Long‐term adherence and changes in the Mediterranean and MIND diets in relation to dementia risk and cognitive function

Abstract

INTRODUCTION

Associations of adherence to and changes in Mediterranean (MedDiet) and Mediterranean–Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay (MIND) diets with cognitive outcomes are unclear.

METHODS

We prospectively followed 86,740 women (Nurses’ Health Study, 1980–2023) and 43,500 men (Health Professionals Follow‐Up Study, 1986–2023). Diet was assessed repeatedly using food frequency questionnaires. Dementia cases were identified from self‐reported physician diagnoses and death records. Cognitive function was assessed by telephone tests, and subjective cognitive decline (SCD) was self‐reported.

RESULTS

Participants in the highest versus lowest MedDiet and MIND categories had 21% (hazard ratio [HR] = 0.79, 95% confidence interval [CI]: 0.73–0.84) and 14% (HR = 0.86, 95% CI: 0.81–0.91) lower dementia risk, respectively. Higher adherence was associated with 0.75 to 1.59 fewer years of cognitive aging, and 40% to 42% lower risk of SCD. Similar inverse associations were observed for improved adherence over 4 or 8 years.

DISCUSSION

Adherence to Mediterranean and MIND diets was associated with cognitive benefits.

Highlights

• We prospectively followed > 130,000 US women and men for up to 43 years.

• Diet was assessed using repeated, validated food frequency questionnaires.

• Multiple cognitive outcomes were assessed across the continuum of Alzheimer's disease and related dementias.

• Long‐term adherence to Mediterranean and Mediterranean‐DASH Intervention for Neurodegenerative Delay diets are linked to cognitive benefits.

• Sustained improvements in adherence are linked to lower dementia risk and better cognition.

1. BACKGROUND

Dementia, characterized by progressive cognitive decline that impairs daily functioning, is among the leading causes of disability and dependence among older adults worldwide, affecting > 55 million people globally. ¹ , ² Cognitive decline typically follows a gradual trajectory along the continuum of Alzheimer's disease (AD), beginning with preclinical pathophysiological changes, such as amyloid beta (Aβ) accumulation and tau pathology, in the absence of symptoms, followed by subjective cognitive decline (SCD), progressing to the prodromal stage marked by objectively measurable mild cognitive impairment (MCI), and ultimately advancing to AD dementia. ³ , ⁴ , ⁵ , ⁶ , ⁷ SCD is defined as a self‐experienced, persistent decline in cognition compared to a previously normal state in the setting of normal objective cognitive performance, and is increasingly recognized as a clinically meaningful early risk stage within the AD continuum, with prior studies linking SCD to an increased risk of subsequent MCI and AD. ⁸ , ⁹ , ¹⁰ Recent evidence further suggests that longitudinal changes in blood‐based biomarkers among individuals with SCD are associated with later cognitive decline. ¹¹ The duration of each stage along this continuum can vary widely across individuals and is influenced by major dementia risk factors such as age, sex, genetic susceptibility (e.g., apolipoprotein E [APOE] ε4 status), and other biological and lifestyle factors. ¹²

Diet has emerged as a promising target for dementia prevention. ¹³ The Mediterranean diet (MedDiet), characterized by high intake of plant‐based foods such as fruits, vegetables, whole grains, and legumes; olive oil as the primary source of fat; limited intake of red and processed meats; and moderate alcohol consumption, has been associated with cognitive benefits, with evidence from the PREDIMED trial showing a causal role in cognitive preservation. ¹⁴ , ¹⁵ Specific dietary components, such as polyphenols from fruits and vegetables, monounsaturated fats from olive oil, and omega‐3 fatty acids from fish, are thought to contribute to the MedDiet's neuroprotective effects. ¹⁶ , ¹⁷ , ¹⁸ Building on the MedDiet, the Mediterranean–Dietary Approaches to Stop Hypertension (DASH) Intervention for Neurodegenerative Delay (MIND) diet, a hybrid of the Mediterranean and DASH diets, was specifically developed to emphasize nutrients and food groups most relevant to brain health, such as leafy greens and berries, which are rich in folate, carotenoids, flavonoids, and other bioactive compounds linked to improved cognitive function. ¹⁹ , ²⁰ , ²¹ , ²²

While studies suggest that adherence to the Mediterranean or MIND diets may reduce dementia risk and preserve cognitive function, their roles across the trajectory of cognitive decline and within specific cognitive function domains remain largely unclear. ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ In addition, many observational studies and randomized controlled trials (RCTs) are limited by small sample sizes and short follow‐up durations. ²⁵ For example, the recent MIND trial did not find significant cognitive benefits, which may be partially due to the relatively short intervention and follow‐up period and reduced adherence during the COVID‐19 pandemic. ²⁹ Additionally, many prior cohort studies relied on a single baseline dietary assessment. However, repeated and long‐term measurement of dietary intake is especially important in studying dietary pattern to capture habitual diet and changes over time, and evaluate their chronic effects on cognitive decline, which develops over decades as a result of slowly accumulating neurodegenerative and vascular changes. ³⁰

Our recent study in two large prospective US cohorts, the Nurses’ Health Study (NHS) and Health Professionals Follow‐Up Study (HPFS), showed that greater long‐term adherence to the MedDiet was prospectively associated with a lower risk of dementia, particularly among APOE ε4 homozygotes. ³¹ However, several important questions remain unanswered. Specifically, it remains unclear how long‐term dietary adherence and mid‐ to long‐term changes in adherence, mimicking dietary intervention and lifestyle modification, are associated with cognitive outcomes at different stages of cognitive decline. Moreover, how these associations may differ across subgroups beyond APOE ε4 status and dose–response relationships have yet to be fully elucidated. In addition, our previous study focused solely on the MedDiet; whether similar associations exist for the more cognitively focused MIND diet has not yet been evaluated in this setting. We therefore expanded upon our previous work by systematically examining the associations of both long‐term adherence and changes in adherence to both the Mediterranean and MIND diets with SCD, objective cognitive function, and dementia risk. We leveraged repeated dietary assessments using validated semi‐quantitative food frequency questionnaires (SFFQs) and up to 43 years of follow‐up in NHS and HPFS. We hypothesized that greater long‐term adherence and sustained improvements in adherence to Mediterranean and MIND diets are associated with moderately lower risk of dementia, slower cognitive decline, and reduced risk of SCD.

2. METHODS

2.1. Study population and design

RESEARCH IN CONTEXT

1. Systematic review: The authors searched PubMed for studies examining the associations of the Mediterranean and Mediterranean‐DASH Intervention for Neurodegenerative Delay (MIND) diets with cognitive outcomes. Existing evidence is limited and mixed, particularly for long‐term adherence and changes in dietary pattern over time in relation to multiple cognitive outcomes across the continuum of Alzheimer's disease and related dementias.

2. Interpretation: We found that both higher long‐term adherence and improvements in adherence to the Mediterranean and MIND diets are associated with lower dementia risk, slower cognitive decline, and a reduced risk of subjective cognitive decline. These findings provide strong evidence for the role of healthy dietary patterns in preserving cognitive function and highlight the potential of dietary interventions to mitigate both early and late stages of cognitive decline.

