Association of Marital Status With Total and Cause-Specific Mortality in Asia

Key Points

Question  What is the association between marital status and mortality outcomes in Asian populations?

Findings  In this cohort study of 623 140 individuals from 16 prospective cohorts participating in the Asia Cohort Consortium, a pooled analysis found that being unmarried was associated with a higher risk of all-cause and cause-specific mortality, compared with being married. The excess risks of death persisted across health conditions and were particularly pronounced among men and people younger than 65 years.

Meaning  These findings highlight that being unmarried substantially increases the risk of death among Asian individuals.

Importance  Marital status has been shown to be associated with mortality, but evidence in Asian populations is limited.

Objective  To examine the association of marital status with total and cause-specific mortality.

Design, Setting, and Participants  This cohort study included individual participant data from 16 prospective studies in the Asia Cohort Consortium conducted between 1963 and 2015. Asian participants with complete information on marital and vital status were included. Study-specific hazard ratios (HRs) and 95% CIs were estimated using Cox proportional hazards model and then pooled using a random-effects meta-analysis. The analysis began in February 2021 and ended in August 2021.

Exposures  Marital status.

Main Outcomes and Measures  All-cause and cause-specific mortality.

Results  Of 623 140 participants (326 397 women [52.4%] and 296 743 men [47.6%]; mean [SD] age, 53.7 [10.2] years; mean [SD] follow-up time, 15.5 [6.1] years), 123 264 deaths were ascertained. Compared with married individuals, those who were unmarried had pooled HRs of 1.15 (95% CI, 1.07-1.24) for total mortality, 1.12 (95% CI, 1.03-1.22) for cerebrovascular disease mortality, 1.20 (95% CI, 1.09-1.31) for coronary heart disease mortality, 1.17 (95% CI, 1.07-1.28) for circulatory system diseases mortality, 1.06 (95% CI, 1.01-1.11) for cancer mortality, 1.14 (95% CI, 1.05-1.23) for respiratory diseases mortality, and 1.19 (95% CI, 1.05-1.34) for external causes of death. Positive associations with total mortality were also observed for those who were single (HR, 1.62; 95% CI, 1.41-1.86), separated (HR, 1.35; 95% CI, 1.13-1.61), divorced (HR, 1.38; 95% CI, 1.13-1.69), and widowed (HR, 1.09; 95% CI, 1.04-1.13). In subgroup analyses, the positive association persisted across baseline health conditions, and the risk of death was more pronounced among men or people younger than 65 years.

Conclusions and Relevance  This large pooled cohort study of individual participant data provides strong evidence that being unmarried, as well as belonging to the unmarried subcategories, was positively associated with total and cause-specific mortality. Investment of targeted social support services might need to be considered in light of the mortality differences between married and unmarried individuals.

Globally, the marriage rate has decreased over the last decades.¹ There is broad agreement on the association of marital status with mortality.^2-4 A recent meta-analysis³ showed that individuals who were unmarried had a higher risk of all-cause and cause-specific mortality than their married counterparts. Achieving socioeconomic support and health-promoting behaviors have been implicated to be the factors contributing to these protective associations of marriage.^5-9 Physiologically, a higher level of cortisol and a flatter diurnal cortisol slope are found in unmarried individuals,¹⁰ which are linked to an increased risk of carotid atherosclerosis,¹¹ poor glycemic control in patients with diabetes,¹² metabolic syndrome,¹³ and shorter survival time in patients with cancer.^14,15

