Sachdev P, Parslow R, Salonikas C, Lux O, Wen W, Kumar R, Naidoo D, Christensen H, Jorm A. Homocysteine and the Brain in Midadult LifeEvidence for an Increased Risk of Leukoaraiosis in Men. Arch Neurol. 2004;61(9):1369-1376. doi:10.1001/archneur.61.9.1369
High serum homocysteine (HCY) levels have been associated with thromboembolic cerebrovascular disease, but their relationship to microvascular disease is uncertain. Homocysteine also has a direct neurotoxic effect and has been linked to brain atrophy and an increased risk of Alzheimer disease.
To examine the relationship of HCY levels to brain and cognitive measures in a healthy community sample.
Individuals residing in Canberra and Queanbeyan, Australia, who were participating in the longitudinal PATH Through Life Project.
Individuals aged 60 to 64 years selected randomly from the community, 196 men and 189 women.
Main Outcome Measures
Regression coefficients with HCY level as the putative determinant and various magnetic resonance imaging measures (brain atrophy index, ventricle-brain ratios, volume of periventricular and deep white matter hyperintensities) and cognitive measures (information processing speed, verbal memory, fine motor speed) as dependent measures.
Homocysteine levels did not have a significant relationship with brain atrophy index or ventricle-brain ratios. High HCY levels were related to increased deep white matter hyperintensities but not periventricular white matter hyperintensities, after correcting for levels of folate, vitamin B12, creatinine, and thyrotropin; hypertension; smoking; and diabetes, the relationship being significant only in men. Homocysteine levels were related to impairment in verbal memory and fine motor speed but not after the previously mentioned correction.
Total HCY level is independently related to leukoaraiosis in middle-aged men, and this may be functionally relevant in the form of mild cognitive impairment. The remediation of hyperhomocysteinemia should begin early in life if its deleterious effects on the brain are to be prevented.
The relationship between homocysteine (HCY) and the brain has received much recent attention.1,2High HCY levels have long been recognized to increase the risk of vascular disease. A meta-analysis of 27 studies reported a summary odds ratio for cerebrovascular disease of 1.9 for every 0.68-mg/L (5 μmol/L) increment in total HCY level (tHcy),3 while another such analysis reported an odds ratio of 3.97 if the tHcy level was higher than the 95th percentile.4 The focus of these studies has been large-vessel atheroembolic disease. Homocysteine has also been implicated in small-vessel disease, with an increased risk of leukoaraiosis attributed to it,5,6but the evidence for this has been less consistent.7Homocysteine has, in addition, been suggested as a risk factor for brain atrophy in healthy subjects8 and alcoholics9 and has been proposed to increase the risk of Alzheimer disease.10,11
Cross-sectional studies have found an inverse relationship between elevated HCY levels and cognitive function in healthy populations,1,12,13 but longitudinal studies have revealed an inconsistent picture.14- 16 The cognitive impairment is not necessarily explained by increased cerebrovascular disease or by lower folate or vitamin B12 levels.13 These findings suggest a direct neurotoxic effect of HCY, which appears to be dose related.
The brain effects described earlier of HCY are biologically plausible. Homocysteine is recognized to be proatherogenic and prothrombotic. It produces these effects possibly by increasing smooth muscle proliferation, decreasing endothelial DNA synthesis,17 inhibiting anticoagulant factors, increasing platelet aggregation,18 and impairing endothelial-dependent arterial vasodilatation.19 In animal models, HCY has been shown to induce N-methyl-D-aspartate receptor–mediated excitotoxicity,20,21 produce mitochondrial dysfunction and thereby apoptosis,22 promote τ-phosphorylation,23 and increase the neurotoxic effect of β-amyloid.24 We report an investigation of the relationship of HCY levels to various brain parameters and cognition in a community sample of individuals aged 60 to 64 years. We wanted to replicate the findings of the relationship of HCY levels with brain atrophy measures, leukoaraiosis, and cognitive function using quantitative measures in a representative community sample. In particular, we wanted to examine the relationships in a relatively younger population than has been previously studied to see if these relationships can be seen in midlife.
