Findings from observational studies suggest that dietary patterns may offer protective benefits against cognitive decline, but data from clinical trials are limited. The Mediterranean–DASH Intervention for Neurodegenerative Delay, known as the MIND diet, is a hybrid of the Mediterranean diet and the DASH (Dietary Approaches to Stop Hypertension) diet, with modifications to include foods that have been putatively associated with a decreased risk of dementia.
We performed a two-site, randomized, controlled trial involving older adults without cognitive impairment but with a family history of dementia, a body-mass index (the weight in kilograms divided by the square of the height in meters) greater than 25, and a suboptimal diet, as determined by means of a 14-item questionnaire, to test the cognitive effects of the MIND diet with mild caloric restriction as compared with a control diet with mild caloric restriction. We assigned the participants in a 1:1 ratio to follow the intervention or the control diet for 3 years. All the participants received counseling regarding adherence to their assigned diet plus support to promote weight loss. The primary end point was the change from baseline in a global cognition score and four cognitive domain scores, all of which were derived from a 12-test battery. The raw scores from each test were converted to z scores, which were averaged across all tests to create the global cognition score and across component tests to create the four domain scores; higher scores indicate better cognitive performance. The secondary outcome was the change from baseline in magnetic resonance imaging (MRI)–derived measures of brain characteristics in a nonrandom sample of participants.
A total of 1929 persons underwent screening, and 604 were enrolled; 301 were assigned to the MIND-diet group and 303 to the control-diet group. The trial was completed by 93.4% of the participants. From baseline to year 3, improvements in global cognition scores were observed in both groups, with increases of 0.205 standardized units in the MIND-diet group and 0.170 standardized units in the control-diet group (mean difference, 0.035 standardized units; 95% confidence interval, −0.022 to 0.092; P=0.23). Changes in white-matter hyperintensities, hippocampal volumes, and total gray- and white-matter volumes on MRI were similar in the two groups.
Among cognitively unimpaired participants with a family history of dementia, changes in cognition and brain MRI outcomes from baseline to year 3 did not differ significantly between those who followed the MIND diet and those who followed the control diet with mild caloric restriction. (Funded by the National Institute on Aging; ClinicalTrials.gov number, NCT02817074. opens in new tab.)
Although therapies have recently been approved for Alzheimer’s disease,1 there is a need for prevention strategies. Lifestyle interventions targeting diet are a possible approach that could have an effect on public health. Most clinical trials have investigated comprehensive diets, in contrast to dietary manipulation of single foods or nutrients, and most have been conducted predominantly with respect to cardiovascular health, which could influence the occurrence of dementia. Such diets include the Dietary Approaches to Stop Hypertension (DASH) or the Mediterranean diet to reduce cardiovascular risk factors. Data from trials of dietary interventions for brain health are limited.
We performed a randomized trial of the Mediterranean–DASH Intervention for Neurodegenerative Delay, known as the MIND diet, which is a hybrid of the DASH and Mediterranean diets that incorporates many components of the two but with modifications to include foods that have been putatively associated with a decreased risk of Alzheimer’s disease,2-4 slower cognitive decline,4 and fewer neuropathologic changes of Alzheimer’s disease.5,6 The MIND diet, like its constituent diets, emphasizes consumption of plant-based foods, including green leafy vegetables, nuts and berries, fish, and olive oil. The diet limits the intake of foods with high levels of saturated fat and sugar, such as red or processed meat, butter and margarine, whole-fat cheese, pastries and sweets, and fried foods. The MIND diet specifically incorporates foods that have been putatively associated with a decreased risk of dementia.7-11 Here, we compared the MIND diet with mild caloric restriction with a control diet with mild caloric restriction to evaluate the effects of a 3-year dietary intervention on cognitive decline and brain-imaging markers of dementia and Alzheimer’s disease in older, cognitively unimpaired adults at risk for dementia because of family history.
TRIAL DESIGN AND OVERSIGHT
In this 3-year, two-site, randomized, controlled trial, the participants were randomly assigned to follow the MIND diet with mild caloric restriction for weight loss (goal of intake was to consume 250 kcal less per day) or their usual diet with the same mild caloric restriction for weight loss (control diet). The executive steering committee conceived the trial design and statistical analysis plan. The data coordinating center staff conducted regular safety checks and created data-monitoring reports and analyses for review by an independent National Institute of Health (NIH)–approved data and safety monitoring board of experts in Alzheimer’s disease, clinical trials, and statistics, who were unaware of the trial-group assignments. The trial followed the ethical standards of the 1964 Declaration of Helsinki and its later amendments and was approved by the institutional review boards at Rush University Medical Center, Harvard School of Public Health, and Brigham and Women’s Hospital. All the participants provided written informed consent. The authors vouch for the completeness and accuracy of the data and reporting of adverse events and for the fidelity of the trial to the protocol, available with the full text of this article at NEJM.org.
