Propensity-score Matching Analysis for Marital Status and Survival of Oral and Oropharyngeal Squamous Cell Carcinoma Based on SEER Database


 ObjectMarital status plays different roles as a risk factor on survival in various cancers . The study is aimed to analyze the impact of marital status on survival of oral and oropharyngeal squamous cell carcinoma(OPSCC) at population level based on SEER database using propensity-score matching method(PSM).Methods37,023 eligible patients were extracted from the Surveillance, Epidemiology, and End Results(SEER) database, and analyzed the impact of various marital status on cancer-specific survival(CSS) of OPSCC by Kaplan-Meier method and Cox regression model. Then we used propensity-score matching analysis to balance baseline characteristics between married, single, divorced and widowed patients. The impact of various marital status after pairwise matching using p-value adjusted and PSM on CSS was re-analyzed by Kaplan-Meier method.ResultsThe age, sex, race, tumor location, pathologic grades, SEER stages, treatments, composite socioeconomic status(C-SES), insurance, and marital status were identified as independent prognostic factors for CSS of OPSCC. Widowed patients presented the worst CSS, compared with married, single, and divorced patients(P<0.001). Subgroup analysis indicated that widowed patients always presented with the significantly decreasing risk of CSS compared with other marital status in different SEER stages(P<0.001), and different C-SES(P<0.001). After propensity-score matching, widowed patients were still found to be associated with significantly decreased CSS compared with other marital groups(P<0.001).Conclusion﻿Marital status was first analyzed after using PSM to balance clinicopathological and socioeconomic confounding factors and identified as an independent prognostic factor for CSS of OPSCC. Widowed patients was significantly associated with a decreasing CSS, which indicated that absence of spousal support and optimal psychosocial coping strategies may explain the phenomenons.


Introduction
Oral and oropharyngeal squamous cell carcinoma(OPSCC) was one of the common malignant tumor in head neck. The incidence of OPSCC continued to increase, due to excessive tobacco, alcohol consumption and HPV infection [1,2]. Furthermore, mortality rates were relentlessly increasing overall by 0.5% per year from 2009 to 2018, although optimized multidisciplinary treatments improved local control [3,4].
As we know that the anatomy of oral and oropharyn is not deeply hidden, primary tumor should been easier to be examined in early stage, and have more opportunities to be cure and better survival. However, the reality of survival of OPSCC was opposite. Therefore, we need to further analyze the predictive factors to establish comprehensive treatment strategies and improve survival of patients with OPSCC.
Prior studies suggested that the confounding risk factors, including age, sex, race, tumor site, pathologic grade, treatment, socioeconomic status, alcohol and tobacco consumption, HPV infection and marital status, played controversial roles in survival of patients with OPSCC [5][6][7][8][9][10]. Many studies also indicated that unmarried status was an increasing risk of mortality in oral cavity or oropharyngeal cancer, compared with married status based on a population studies [11][12][13][14]. However, the impact of more detailed unmarried status on survival has not been previously reported. The effect of socioeconomic and marital status on survival of OPSCC had been reported in seldom. Therefore, it is important to explore the effect of different marital status on survival of OPSCC patients and formulate more reasonable strategies.
In the present study, we extracted data from the Surveillance, Epidemiology, and End Results cancer registry database to assess and compare the effects of married, single, divorced and widowed status on cancer-speci c survival using PSM and explored the possible mechanisms.

Study variables
Variables about clinicopathological and composite socioeconomic characteristics were shown in Table 1. The tumor location was divided into tongue(anterior and basal), oroharynx(tonsils) and others(mucous of buccal, mouth oor, lip, palate and others). Pathological grades were divided into 2 categories: low grade (well and moderately differentiated) and high grade (poorly and undifferentiated). In our study, we merged seperated and divorced patients into divorced group.
Three standard 2000 US Census variables from SEER database, including median family income, the percentage of persons who have less than high school education, and the percentage of persons with income below the poverty level, were extracted.
A composite SES variable was set up according to the three SES variables as described in previous studies [15][16][17]. The composite SES variables was further classi ed as low C-SES (C-SES score <5) and high C-SES (C-SES score ≥5) [18].
Propensity-score matching: Marital status was divided into married, single, divorced, and widowed groups from SEER database. To precisely assess the roles of marital status in prognosis of OPSCC patients, propensity-score matching method was performed to balance clinicopathological and socioeconomic characteristics between different marital pairwise groups. After p-value adjusted and pairwise matching, the prognostic importance of marital status on CSS were re-analyzed within the matching pair of groups.

