Context: Randomized controlled trials (RCTs) testing supplementation with eicosapentaenoic (EPA) and docosahexaenoic (DHA) fatty acids have failed to provide evidence supporting a suggested inverse association between fish intake and dementia risk.

Objective: Dose-response analyses were conducted to evaluate associations between fish intake, all-cause dementia or Alzheimer's Disease (AD), and the effect of EPA/DHA supplementation on cognitive performance.

Data sources: PubMed, Scopus and Web of Science databases were searched for original research evaluating either associations between fish intake and dementia or AD, or the impact of EPA and/or DHA supplementation on the risk of cognitive decline.

Data extraction: Data were collected on study characteristics and methods; number of cases/deaths (for observational studies); categories of exposure; model covariates; risk estimates from the most-adjusted model; type and dosage of supplementation (from RCTs); fatty acid levels in blood; and differences in cognition test results before and after supplementation. Risk of bias was assessed through the ROBINS-E and RoB2.0 tools for observational and experimental studies, respectively.

Data analysis: Weighted mixed-effects models were applied, allowing for the inclusion of studies with 2 levels of exposure. Based on findings with low/moderate risk of bias, fish intake of up to 2 portions (250 g) per week was associated with a 10% reduction (95% confidence interval [CI]: 0.79, 1.02, Ν = 5) in all-cause dementia and a 30% reduction (95% CI: 0.54, 0.89, Ν = 3) in AD risk. Changes in EPA and DHA body status had a positive impact on participants' executive functions, but not on their overall cognitive performance.

Conclusion: The protection offered by fish intake against cognitive decline levels off at intakes higher than 2 portions/week and likely relates to the impact of EPA and DHA on the individual's executive functions, although there remain questions about the mechanisms linking the short- and long-term effects.

Systematic review registration: PROSPERO registration no. CRD42019139528.