Novau-Ferré · International journal of epidemiology 2025 · prospective cohort study · n=202302

Glycemic index, glycemic load, and risk of dementia: a prospective analysis within the UK Biobank cohort.

Cited 2 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective observational cohort study

PubMed 41177554 · doi:10.1093/ije/dyaf182 · record verified 2026-08-27

What was done

A prospective cohort analysis evaluated 202,302 dementia-free participants in the UK Biobank. Dietary glycemic index (GI) and glycemic load (GL) were estimated via the Oxford WebQ 24-hour web-based dietary questionnaire. Cox proportional hazards regression models with restricted cubic splines and two-piece Cox models were used to assess nonlinear relationships, identify inflection points, and calculate hazard ratios (HRs) for all-cause dementia, Alzheimer's disease (AD), vascular dementia (VD), and frontotemporal dementia after adjusting for confounders.

What was found

Dementia incidence was 0.89 per 1000 person-years. - Dietary GI (mean ± SD: 48.71 ± 7.68) showed an inverse J-shaped association with dementia risk (inflection point 49.30); GI < 49.30 was associated with lower dementia risk (HR 0.838; 95% CI, 0.758–0.926). - Dietary GL (mean ± SD: 121.49 ± 39.00) showed a J-shaped association (inflection point 111.01); GL > 111.01 was associated with higher dementia risk (HR 1.145; 95% CI, 1.048–1.251). - Similar associations were observed for AD and VD.

Why it matters

This large-scale study suggests that both carbohydrate quality (lower GI) and carbohydrate quantity (lower GL) are independently associated with reduced risk of developing all-cause dementia, AD, and VD.

Limits

Dietary intake relied on self-reported 24-hour web questionnaires, which are prone to recall error and misclassification. As an observational design, residual confounding cannot be ruled out, and causality cannot be determined. Specific risk estimates and case numbers for frontotemporal dementia and other subtypes were not detailed in the abstract. UK Biobank participants exhibit a healthy-volunteer selection bias, which may limit generalizability.

Cited by