Statistical analysis
Pearson's Chi-squared test was performed to analyze relationship of different clinicopathological and socioeconomic characteristics between various marital groups. CSS was estimated by Kaplan-Meier methods. Log rank test was performed to evaluate the statistical signi cance of differences between survival curves of various marital groups. Multivariable Cox proportional-hazards regression model was performed to analyze the hazard ratios of risk factors. All statistical tests were twotailed. Statistical signi cance was identi ed as p value less than 0.05. The statistical analysis was conducted using SPSS(version 20.0) and R(version4.1.1).

Results
Baseline clinicopathological and socioeconomic characteristics A total of 37,023 eligible patients (27,776 males and 9,247 females) were included in our studies from the SEER database. Of these patients, there were 21,049(56.9%) married patients, 7,554(20.4%) single, 5,428(14.7%) divorced and 2,992(8.1%) widowed patients. Detailed clinicopathological and composite socioeconomic characteristics were shown in Table 1.
Statistically signi cant differences were identi ed in age, sex, race, tumor location, pathological grades, SEER stages, treatments, insurance, C-SES between different marital groups(P<0.001). Among the comparisons of risk factors, widowed patients had twice or three times higher proportion in the group of age ≥60-year old(87.9%), and female(62.6%), than that of other corresponding marital groups.

The roles of marital status on CSS of OPSCC
The detailed univariate and Cox-regression multivariate analysis were shown in Table 2. The groups of age ≥60-year old, female, black race, other tumor location, low pathological grade group, the distant SEER stage, low C-SES, non-insurance, and non-treatment had signi cant worse 5-year CSS (P<0.001). Furthermore, these variables were also determined as independent risk factors for CSS of OPSCC by multivariate analysis. Forest plot was showed in Figure 1 according to multivariate Coxproportional hazard analysis.
Univariate analysis by Kaplan-Meier method showed that signi cant differences were observed in CSS among different marital groups. In a crude survival analysis, the widowed group was associated with a signi cant decreasing risk of CSS( P>0.001), as shown in Figure 2A. The 5-year CSS of widowed group was signi cantly worse than that of married, single and divorced groups. Furthermore, the widowed patients had a signi cant lower risk of cancer-speci c survival compared with married, single and divorced patients after controlling for the confounding risk factors(P<0.001), as showed in Figure 2B.

Subgroup analysis of the roles of marital status by SEER stage
The effects of marital status on CSS were evaluated regarding the SEER stage, as shown in Table 3. We found that widowed patients invariably had the worse 5-year CSS compared with married, single and divorced patients in the localized, regional and distant groups( Figure 3

Subgroup analysis the role of marital status by C-SES
The effects of marital status on CSS at different C-SES were assessed, as shown Table 4. Subgroup ananlysis showed that marital status was invariably determined to be an independent risk factors of CSS among OPSCC patients with low and high C- Propensity -score matching for different pairwise marital groups P-values adjusted method(BH) was used to test the statistically differences between pairwise marital groups. Then we found that there were signi cant differences between widowed and other marital groups(P<0.001), between married and other marital groups(P<0.001) respectively. Since the several baseline characteristics among different marital groups were signi cantly different, we performed propensity-score matching analysis to balance the confounding factors. If there was statistically signi cant differences of survival analysis between different marital status using Kaplan-mier methods, each group was matched a 1:1 ratio. The detailed variables after propensity-score matching were shown in Table 5.
Survival analysis after propensity-score matching Figure 5A and 5B showed that before and after PSM, CSS of widowed patients was always signi cantly worse than that of married patients (P<0.001). After adjusting for confounding factors (age, sex, race, pathological grade, tumor location, treatments, insurance, C-SES), widowhood was still statistically signi cant increasing risk of mortality (HR 1.366, 95%CI:1.246-1.497, P<0.001). Figure 5C and 5D showed before and after PSM of widowed and divorced patients, CSS of widowed patients was also signi cantly worse than that of divorced patients (P<0.001). After adjusting for confounding factors, widowed status was statistically signi cant risk factor of CSS (HR 1.255, 95%CI:1.145-1.376, P<0.001).
Pairwise comparison of different marital status effect on CSS before and after PSM, as such married and divorced (HR  Figure 5E to Figure 5J. After balancing the confounding prognostic factors, marital status was statistically signi cant risk factor for CSS of OPSCC patients.