3. Future directions: Future work should investigate the neuropathological mechanisms underlying these associations and whether individual components of these dietary patterns may drive them. The generalizability of these findings to other populations should also be examined.

The NHS is an ongoing prospective cohort that began in 1976, enrolling 121,700 registered female nurses aged 30 to 55 years. The HPFS, established in 1986, enrolled 51,529 male health professionals aged 40 to 75 years. Participants in both cohorts completed a baseline questionnaire and are followed biennially to update information on lifestyle and health conditions. In this study, baseline was defined as 1980 for NHS and 1986 for HPFS, when the first dietary assessment was completed. We excluded participants with a history of dementia, Parkinson's disease, stroke, or cancer at baseline, as well as those with missing dietary scores or implausible total energy intake (< 500 or > 3500 kcal/day for women; < 800 or > 4200 kcal/day for men). For the analysis of the dementia endpoint, we included 86,740 women from NHS and 43,500 men from HPFS. For objective cognitive function, the analysis included 15,610 women from NHS, as this was not assessed in HPFS. For subjective cognitive decline, a subset of 41,553 women from NHS and 19,255 men from HPFS who answered those questionnaires were included. The study protocol was approved by the institutional review boards of the Brigham and Women's Hospital and Harvard T.H. Chan School of Public Health and was conducted in accordance with the ethical standards of the 1964 Declaration of Helsinki and its later amendments.

2.2. Dietary assessment and calculation of dietary scores

Dietary intake was assessed using validated SFFQs administered at baseline and every 2 to 4 years during follow‐up. Participants reported how often, on average, they consumed specified portion sizes of various foods over the previous year. The validity and reproducibility of the SFFQs have been established in prior studies. ³² , ³³ , ³⁴ , ³⁵ Adherence to the MedDiet was evaluated using the alternate Mediterranean diet (aMED) score, ³⁶ , ³⁷ which included nine dietary components. Participants received 1 point (0 otherwise) for consumption at or above the population‐specific median for vegetables, fruits, nuts, whole grains, legumes, fish, and the ratio of monounsaturated to saturated fat, and 1 point for consumption below the median for red and processed meats. For alcohol, 1 point was assigned for intake between 5 and 15 g/day, and 0 otherwise. The total aMED score ranged from 0 to 9, with higher scores indicating greater adherence. The MIND diet score was calculated based on 10 brain‐healthy food groups and five unhealthy food groups. ²⁴ , ³⁸ Brain‐healthy components included green leafy vegetables (spinach, kale, and lettuce), other vegetables, berries (blueberries and strawberries), nuts, whole grains, fish, beans, poultry, moderate wine intake, and use of olive oil as the primary fat source. Unhealthy components included butter and margarine, cheese, red meat and products, fast or fried foods, and pastries and sweets. For each brain‐healthy component, 0, 0.5, or 1 point was assigned based on intake level, while reverse scoring was applied to unhealthy components (Table S1 in supporting information). The total MIND score ranged from 0 to 15, with higher scores reflecting greater adherence.

2.3. Ascertainment of dementia

Participants were followed from baseline through 2023 for a composite dementia endpoint, defined as either self‐reported dementia or death attributed to dementia. Every 2 years, participants reported whether they had received a physician diagnosis of dementia through mailed questionnaires. Deaths were identified through multiple sources, including state vital records, the National Death Index, reports from family members, and the postal system. A study physician reviewed medical records and death certificates to determine whether dementia was a primary or contributing cause of death. Death identification is > 98% complete. ³⁹ , ⁴⁰ In a subset of participants with medical records obtained for other conditions (e.g., myocardial infarction or stroke) after a dementia report, all physician‐reviewed cases (n = 25) included documentation consistent with dementia, supporting the validity of self‐reported diagnoses. The composite dementia endpoint was further validated using the APOE genotype and plasma phosphorylated tau (p‐tau217) levels, with a 2‐fold increase in dementia risk for APOE ε4 heterozygotes and 5‐fold for homozygotes compared to non‐carriers, and with a 2.8‐fold higher risk comparing the highest to the lowest quartile of p‐tau217 levels (p < 0.001). ³¹

2.4. Assessment of objective cognitive function

Objective cognitive function was assessed in a subset of NHS participants aged ≥ 70 years through telephone‐based cognitive interviews conducted four times between 1995 and 2008. The initial assessment included the Telephone Interview for Cognitive Status (TICS), a validated telephone adaptation of the Mini‐Mental State Examination. ⁴¹ Five additional tests were later incorporated, including immediate and delayed recall of the East Boston Memory Test (EBMT), delayed recall of the TICS 10‐word list, verbal fluency, and the digit span backward test. The high validity and reliability of this telephone‐based cognitive battery have been demonstrated previously. ⁴² In this study, we derived three cognitive outcomes: global cognition, verbal memory, and the TICS score. Global cognition included all six tests, while verbal memory was based on four recall‐based tests from the EBMT and TICS. For each outcome, test‐specific z scores were calculated at each measurement using the baseline mean and standard deviation, and then averaged across relevant tests within each measurement. ⁴³

2.5. Assessment of SCD

SCD was assessed through a set of memory‐related questionnaire items covering four cognitive domains: general memory, executive function, attention, and visuospatial skills (Table S2 in supporting information). Participants answered six or seven yes/no questions, with “yes” scored as 1 and “no” as 0. Responses were summed to generate a SCD score, treated as a count outcome in the analysis. For NHS, SCD data were collected in 2012 and 2014; for HPFS, data were collected in 2008, 2012, 2016, 2018, and 2020. The validity of the SCD assessment has been previously established. ⁴³

2.6. Assessment of covariates

Participants reported age, body weight, education attainment (nurses’ and husbands’ education in NHS; professions in HPFS), family history of dementia, menopausal and hormone therapy status (NHS only), neighborhood socioeconomic status (nSES, measured using Census tract data ⁴⁴ ), marital status, living arrangement, smoking status, history of depression, history of hypertension, history of diabetes, regular use of antihypertensive medications, and regular use of antidepressant medications through biennial questionnaires. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared, using height reported in 1976 and the updated weights. Leisure‐time physical activity was assessed using extensively validated questionnaires. ⁴⁵ , ⁴⁶