Although the pattern of marriage differs among societies, few studies have reported the linkage between marriage and mortality in non-Western populations.^16-20 Marriages in East Asia have distinct features, including a strong tradition of extended family coresidence, an emphasis on family lineage that stabilizes the institution of marriage, and strong family ties,²¹ which have been associated with lower mortality risk.²² In addition, the pronounced marital selection outcomes and financial burden of single-earner households in Asian society have been suggested as factors associated with higher mortality among single Asian individuals.²³ Together, we hypothesized that being married would be associated with survival benefits in Asian populations. Nevertheless, previous studies^16-20 of Asian cohorts have published inconsistent results. It is also noteworthy that, in previous investigations, there were concerns regarding the potential effects of reverse causation, where initial presentation of undiagnosed diseases may influence the likelihood of marriage dissolution.²⁴ In addition, the association may differ across participant subgroups. Prior studies investigating potential modifying effects of age and sex demonstrated mixed results.^3,25-29 With most studies focused on the generally healthy population,³ less is known regarding the mortality differences by marital status among individuals with chronic diseases. Given the increasing prevalence of noncommunicable diseases in Asian countries,³⁰ it is imperative to understand whether the association between marital status and mortality differs among people with underlying comorbidity. We therefore conducted a cohort study using individual-level data from the Asia Cohort Consortium (ACC), with more than 1 million participants, to ascertain the association between marital status and cause-specific mortality.

The ACC was established to foster collaborative epidemiological research in Asian countries using pooled multicenter cohort data. The ACC currently contains individual-level data from more than 1 million participants. Of 23 prospective cohort studies in the ACC, we used data from 16 cohorts with available data on marital status. The study designs of the included cohorts have been described elsewhere.³¹ Participating cohorts from 5 countries were included: Mainland China, Shanghai Cohort Study, Shanghai Men’s Health Study, and Shanghai Women’s Health Study; Japan, Japan Collaborative Cohort Study, Japan Public Health Center-Based Prospective Study, Miyagi Cohort Study, Ohsaki National Health Insurance Cohort Study, Life Span Study Cohort–Radiation Effects Research Foundation, and Takayama Study; Republic of Korea, Korea Multi-Center Cancer Cohort Study, Korea National Cancer Center Cohort, and Seoul Male Cancer Cohort; Singapore, Singapore Chinese Health Study; and Taiwan, Community-Based Cancer Screening Project and Cardiovascular Disease Risk Factor Two-Township Study. Of the 681 533 potential participants, we excluded those with missing information on marital status (37 600 individuals) or vital status (20 793 individuals). This study was approved by the ethics committees of each cohort study and by the institutional review board of the ACC coordinating center. Informed consent (either written or oral, depending on the cohort) was obtained from all participants per study cohort. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Cohort questionnaires at baseline assessed information on lifestyle factors, sociodemographic characteristics, diet, and medical history of type 2 diabetes, hypertension, cancer, coronary heart disease, and cerebrovascular disease. Among 16 cohorts included, participants from Korea National Cancer Center Cohort, Korea Multi-Center Cancer Cohort Study, Community-Based Cancer Screening Project, and Cardiovascular Disease Risk Factor Two-Township Study were interviewed by trained interviewers, whereas the other cohorts used self-administered questionnaires. Brief descriptions of participating cohorts are presented in eTable 1 in the Supplement. Information on marital status from each cohort was harmonized and classified into 5 categories: married, single, separated, widowed, and divorced. Analyses for subcategories of unmarried status (single, separated, widowed, and divorced) were performed where data were available (eAppendix in the Supplement).

Our primary end point was death from all causes. Deaths were ascertained via linkage to the death certificates or active follow-up. The causes of death were classified using the International Classification of Diseases, Ninth Revision or International Classification of Diseases, Tenth Revision (eAppendix in the Supplement).

In this study, we performed a 2-stage meta-analysis of individual participant data to estimate the pooled hazard ratios (HRs) and 95% CIs.³² First, a Cox proportional hazards regression model with age as the time scale was used to estimate HRs and 95% CIs across marital status categories by cohorts. Married individuals were the reference group. Time of entry was age at the baseline interview, and exit time was age at death or last follow-up, whichever came first. The multivariable model was adjusted for potential confounders, including sex, age at baseline (year, continuous), education (no formal education or primary education, secondary education, trade or technical education, university education or above, and missing), smoking status (never, former, current, and missing), alcohol intake (nondrinker, drinker, and missing), physical activity (none or almost none, low, intermediate, high, and missing), and baseline health conditions (cerebrovascular disease, coronary heart disease, cancer, hypertension, and diabetes). We used missing indicator variables to code missing values for potential confounders. Second, we obtained the overall estimates using random-effects meta-analysis. To minimize the bias due to reverse causation, we repeated the analyses and excluded deaths occurring within the first 5 years of follow-up.

