Dietary diversity is associated with longitudinal changes in hippocampal volume among Japanese community dwellers

Dementia, including Alzheimer’s disease, is a significant and growing global health issue with no curative treatments, highlighting the need for preventive strategies. Lifestyle factors, including diet, are implicated in dementia risk. While some studies suggest the Mediterranean diet may help prevent dementia, a systematic review found insufficient randomized trial evidence to confirm benefits on cognitive impairment or dementia. The Japanese diet features a wide variety of foods and high dietary diversity, which has been associated internationally with longer healthy life expectancy independent of GDP. Prior work in older Japanese adults linked higher dietary diversity to better cognitive function measured by the Mini-Mental State Examination, but objective measures less susceptible to retest effects are needed. Structural neuroimaging allows objective assessment of brain regions vulnerable in dementia, particularly the hippocampus, which is central to memory and shows atrophy in Alzheimer’s disease and other dementias. Although prior studies associated diet quality (e.g., Mediterranean or Western patterns, Healthy Eating Score) with hippocampal volume, these studies were not large-scale, only one assessed longitudinal change, and none examined dietary diversity. This study investigates, in a large sample of 1683 middle-aged and older Japanese community dwellers, the longitudinal relationship between dietary diversity and hippocampal volume to evaluate effects of daily diet on structural brain changes.

Evidence on diet and brain health indicates potential benefits of diet quality on cognitive outcomes, but randomized trial evidence for the Mediterranean diet’s benefits on cognitive impairment or dementia remains insufficient. Japan exhibits high dietary diversity and healthy life expectancy, and higher dietary diversity has been linked to better cognitive function among older Japanese in prior work using the MMSE. Neuroimaging studies have associated greater adherence to the Mediterranean diet with larger hippocampal volume, Western diet with smaller hippocampal volume over time, and higher long-term Healthy Eating Scores with larger hippocampal volumes. However, these studies involved smaller samples (n<1000), only one examined changes in brain volume longitudinally, and none focused on dietary diversity. This work addresses these gaps by evaluating longitudinal associations between dietary diversity and hippocampal volume.

Design and participants: Data came from the National Institute for Longevity Sciences-Longitudinal Study of Aging (NILS-LSA), a population-based cohort of community-dwelling residents near the National Center for Geriatrics and Gerontology (Obu City and Higashiura Town, Aichi, Japan). The first wave (1997–2000) enrolled 2267 participants aged 40–79 years, with follow-ups every 2 years and replenishment of younger age groups. For this study, participants from waves 6 (July 2008–July 2010) and 7 (July 2010–July 2012), where 3D MRI data were available, were included. Exclusions among wave 6 participants (n=2302) were: non-participation in wave 7 (n=315), missing/defective MRI (n=153), history of dementia (n=4), history of head surgery (n=21), new cerebrovascular lesions on MRI during follow-up (n=2), incomplete nutritional (n=113) or lifestyle questionnaire data (n=11). The final analytic sample was 1683 individuals (851 men, 832 women) aged 40–89 years. Dietary assessment and dietary diversity: After wave 6 participation, subjects completed 3-day dietary records (2 weekdays, 1 weekend day) at home, weighing foods on a 1-kg kitchen scale or estimating portions, and photographed meals before/after eating with a disposable camera. Dietitians used photos to verify and complete records, with follow-up calls as needed. Intakes of foods and nutrients (including alcohol) were computed based on the Standard Tables of Food Composition in Japan 2010 and other sources. An original supplement database was created, but due to limited nutrient information and omission of some compounds (e.g., phosphatidyl serine, choline) and limited published values from companies, nutrient intakes from supplements were not considered. Dietary diversity at baseline was quantified using the Quantitative Index for Dietary Diversity (QUANTIDD), calculated from the distribution across 13 food groups (cereals, potatoes, beans, nuts and seeds, non-green-yellow vegetables, green-yellow vegetables, fruit, mushrooms, seaweed, fish and shellfish, meat, eggs, milk and dairy). The index ranges from 0 (unbalanced) to 1 (even distribution). Quintiles of the dietary diversity score were computed by sex. MRI acquisition and processing: MRI scans used a 3.0-T Siemens Magnetom Tim Trio with an MPRAGE sequence (TR/TE/TI=1800/1.98/800 ms, 9° flip angle, 0.98×0.98×1.1 mm³, 256×256 matrix). FreeSurfer v5.3 performed cortical surface reconstruction and regional gray matter volume estimation using standard preprocessing, GM/WM segmentation, surface tessellation, topology correction, cortical parcellation (Desikan-Killiany atlas), and subcortical segmentation. The longitudinal processing stream was applied to reduce inter-individual variability. Cases with failed processing (e.g., segmentation failures) were excluded. Volumes (cm³) for bilateral hippocampus and total gray matter were derived and normalized by estimated total intracranial volume (eTIV) by dividing individual volumes by individual eTIV and multiplying by the cohort mean eTIV. Other measurements and covariates: Baseline assessments included years of education (≤9, 10–12, ≥13 years), current smoking status, and medical history (hypertension, stroke, heart disease, diabetes mellitus, dyslipidemia) via questionnaire. Physical activity was assessed as 24-hour MET-hours/day via interviewer-administered semi-quantitative assessment. Outcomes and statistical analysis: Brain volume change from baseline (wave 6) to follow-up (wave 7) was expressed as absolute difference (cm³) and percent decrease [(baseline−follow-up)/baseline ×100]. Group differences across dietary diversity quintiles used chi-squared tests (categorical) and ANOVA (continuous); trends used Cochran-Armitage or general linear models. General linear models estimated mean brain volumes and changes across quintiles with adjustments: Model 1 adjusted for sex, age (continuous), education, current smoking, alcohol intake (mL/day, continuous), physical activity (MET-hours/day, continuous), and histories of stroke, dyslipidemia, diabetes, hypertension, and heart disease; Model 2 additionally adjusted for the corresponding baseline brain volume (total gray matter or hippocampal volume). SAS 9.3 was used. Significance thresholds were p<0.05 (significant) and p<0.1 (marginal).

