7 Signet ring cell carcinoma.

Marital Status and Gastric Cancer
Impact of marital status on incidence of metastasis at diagnosis in GC Single (never married) GC patients displayed a higher incidence of metastasis at diagnosis than married GC patients (odds ratio [OR] 1.138, 95% CI: 1.040-1.245; P = 0.005; Table 2). While difference between the divorced/separated patients and married patients was not significant (P = 0.064, Table 2), difference between widowed patients and married patients was not significant either (P = 0.085, Table 2). Black GC patients had a lower incidence of metastasis at diagnosis compared with the white (OR 0.825, 95% CI: 0.748-0.911; P < 0.001; Table 2). Year of diagnosis had no significant impact on incidence of metastasis at diagnosis. Patients with cardia cancer had a lower rate of distant metastasis compared with noncardia GC patients (OR 0.716, 95% CI: 0.667-0.769; P < 0.001; Table 2). Poorly differentiated/undifferentiated GC patients had a higher incidence of metastasis compared with the well/moderately differentiated GC (OR 1.428, 95% CI: 1.322-1.543; P < 0.001; Table 2). GC patients with signet ring cell carcinoma and mucinous cell adenocarcinoma had a lower incidence with metastasis compared with patients with adenocarcinoma (OR 0.621, 95% CI: 0.502-0.768, P < 0.001; OR 0.775, 95% CI: 0.720-0.835, P < 0.001; respectively).

Impact of marital status on receipt of surgery in GC
To determine the differences in receipt of surgery according to marital status, we excluded patients with metastasis  Table 2). Patients with cardia cancer were much less likely to accept surgery compared with noncardia GC patients (OR 0.441, 95% CI: 0.395-0.493, P < 0.001, Table 2).

Impact of marital status on CSS in GC
With regard to the association between marital status and CSS of GC patients, Cox proportional hazards regression model was adopted in total of 18,196 (Table 3). Further analysis was conducted according to TNM stage, the association between marital status and CSS length remained significant as well. Details were shown in Table 4 and Figure 1. Among other clinical parameters, we found that female GC patients had better CSS than the male GC patients (OR 0.956, 95% CI: 0.920-0.992, P = 0.018,     Table 4). GC patients diagnosed with mucinous cell adenocarcinoma displayed better CSS than those diagnosed with adenocarcinoma (OR 0.825, 95% CI: 0.734-0.928, P = 0.001, Table 4). GC patients at stage II/III/IV had significantly worse CSS compared with those at stage I (details at Table 4).

Discussion
In this study, we find that marriage has a protective effect on GC patients. Married GC patients have a lower incidence of metastasis at diagnosis than single patients. Married GC patients are more likely to accept surgery than the single, the divorced/separated, and the widowed. In addition, married patients have a lower GCSM than the unmarried, even in each of the TNM stage. It is the first study to demonstrate the significant protective impact that marriage can have on incidence of metastasis at diagnosis, surgery receipt, and CSS of GC patients.
Incidences of metastasis at diagnosis in each group are as follows: 44.2% for the married, 50.4% for the single (never married), 47.2% for the separated/divorced, 41.6% for the widowed. The incidences of metastasis in each group may be affected by age, a previous study demonstrated that GC tends to exhibit more aggressive tumor behavior in young patients (40 years or younger) than in old patients [23]. After adjusted for age, race, tumor location, differentiated grade, histological type, and year of diagnosis, only single patients displayed a higher incidence of metastasis than married GC patients; differences between the separated/divorced or the widowed and the married are not significant. To explain this phenomenon, married people may have better access to care than the unmarried [24]. Reports have demonstrated that even in nations with universal access to free care, sociodemographic factors influence outcomes in various health conditions [24][25][26][27]. Additionally, married people may benefit from encouragement by spouses to seek medical attention for worrisome symptoms.
The association between marital status and the receipt of surgery is valid in our study in GC cases without distant metastasis. Spouses of these married patients may encourage them to perform surgery versus expectant management [28], which could partly account for the discrepancies. Studies showed diagnosis of cancer caused more distress than other diseases [29]. Married people were easier to benefit from social support from their friends and family and displayed less distress and depression after the cancer diagnosis [30]. Patients with depression displayed three times greater odds to be noncompliant with medical treatment recommendations compared with those who were not depressed [31]. And a study in breast cancer demonstrated that women patients with depression were less likely to accept surgery [32]. Physicians should pay more attention to those unmarried and diagnosed with GC, and recommen them for psychologist's help if necessary. Adequate support and timely psychological interference may contribute to more possibility of receiving surgery in unmarried GC patients.
Partly resulting from the advantage in treatment selection, the married enjoy a much better CSS than those unmarried. There are explanations for the survival advantage in other aspects. Studies suggested that the unmarried may be at greater risk of smoking and alcohol use [33,34], which could do additional harm to the patients' health. Physiologically, abnormal diurnal cortisol rhythm predicts earlier cancer death [35][36][37], and the abnormal profiles might be associated with quality of social support from friends and marriage. Suppression of natural killer (NK) cell count and NK function may be involved in the progression [38]. Adverse results exist [39] and further investigations on this subject are warranted.
This study gives conclusive results of the association between marital status and outcomes of GC. There are some potential limitations we should consider. Firstly, risk factors included in this study is limited. Risk factors such as smoking, alcohol consumption, above normal body weight, high salt/fat consumption, low vegetable and fruits consumption, low economic status, other chronic gastric diseases, and HP infection are not recorded in the SEER database [40,41]. Yet, health behavior variables including smoking, diet, and physical activity, were reported to have no indirect effect on the association between living arrangements and mortality [42]. Secondly, some unmarried patients may cohabit with a partner other than a spouse which could provide support to the patients. Data from the 2010 US Census indicate that about 90 million unmarried Americans more than 15 years old live "with other persons", whereas, only approximately 30 million live alone [3]. Neglect of the cohabiting patients may lessen the variation in mortality. Thirdly, information on comorbidities besides GC is not available from the SEER database' this is a possible limitation to this study.
Despite the stated limitations, our study demonstrates that, unmarried GC patients are less likely to accept surgery and have worse CSS than married GC patients. Spousal support may contribute to higher rate of surgery receipt and better survival in GC. Special attention should be paid to the unmarried GC patients; social support may help improve their prognosis.