PT  - JOURNAL ARTICLE
AU  - Cécilia Samieri
AU  - Abhijeet Rajendra Sonawane
AU  - Sophie Lefèvre-Arbogast
AU  - Catherine Helmer
AU  - Francine Grodstein
AU  - Kimberly Glass
TI  - Using network science tools to identify novel diet patterns in prodromal dementia
AID  - 10.1212/WNL.0000000000009399
DP  - 2020 May 12
TA  - Neurology
PG  - e2014--e2025
VI  - 94
IP  - 19
4099  - http://n.neurology.org/content/94/19/e2014.short
4100  - http://n.neurology.org/content/94/19/e2014.full
SO  - Neurology2020 May 12; 94
AB  - Objective To use network science to model complex diet relationships a decade before onset of dementia in a large French cohort, the 3-City Bordeaux study.Methods We identified cases of dementia incident to the baseline food frequency questionnaire over 12 years of follow-up. For each case, we randomly selected 2 controls among individuals at risk at the age at case diagnosis and matched for age at diet assessment, sex, education, and season of the survey. We inferred food networks in both cases and controls using mutual information, a measure to detect nonlinear associations, and compared food consumption patterns between groups.Results In the nested case-control study, the mean (SD) duration of follow-up and number of visits were 5.0 (2.5) vs 4.9 (2.6) years and 4.1 (1.0) vs 4.4 (0.9) for cases (n = 209) vs controls (n = 418), respectively. While there were few differences in simple, average food intakes, food networks differed substantially between cases and controls. The network in cases was focused and characterized by charcuterie as the main hub, with connections to foods typical of French southwestern diet and snack foods. In contrast, the network of controls included several disconnected subnetworks reflecting diverse and healthier food choices.Conclusion How foods are consumed (and not only the quantity consumed) may be important for dementia prevention. Differences in predementia diet networks, suggesting worse eating habits toward charcuterie and snacking, were evident years before diagnosis in this cohort. Network methods, which are designed to model complex systems, may advance our understanding of risk factors for dementia.DASH=Dietary Approaches to Stop Hypertension; DSM-IV=Diagnostic and Statistical Manual of Mental Disorders, 4th edition; FFQ=food frequency questionnaire; MeDi=Mediterranean diet; MI=mutual information; MIND=Mediterranean-DASH Intervention for Neurodegenerative Delay; 3C=3-City