We performed prespecified stratified analysis according to age at baseline (<65 and ≥65 years), sex (men and women), health status (prior diagnosis of cancer, cerebrovascular disease, or coronary heart disease; prior diagnosis of type 2 diabetes or hypertension but not cancer, cerebrovascular disease, or coronary heart disease; and healthy participants with none of the aforementioned diseases), birth year (before and after 1940), and country. Metaregression was used to test for effect modification. We used Stata statistical software version 16.0 (StataCorp) for analysis. All reported P values were 2-sided, and P < .05 was considered to be significant. Data analysis was performed from February to August 2021.

A total of 623 140 participants were included in the analyses: 296 743 men (47.6%) and 326 397 women (52.4%). The mean (SD) age at baseline was 53.7 (10.2) years. During a mean (SD) follow-up of 15.5 (6.1) years, 123 264 deaths were ascertained, of which 37 394 were classified as death from circulatory system diseases, 8013 were from coronary heart disease, 14 563 were from cerebrovascular disease, 41 362 were from cancer, 13 583 were from respiratory diseases, and 7795 were from external causes. The mean prevalence of being married was 86.4% (538 377 individuals), ranging from 76.4% (36 046 individuals) in the Life Span Study Cohort–Radiation Effects Research Foundation to 98.7% (13 777 individuals) in the Seoul Male Cancer Cohort. Baseline characteristics of participants of each cohort are shown in Table 1 and eTable 2 in the Supplement.

In multivariable analyses, being unmarried was associated with a 15% greater risk of all-cause mortality (pooled multivariable HR, 1.15; 95% CI, 1.07-1.24) compared with being married (Figure and eTable 3 in the Supplement). We further examined associations of subcategories of unmarried status with mortality. Compared with married participants, positive associations for death from all causes were found among those who were single (pooled HR, 1.62; 95% CI, 1.41-1.86), separated (pooled HR, 1.35; 95% CI, 1.13-1.61), divorced (pooled HR, 1.38; 95% CI, 1.13-1.69), and widowed (pooled HR, 1.09; 95% CI, 1.04-1.13). The associations remained positive when we excluded participants who died during the first 5 years of follow-up (Figure and eTable 3 in the Supplement).

The pooled multivariable HRs for unmarried compared with married individuals were 1.12 (95% CI, 1.03-1.22) for cerebrovascular disease mortality, 1.20 (95% CI, 1.09-1.31) for coronary heart disease mortality, 1.17 (95% CI, 1.07-1.28) for circulatory system diseases mortality, 1.06 (95% CI, 1.01-1.11) for cancer mortality, 1.14 (95% CI, 1.05-1.23) for respiratory diseases mortality, and 1.19 (95% CI, 1.05-1.34) for external causes of death (Table 2). Compared with those who were married, being single was associated with a greater risk of death from cerebrovascular disease, coronary heart disease, circulatory system diseases, cancer, respiratory diseases, and external causes (Table 2). Being widowed was associated with higher risks of death from cerebrovascular disease, coronary heart disease, circulatory system diseases, and external causes. There were higher risks of coronary heart disease mortality and external causes of death among separated participants. Being divorced was positively associated with risk of death from cerebrovascular disease, with similar magnitude of associations observed with risk of death from circulatory system diseases, respiratory diseases, and external causes (Table 2). Similar patterns were observed after the exclusion of mortality cases with the first 5 years of follow-up.