The sample was drawn from the PATH Through Life Project, which was designed to study the risk and protection factors for normal aging, dementia, and other neuropsychiatric disorders. The study cohort comprised 2551 individuals who were residents of the city of Canberra and the adjacent town of Queanbeyan, Australia. They had a mean education of 13.72 years and were recruited randomly through the electoral roll. Enrollment to vote is compulsory for Australian citizens. The response rate was 58.3% for the total sample. About 1 subject in 5 was selected at random for participation in the magnetic resonance imaging (MRI) study. Of 622 participants approached, 478 (77%; 252 men) agreed to undergo MRI of the brain and provided written informed consent. Of the latter, 385 subjects (196 men, 189 women) agreed to provide a fasting blood sample to test HCY levels and other biochemical measurements. To assess for any systematic selection biases, we compared the 385 subjects in the study with the 93 not included. The included subjects had slightly more education (14.00 years vs 13.45 years; P = .03) and performed better on the Spot-the-Word Test25 as a measure of “premorbid” intelligence (52.2 vs 50.4; P = .002) but did not differ in age, sex, or any brain measures examined. Approval for the study was obtained from the ethics committees of the Australian National University, Canberra, and the University of New South Wales, Sydney, Australia.
All subjects were imaged with a 1.5-T Philips Gyroscan scanner for T1-weighted 3-dimensional structural and T2-weighted fluid-attenuated inversion recovery sequence MRI. A scout midsagittal cut for anterior commissure0dash;posterior commissure plane alignment was first acquired. The 3-dimensional structural MRI was acquired in coronal orientation using a T1-weighted free-flow electrophoresis sequence with the following parameters: repetition time/echo time = 28.05 milliseconds/2.64 milliseconds; flip angle = 30°; matrix size = 256 × 256; field of view = 260 × 260 mm; section thickness-2 mm; and intersection distance = 1 mm, yielding overcontiguous coronal sections and an in-plane spatial resolution of 1.016 × 1.016 mm/pixel. The fluid-attenuated inversion recovery sequence was acquired in coronal orientation with repetition time/echo time/inversion time = 11 000 milliseconds/140 milliseconds/2600 milliseconds; matrix size = 256 × 256; field of view = 230 × 230 mm; and section thickness = 4 mm with no gap between sections and an in-plane spatial resolution of 0.898 × 0.898 mm/pixel.
Magnetic resonance images were transferred to an independent Windows NT workstation and analyzed using the software packages Analyze (Mayo Foundation, Rochester, Minn) and SPM99 (Cognitive Neuroscience Group, National Hospital for Nervous Diseases, London, England). The intracranial and total brain volumes were computed automatically using an algorithm within SPM99. The difference of the 2 was divided by the intracranial volume to yield the brain atrophy index. The anterior and midventricular ventricle-brain ratios were measured using the method of Victoroff et al.26 High intrarater and interrater reliability (intraclass correlation coefficients >0.9) were established for each of the previously mentioned measures. White matter hyperintensities (WMHs) were identified on fluid-attenuated inversion recovery sequences. A special computer program was written by one of us (W.W.) to automatically delineate WMHs in both the periventricular and deep white matter regions.27 The absolute volume of total white matter was determined, and the percentage of white matter with a hyperintense signal was calculated for each subject. Since both linear and nonlinear transforms were applied onto each individual MRI when calculating WMH volume, the WMH thus measured is not exactly the absolute volume. It is a measure that would be a true absolute WMH volume if all subjects’ intracranial volumes were the same. However, the “density” (ratio of WMH against white matter for the individual) of the WMH of each individual should be relatively reliable. Twenty images were processed twice to determine the retest reliability of the procedure, and 100% correspondence was noted. For concurrent validity, the images were visually rated by 2 independent physicians experienced in examining MRIs on a scale modified by Fazekas et al.28 One hundred images were visually rated, and the intraclass correlation test between the automated measures and the visual ratings showed a modest correspondence, with intraclass correlation = 0.43 (F99 = 1.76; P = .003) for the whole brain WMH, intraclass correlation = 0.63 (F99 = 2.74; P<.001) for WMH in the deep white matter areas, and intraclass correlation = 0.59 (F99 = 2.44; P<.001) for periventricular WMH. Pearson correlations were also performed, and the results were whole brain, ρ = 0.791 (P<.001); deep white matter, ρ = 0.724 (N = 100; P<.001); periventricular, ρ = 0.717 (N = 100; P<.001).