2.7. Statistical analysis

In the analyses of long‐term adherence, participants were categorized into approximate quintiles of the aMED and MIND diet scores based on their respective distributions in the NHS and HPFS. For the dementia risk analysis, hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards (PH) models, comparing each higher category of the diet scores to the lowest within each cohort. Person‐years were calculated from baseline (the return date of the 1980 questionnaire for NHS and the 1986 questionnaire for HPFS) until the earliest of dementia diagnosis, death, or the end of follow‐up (January 31, 2023, for both cohorts). We used a counting process data structure to handle left truncation and incorporate time‐varying covariates. To control as precisely as possible for confounding by age, calendar time, and their possible two‐way interactions, the analysis was jointly stratified by age (in months) at the start of follow‐up and calendar year of the current questionnaire cycle. The time scale for the analysis was defined as months since the start of the current questionnaire cycle, which is equivalent to age in months given the structure of the data and model formulation. The multivariate models were adjusted for family history of dementia; nSES index; education attainment; smoking status; physical activity; BMI; history of hypertension, diabetes, and depression; regular use of antihypertensive and antidepressant medications; marital status; living arrangement; total energy intake; and menopausal status with hormone use, with covariates updated at each questionnaire cycle and treated as time varying when appropriate. For the analyses of objective cognitive function in NHS, we used generalized estimating equations (GEE) with a Gaussian distribution to estimate mean differences and 95% CIs for the three cognitive function outcomes, comparing higher categories of the diet scores to the lowest category. To facilitate interpretation, mean differences were converted into years of cognitive aging by dividing the regression coefficients for diet scores by the corresponding coefficients for age from the same model. For the analyses of SCD, GEE models with a Poisson distribution and a log link were used to estimate relative risks (RRs) and 95% CIs per 3‐unit increment in the SCD score, comparing higher categories of the diet scores to the lowest category within each cohort. A 3‐unit increment corresponds to reporting three additional cognitive concerns on the SCD questionnaire and has been used consistently in previous NHS and HPFS studies, in which reporting three or more SCD items reflects a meaningful level of perceived cognitive difficulty and has been associated with poorer objective cognitive function. ⁴⁷ , ⁴⁸ , ⁴⁹ In NHS, dietary scores were averaged from SFFQ assessments in 2006 and 2010, preceding the first SCD assessment in 2012; in HPFS, dietary scores were averaged from SFFQ assessments in 2002 and 2006, preceding the first SCD assessment in 2008, thereby defining dietary exposure prior to cognitive assessments with an inherent 2‐year lag to reduce potential reverse causation. Models for objective cognitive function and SCD were adjusted for the same covariates as in the dementia risk analysis. Linear trends were assessed by modeling the median value of each category as a continuous variable. p values for linear trend were obtained from the Wald test. Fixed‐effect meta‐analyses were performed to combine estimates in both cohorts for the dementia risk and SCD analyses.

To assess the impacts of changes in dietary patterns, we examined 4‐year and 8‐year changes in aMED and MIND diet scores in relation to subsequent cognitive outcomes. For the dementia analysis, person‐years were accrued from the end of each 4‐year or 8‐year interval until dementia diagnosis, death, or the end of follow‐up. Participants were categorized into approximate quintiles of change in diet scores, and HRs and 95% CIs were estimated using Cox PH models, comparing each category to the group that maintained relatively stable diet adherence. For the analyses of objective cognitive function and SCD, changes in diet scores over 4‐year and 8‐year intervals were modeled as main exposures in GEE models. All models were additionally adjusted for the initial diet score and total energy intake at the beginning of the change interval, as well as change in total energy intake.

We further assessed potential non‐linear dose–response relationships between diet scores and cognitive outcomes by fitting restricted cubic spline models with three knots in a pooled dataset from both cohorts and adjusting for the same covariates as in the primary analysis, with additional stratification by cohort in the Cox PH models and adjustment for cohort in the GEE models. ⁵⁰ Non‐linearity was assessed using a likelihood ratio test comparing a model with only a linear term to one that included additional cubic spline terms. We further conducted subgroup analyses in the pooled dataset, with subgroups defined by age, BMI, smoking status, family history of dementia, and histories of hypertension, diabetes, and depression. Interaction p values were derived from Wald tests of the cross‐product term between the diet score and each stratified variable in the full dataset. To mitigate potential reverse causation, we conducted lag analyses excluding dietary exposures from 4 or 8 years immediately preceding each time point in the dementia risk analysis. All analyses were conducted using SAS software, version 9.4 (SAS Institute Inc.), with a two‐tailed significance level of p < 0.05.

3. RESULTS

3.1. Population characteristics

Table 1 presents the age‐standardized characteristics of NHS and HPFS participants by the five categories of the aMED and MIND diet scores in 2002, the midpoint of follow‐up. Because of the distribution of scores, the number of individuals within each category varies. Women with higher adherence to both dietary patterns tended to be younger and had higher education attainments. This pattern of younger age was observed only for the MIND score among men. In both cohorts, participants with higher adherence to both dietary patterns tended to have greater physical activity, higher nSES index, were more likely to be married, and reported a family history of dementia, while those with lower diet scores tended to have higher BMI and were more likely to be current smokers.

TABLE 1.

Age‐standardized characteristics by aMED and MIND diet scores in 2002 in NHS (n = 77,183) and HPFS (n = 36,947).

	Categories of aMED score			Categories of MIND score
	Q1	Q3	Q5	Q1	Q3	Q5
Nurses’ Health Study
N	12,882	28,209	12,005	11,447	15,587	22,340
Age, years	68.4 (7.3)	67.7 (7.1)	67.0 (6.7)	69.3 (7.2)	67.9 (7.2)	66.6 (6.7)
Physical activity, hours/week	1.0 (1.9)	1.6 (2.6)	2.5 (3.2)	1.0 (2.1)	1.5 (2.5)	2.2 (3.0)
Neighborhood SES index	−0.5 (3.5)	0 (3.8)	0.6 (4.3)	−0.6 (3.5)	−0.2 (3.7)	0.6 (4.1)
Body mass index, kg/m2	27.3 (5.8)	26.9 (5.4)	25.7 (4.8)	27.2 (5.9)	27.1 (5.5)	26.2 (5.1)
Total energy intake (kcal/day)	1443 (484.3)	1693.2 (530.9)	1968.7 (518.2)	1584.5 (544.5)	1689.6 (556.3)	1769.9 (528.5)
Smoking status, %
Never	41.5	44.8	46.1	42.5	44.9	44.2
Past	43.3	45.8	49.7	41.0	45.0	50.6
Current	15.2	9.3	4.2	16.5	10.1	5.3
Nurses’ education level, %
Registered nurse degree	76.7	69.9	61.9	77.4	72.0	63.6
Bachelor's degree	16.6	20.1	24.7	15.9	19.0	23.8
Graduate degree	6.7	10.0	13.4	6.6	9.0	12.6
Spouse's education level, %
High school or less	55.6	46.2	34.9	57.1	48.6	37.5
College degree	25.5	29.9	32.4	25.5	29.0	31.7
Graduate school	18.9	23.9	32.7	17.5	22.4	30.9
Postmenopausal hormone use, %
Pre‐menopause	1.6	1.6	1.1	3.3	1.4	0.8
Postmenopausal hormone use never/unknown	29.6	27.3	23.8	30.2	27.6	24.4
Postmenopausal hormone use current	31.8	33.2	35.3	30.4	33.7	35.4
Postmenopausal hormone ever user	37.0	37.9	39.7	36.1	37.3	39.5
Family history of dementia, %	17.9	19.3	20.4	16.6	18.9	20.7
Hypertension, %	57.1	55.0	50.0	55.0	54.9	52.7
Use of antihypertensive medications, %	40.9	40.1	37.6	37.7	40.0	40.7
Depression, %	11.9	10.8	9.6	12.1	10.6	9.8
Use of antidepressant medications, %	10.9	9.7	8.4	10.5	9.6	8.9
Diabetes, %	11.6	11.3	8.0	10.8	10.9	9.8
Married, %	58.5	62.3	66.3	54.9	61.7	66.5
Living alone, %	18.1	16.9	16.8	17.7	17.0	17.2
Health Professionals Follow‐Up Study
N	8037	6629	5110	6731	8492	7400
Age, years	67.0 (9.0)	68.2 (9.1)	68.7 (8.9)	68.5 (9.4)	68.1 (9.1)	67.7 (8.7)
Physical activity, hours/week	2.3 (4.4)	3.3 (5.5)	5.6 (6.8)	2.3 (4.3)	3.5 (5.7)	5.1 (6.5)
Neighborhood SES index b	−0.4 (5.3)	0 (5.4)	0.7 (5.7)	−0.6 (5.3)	0.1 (5.5)	0.8 (5.8)
Body mass index, kg/m2	26.8 (4.0)	26.3 (3.8)	25.4 (3.7)	26.7 (4.2)	26.4 (3.8)	25.6 (3.6)
Total energy intake (kcal/day)	1735.9 (557.7)	1975.4 (613.8)	2336 (608.7)	1948.1 (641.5)	2002.2 (641.1)	2053.1 (609.4)
Smoking status, %
Never	36.2	39.0	44.8	36.7	39.8	42.5
Past	55.1	56.1	53.0	53.9	55.5	55.2
Current	8.7	5.0	2.2	9.4	4.7	2.3
Profession, %
Dentist	51.2	56.3	66.2	47.8	58.2	66.0
Veterinarian	23.4	21.4	15.2	26.7	20.2	14.8
Pharmacist/optometrist/osteopath/podiatrist	25.4	22.3	18.7	25.5	21.7	19.2
Family history of dementia, %	14.8	15.1	17.4	13.9	15.4	16.2
Hypertension, %	49.3	47.4	42.3	46.1	46.2	44.8
Use of antihypertensive medications, %	27.5	27.4	27.2	24.9	27.2	29.5
Depression, %	11.8	10.9	11.2	11.8	11.3	11.8
Use of antidepressant medications, %	4.6	3.9	3.5	4.6	4.0	3.8
Diabetes, %	10.8	10.2	7.7	9.8	10.0	9.7
Married, %	64.7	67.5	71.1	62.1	67.5	72.8
Living alone, %	8.3	6.8	6.0	8.5	6.3	6.1

