To identify distinct behavioral patterns of diet, exercise, social interaction, church attendance, alcohol consumption, and smoking and to examine their association with subsequent dementia risk.DESIGN:
Longitudinal, population-based dementia study.SETTING:
Rural county in northern Utah, at-home evaluations.PARTICIPANTS:
Two thousand four hundred ninety-one participants without dementia (51% male, average age 73.0 ± 5,7; average education 13.7 ± 4.1 years) initially reported no problems in activities of daily living and no stroke or head injury within the past 5 years.MEASUREMENTS:
Six dichotomized lifestyle behaviors were examined (diet: high ≥ median on the Dietary Approaches to Stop Hypertension scale; exercise: ≥5 h/wk of light activity and at least occasional moderate to vigorous activity; church attendance: attending church services at least weekly; social Interaction: spending time with family and friends at least twice weekly; alcohol: currently drinking alcoholic beverages ≥ 2 times/wk; nonsmoker: no current use or fewer than 100 cigarettes ever). Latent class analysis (LCA) was used to identify patterns among these behaviors. Proportional hazards regression modeled time to dementia onset as a function of behavioral class, age, sex, education, and apolipoprotein E status. Follow-up averaged 6.3 ± 5.3 years, during which 278 cases of incident dementia (200 Alzheimer's disease (AD)) were diagnosed.RESULTS:
LCA identified four distinct lifestyle classes. Unhealthy–religious (UH-R; 11.5%), unhealthy–nonreligious (UH-NR; 10.5%), healthy–moderately religious (H-MR; 38.5%), and healthy–very religious (H-VR; 39.5%). UH-NR (hazard ratio (HR) = 0.54, P = .028), H-MR (HR = 0.56, P = .003), and H-VR (HR = 0.58, P = .005) had significantly lower dementia risk than UH-R. Results were comparable for AD, except that UH-NR was less definitive.CONCLUSION:
Functionally independent older adults appear to cluster into subpopulations with distinct patterns of lifestyle behaviors with different levels of risk for subsequent dementia and AD.