維生素 B2（核黃素 / Riboflavin） Vitamin B2 (Riboflavin) × 認知功能

結論：Poorer riboflavin status (highest EGRac quartile, worst riboflavin) was independently and strongly associated with cognitive dysfunction: ORadj 1.73 (95% CI 1.44-2.09, p<0.001) - the largest effect among B-vitamins examined. Lower B12 ORadj 1.30 (1.08-1.58, p=0.007), lower B6 ORadj 1.37 (1.12-1.67, p=0.002), higher tHcy ORadj 1.50 (1.22-1.83, p<0.001) were also significant. NOVEL FINDING: ApoE epsilon4 carriage interacted adversely with low B12 (p_int=0.030) and elevated tHcy (p_int=0.008) but NOT with riboflavin (EGRac x ApoE p=0.789). This is the most recent and largest TUDA cognitive analysis and confirms the riboflavin-cognitive dysfunction signal at scale.

結論：In TUDA, low biomarker status of vitamin B6 and riboflavin were each independent predictors of cognitive dysfunction; combined low B6 + low riboflavin status was associated with a 30-50% increased risk of cognitive dysfunction after adjustment. The MTHFR 677C>T polymorphism alone did NOT associate with cognitive dysfunction, but the homozygous 677TT genotype combined with low riboflavin status was associated with a 53% increased risk of cognitive dysfunction on the RBANS battery - the only gene-nutrient interaction identified for cognition in this cohort. Authors emphasised that no targeted RCT of riboflavin (or B6) for cognitive endpoints in 677TT carriers has yet been performed.

結論：Mean MMSE declined from 29.1+/-1.3 to 27.5+/-2.4 (p<0.001); ~27% experienced accelerated decline. Lower B6 status (PLP <43 nmol/L) was the strongest predictor of accelerated decline: ORadj 3.48 (95% CI 1.58-7.63, p<0.05); lower dietary B6 (0.9-1.4 mg/d) ORadj 4.22 (1.28-13.90, p<0.05). CRITICALLY: NO significant prospective association was found between baseline riboflavin status (EGRac) or dietary riboflavin and 4-year cognitive decline. The riboflavin-cognition signal in TUDA therefore appears to be cross-sectional / contemporaneous rather than predictive of within-individual decline in a smaller longitudinal sub-sample.

結論：Higher dietary riboflavin was associated with better 2-year cognitive performance: global cognition (MoCA) beta = 1.31 (95% CI 0.26-2.35, p=0.015); verbal memory beta = 0.37 (95% CI 0.02-0.71, p=0.038). High-riboflavin subgroup showed less cognitive decline (p=0.004). Unsaturated fatty acid intake gave parallel protective signals. Result is consistent with TUDA cross-sectional findings (PMID 40717068, PMID 27854316) and extends them to a non-Western, non-fortified Chinese population.

結論：Higher dietary riboflavin intake (highest vs lowest quartile) was associated with significantly LOWER odds of cognitive decline across multiple domains: DSST ORadj 0.53 (95% CI 0.37-0.77); composite Z ORadj 0.56 (0.39-0.80); AFT ORadj 0.68 (0.49-0.96); DR ORadj 0.73 (0.53-1.00, borderline). Dose-response was L-shaped with an inflection point near 2.98 mg/d - close to but slightly above U.S. RDA (1.3 mg/d for men, 1.1 mg/d for women), suggesting most protection accrues when intake exceeds RDA but plateaus thereafter.

結論：Higher B2 intake (Q4 vs Q1) was associated with better DSST performance with an implausibly large point estimate (OR 45.1, 95% CI 3.99-510, p=0.004), indicating sparse-data instability - the very wide CI signals that this OR should NOT be interpreted as a magnitude, only as a directional signal. Per-mg increase in B2 intake had a much more credible association with CERAD total score (OR 1.03, 95% CI 1.01-1.04, p=0.003). Associations were stronger in males, non-Hispanic whites, and BMI 18.5-30 kg/m^2. Direction is consistent with PMID 39012764 (also NHANES) and PMID 31555120 (Chinese cohort).

結論：B-vitamin allocation lowered tHcy by ~25% but produced NO significant effect on any cognitive domain (memory, processing speed, executive function) or on global cognitive function; estimated effect equivalent to 0.02 years of cognitive aging per year of treatment - clinically meaningless. RELEVANCE TO B2: very few included trials separated riboflavin from the folic-acid/B12 backbone, so this meta-analysis does NOT directly test riboflavin monotherapy. It does, however, establish that the homocysteine-cognition pathway is not a reliable causal route in unselected older adults - which is the proposed mechanism for B2 benefit via the FAD-dependent MTHFR step.

結論：Very small pooled benefit on global cognition after outlier exclusion (Hedges g = 0.110, 95% CI 0.034-0.186, I^2 = 15.4%, GRADE high in refined analysis); clinically modest, absent in cognitively healthy subgroup, and disappears with trim-and-fill. The MA is the most recent comprehensive pooled estimate but does NOT isolate riboflavin; it provides a ceiling on plausible B-vitamin cognitive effects relevant to the riboflavin discussion.