Dietary Interventions and Cognitive Function across the Dementia Continuum: A Systematic Review, Meta-Analysis, Meta-Regression and Call to Action for Research Reform

Background: Dementia cases are projected to rise to 153 million worldwide by 2050. Diet is a modifiable risk factor, and interventions may delay cognitive decline. No review has synthesised the full randomised controlled trial (RCT) evidence across all dietary interventions and dementia stages. Objective: To critically appraise and synthesise evidence of dietary interventions on cognitive function across the dementia disease continuum (healthy, at-risk/preclinical, mild cognitive impairment, clinical). Methods: Following PRISMA guidelines, seven databases and three trial registries were searched for RCTs from inception to 28 January 2025. Eligible RCTs evaluated multidomain (e.g., diet plus exercise), whole dietary pattern (e.g., Mediterranean diet), or single food (e.g., blueberries) interventions of 2-weeks in adults, 18+ years. Data extraction was triple coded; risk of bias was assessed using the Cochrane Risk of Bias (RoB2) tool. Narrative synthesis was complemented by pooled random-effects meta-analysis. Prespecified meta-regression explored reasons for heterogeneity. Sensitivity analyses were performed. Results: Eighty-three RCTs (110 diet intervention comparisons; n = 24 063 participants) met inclusion criteria. Overall, 190 cognitive measures were extracted and mapped onto the six DSM-V neurocognitive domains. A total of 885 cognitive, neuroimaging, and blood biomarker outcomes were assessed. Pooled meta-analysis (k = 35) showed a small, significant improvement in global cognition (g = 0.25, 95%CI: 0.15 to 0.36). Multidomain (g = 0.25, 95%CI: 0.10 to 0.41) and total diet (g = 0.27, 95%CI: 0.10 to 0.44) interventions showed benefit; single food interventions were non-significant. Trial durations less than 12-weeks (Beta = 0.59) and presence of cardiometabolic comorbidities at baseline (Beta = 0.45) predicted greater cognitive improvements in multidomain interventions; for total diet interventions, lower risk of bias (Beta = 0.51) moderated effects. Neuroimaging revealed benefits in MCI across all intervention types. Certainty of evidence was rated very low. Discussion: Dietary interventions modestly improve cognitive function across dementia stages, with multidomain and total diet interventions showing the most benefit. RCTs during the preclinical phase (e.g., MCI) of dementia in adults with co-occurring cardiometabolic risk are recommended. Given substantial heterogeneity, risk of bias, and measurement inconsistency, international consensus is urgently needed to standardise intervention protocols (e.g., duration, type, dosage) and cognitive outcome measures to strengthen evidence for dementia prevention and care. Funding: NIHR Applied Research Collaboration Oxford and Thames Valley, NIHR Oxford Health Biomedical Research Centre (BRC) Registration: NIHR PROSPERO database (registration: CRD42023488336)