After overnight fasting, blood samples were obtained for biochemical measurements. For tHcy level, blood was centrifuged within minutes of collection and the plasma stored at −70°C for later measurement. Total HCY level was measured by reverse-phase high-performance liquid chromatography with fluorometric detection after derivatization with 4-(aminosufonyl)- 7-fluorobenzo-2-oxa-1, 3-diazole (reference range, women, <12 μmol/L; men, <15 μmol/L). Homocysteine by high-performance liquid chromatography reagent kits (Bio-Rad Laboratories, Sydney) were used on a high-performance liquid chromatography system (Shimadzu Scientific Instruments, Sydney). Levels of serum folic acid (reference range, 2.4-14.7 ng/mL [5.5-33.3 nmol/L]), vitamin B12 (reference range, 131-534 pg/mL [97-394 pmol/L]), and thyrotropin (reference range, 0.1-3.8 mIU/L) were determined by immunoassay on an IMMULITE 2000 analyzer (Bio-Mediq DPC, Melbourne, Australia). Levels of serum creatinine (reference range, 0.68-1.24 mg/dL [60.00-110.00 μmol/L]) were measured by the Beckman LX20 analyzer (Beckman Coulter, Sydney).
A brief battery of cognitive tests was administered to the subjects within 3 months of the MRI and blood examinations. They comprised the following: Reaction Time (“simple” measured more than 80 trials, and “choice” measured more than 40 trials),29 Symbol Digit Modalities Test,30 California Verbal Learning Test (immediate recall and delayed recall after 1 trial),31 Purdue Pegboard Test (both hands),32 and the Mini-Mental State Examination.33
We examined simple correlations between tHcy levels and various MRI and cognitive measures. Linear and logistic regression analyses were then performed to determine the significance of HCY level as a predictor for 21 variables of interest. These were performed in 2 stages: first, correcting for age and sex, and second, for folate, vitamin B12, creatinine, and thyrotropin levels; hypertension; smoking; and diabetes. The analyses were repeated for men and women separately. Where required, non-normally distributed dependent variables were first transformed using a logarithmic function.
The descriptive characteristics of the sample are given in Table 1. The sample mean ± SD tHcy level was 1.64 ± 0.54 mg/L (12.15 ± 3.98 μmol/L), with 1.78 ± 0.59 mg/L (13.14 ± 4.40 μmol/L) for men and 1.50 ± 0.43 mg/L (11.12 ± 3.19 μmol/L) for women (P<.001). Thirty-six men (18.4%) and 64 women (33.9%) had levels higher than the manufacturer’s reference range. The cut-offs for quintiles for the levels of tHcy in the whole group were 1.25, 1.46, 1.69, and 1.95 mg/L (9.22, 10.80, 12.50, and 14.40 μmol/L), respectively. The tHcy levels were higher in men, but on regression analysis, when folate, vitamin B12, and creatinine levels were included in the analysis, sex differences were not significant. The men and women did not differ on most MRI parameters except women had significantly higher absolute (total volume) and relative (as percentage of total white matter) WMH volumes (Table 2). The cognitive performance was not different for the 2 sex groups. Women had significantly higher vitamin B12 and thyrotropin levels and lower creatinine levels, but the folate levels were not different in the 2 groups.
These correlations are presented in Table 3. Total HCY levels did not have any significant correlation with indices of total brain atrophy or subcortical atrophy. Total HCY levels were, however, significantly related to the volume of deep WMHs, although the correlation was low (r = 0.135). Total HCY levels explained 1.2% of the variance in total WMH levels and 1.8% in deep WMH levels. Total HCY levels also had significant correlation with indices of verbal memory (both immediate and delayed recall) and fine motor speed (as measured by the Purdue Pegboard Test),32 although the correlations were again low (−0.11 to −0.16).