Notes: Unless otherwise indicated, data are expressed as means. All variables except age were age standardized.

Abbreviations: aMED, alternate Mediterranean diet; HPFS, Health Professionals Follow‐Up Study; MIND, Mediterranean–Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay; SES, socioeconomic status.

3.2. Dementia risk

During up to 43 years of follow‐up (4,282,833 person‐years), we documented 10,924 incident dementia cases (NHS: 7986; HPFS: 2938). In the meta‐analyzed multivariate‐adjusted analyses of long‐term adherence in both cohorts, participants in the highest category of the aMED score had a 21% lower risk of dementia compared to those in the lowest category (HR _{Q5 vs. Q1} = 0.79; 95% CI: 0.73, 0.84; p _trend < 0.001; Table 2). Similarly, for the MIND score, the HR for dementia risk comparing the highest to the lowest category was 0.86 (95% CI: 0.81, 0.91; p _trend < 0.001). The results were broadly consistent in both women and men. Lag analyses yielded similar results, although slightly attenuated (Table S3 in supporting information). A nominally significant non‐linear association was observed only for the aMED score, with a greater rate of risk reduction beginning around a score of 3.5 (p _{non‐linearity} = 0.04; Figure S1 in supporting information); each one‐unit increase in the aMED and MIND scores was associated with a 5% and 3% lower risk of dementia in linear trend analyses, respectively. The protective associations were generally consistent across subgroups defined by major risk factors of dementia, with only suggestive interactions observed between smoking status and both diet scores, and between history of diabetes and the aMED score (Table S4 in supporting information), for which the protective associations were only observed among non‐current smoker (p _interaction = 0.03) and more pronounced among participants without a history of diabetes (p _interaction = 0.05).

TABLE 2.

Associations of aMED and MIND diet scores with dementia risk (HR and 95% CI) in NHS (1980–2023) and HPFS (1986–2023).

	Categories of diet scores					p trend	HR (95% CI) per unit increment a
	Q1	Q2	Q3	Q4	Q5
aMED
NHS
Median	2.0	3.0	4.0	6.0	7.0
Cases/PYs	1716/488089	1383/481839	2869/1148669	1029/478639	989/474829
Model 1 b	Ref.	0.94 (0.87, 1.01)	0.89 (0.84, 0.95)	0.80 (0.74, 0.87)	0.73 (0.68, 0.79)	<0.001	0.94 (0.93, 0.95)
Model 2 c	Ref.	0.97 (0.90, 1.04)	0.92 (0.87, 0.98)	0.83 (0.77, 0.90)	0.77 (0.70, 0.83)	<0.001	0.95 (0.94, 0.96)
HPFS
Median	2.0	3.0	4.0	5.0	7.0
Cases/PYs	665/242992	499/198395	530/218500	849/377519	396/173360
Model 1 b	Ref.	0.94 (0.84, 1.06)	0.90 (0.81, 1.01)	0.84 (0.76, 0.93)	0.80 (0.71, 0.91)	<0.001	0.95 (0.93, 0.98)
Model 2 c	Ref.	0.97 (0.86, 1.09)	0.94 (0.83, 1.05)	0.87 (0.78, 0.97)	0.84 (0.74, 0.96)	0.003	0.96 (0.94, 0.99)
Pooled d
Model 1 b	Ref.	0.94 (0.88, 1.00)	0.90 (0.85, 0.94)	0.82 (0.77, 0.87)	0.75 (0.70, 0.80)	<0.001	0.95 (0.93, 0.96)
Model 2 c	Ref.	0.97 (0.91, 1.03)	0.92 (0.87, 0.98)	0.84 (0.79, 0.90)	0.79 (0.73, 0.84)	<0.001	0.95 (0.94, 0.96)
MIND
NHS
Median	5.0	6.5	7.5	8.0	9.5
Cases/PYs	1665/527266	1431/540690	1659/659765	1498/611576	1733/732768
Model 1 b	Ref.	0.89 (0.83, 0.96)	0.89 (0.84, 0.96)	0.88 (0.82, 0.94)	0.79 (0.74, 0.85)	<0.001	0.95 (0.94, 0.97)
Model 2 c	Ref.	0.93 (0.86, 1.00)	0.95 (0.89, 1.02)	0.94 (0.87, 1.01)	0.88 (0.82, 0.94)	<0.001	0.97 (0.96, 0.99)
HPFS
Median	5.5	7.0	8.0	8.5	10.0
Cases/PYs	633/224163	558/227702	651/277048	549/241042	547/240812
Model 1 b	Ref.	0.94 (0.84, 1.05)	0.90 (0.81, 1.00)	0.86 (0.77, 0.96)	0.76 (0.68, 0.85)	<0.001	0.94 (0.92, 0.96)
Model 2 c	Ref.	0.95 (0.85, 1.06)	0.92 (0.82, 1.03)	0.87 (0.78, 0.98)	0.80 (0.71, 0.90)	<0.001	0.95 (0.93, 0.98)
Pooled d
Model 1 b	Ref.	0.90 (0.85, 0.96)	0.90 (0.85, 0.95)	0.87 (0.82, 0.93)	0.78 (0.74, 0.83)	<0.001	0.95 (0.94, 0.96)
Model 2 c	Ref.	0.93 (0.88, 0.99)	0.94 (0.89, 1.00)	0.92 (0.87, 0.98)	0.86 (0.81, 0.91)	<0.001	0.97 (0.96, 0.98)

Abbreviations: aMED, alternate Mediterranean diet; CI, confidence interval; HPFS, Health Professionals Follow‐Up Study; HR, hazard ratio; MIND, Mediterranean–Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay; NHS, Nurses’ Health Study; PY, person‐year.

