Homocysteine and Cognitive Function in a Population-based Study of Older Adults

Jyme H. Schafer, MD, MPH; Thomas A. Glass, PhD; Karen I. Bolla, PhD; Margaret Mintz, MS; Anne E. Jedlicka, MS; Brian S. Schwartz, MD, MS


J Am Geriatr Soc. 2005;53(3):381-388. 

In This Article


Study subjects had a mean age of 59.3 and were 65.9% female and mostly of non-Hispanic African-American (41%) and non-Hispanic white (54.2%) race/ethnicity ( Table 1 ). Although there was a range of educational levels, with 52.4% with a high school or trade school degree or less, the population included large numbers of subjects with higher educational attainments; 41.7% of subjects had a 4-year college degree or more education. A significant proportion of subjects had chronic diseases and the ApoE-ε4 allele, and there was a wide range of cholesterol levels and BMIs among study subjects. Homocysteine levels ranged from 4.4 µmol/L to 48.6 µmol/L (>15 µmol/L is generally accepted as elevated).

In unadjusted analysis, homocysteine levels were associated with worse neurobehavioral test scores for all of the 20 measures (data not shown). Linear regression models with increasing control for covariates are presented in Table 2 . In all models, for all test scores, the signs of the beta coefficients were negative, indicating that higher homocysteine levels were associated with lower test scores. Because the test scores have been standardized, the magnitude of the beta coefficients can be compared across all models. In Model 1, the largest associations are in the domains of eye-hand coordination/manual dexterity and verbal memory and learning. As an example, for every 1 µmol/L increase in homocysteine, the score for the Purdue pegboard dominant hand test decreased 0.051 units (mean number of pegs for three trials), indicating worse performance. The magnitude of the associations across models for each test score are reasonably nonvarying; only tests in the domain of eye-hand coordination/manual dexterity evidenced consistent declines in the magnitude of the associations with homocysteine across models of more than 15%. The CIs from model to model generally widen, as reflected by 16 coefficients in Model 1 whose CIs excluded 1.0 (thus indicating an associated two-tailed P -value <.05), 17 coefficients in Model 2 whose CIs excluded 1.0, and 15 coefficients in Model 3 whose CIs excluded 1.0. The magnitude of the homocysteine relations was large; on average, for all neurobehavioral tests that were associated with homocysteine and age in the base model, an increase of homocysteine from the 25th to the 75th percentile (7.6-11.3 µmol/L) was equivalent in its associations with test scores to an increase of age at the first study visit of 4.2 years.

The analysis to evaluate whether sex, race/ethnicity, age, blood lead levels, or LDL-C modified the effect of homocysteine did not reveal evidence of effect modification (data not shown), but evidence was found that ApoE genotype may modify these relations. The effect of homocysteine was significantly stronger in subjects with the ε4ε4 haplotype (n = 29) (Figure 1). Regression coefficients were negative for 16 of 20 cross-product terms, with five coefficients achieving statistical significance ( P <.05) and one additional coefficient achieving borderline statistical significance ( P <.10) (data not shown).

Adjusted relationship between standardized Rey Auditory Verbal Learning Test (RAVLT) immediate recall score and homocysteine levels (µmol/L) in 978 subjects in the Baltimore Memory Study, 2001-2002. The predicted linear fit for subjects with the apolipoprotein E ε4ε4 haplotype (dashed line) and all other subjects (solid line) are displayed. Subjects with homocysteine of less than 5.2 µmol/L or greater than 16.2 µmol/L are not included in the plot so that the range of plotted levels in the two haplotype groups is the same.

Logistic regression models revealed that homocysteine was associated with increased odds of worse test performance independent of confounders ( Table 3 ). For all tests, the OR for poor test performance exceeded 1.0 for those in the top quartile of homocysteine compared with the lowest quartile. Ten of the CIs for these associations excluded 1.0 (and thus had an associated two-tailed P -value<.05). For example, for Raven's Coloured Progressive Matrices, the OR for being in the lowest quartile of scores (compared with the higher three quartiles) was 1.83 (95% CI = 1.14-2.95) for those with a homocysteine level greater than 11.3 µmol/L (compared with those with a homocysteine level <7.6 µmol/L). This effect of homocysteine is comparable with the magnitude of other known correlates of cognitive function in late life including age and education (college degree or greater). For example, on average, the ORs for homocysteine were 71% of those for age for tests for which both were significant independent risk factors.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.