The results of the linear regression analyses with tHcy levels as the predictor variable are presented in Table 4. The only MRI parameter with a significant relationship with tHcy levels was deep WMH, after controlling for age and sex, and this remained significant after entering further covariates: serum folate, vitamin B12, creatinine, and thyrotropin levels; hypertension; smoking; and diabetes. The result was unchanged when WMH was transformed using a natural logarithmic function (R2 = 0.014; β = 0.139; P = .02). The association was maintained in a logistic regression analysis with a tHcy level ≤2.03 and >2.03 mg/L (≤15 and >15 μmol/L).
Fine motor speed on the Purdue Pegboard Test32 was a significant correlate of tHcy level after controlling for age and sex but was not significant once levels of folate, vitamin B12, and creatinine were entered into the analysis. The relationship was low and nonsignificant once WMH level was added as a possible mediating variable between tHcy levels and fine motor speed.
The sex differences in tHcy level and other variables have already been alluded to. When the regression analyses were repeated for men and women separately, the results were largely the same as previously mentioned for men, but none of the MRI and cognitive variables was significantly related to tHcy level in women (Table 5).
We found that in a representative sample of adults in mid-adult life, tHcy levels were elevated in about 15% of individuals. While these subjects were not deficient in folate and vitamin B12 and generally had normal renal function, the tHcy levels had significant negative correlations with folate and vitamin B12 levels and a positive correlation with creatinine levels. We do not yet have genetic information on this population but speculate that a polymorphism of genes for enzymes such as MHTFR may explain much of the variance in the tHcy levels.
Contrary to expectation and based on our finding in a previous study involving somewhat older individuals, tHcy levels did not relate with brain atrophy.8 The Rotterdam Scan Study of 1077 healthy elderly individuals (>65 years) found that those with higher tHCY levels had more cortical atrophy.34 In 156 healthy elderly individuals, Williams et al 35 found that higher HCY levels were independently related to smaller hippocampal widths on MRI. We consider 2 possible reasons for the discrepant findings: age of subjects and methodological differences. The subjects in our current study were younger than in the previously reported studies, and it is possible that the effect of tHcy levels on brain structure does not manifest until later in life, possibly at age 70 years and older. Furthermore, the quantification of brain atrophy on MRI in the present study was much more rigorous. Total brain atrophy in the previous studies was based on visual ratings, whereas we obtained absolute measures. The narrow age range of our sample is noteworthy, since the results of our study are not subject to an artifact of a nonlinear relationship between age and brain atrophy or age and tHcy levels. If it is true that HCY produces obvious brain atrophy only in the later years of life, it presents a window of opportunity for intervention up to midadult life to lower HCY levels.
The important positive finding of our study is the association of tHcy levels with deep WMHs, with tHcy levels accounting for 1.2% to 1.8% of the variance in WMHs. Levels of HCY have previously been reported to be associated with subcortical vascular encephalopathy by 1 group,5 but we failed to find this association in a stroke sample.7 In otherwise healthy individuals, WMHs are most likely to have an ischemic etiology and are related to microvascular disease.36 Homocysteine levels have also been linked to leukoaraiosis in patients with Alzheimer disease,6 but the pathophysiological features of WMHs in these patients may only be partly explained by vascular factors. There are various mechanisms by which HCY could cause or exacerbate WMHs. High HCY levels can cause endothelial dysfunction17,19 and therefore exacerbate microvascular disease. The WMHs may, on the other hand, be the consequence of direct neurotoxicity of HCY by excitotoxic and apoptotic mechanisms.20,21 High HCY levels may also be a marker for hypomethylation, which may affect the integrity of myelin.37 The association was present for deep WMHs but not periventricular lesions. This finding is consistent with the published literature.6 Even though periventricular and deep WMHs were highly correlated (r = 0.81), the pathological basis of these lesions may be different, with ischemic etiology likely to be more important for deep WMHs, as suggested by MRI-histopathological correlative studies.37,38 Periventricular hyperintensities are often related to myelin pallor or rarefaction without other convincing evidence for ischemia. However, the contribution of nonischemic white matter disease to all WMHs cannot be ruled out.38
Homocysteine levels were also related in our study to measures of cognition, but the relationship after adjusting for covariates was not significant. The interpretation of this finding is difficult in a cross-sectional study, but we favor the possibility that nutritional factors lead to cognitive dysfunction through HCY as a mediating factor. The relationship was also nonsignificant if WMHs were entered as a covariate because of a significant negative correlation between WMHs and cognitive function. This prompts us to suggest that the cognitive impairment linked to high HCY levels was possibly mediated through WMHs as an intervening variable. Homocysteine levels correlated with performance on verbal memory and fine motor speed. While the latter is not unexpected, considering the effect of WMHs on psychomotor speed, the relationship with verbal memory was a surprising result. Our brief cognitive test battery did not permit further refinement of the nature of memory disturbance. The distinction between spontaneous recall and recognition memory has been used to distinguish memory disturbance due to frontosubcortical system and hippocampal disturbance, and it would be interesting to have information on recognition memory in this sample. Moreover, the relationship of HCY levels with hippocampal volume is of interest and is currently being examined in our sample.