^a Hazard ratios (95% CIs) for dementia risk per unit increase in the diet score.

^b Model 1 stratified by age (months) and calendar time (years), and adjusted for family history of dementia (yes vs. no).

^c Model 2 further adjusted for neighborhood socioeconomic status index (in tertiles), education attainments (nurses’ and husbands’ education in NHS, professions in HPFS), smoking status (never, past, current 1–14 cigarettes/day, current 15–24 cigarettes/day, current ≥ 25 cigarettes/day), physical activity (< 0.5, 0.5–2, 2–3.5, 3.5–5.5, ≥ 5.5 hours/week), body mass index (< 23, 23–24.9, 25–29.9, 30–34.9, ≥ 35 kg/m²), history of hypertension (yes vs. no), history of diabetes (yes vs. no), history of depression (yes vs. no), regular antihypertensive medications (yes vs. no), regular antidepressant drug use (yes vs. no), marriage status (yes vs. no), living arrangement (alone/not alone), total energy intake (kcal/day), and menopausal status and hormone use for NHS.

^d Fixed‐effect meta‐analysis of the results from NHS and HPFS.

3.3. Objective cognitive function

Overall, higher adherence to both the aMED and MIND diets was associated with slower cognitive aging in NHS (all p _trend ≤ 0.005; Table 3). Specifically, in the multivariate‐adjusted analyses of long‐term adherence, compared to those in the lowest category (Q1), participants in the highest category (Q5) of the aMED score showed a delay in cognitive aging of 1.52 years for global cognition (95% CI: −2.17, −0.86), 1.46 years for verbal memory (95% CI: −2.16, −0.77), and 1.18 years for the TICS score (95% CI: −1.86, −0.51). Similarly, higher adherence to the MIND diet (Q5 vs. Q1) was associated with 1.17 fewer years of cognitive aging for global cognition (95% CI: −1.78, −0.55), 1.59 years for verbal memory (95% CI: −2.26, −0.93), and 0.75 years for the TICS score (95% CI: −1.39, −0.11). Suggestive non‐linear relationships were observed for the associations of MIND diet with global cognition and verbal memory, with little cognitive benefit at lower scores and a sharper reduction in cognitive aging beginning around a score of 7 (p _{non‐linearity} ≤ 0.04; Figure S2 in supporting information). The protective associations were generally consistent across subgroups, with a suggestive more pronounced association between the aMED score and global cognition among non‐current smokers (p _interaction = 0.04), consistent with findings for dementia risk, and between the MIND score and global cognition among women aged > 75 years (p _interaction = 0.03; Table S5 in supporting information).

TABLE 3.

Associations of aMED and MIND diet scores with objective cognitive function (MD and 95% CI) in NHS (1995–2008).

	Categories of diet scores					p trend	MD (95% CI) per unit increment a
	Q1	Q2	Q3	Q4	Q5
aMED
Median	2.5	4.0	5.0	5.5	7.0
Global cognition
Model 1 b	Ref.	−0.64 (−1.18, −0.10)	−1.28 (−1.81, −0.76)	−1.40 (−1.95, −0.85)	−2.01 (−2.58, −1.44)	<0.001	−0.46 (−0.58, −0.34)
Model 2 c	Ref.	−0.54 (−1.11, 0.04)	−1.11 (−1.68, −0.54)	−1.06 (−1.67, −0.45)	−1.52 (−2.17, −0.86)	<0.001	−0.35 (−0.49, −0.21)
Verbal memory
Model 1 b	Ref.	−0.61 (−1.19, −0.04)	−1.19 (−1.76, −0.62)	−1.41 (−2.00, −0.81)	−1.76 (−2.37, −1.15)	<0.001	−0.41 (−0.54, −0.28)
Model 2 c	Ref.	−0.55 (−1.17, 0.06)	−1.10 (−1.71, −0.49)	−1.22 (−1.87, −0.56)	−1.46 (−2.16, −0.77)	<0.001	−0.35 (−0.49, −0.20)
TICS score
Model 1 b	Ref.	−0.47 (−1.04, 0.10)	−1.24 (−1.80, −0.68)	−1.44 (−2.02, −0.87)	−1.98 (−2.58, −1.39)	<0.001	−0.47 (−0.59, −0.34)
Model 2 c	Ref.	−0.27 (−0.87, 0.34)	−0.93 (−1.53, −0.33)	−0.90 (−1.53, −0.27)	−1.18 (−1.86, −0.51)	<0.001	−0.29 (−0.43, −0.15)
MIND
Median	6.0	7.0	7.8	8.5	9.8
Global cognition
Model 1 b	Ref.	−0.51 (−1.06, 0.05)	−0.12 (−0.67, 0.42)	−1.22 (−1.74, −0.71)	−1.80 (−2.36, −1.24)	<0.001	−0.48 (−0.62, −0.35)
Model 2 c	Ref.	−0.32 (−0.91, 0.28)	0.19 (−0.39, 0.77)	−0.85 (−1.42, −0.29)	−1.17 (−1.78, −0.55)	<0.001	−0.32 (−0.47, −0.17)
Verbal memory
Model 1 b	Ref.	−0.77 (−1.37, −0.17)	−0.33 (−0.92, 0.26)	−1.33 (−1.89, −0.77)	−2.04 (−2.64, −1.44)	<0.001	−0.53 (−0.67, −0.38)
Model 2 c	Ref.	−0.65 (−1.29, −0.01)	−0.11 (−0.74, 0.52)	−1.07 (−1.68, −0.46)	−1.59 (−2.26, −0.93)	<0.001	−0.41 (−0.57, −0.24)
TICS score
Model 1 b	Ref.	−0.37 (−0.96, 0.21)	−0.34 (−0.91, 0.23)	−1.16 (−1.71, −0.62)	−1.63 (−2.21, −1.05)	<0.001	−0.45 (−0.59, −0.30)
Model 2 c	Ref.	−0.08 (−0.71, 0.54)	0.07 (−0.54, 0.67)	−0.64 (−1.23, −0.05)	−0.75 (−1.39, −0.11)	0.005	−0.23 (−0.38, −0.07)

Abbreviations: aMED, alternate Mediterranean diet; CI, confidence interval; MD, mean difference; MIND, Mediterranean–Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay; NHS, Nurses’ Health Study; TICS, Telephone Interview for Cognitive Status.

