Dietary Patterns Derived by Hybrid Clustering Method in Older People: Association with Cognition, Mood, and Self-Rated Health

BackgroundSeveral nutritional factors, including dietary fatty acids, antioxidants, and folates, have been related to pathological brain aging. Dietary patterns that represent a combination of foods may better predict disease risk than single foods or nutrients.ObjectiveTo identify dietary patterns by a mixed clustering method and to analyze their relationship with cognitive function, depressive symptoms, and self-rated health in older people.DesignCross-sectional population-based study.Subjects/settingSubjects included 1,724 elderly community dwellers living in Bordeaux, France from 2001 to 2002.Statistical analysisCluster analysis, combining hybrid clustering, and research for stable groups during the k-means step on mean number of weekly servings of 20 predetermined food groups, separately in men and women.ResultsFive dietary clusters were identified in each sex. A “healthy” cluster characterized by higher consumption of fish in men (n=157; 24.3%) and fruits and vegetables in women (n=267; 24.8%) had significantly lower mean number of errors to Mini Mental State score after adjustment for socio-demographic variables (β=−0.11; 95% confidence interval [CI], −0.22 to −0.004 in men; β=−0.13; 95% CI, −0.22 to −0.04 in women). The same cluster was associated with borderline significance with lower depressive symptoms in women (β=−0.16; 95% CI, −0.33 to 0.007). Men in the “pasta eaters” cluster (n=136; 21%) had higher depressive symptoms (β=0.26; 95% CI, 0.06 to 0.46) and higher risk to report poor health (polytomous regression, odds ratio [OR]=1.91; 95% CI, 1.21 to 3.01) than the “healthy” cluster. Women in the “biscuits and snacking” cluster (n=162; 15%) had greater risk of poor perceived health (OR=1.69; 95% CI, 1.15 to 2.48) compared to “healthy” eaters. Additional adjustment for body mass index and medication use strengthened these associations.ConclusionsSex-specific dietary patterns derived by hybrid clustering method are associated with fewer cognitive and depressive symptoms and better perceived health in older people.