A noteworthy finding of our study was the sex difference in the relationships. When examined separately, the relationships were significant for men but not for women. Women had slightly but significantly more WMHs as a proportion of total white matter, and their serum folate levels were not different from men. Yet HCY levels were not a significant determinant of WMHs in women. This finding extended to other risk factors of WMHs, such as hypertension, diabetes, smoking, and physical health, which also, except for hypertension, were significant for men but not women in this study (data not reported here). This raises the question whether the pathogenesis of WMHs is different in men and women, an issue that has not been previously examined. Neuropathological studies of WMHs must pay particular attention to the sex of the subjects in future investigations.
This study has some limitations. First, it was cross-sectional. Although we hypothesized about the mechanisms driving the associations among our variables, we recognize that relationships will need to be demonstrated at the individual level through longitudinal follow-up. Second, although the narrow age range permits the ruling out of spurious associations, it had the disadvantage that we were unable to examine the effect of high HCY levels in older individuals. It is possible that the brain effects of HCY, in particular brain atrophy, do not manifest until much later in life, thus explaining the lack of this relationship in our subjects. A third limitation was the exclusion of a number of potential participants, although we did not consider this to have systematically biased our sample.
In conclusion, tHcy levels are independently related to leukoaraiosis in middle-aged men, and this may be functionally relevant in the form of mild cognitive impairment. High HCY levels are therefore a risk factor for microvascular disease much before they contribute to increased strokes or brain atrophy and Alzheimer disease. The remediation of hyperhomocysteinemia should begin early in life if its deleterious effects on the brain are to be prevented.
Correspondence: Perminder Sachdev, MD, PhD, Neuropsychiatric Institute, The Prince of Wales Hospital, Barker St, Randwick NSW 2031, Australia (firstname.lastname@example.org).
Accepted for Publication: March 26, 2004.
Author Contributions:Study concept and design: Sachdev, Naidoo, Christensen, and Jorm. Acquisition of data: Sachdev, Salonikas, Lux, Wen, and Kumar. Analysis and interpretation of data: Sachdev, Parslow, Salonikas, and Jorm. Drafting of the manuscript: Sachdev and Kumar. Critical revision of the manuscript for important intellectual content: Sachdev, Parslow, Salonikas, Lux, Wen, Naidoo, Christensen, and Jorm. Statistical expertise: Parslow and Kumar. Obtained funding: Sachdev, Naidoo, and Jorm.Administrative, technical, and material support: Salonikas, Naidoo, Christensen, and Jorm. Study supervision: Sachdev, Kumar, and Jorm.
Funding/Support: This study was supported by Project Grant ID 157125 and Program Grant 179805 from the National Health and Medical Research Council of Australia, Canberra.
Acknowledgment: We thank the following for their contribution to the PATH Through Life Project: Kaarin Anstey, PhD; June Cullen, RN; Trish Jacomb, MSc; Jerome Maller, MSc(Psych); Karen Maxwell; Chantal Meslin, MD; Jeremy Price, FRACR; Bryan Rodgers, PhD; and the PATH interviewing team. Angie Russell, BSocSc(Hon) prepared the manuscript.