^a Mean differences (95% CIs) in cognitive function, expressed in years of cognitive aging per unit increase in the diet score.

^b Model 1 adjusted for age at cognitive function assessment (months) and family history of dementia (yes vs. no).

^c Model 2 further adjusted for neighborhood socioeconomic status index (in tertiles), education attainments (nurses’ and husbands’ education), smoking status (never, past, current 1–14 cigarettes/day, current 15–24 cigarettes/day, current ≥ 25 cigarettes/day), physical activity (< 0.5, 0.5–2, 2–3.5, 3.5–5.5, ≥ 5.5 metabolic equivalent hours/week), body mass index (< 23, 23–24.9, 25–29.9, 30–34.9, ≥ 35 kg/m²), history of hypertension (yes vs. no), history of diabetes (yes vs. no), history of depression (yes vs. no), regular antihypertensive medications (yes vs. no), regular antidepressant drug use (yes vs. no), marriage status (yes vs. no), living arrangement (alone/not alone), total energy intake (kcal/day), and menopausal status and hormone use.

3.4. SCD

Similar inverse associations of adherence to both dietary patterns were observed in relation to SCD (all p _trend < 0.001; Table 4). In the meta‐analyzed multivariate‐adjusted analyses of long‐term adherence in both cohorts, participants in the highest category of the aMED score had a 40% lower risk of SCD compared to those in the lowest category (RR _{Q5 vs. Q1} = 0.60; 95% CI: 0.55, 0.67). The protective association was more pronounced for the MIND score, with a 42% lower risk of SCD comparing the highest to the lowest category (RR _{Q5 vs. Q1} = 0.58; 95% CI: 0.52, 0.64). These inverse associations were consistent in women and men. Non‐linear relationships were observed between both diet scores and SCD, with overall decreasing trends indicating lower risk with higher adherence (Figure S3 in supporting information). For the aMED score, the sharp decline in SCD risk began at a score of ≈ 3.5. Associations were broadly consistent across subgroups defined by dementia risk factors (Table S6 in supporting information).

TABLE 4.

Associations of aMED and MIND diet scores with subjective cognitive decline (RR and 95% CI) in NHS (2012 and 2014) and HPFS (2008, 2012, 2016, 2018, and 2020).

	Categories of diet scores					p trend	RR (95% CI) per unit increment a
	Q1	Q2	Q3	Q4	Q5
aMED
NHS
Median	2.0	3.5	5.0	6.0	7.0
Model 1 b	Ref.	1.01 (0.91, 1.12)	0.89 (0.79, 1.00)	0.79 (0.70, 0.89)	0.63 (0.56, 0.71)	<0.001	0.91 (0.89, 0.93)
Model 2 c	Ref.	0.98 (0.88, 1.08)	0.86 (0.76, 0.96)	0.74 (0.65, 0.84)	0.60 (0.53, 0.69)	<0.001	0.91 (0.88, 0.93)
HPFS
Median	2.0	3.0	4.0	5.0	6.5
Model 1 b	Ref.	0.87 (0.74, 1.03)	0.80 (0.67, 0.94)	0.75 (0.64, 0.90)	0.63 (0.54, 0.74)	<0.001	0.91 (0.88, 0.94)
Model 2 c	Ref.	0.87 (0.74, 1.03)	0.78 (0.66, 0.92)	0.72 (0.60, 0.85)	0.61 (0.51, 0.72)	<0.001	0.90 (0.87, 0.93)
Pooled d
Model 1 b	Ref.	0.97 (0.88, 1.06)	0.86 (0.78, 0.94)	0.78 (0.70, 0.86)	0.63 (0.57, 0.69)	<0.001	0.91 (0.89, 0.93)
Model 2 c	Ref.	0.94 (0.86, 1.03)	0.83 (0.75, 0.91)	0.73 (0.66, 0.81)	0.60 (0.55, 0.67)	<0.001	0.90 (0.89, 0.92)
MIND
NHS
Median	5.5	7.0	7.8	8.8	10.0
Model 1 b	Ref.	0.88 (0.79, 0.98)	0.86 (0.77, 0.95)	0.64 (0.57, 0.72)	0.52 (0.47, 0.59)	<0.001	0.87 (0.85, 0.89)
Model 2 c	Ref.	0.90 (0.81, 1.00)	0.89 (0.81, 0.99)	0.69 (0.61, 0.77)	0.57 (0.51, 0.64)	<0.001	0.88 (0.86, 0.91)
HPFS
Median	6.0	7.3	8.0	8.8	10.0
Model 1 b	Ref.	0.76 (0.65, 0.89)	0.77 (0.65, 0.92)	0.66 (0.56, 0.78)	0.52 (0.44, 0.62)	<0.001	0.86 (0.82, 0.89)
Model 2 c	Ref.	0.76 (0.65, 0.89)	0.79 (0.67, 0.94)	0.70 (0.59, 0.83)	0.59 (0.49, 0.70)	<0.001	0.88 (0.85, 0.92)
Pooled d
Model 1 b	Ref.	0.84 (0.77, 0.92)	0.83 (0.76, 0.91)	0.65 (0.59, 0.71)	0.52 (0.47, 0.58)	<0.001	0.86 (0.85, 0.88)
Model 2 c	Ref.	0.86 (0.78, 0.93)	0.86 (0.79, 0.95)	0.69 (0.63, 0.76)	0.58 (0.52, 0.64)	<0.001	0.88 (0.87, 0.90)

Abbreviations: aMED, alternate Mediterranean diet; CI, confidence interval; HPFS, Health Professionals Follow‐Up Study; MIND, Mediterranean–Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay; NHS, Nurses’ Health Study; RR, relative risk.

^a Relative risk (95% CIs) for a 3‐unit increment in subjective cognitive decline per unit increase in the diet score.

^b Model 1 adjusted for age (months) and family history of dementia (yes vs. no).

^c Model 2 further adjusted for neighborhood socioeconomic status index (in tertiles), education attainments (nurses’ and husbands’ education in NHS, professions in HPFS), smoking status (never, past, current 1–14 cigarettes/day, current 15–24 cigarettes/day, current ≥ 25 cigarettes/day), physical activity (< 0.5, 0.5–2, 2–3.5, 3.5–5.5, ≥ 5.5 hours/week), body mass index (< 23, 23–24.9, 25–29.9, 30–34.9, ≥ 35 kg/m²), history of hypertension (yes vs. no), history of diabetes (yes vs. no), history of depression (yes vs. no), regular antihypertensive medications (yes vs. no), regular antidepressant drug use (yes vs. no), marriage status (yes vs. no), living arrangement (alone/not alone), total energy intake (kcal/day), and menopausal status and hormone use.

^d Fixed‐effect meta‐analysis of the results from NHS and HPFS.