Persons 65 years of age or older were eligible for inclusion if they had a score of at least 22 on the 30-item Montreal Cognitive Assessment12 (range, 0 to 30, with lower scores indicating greater impairment), were overweight (defined as a body-mass index [the weight in kilograms divided by the square of the height in meters] of ≥25), reported a family history of Alzheimer’s dementia (defined as having a first-degree blood relative with dementia), and had suboptimal diets (defined as a MIND-diet score of ≤8, as based on a 14-item diet questionnaire that was devised by the investigators and designed to detect inadequate diet with respect to brain health [scores range from 0 to 14, with lower scores indicating a less adequate diet with respect to putative brain health; see the Supplementary Appendix, available at NEJM.org]).
Participants were recruited from January 2017 through April 2018 at two sites in the United States (Chicago and Boston metropolitan areas) by means of mass mailings to commercially available lists, recruitment flyers distributed to local health care establishments, and advertisements in newspapers, buses, and subway cars and stations and on the radio. Interested participants contacted the team and underwent a two-stage prescreening protocol for the assessment of eligibility. The initial screening was conducted by telephone to collect information on age, family history of dementia, diet quality, food allergies, use of nutritional supplements, and use of medications for Alzheimer’s disease or Parkinson’s disease. Eligible persons had to agree to not take vitamins (e.g., folate, vitamin B12, or multivitamins) and complete a 3- to 4-week run-in period to assess their likelihood of complying with the trial intervention and activities. During the run-in period, participants had to answer all telephone calls pertaining to the trial and complete dietary records for at least 3 of the 4 weeks. Those who completed the 3- to 4-week run-in period were evaluated by dietitians and trial investigators for suitability for trial participation. The rationale for the trial has been published,13 and full inclusion and exclusion criteria are provided in the protocol.
RANDOMIZATION AND INTERVENTION
Participants were randomly assigned in a 1:1 ratio to the MIND-diet group or the control-diet group. Randomization was performed with the use of permuted blocks created with R software, version 3.3 (R Project for Statistical Computing). Assignments were stratified according to trial site, sex, and age categories (65 to 69 years, 70 to 74 years, 75 to 80 years, and 81 to 84 years). At baseline, measurements (e.g., height, weight, and blood pressure) were obtained, tests of cognition were performed, and questionnaires (e.g., 24-hour dietary recall and food-frequency questionnaire) were completed, and then participants were assigned to a diet group by research staff at the data coordinating center. Participants were instructed to follow their assigned diet for 3 years. Data for the key end points and covariates were collected at month 6 and then yearly until the end of the trial.
Dietary counseling for all participants, led by registered dietitians, was conducted by telephone in a planned schedule of weekly contact during the first 6 months and every other week during the second 6 months; during years 2 and 3, the participants were contacted by telephone at least twice monthly. Individualized guidance was available to participants in both groups. In the MIND-diet group, foods that were part of that diet but may have been lacking, as suspected on the basis of past dietary recall, were identified in order for their consumption to be increased. In the control-diet group, goals that were focused on portion control were suggested. These modifications were intended to alter the participants’ short-term plans to achieve expected targets but did not change the structure of the diet assignment. Dietary counseling for the MIND-diet group included instructions to incorporate foods from the MIND diet and to use MIND diet recipes for behavioral strategies to lose weight, to keep exercise levels the same as at baseline, and to monitor one’s own accountability for dietary intake through coaching and goal setting. Dietary counseling for the control-diet group consisted of an equivalent frequency of consultation but focused on calorie tracking, portion control, and behavioral strategies to lose weight (e.g., coaching, goal setting, and mindful eating techniques) without changing the types of foods consumed. Both groups received equivalent amounts of support with respect to mild caloric restriction with the aim of achieving a weight loss of 3 to 5% by year 3.
Participants were incentivized regularly to follow their assigned diets. The MIND-diet group received a monthly supply of blueberries (2.5 cups, or 15 oz [425 g], per week), mixed nuts (5 oz [142 g] per week), and extra virgin olive oil (14 tablespoons, or 7 fl oz [207 ml] per week) that was donated by nonprofit food-grower organizations. The participants in the control-diet group received $30 gift cards at the same frequency that the MIND-diet group received food. All the participants had at least five opportunities to connect with other members in their assigned trial group in group sessions that took place at months 3, 9, 18, and 30 to cultivate social support. Group sessions included education about their assigned diet, tips to promote mild weight loss, and other motivational activities such as cooking sessions, trivia games, competitions, and holiday celebrations.
Information was collected on several short-term measures of dietary adherence, such as weight assessments, participant-reported dietary assessments with the MIND-diet questionnaire (a 14-item questionnaire designed to capture adherence to the MIND diet) and food-frequency questionnaire, and individualized goal setting. Serum biomarkers of adherence related to key dietary components, including carotenoids (lutein and zeaxanthin, which reflect intake of green leafy vegetables and other cruciferous vegetables) and alpha and beta carotene (e.g., carrots, squash, and peppers), were measured in a subgroup of participants. These measures and biomarkers are summarized in the Supplementary Appendix.