We further evaluated the association between marital status and mortality risk stratified by health status at baseline (Table 3). In subgroup analyses defined by variables (e.g. disease status at baseline), there were positive associations between unmarried and total mortality in each subgroup (Table 3). An insignificant P value for interaction indicated that there was no difference between 3 groups: (1) cancer, coronary heart disease, or cerebrovascular disease; (2) diabetes or hypertension only; and (3) healthy. We found that being single was associated with higher risk of mortality among healthy participants than among those with a history of cancer, cerebrovascular disease, or coronary heart disease (P for interaction = .01). The association for the separated was also modified by health status (P for interaction = .01). In subgroup analyses, we found that the associations between being unmarried, widowed, or divorced and risk of death from all causes were modified by sex (Table 3). Positive associations were found among unmarried (P for interaction = .003), widowed (P for interaction =. 049), and divorced (P for interaction = .047) men, but no association was found among women. The associations between marital status (unmarried or single) and risk of total mortality were modified by age (P for interaction < .001) (Table 3). The HRs were higher among single participants younger than 65 years (HR, 1.79; 95% CI, 1.54-2.08) than those who were aged 65 years and older (HR, 1.11; 95% CI, 1.02-1.22) (P for interaction < .001). Subgroup analyses according to birth year and country are presented in eTable 4 in the Supplement.

In this cohort study using pooled data for 623 140 Asian individuals from 16 prospective cohorts, we demonstrated that unmarried individuals had a 15% increased risk of total mortality, after adjusting for potential confounders. Compared with married participants, those who were single, separated, widowed, and divorced had a greater risk of death from all causes. The results were similar after excluding deaths occurring within the first 5 years, which confirmed the robustness of our findings. This pooled analysis further showed that being unmarried was associated with a greater risk of death from 6 specific causes. The large sample size of the ACC allowed us to examine whether the association between marital status and mortality risks differed across subgroups of sex, age, and health condition at baseline.

Our results confirmed the findings in a recent meta-analysis,³ which reported pooled risk ratios (RRs) of 1.33 (95% CI, 1.24-1.43) for death from all causes, 1.37 (95% CI, 1.21-1.55) for cancer, and 1.11 (95% CI, 1.07-1.15) for circulatory system diseases for unmarried vs married individuals. The evidence for respiratory disease mortality has been sparse, however. In accordance with our findings, data from the Japan Collaborative Cohort Study¹⁶ and the National Cohort Study in Denmark²⁵ showed a greater risk of death from respiratory diseases among those who were unmarried compared with married individuals.

Although the association between unmarried subcategories with mortality has been extensively investigated, the results are inconclusive.^2-4 An early meta-analysis² found that the positive association was greater among divorced or separated individuals (RR, 1.16; 95% CI, 1.09-1.23) than among widowed individuals (RR, 1.11; 95% CI, 1.08-1.14) or those who were never married (RR, 1.11; 95% CI, 1.07-1.15); however, only 5 of the 53 included studies were from Asian populations, and they were limited by small samples. More studies have been conducted since but with conflicting results (HR for single individuals, 1.54; 95% CI, 1.46-1.63; HR for divorced or separated individuals, 1.43; 95% CI, 1.34-1.52; HR for widowed individuals, 1.23; 95% CI, 1.19-1.26).³ In our study of 623 140 Asian individuals with a mean follow-up of 15.5 years, we found that, among all unmarried subcategories, being single was most associated with death from all causes. These findings could be explained by self-selection of single individuals with less favorable socioeconomic and psychosocial status to enter into marriage³³ and by the disparity in social support across subgroups of unmarried individuals. The duration of living with a spouse may also play an important role in mortality risks, because individuals who were divorced, separated, or widowed had been cohabitating with a spouse for a time during marriage. Similar findings were seen for deaths from other major causes, except for coronary heart disease mortality and external causes of death, for which separation was associated with greater risks than other unmarried status. The observed associations may be a consequence of the reduction in the ability to cope with stress accrued from separation.⁴

Prior literature reported decreased survival for unmarried patients with known cardiovascular disease^4,24 and cancer.³⁴ Although marital status has been associated with the incidence of hypertension³⁵ and type 2 diabetes,³⁶ none of the previous studies has explored mortality outcomes by marital status after the onset of hypertension or diabetes, rendering interpretation of long-term outcomes difficult. To our knowledge, the current study is the first to examine the association between marital status and mortality risks among participants with various health statuses within the same cohort. Our results suggest that there is a survival advantage not only for married individuals with life-threatening diseases (cancer, coronary heart disease, and cerebrovascular disease) but also for those with diabetes or hypertension. The protective outcomes of marriage among people with chronic diseases may be associated with the encouragement from partners to seek medical help and adhere to treatment.^37,38 Given the wealth of evidence indicating the positive association between marital status and mortality, the importance of adequate support to unmarried individuals with chronic diseases cannot be neglected.