3.5. Change in dietary adherence and cognitive outcomes

In change analyses, mid‐ to long‐term improvements in adherence to the aMED and MIND diets were also associated with cognitive benefits. For dementia risk, the linear trends were more pronounced for changes in both scores compared to long‐term adherence, with each one‐unit increase in the aMED change over 4 years associated with a 7% lower risk of dementia (p _trend < 0.001; Table 5 and Tables S7 and S8 in supporting information). Greater adherence to the aMED over 4 years was associated with slower cognitive aging across all domains (p _trend ≤ 0.005; Table 5 and Table S9 in supporting information). The associations were directionally similar for the MIND score and 8‐year changes but generally weaker (Table S10 in supporting information). Greater 4‐year improvement in aMED and MIND adherence was associated with a lower risk of SCD (p _trend < 0.001; Table 5 and Table S11 in supporting information), with similar beneficial associations observed for 8‐year improvement (Table S12 in supporting information).

TABLE 5.

Associations of 4‐year changes in aMED and MIND diet scores with dementia risk, objective cognitive function, and subjective cognitive decline in NHS and HPFS.

	Categories of change in diet scores					p trend
	Q1	Q2	Q3	Q4	Q5
Dementia risk							HR (95% CI) per unit increment a
aMED	1.23 (1.16, 1.31)	1.09 (1.02, 1.16)	Ref.	0.96 (0.89, 1.03)	0.92 (0.86, 0.99)	<0.001	0.93 (0.91, 0.95)
MIND	1.23 (1.15, 1.31)	1.00 (0.94, 1.07)	Ref.	0.97 (0.89, 1.05)	0.98 (0.92, 1.06)	<0.001	0.94 (0.93, 0.96)
Objective cognitive function							MD (95% CI) per unit increment b
Global cognition
aMED	0.58 (−0.08, 1.24)	−0.01 (−0.60, 0.58)	Ref.	−0.18 (−0.77, 0.40)	−0.76 (−1.37, −0.15)	0.001	−0.27 (−0.44, −0.11)
MIND	0.71 (−0.01, 1.42)	−0.29 (−0.94, 0.35)	Ref	−0.53 (−1.18, 0.12)	−0.06 (−0.76, 0.64)	0.03	−0.18 (−0.35, −0.01)
Verbal memory
aMED	0.58 (−0.12, 1.28)	0.02 (−0.61, 0.65)	Ref.	−0.09 (−0.71, 0.53)	−0.66 (−1.31, −0.00)	0.005	−0.25 (−0.43, −0.07)
MIND	0.62 (−0.14, 1.39)	−0.30 (−1.00, 0.39)	Ref.	−0.74 (−1.44, −0.05)	−0.16 (−0.92, 0.59)	0.03	−0.20 (−0.38, −0.02)
TICS score
aMED	0.70 (0.01, 1.39)	0.15 (−0.47, 0.78)	Ref.	−0.07 (−0.69, 0.55)	−0.88 (−1.51, −0.25)	<0.001	−0.33 (−0.50, −0.16)
MIND	0.70 (−0.06, 1.45)	−0.16 (−0.83, 0.52)	Ref.	−0.41 (−1.09, 0.27)	0.08 (−0.66, 0.82)	0.10	−0.15 (−0.33, 0.03)
Subjective cognitive decline							RR (95% CI) per unit increment c
aMED	1.25 (1.13, 1.38)	0.96 (0.87, 1.06)	Ref.	0.90 (0.82, 0.99)	0.84 (0.75, 0.93)	<0.001	0.92 (0.89, 0.94)
MIND	1.18 (1.06, 1.33)	1.01 (0.91, 1.13)	Ref.	0.93 (0.83, 1.03)	0.91 (0.81, 1.02)	<0.001	0.93 (0.91, 0.96)

Abbreviations: aMED, alternate Mediterranean diet; CI, confidence interval; HPFS, Health Professionals Follow‐Up Study; HR, hazard ratio; MD, mean difference; MIND, Mediterranean–Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay; NHS, Nurses’ Health Study; TICS, Telephone Interview for Cognitive Status.

^a Hazard ratios (95% CIs) for dementia risk per unit increase in the 4‐year change in diet score.

^b Mean differences (95% CIs) in cognitive function, expressed in years of cognitive aging per unit increase in the 4‐year change in diet score.

^c Relative risk (95% CIs) for a 3‐unit increment in subjective cognitive decline per unit increase in the 4‐year change in diet score.

Notes: All models stratified by age (months) and calendar time (years), and adjusted for baseline diet score during the 4‐year interval, family history of dementia (yes vs. no), baseline total energy intake and 4‐year change (kcal/day), neighborhood socioeconomic status index (in tertiles), education attainments (nurses’ and husbands’ education in NHS, professions in HPFS), smoking status (never, past, current 1–14 cigarettes/day, current 15–24 cigarettes/day, current ≥ 25 cigarettes/day), physical activity (< 0.5, 0.5–2, 2–3.5, 3.5–5.5, ≥ 5.5 hours/week), body mass index (< 23, 23–24.9, 25–29.9, 30–34.9, ≥ 35 kg/m²), history of hypertension (yes vs. no), history of diabetes (yes vs. no), history of depression (yes vs. no), regular antihypertensive medications (yes vs. no), regular antidepressant drug use (yes vs. no), marriage status (yes vs. no), living arrangement (alone/not alone), and menopausal status and hormone use for NHS at the end of the 4‐year interval (start of follow‐up). Fixed‐effect meta‐analysis of the results from NHS and HPFS are shown for dementia risk and subjective cognitive function analyses. Results from NHS are shown for objective cognitive function analysis.

4. DISCUSSION

In two large prospective cohorts of > 130,000 women and men followed for up to four decades, we found that higher long‐term adherence to the Mediterranean and MIND diets was consistently associated with lower risk of dementia, slower cognitive aging, and reduced risk of SCD. By leveraging the uniquely comprehensive data on repeated dietary assessments, diverse cognitive outcomes, and extended follow‐up in NHS and HPFS, our study provides strong evidence for the potential benefits of the Mediterranean and MIND diets in preserving cognitive health across the trajectory of cognitive decline. Importantly, increase in adherence to these healthy diets over time was also robustly associated with better cognitive outcomes, reinforcing the benefits of mid‐life dietary changes in promoting cognitive health.