Discussion
The effect of marital status on prognostic of various diseases was inconsistent. Some studies indicated married status had a better life expectancy and quality in cardiovascular disease and carcinomas [19,20,21], while no signi cant effects on survival of cancer patients were suggested in other studies [22,23]. Prior studies on marital status on survival of oral or oropharyngeal squamous cell carcinoma showed that marital status was an independent prognostic factor, and marriage had an protective effect on survival for for oral or oropharyngeal cancer [11,24]. However, these studies had not been combined clinicalpathological characteristics with socioeconomic characteristics of OPSCC, and analyzed survival of the pairwise marital patients after using prospensity-score matching without p adjusted method. Therefore, it is necessary to clarity the holistic impact of marital status on survival.
Based on clinicopathological and socioeconomic characteristics of OPSCC, marital status is de nitely identi ed as an independent prognostic risk factor for CSS after balancing age, sex, race,pathologic grade, SEER stage, tumor location, treatment modalities, insurance, C-SES using PSM and p-values adjusted method. Furthermore, we found that in both in multivariate survival analysis and pairwise matching survival after p-value adjusted method and propensity-scores matching, widowed patients invariably had the worst cancer-speci c survival compared with married, single and divorced patients. Our studies also indicated that prognostic factors of non-insurance and low composite socioeconomic status were determined to be decreasing risk factors of survival. Therefore, evaluating the socioeconomic status was necessary for the prognosis of OPSCC patients.
Marital status has been indicated to be associated with the prognosis in various malignant tumors. Prior studies have demonstrated that marriage had an independent protective effects on survival, and married patients had lower risk for metastatic caners, and were more inclined to receive de nite treatment. Perhaps, there were the strong social support and insurance coverage, better living habits and psychology state in the married populations [11,23,25,26,27]. Our ndings also con rmed the protective roles of marriage in OPSCC. Furthermore, the detailed classi cation of marital status indicated that widowed patients had the worst CSS compared with married, single and divorced patients.
Some studies found that the worse prognosis in widowed patients was due to delayed diagnosis and insu cient treatments.
Lack of spousal support and nancial assistance were mainly disadvantages in widowed patients [28,29,30]. Our studies also showed that the higher percentage of low composite socioeconomic status was in widowed patients, compared with married and divorced patients. Moreover, subgroup analysis in low and high C-SES indicated that widowed patients always had the worst CSS, compared with other marital status. Even if balancing the confounding factors using PSM, widowed patients still had the worst CSS, compared with others. The association between widowhood and poor survival may be attributed to the psychosocial factors [31]. The spousal death and the need of adaption to new social roles were very depressed and stressful for the surviving companion. Psychological depression could lead to poor medical compliance, and increasing the mortality risk of malignant tumor. [32] Marriage had protective roles for survival of oral cancer. However, widowed status was faced by almost everyone in the course of life. To explore the disadvantaged prognostic factors for survival seemed to be more pressing.Our studies also indicated that the importance of marital status in survival of OPSCC patients. Marriage had an advantage on cancer prognosis, and more compliant to proper treatments under their spousal support and encouragements. However, widowed patients seemed to be particularly prone to depression [33], and widowhood had a negative impact on immune status and hormone level.
However, a few researchers suggested that there were no signi cant differences in prognosis between malignant tumor and marital status. These authors suggested that the phenomenons may be originated from the limitations of the database, including the number of confounding prognostic factors, different clinicopathological and socioeconomic characteristics, and study design type [23,29]. In our studies, we extracted 3 standard 2000 US Census variables to set up a composite socioeconomic status variable, and explored the association between marital status and CSS after controlling socioeconomic and clinicopathologic variables using propensity-score matching method.
There are some limitations in our studies. First, the information of marital status in SEER databse was acquired only at the time of tumor diagnosis, and there were no changes during the course of therapy, even till death. Second, psychosocial factors, tobacco and alcohol consumption, and HPV infection were not including in SEER database, the association of psychosocial and survival should be further performed and validated. Third, although we utilized the PSM methods to reduce the bias, there were inherent bias of retrospective research. The evaluation of life quality was also an important goal missing from the SEER database.

Conclusion
Based on population level of SEER database, widowed patients with oral and oropharyngeal carcinoma were rst identi ed to have the worst cancer-speci c survival compared with married, single and divorced patients, and marital status was shown as an independent prognostic risk factor for survival after balancing the confounding prognositic factors using p-value adjusted method and propensity-score matching method. Psychosocial depression and loss of spousal support leading to the negative effects on the widowed patients should be further studied in multicenter, large samples, randomized control study.