The current analysis found a significant interaction between age and all-cause mortality among unmarried groups, especially for the single and the separated. Sorlie et al²⁶ reported that unmarried participants younger than 65 years experienced a higher risk of mortality than those who were older. The smaller association in elderly populations may be due to extramarital support from relatives or community, diminished health-promoting outcomes of marriage, and reduction of social support disparity between married and unmarried groups.^2,37 The exact reasons still need to be explored. Some previous studies found that sex is a modifier of the association of marriage with death,^27,28 whereas others failed to demonstrate the disparity.^25,29 One meta-analysis³ summarizing 3 studies from Asia and 12 studies from America and Europe reported a women-to-men relative RR of 0.86 (95% CI, 0.79-0.94) for total mortality associated with being unmarried. In our study, we found that being unmarried was associated with higher mortality among men but not among women, reflecting the possibility that, in Asian marriages, gender inequality in the division of domestic labor and high expectations of childcare for women²¹ may counteract the health benefits of marriage among women. Another possible explanation is that, compared with married women, a higher employment rate among unmarried women may have contributed to financial security and favorable health outcomes.^16,39 Furthermore, unmarried men may receive less social support and financial protection from the government than unmarried women in Asia.^16,39

The strength of our pooled analyses includes the prospective design of each study, which minimized potential recall bias and selection bias. In addition, to our knowledge, the present study is the largest pooled analysis of Asian populations from 16 cohorts, which allowed us to provide estimates with sufficient precision and to examine reverse causation.

However, our study has limitations that merit further discussion. First, given the observational nature of the study and the lack of data on income, residual confounding could exist. Second, we excluded participants with missing information on marriage and vital status, and this may lead to selection bias if participants with complete information are not representative of the cohort population. Also, individuals with higher socioeconomic status may be more likely to participate in the study and less likely to be lost to follow-up. Third, information regarding cohabitation and changes in marital status was not captured in our data set, and misclassification is possible. However, later life marital dissolution may be more dominant than remarriage, which could diminish the observed associations. In addition, the rate of nonmarital cohabitation is likely to be small in the included countries of the present analysis.²¹ Such participants would be grouped as unmarried in the analyses and were expected to show better survival outcomes, which, in turn, biased the association toward the null. Fourth, the likelihood of alterations in marital status before the baseline survey may be influenced by initial symptoms of undiagnosed diseases. To minimize bias from reverse causation, we performed analysis excluding deaths within the first 5 years of follow-up. Similar results were observed, suggesting these biases could be minimal in our findings. Fifth, covariates were assessed at baseline; therefore, the changes in these variables over time were not considered. Although we have included 16 population-based cohorts in Asia, we acknowledge that the heterogeneity within the region may not be fully represented by this study. Sixth, the mean age of 53.7 years at baseline was young with regard to the outcome of mortality, despite the long follow-up of 15.5 years.

In this large-scale cohort study pooling data from multiple cohorts, being unmarried was associated with a higher risk of total and cause-specific mortality. The association persisted across baseline health conditions and was particularly evident in men and participants who were younger than 65 years at baseline. Our findings underscore the potential to reduce unfavorable mortality outcomes among unmarried individuals in Asia.

Accepted for Publication: March 27, 2022.

Published: May 31, 2022. doi:10.1001/jamanetworkopen.2022.14181

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Leung CY et al. JAMA Network Open.