During the coronavirus disease 2019 (Covid-19) pandemic, we were obliged to temporarily halt trial operations for approximately 3 months (March 19, 2020, to July 15, 2020; Fig. S1 in the Supplementary Appendix). During this time, participants in both trial groups continued to receive dietary counseling by telephone, but no in-person testing (e.g., cognition, weight measurements, and blood draws) was conducted. After stay-at-home orders were lifted, 44 participants (24 in the MIND-diet group and 20 in the control-diet group) declined to return to the office for the in-person cognitive testing at the year 3 assessment. We revised the protocol, with guidance from the data and safety monitoring board, to assess cognition (episodic and semantic memory only) by telephone and included these assessments in the primary analysis.
The primary end point was the change from baseline in global cognition and in specific cognitive domains through year 3. Cognition was assessed with an established battery of 12 publicly available cognitive function tests, details of which have been published.14 There were 5 tests of episodic memory (Word List Memory, Word List Recall, Word List Recognition,15 East Boston Story Immediate Recall, and East Boston Story Delayed Recall16); 2 tests of semantic memory (Category Fluency [animals and fruits and vegetables]15 and the Multilingual Naming Test17); 2 tests of executive function (Trail Making Test B18 and the Flanker Inhibitory Control and Attention Test from the NIH toolbox19); and 3 tests of perceptual speed (Oral Symbol Digit Modality Test,20 Pattern Comparison Test,19 and Trail Making Test A18). Raw scores from each test were converted to z scores with the use of the mean and standard deviation values at baseline, and the resulting z scores were averaged across all tests to create a global composite score and across component tests to create the four domain scores. Higher scores on the composite measures reflect better cognitive performance.14 Cognition was assessed at baseline and at months 6, 12, 24, and 36. Cognitive testing was conducted by research assistants who were certified in cognitive testing and unaware of the trial-group assignments.
The secondary end points were changes from baseline in magnetic resonance imaging (MRI)–derived measures of total brain volume, hippocampal volume, and volume of white-matter hyperintense lesions. At trial inception, all the participants were offered the opportunity of undergoing MRI (a 3-Tesla Philips Achieva MRI scanner was used at the Chicago site and a Siemens Skyra MRI scanner at the Boston site). After enrolling the first 100 participants, we estimated the proportion of persons in each diet group who would participate in undergoing MRI, and a nonrandom subgroup of participants were offered MRI to reach the target sample size of 300 (approximately half the enrolled participants). MRI scans were obtained only at baseline and at month 36. Details on data collection and postprocessing are provided in the Supplementary Appendix. The scans were evaluated by reviewers who were unaware of the trial-group assignments.
The trial was powered to test the effect of the MIND diet over the control diet on the annual rate of change in the global cognition score over 3 years. Using a simulation approach in PASS (Power Analysis and Sample Size) 2008 software (NCSS), designed for mixed-model analyses, we analyzed data from two observational cohort studies (Rush Memory and Aging Project7 and Chicago Health and Aging Project21) and estimated the effect size of the MIND-diet score on cognitive decline in persons without cognitive impairment at baseline. We determined that a sample of 300 participants per trial group would provide the trial with more than 90% power to detect a between-group difference of 0.02 standardized units per year in the annual rate of cognitive decline, assuming a dropout rate of 20%. The determination of power for the secondary MRI outcomes was not part of the sample-size estimations. We conducted an intention-to-treat analysis that included all the participants who had undergone randomization.
Because the trajectories of global cognition across time were nonlinear, we estimated the mean change from baseline at each follow-up visit and then compared the changes between the trial groups over time. We used linear mixed-effect models with random intercept, as specified in the protocol, but instead of modeling time linearly, we added time as an indicator variable for each visit. The effect of the MIND diet, as compared with the control diet, on global cognition at each visit was estimated with the use of the product of the indicator variables for trial-group assignment and trial visit. Within-person residuals were assumed to follow a first-order autoregressive model that was conditional on the random intercepts; the autocorrelation coefficient for a lag of 1 year was estimated to be 0.28. A similar approach was used for the cognitive domain scores and other outcomes (e.g., weight change) when nonlinearity was observed. For the analysis of brain morphologic characteristics, we also adjusted models for the clinical site (i.e., Boston and Chicago) to account for differences in the scanners. The widths of confidence intervals for multiple outcomes, such as cognitive domains (four outcomes) and MRI-derived measures (three outcomes), were not corrected for multiple testing, so these confidence intervals should not be used to infer treatment effects for these outcomes.
To account for missing data, we performed multiple imputation using the R package Multiple Imputation by Chained Equations (MICE) algorithm,22 incorporating a wide range of cognition-related predictors such as age, sex, education, cognitive activities, clinical site, and baseline cognitive scores. We considered the analysis that included imputed data as supplementary to the primary analysis. To test whether the inclusion of assessments made by telephone after the Covid-19 pandemic affected primary results, we conducted additional analyses that excluded these telephone assessments.
We standardized brain MRI-derived volumes to intracranial volume and followed standard procedures to normalize the data. We then computed the mean and 95% confidence interval using estimated marginal means, with brain imaging outcomes and dietary assignment and clinical site as independent variables. We computed the difference from baseline to year 3 and then performed pairwise comparisons to evaluate differences between the MIND-diet group and the control-diet group. All analyses were performed with the use of R statistical computing, version 4.0 (R Foundation for Statistical Computing).