Older adults whose blood already carried a marker of Alzheimer's pathology were still less likely to develop dementia if they followed a diet with lower inflammatory potential, according to a population-based cohort study published in JAMA Network Open on June 25, 2026. Researchers led by Anja Mrhar and Davide Liborio Vetrano at the Aging Research Center, Karolinska Institutet and Stockholm University, followed 1,865 dementia-free adults aged 60 and over from the Swedish National study on Aging and Care in Kungsholmen (SNAC-K), recruited between March 2001 and August 2004 and tracked for up to 15 years, with diet measured repeatedly over the first six years. Some 240 participants developed dementia. Among those with elevated serum p-tau217, each one standard deviation increase in adherence to the reversed Empirical Dietary Inflammatory Index was associated with a 29 percent lower risk of dementia (hazard ratio 0.71, 95 percent confidence interval 0.58 to 0.88). Similar associations appeared in participants with elevated neurofilament light chain (HR 0.79) and glial fibrillary acidic protein (HR 0.73), two markers of nerve cell injury and glial change.
The question the study addresses is now a practical one for families. Blood tests for p-tau217, one of the more accurate blood markers for early detection of Alzheimer's disease, can flag Alzheimer's-related biology in people with no symptoms, and the first thing many ask is whether anything they do still counts. Until now most diet and dementia cohorts had not sorted participants by biomarker status, so it was unclear whether the association survived once pathology was already detectable. In this cohort it did, for one of the three dietary patterns tested.
The pattern that held up at higher biomarker levels was not the Mediterranean one. The team compared three indices: the Alternate Mediterranean Diet, the Alternative Healthy Eating Index and the reversed Empirical Dietary Inflammatory Index, which scores a diet by how its food combinations have tracked with inflammatory markers in earlier research. The Mediterranean and healthy eating indices were generally associated with lower dementia risk only among participants whose biomarker levels were lower; for the Alternative Healthy Eating Index the association differed significantly by neurofilament light chain level (P for interaction 0.008). The MIND diet, widely discussed in consumer coverage, was not among the patterns tested here.
An invited commentary published alongside the paper by Sokratis Charisis and Nikolaos Scarmeas set the findings in the context of an evidence base they described as promising but not definitive. Mrhar has said publicly that the findings do not show that diet prevents dementia. The design is observational: people who eat well differ from those who do not in education, income, activity and health, and the analysis adjusted for such factors rather than eliminating them. Biomarkers were measured in serum at a single baseline point, and elevated levels indicate raised biological risk rather than a diagnosis; many people with them never develop dementia.