- Cohort-wide change: Over 2.01 (±0.12) years, mean percent decreases were 1.00% (±2.27%) for hippocampal volume and 0.78% (±1.83%) for total gray matter, corresponding to annual decreases of ~0.50% and ~0.39%, respectively. - Baseline characteristics by dietary diversity quintile: Higher dietary diversity was associated with older age, lower physical activity, lower prevalence of current smoking, and higher prevalence of hypertension, dyslipidemia, and diabetes. Crude baseline brain volumes (hippocampal and total gray matter) were lower in higher dietary diversity groups, but after adjustment for age and covariates, baseline hippocampal and total gray matter volumes were not significantly associated with dietary diversity. - Total gray matter change: Difference in total gray matter volume from baseline to 2 years was associated with baseline dietary diversity (Model 1: p=0.064; trend p=0.018). After additional adjustment for baseline total gray matter volume (Model 2), associations were attenuated (p=0.102; trend p=0.028). Percent decrease showed similar associations (Model 1: p=0.065; trend p=0.017). - Hippocampal change: Greater dietary diversity was significantly associated with smaller hippocampal volume loss over 2 years. Estimated mean (±SE) absolute decreases across ascending dietary diversity quintiles were −0.109 cm³ (±0.010), −0.088 cm³ (±0.010), −0.082 cm³ (±0.010), −0.070 cm³ (±0.010), and −0.072 cm³ (±0.010) (Model 1 p=0.042; trend p=0.004). The association remained after adjustment for baseline hippocampal volume (Model 2). Percent decreases across quintiles were 1.31% (±0.12%), 1.07% (±0.12%), 0.98% (±0.12%), 0.81% (±0.12%), and 0.85% (±0.12%) (Model 1 p=0.030; trend p=0.003). - Dietary intake patterns: Higher dietary diversity was characterized by lower cereal intake and higher intakes of the other 12 food groups. Energy intake did not differ by diversity, but protein and several micronutrients (e.g., sodium, calcium, magnesium, iron, zinc, vitamins A and C) increased with higher diversity.

In this large longitudinal cohort of Japanese community dwellers, higher baseline dietary diversity was associated with a smaller subsequent decrease in hippocampal volume over two years, suggesting that a more varied diet may help preserve hippocampal structure, a region crucial for memory and often affected early in dementia. Associations with total gray matter change were weaker but showed similar trends. The observed annualized hippocampal and gray matter volume decreases (~0.50% and ~0.39%, respectively) align with prior reports of age-related brain atrophy, though hippocampal atrophy in this cohort appears somewhat lower than meta-analytic estimates. The findings extend prior research linking diet quality to hippocampal volume by demonstrating, in a larger sample and with longitudinal imaging, that dietary diversity itself relates to reduced hippocampal atrophy. These results underscore the potential relevance of diverse dietary patterns to brain aging and dementia risk.

This study is the first large-scale longitudinal investigation to show that higher dietary diversity is associated with smaller declines in hippocampal volume among community-dwelling middle-aged and older Japanese adults. The findings suggest that increasing dietary diversity may serve as a practical nutritional strategy to help prevent hippocampal atrophy and potentially support healthy cognitive aging.

- Nutrient intake from dietary supplements was not included due to limitations of the supplement database (e.g., omission of compounds such as phosphatidyl serine and choline and limited nutrient information from manufacturers), which may have led to underestimation or misclassification of total nutrient exposure. - MRI datasets with processing failures (e.g., segmentation failures in FreeSurfer) were excluded from analyses, which could introduce selection bias.