Midlife Marital Status and Subsequent Cognitive Decline over 20 Years: Discovery from ARIC

Background — Recent studies show that marriage is associated with a protective effect against cognitive decline among older adults. However, definite evidence from large prospective cohorts is lacking. Even less is known about the association between marital status during midlife and domain-specific cognitive decline in later life. Objective — To investigate the effect of midlife marital status on subsequent domain specific cognitive decline over 20 years among community dwelling older adults. Methods — Prospective cohort study of a biracial population of 14148 participants in the Artherosclerosis Risk in Communities (ARIC) study. Participants were followed from 1990-1992 through 2011-2013 and underwent three cognitive assessments at visit 2 (1990-1992), visit 4 (2004-2006) and visit 5 (2011-2013). Factor scores of cognitive functions were developed from neurocognitive tests during three visits and harmonized into three cognitive domains including executive, memory and language as well as general cognitive performance. Marital status was collected at visit 2 and categorized into four groups: married, divorced or separated, widowed, and never married. Multiple imputation using chained equations (MICE) was applied to address the challenge of informative attrition. Multivariable mixed effect linear models were applied after imputation with a spline term at year six to represent non-linear cognitive decline and a random slope to allow individual variations in cognitive decline after year six. Models were adjusted for age, sex, race, education, hypertension, diabetes, alcohol consumption, smoking status, and BMI measured at baseline. Results — The median age of 14148 participants was 56.8; 24.2% of participants were black and 55.6% were female. Median follow-up time was 20.78 years. For the first six years, there was no statistically significant cognitive decline difference comparing groups without partners at baseline to the married group. After year six, models suggested that being widowed at baseline was associated with greater cognitive decline compared to the married group across all cognitive domains. In the general cognitive domain, the difference in annual cognitive decline comparing the widowed group to the married group was -0.019 standard deviation (SD) units [95%CI (-0.023, -0.015)] (p < 0.001). Similarly, the corresponding difference in cognitive decline in the memory domain was -0.017 [95%CI (-0.023, -0.012), p< 0.001] unit, and for language was -0.028 [ 95%CI (-0.032, -0.024) , p<0.001 ]. For participants who were divorced or separated, annual cognitive decline only presented statistically significant difference in language domain with -0.006 [95%CI (-0.010, -0.003) , p=0.001] unit greater decline but not in other cognitive domains. There was no difference in the rate of cognitive decline between the married and never-married group.Conclusion — Widowhood by midlife may increase the risk of cognitive decline across all cognitive domains, while being divorced or separated in midlife was associated with greater annual cognitive decline in language domain. Being never married is not associated with greater cognitive decline.