Our findings expand upon prior evidence from observational studies and RCTs suggesting protective associations between adherence to the Mediterranean or MIND diets and cognitive outcomes. ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ Among dietary patterns, MedDiet is the only one to date that has demonstrated causal benefits for cognitive function in RCTs, most notably in the PREDIMED trial and its cognitive substudies. ¹⁵ , ⁵¹ However, previous findings have been mixed across specific cognitive domains and stages of cognitive decline. A recent meta‐analysis of 26 cohort studies and two RCTs reported that greater adherence to the MedDiet was associated with a reduced risk of AD and MCI, but not with dementia risk or other cognitive domains in cohort studies. ²⁶ RCTs included in the analysis showed benefits for global cognition, working memory, and episodic memory, but an adverse effect was observed for attention. Similarly, although the MIND diet was designed specifically to promote brain health and has shown promising protective associations with dementia risk in observational studies, ⁵³ evidence from RCTs has been mixed. ²⁹ , ⁵⁴ The MIND trial, a 3‐year RCT of 604 older adults at risk for dementia, reported no significant effect of the MIND diet on global cognitive change compared to a control diet with mild caloric restriction. ²⁹ It is possible that the relatively short duration, and the reduction of adherence during the COVID‐19 pandemic, may partially explain those null findings. Moreover, biomarker data (e.g., serum lutein/zeaxanthin and carotenoids) in the MIND trial indicated only modest increases, suggesting limited dietary change despite reported adherence. In contrast, food‐provision trials such as Women's Healthy Eating and Living Study and PREDIMED observed substantially larger biomarker changes, consistent with stronger adherence and greater metabolic responses. ⁵⁵ , ⁵⁶ A recent systematic review of 32 cohort studies on the MIND diet also highlighted inconsistencies in its associations across different cognitive domains. ⁵²

A major limitation of prior studies is the lack of long‐term assessment of dietary intake and cognitive outcomes across the trajectory of cognitive decline during extended follow‐up. Few studies have examined both the Mediterranean and MIND diets in parallel or explored their associations across multiple cognitive outcomes within the same population. Our prior study in NHS and HPFS provided evidence for the interplay among genetics, diet, and metabolites in relation to dementia risk and cognitive function. ³¹ The present work extends these findings, including incorporating SCD, thereby capturing earlier and more subtle stages of cognitive decline that extend beyond genetically high‐risk groups to the general population. We observed a > 40% lower risk of a 3‐unit increment in SCD with higher adherence to both diets. Importantly, SCD reflects self‐perceived difficulties across everyday cognitive domains, such as remembering recent events or short lists, following conversations or spoken instructions, and navigating familiar environments. A 3‐unit difference corresponds to reporting three additional cognitive concerns, indicating more noticeable cognitive difficulties in day‐to‐day activities across domains. This pronounced association underscores the importance of dietary adherence not only for dementia prevention in genetically high‐risk groups but also for mitigating early, subjectively experienced cognitive difficulties that affect daily functioning and often prompt clinical attention, even before clinically diagnosed cognitive impairment. By systematically evaluating long‐term adherence and changes in both Mediterranean and MIND diets over time, our study provides robust evidence supporting the cognitive benefits of sustained dietary adherence across multiple stages and domains of decline.

MedDiet reflects traditional dietary patterns rooted in Mediterranean cultures and has been widely adopted for chronic disease prevention, ⁵⁷ while the MIND diet was developed specifically to support cognitive health, incorporating dietary components with strong evidence for protecting against neurodegeneration. ¹⁹ Although both diets were similarly linked to cognitive benefits, the MIND diet appeared to be more strongly related to early cognitive changes. This may reflect the additional emphasis on brain‐targeted foods such as leafy greens and berries. ¹⁹ , ²⁰ , ²¹ , ⁵⁸ Furthermore, our dose–response analyses showed that the protective associations were more pronounced above certain thresholds: a MIND score of 7 and an aMED score of 3.5. These findings highlight that moderate adherence may not be sufficient to confer measurable cognitive benefit, reinforcing the importance of promoting high‐quality dietary habits. The change analysis adds novel strength to these results; because it examines change within individuals, the impact of confounding factors is substantially mitigated. Notably, the change analyses, which mimic dietary interventions, suggests that improvements in dietary adherence contributed to meaningful cognitive benefits over a relatively short time frame, underscoring the importance of mid‐life dietary modification as an effective strategy to preserve cognitive health. A potential rapid effect of dietary intervention to improve cognitive function is also supported by the COSMOS trial, which showed a significant benefit in the multivitamin arm after just 3 years. ⁵⁹ Lag analyses for dementia risk yielded results consistent with the main analysis, supporting the robustness of the findings to reverse causation. ⁶⁰ Subgroup analyses showed broadly consistent associations across major risk factors for dementia and cognitive decline, suggesting that the cognitive benefits of both diets are largely independent of these factors.

The cognitive benefits of dietary patterns are supported by several potential biological mechanisms. Both diets are rich in bioactive compounds with antioxidant and anti‐inflammatory properties that may modulate key pathways implicated in the pathophysiology of dementia and cognitive decline. For example, polyphenols and flavonoids found in berries, leafy greens, and olive oil have been shown to reduce oxidative stress and neuroinflammation, thereby mitigating Aβ accumulation and tau hyperphosphorylation. ⁶¹ , ⁶² , ⁶³ , ⁶⁴ Omega‐3 fatty acids from fish may modulate microglial activation and promoting neuronal function and synaptic plasticity. ¹⁸ , ⁶⁵ , ⁶⁶ B vitamins, particularly folate, that are abundant in leafy greens, legumes, and whole grains, have been linked to lower homocysteine levels, a known risk factor for cerebrovascular damage and cognitive impairment. ⁶⁷ , ⁶⁸ Moreover, diets rich in fiber and plant‐based foods may influence the gut–brain axis, potentially modifying neuroinflammation and Aβ pathology through the production of short‐chain fatty acids and other microbial metabolites. ⁶⁹ , ⁷⁰ , ⁷¹

Our study has several limitations. First, as an observational study, unmeasured and residual confounding cannot be ruled out, including potential healthy user bias, as individuals with higher adherence to healthy dietary patterns may also differ in other healthy behaviors, although we adjusted for a broad range of potential confounders to mitigate the confounding effects. In addition, although dietary assessments consistently preceded cognitive assessments across repeated measurements and lag analyses showed robust associations, the potential for reverse causation cannot be fully excluded in an observational study, particularly during early stages of cognitive decline. Second, the dietary patterns were derived from SFFQs and were subject to measurement errors. Nonetheless, we have previously demonstrated the reproducibility and validity of our SFFQs for quantifying dietary patterns. ³⁵ Third, the study population consisted primarily of White health professionals with high education attainment and health literacy, which may limit generalizability to other populations. However, this relatively homogenous population also helps reduce confounding and improves internal validity. Fourth, dementia subtype information was not available; future studies are warranted to examine dietary pattern associations with specific dementia subtypes. In addition, SCD is heterogeneous and may be influenced by psychological or other contextual factors. Despite these limitations, our study is uniquely strengthened by its large sample size, extended follow‐up, repeated assessments of diet and dietary change, and the inclusion of multiple cognitive outcomes, including both subjective and objective measures and dementia incidence.

In conclusion, we found that both higher long‐term adherence and increases in adherence to the Mediterranean and MIND diets were associated with a lower risk of dementia, slower cognitive decline, and reduced risk of SCD. These findings provide strong evidence for the role of healthy dietary patterns in preserving cognitive function and highlight the potential of dietary interventions to mitigate both early and late stages of cognitive decline. By evaluating dietary patterns in relation to multiple cognitive outcomes across the cognitive trajectory, our results support the development of dietary guidelines and public health strategies aimed at promoting cognitive resilience in aging populations.

CONFLICT OF INTEREST STATEMENT

The authors report no conflicts of interest.

CONSENT STATEMENT

All human subjects provided informed consent.

Supporting information

Supporting Information

Supporting Information