Which Brain-Healthy Diet Scores Track Best With Dementia Risk?

The short answer

If you want a clean ranking from human data, Mediterranean-style scoring came out on top, with MIND close behind. In the best head-to-head comparison here, a 2025 NHANES analysis by Gong et al. found that the alternate Mediterranean diet score, or aMED, had the strongest inverse association with Alzheimer's disease mortality, and the MIND score also performed well. DASH and HEI-2015 looked favorable but weaker. The healthy plant-based diet index, hPDI, was less consistent.

That does not mean a Mediterranean or MIND diet has proved it prevents Alzheimer's disease. It means that, in observational data, those score patterns tracked brain-related outcomes better than the other diet indices tested in the same dataset. That is useful. It is also a long way from a prevention claim.

Why this NHANES comparison matters

Diet studies often suffer from a simple problem: one paper tests one score in one cohort against one endpoint, and then people act as if all healthy eating scores are interchangeable. The Gong et al. paper is more informative because it compared five common diet-quality scores inside the same U.S. population and against two brain-relevant outcomes.

The study used NHANES baseline waves from 1999 through 2016 for adults aged 40 and older, with mortality follow-up through December 31, 2019. It also used a cognitive subsample from NHANES 2011-2014 in adults aged 60 and older. The cognition outcome was psychometric mild cognitive impairment, or p-MCI, defined as performance more than 1 standard deviation below the age-, sex-, and education-adjusted mean on a composite drawn from CERAD, Animal Fluency, and the Digit Symbol Substitution Test. Within that setup, aMED and MIND tracked best overall with lower Alzheimer's disease mortality and lower odds of p-MCI. DASH and HEI-2015 were weaker but still directionally favorable. hPDI lagged.

The missing piece is exact effect size detail. The published article contains the hazard ratios, odds ratios, and confidence intervals, but those numbers are not reproduced in the material summarized here. That means the rank order is fair to report, while any precise magnitude claim would need to be checked against the full tables first.

The broader literature mostly points the same way

The NHANES result would be less interesting if it stood alone. It does not. A 2025 systematic review by Tse et al. concluded that stronger adherence to the MIND diet is generally associated with better cognitive health in middle-aged and older adults, while also emphasizing heterogeneity across studies and outcomes. That is the right level of confidence: repeated observational signal, not settled proof.

Other cohort-style studies push in the same direction. A 2025 Multiethnic Cohort analysis linked healthier dietary patterns with lower risk of Alzheimer's disease and related dementias across five racial and ethnic groups. In the UK Biobank, Cornelis et al. in 2023 reported that the MIND dietary pattern was associated with cognition and incident dementia. Different cohorts, different demographics, same general theme: MIND-like and Mediterranean-like patterns keep showing up.

Some supportive studies also move beyond a single outcome. In the 3-year DO-HEALTH study, Sager et al. found that higher MIND adherence was associated with lower odds of mild cognitive impairment in generally healthy older adults. In another NHANES analysis, Derdiger et al. in 2024 studied 2,598 adults with two 24-hour recalls and cognitive testing and reported that both MIND and a Mediterranean-like score called MEPA III were linked to better cognitive performance. Those papers do not test Alzheimer's mortality, but they reinforce the idea that brain-related associations are stronger for some diet patterns than for generic "eat better" scores.

Score construction matters

The interesting part is not just that diet quality matters. It is that different ways of scoring diet quality do not behave the same way. That is the real lesson from the NHANES comparison.

MIND and aMED reward food combinations that look more brain-specific than "healthy eating" in the abstract. Gong et al. found that hPDI, despite being plant-forward, was less consistent than aMED or MIND. That matters because people often flatten diet science into a moral fable about plants. The data here are fussier than that. A scoring system can favor plant foods and still miss whatever combination better tracks cognitive outcomes.

That point lines up with other evidence. Bhave et al. in 2024 linked higher ultra-processed food consumption with adverse brain health outcomes, which fits the broader idea that food pattern quality is not captured equally well by every index. And Carrasco-Marín et al. in 2025 compared four dietary indices against all-cause and cause-specific dementia in UK Biobank, a reminder that rankings can shift depending on what the score rewards and which endpoint you choose.

Why this still falls short of prevention proof

The strongest brake on hype is the 2023 randomized MIND trial by Barnes et al. Randomized evidence should carry more weight than observational ranking, and this trial did not clearly show that prescribing the MIND diet beat an active control enough to justify bold prevention claims. That does not erase the observational literature. It does mean you should stop short of saying, "MIND prevents cognitive decline," because the best intervention evidence here did not cleanly deliver that result.

A later 2026 randomized follow-up analysis by Cornelis et al. adds another wrinkle: response to the MIND intervention varied by genetic background. If the intervention effect depends partly on genotype, then a population-level association in cohort studies may not translate into a uniform benefit for every person. Biology, as usual, declined to be simple.

Longitudinal observational work also tempers any neat ranking. Seago et al. in 2025 examined neuroprotective dietary patterns against longitudinal changes in cognitive function in older adults. That is a harder test than a cross-sectional snapshot, and it does not make Alzheimer's mortality interchangeable with cognitive trajectory. A diet score can look good against one outcome and less impressive against another without anyone committing fraud or witchcraft. The endpoint changed.

Important limitations

The NHANES paper has real strengths, but its limits matter. First, the Alzheimer's outcome was death-certificate Alzheimer's disease mortality coded as ICD-10 G30. That can misclassify dementia subtype and tells you nothing about nonfatal cognitive decline. Second, the p-MCI analysis was cross-sectional and based on neuropsychological testing, not clinician-adjudicated mild cognitive impairment or incident Alzheimer's disease. Useful for population signals, imperfect for disease diagnosis.

Diet measurement also adds noise. In Gong et al., dietary exposure came from one 24-hour recall in 1999-2002 and the mean of two recalls in 2003-2016. That can rank population diet quality reasonably well, but it is still a rough proxy for long-term eating patterns over the many years that matter for neurodegeneration.

Residual confounding is another obvious concern. The NHANES models adjusted for age, sex, survey wave, education, race or ethnicity, income-poverty ratio, smoking, physical activity, total energy intake, BMI, hypertension, diabetes, and cardiovascular disease. That is substantial. It still cannot fully remove confounding from healthcare access, frailty, medication use, depression, food insecurity, social isolation, or preclinical disease changing the way people eat.

Generalizability is also limited. The NHANES mortality cohort covered U.S. adults aged 40 and older, the cognitive subsample covered adults aged 60 and older, DO-HEALTH focused on generally healthy older adults, and several MIND-related studies examined populations already at risk for decline. None of that maps neatly onto a 35-year-old supplement maximalist who thinks blueberries can out-negotiate genetics.

What a careful reader should conclude

If the question is which healthy diet scores best tracked brain outcomes in the evidence here, aMED ranked first and MIND was close behind. DASH and HEI-2015 were generally favorable but weaker. hPDI was less consistent. That conclusion comes mainly from the NHANES head-to-head analysis and fits the surrounding observational literature.

If the question is whether any of these diets has proved it prevents Alzheimer's disease, the answer is no. The observational signal is repeated, but randomized evidence remains limited and mixed, and some intervention effects may vary by genotype. The sober verdict is still the useful one: Mediterranean-style and MIND-style patterns look like the best bets for tracking better cognitive outcomes at the population level, but the word "prevents" is doing work the evidence has not earned.