Corresponding Author: Manami Inoue, MD, PhD, Division of Prevention, National Cancer Center Institute for Cancer Control, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, Japan (mnminoue@ncc.go.jp).

Author Contributions: Drs Leung and Huang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Leung and Huang share co–first authorship.

Concept and design: Leung, Huang, Abe, Boffetta, Chia, Kang, Inoue.

Acquisition, analysis, or interpretation of data: Leung, Huang, Abe, Saito, Islam, Rahman, Ikeda, Sawada, Tamakoshi, Gao, Koh, Shu, Sakata, Tsuji, Kim, Park, Nagata, You, Yuan, M.-H. Shin, Pan, Tsugane, Kimura, Wen, Cai, Ozasa, Matsuyama, Kanemura, Sugawara, A. Shin, Wada, Chen, Wang, Ahn, Ahsan, Boffetta, Matsuo, Qiao, Rothman, Zheng, Inoue.

Drafting of the manuscript: Leung, Huang.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Leung, Huang, Tsuji, Wen, Sugawara, Wada, Wang.

Obtained funding: Shu, Tsuji, Yuan, Ahsan, Matsuo, Zheng, Inoue.

Administrative, technical, or material support: Abe, Islam, Rahman, Ikeda, Sawada, Gao, Tsuji, Nagata, You, Yuan, M.-H. Shin, Tsugane, Cai, Kanemura, A. Shin, Chen, Ahn, Ahsan, Boffetta, Chia, Matsuo, Qiao, Rothman, Zheng.

Supervision: Abe, Gao, Park, Ozasa, A. Shin, Boffetta, Kang, Inoue.

Conflict of Interest Disclosures: Dr Shu reported receiving grants from the National Cancer Institute during the conduct of the study. Dr Chen reported receiving grants from Ministry of Science and Technology and Ministry of Health and Welfare during the conduct of the study. No other disclosures were reported.

Funding/Support: The cohorts participating in the pooled analysis were supported by the following grants: Shanghai Cohort Study, National Institutes of Health (R01CA0403092, R01CA144034, UM1CA182876); Shanghai Men’s Health Study, US National Cancer Institute (UM1 CA173640, Principal Investigator, Xiao-Ou Shu); Shanghai Women’s Health Study, US National Cancer Institute (R37 CA070867 and UM1 CA182910, Principal Investigator: Wei Zheng); Japan Public Health Center-Based Prospective Study, National Cancer Center Research and Development Fund (23-A-31[toku], 26-A-4 and 2020-A-4, since 2011) and a Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (from 1989 to 2010); Japan Collaborative Cohort Study, National Cancer Center Research and Development Fund (a grant-in-aid for cancer research), Ministry of Health, Labour and Welfare, Japan (grant for health services and grant for comprehensive research on cardiovascular and life-style related diseases), and Ministry of Education, Culture, Sports, Science and Technology, Japan (grant for the scientific research); Miyagi Cohort Study, National Cancer Center Research and Development Fund; Ohsaki National Health Insurance Cohort Study, National Cancer Center Research and Development Fund; Life Span Study Cohort–Radiation Effects Research Foundation, The Japanese Ministry of Health, Labour and Welfare and the US Department of Energy; Takayama Study, National Cancer Center Research and Development Fund; Korea Multi-Center Cancer Cohort Study, National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (2016R1A2B4014552); Korea National Cancer Center Cohort, National Cancer Center Research Grant (1510040, 1810090, and 1910330); Seoul Male Cancer Cohort, National R&D Program for Cancer Control, Ministry of Health & Welfare, Republic of Korea (0520160-1); Singapore Chinese Health Study, US National Cancer Institutes (R01CA144034 and UM1CA182876); Community-Based Cancer Screening Project, Ministry of Health and Welfare and Ministry of Science and Technology, Taiwan; and CardioVascular Disease Risk Factor Two-Township Study, Department of Health, Taiwan (DOH80-27, DOH81-021, DOH8202-1027, DOH83-TD-015, and DOH84-TD-006).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank all research team members and participants of each cohort study for their contribution to this research.