Plasma Total Homocysteine Levels and Cranial Magnetic Resonance Imaging Findings in Elderly Persons: The Cardiovascular Health Study

Background  An elevated plasma total homocysteine (tHcy) level is associated with an increased risk of vascular disease. Some studies have shown associations between tHcy level and small-vessel disease of the brain on magnetic resonance imaging (MRI).

Design  In the Cardiovascular Health Study, 622 elderly participants without a history of transient ischemic attack or stroke had results for tHcy level and cranial MRI. We sought associations between tHcy level and MRI findings of ventricular grade, sulcal grade, white matter grade, and infarcts. We controlled for other factors, including levels of creatinine, folate, and vitamins B₆ and B₁₂ and methylenetetrahydrofolate reductase genotype.

Results  After controlling for age and sex, tHcy level was not associated with the individual MRI findings. Further adjustments for other factors and other blood tests had little effect on these findings. The only significant finding was a linear trend across quintiles of tHcy level and a pattern of MRI findings combining infarcts and high white matter grade. The linear trend remained significant after controlling for other risk factors and atherosclerotic markers (top quintile vs bottom quintile odds ratio, 3.3; 95% confidence interval, 0.96-11.20; P = .04 for linear trend) but was slightly diminished after further controlling for creatinine, folate, and vitamins B₆ and B₁₂ (odds ratio, 3.2; 95% confidence interval, 0.81-13.10; P = .07 for linear trend).

Conclusion  We were unable to confirm the results of previous studies with respect to tHcy level and individual MRI findings, although an association was seen for an MRI pattern combining infarcts and high white matter grade.

AN ELEVATED plasma total homocysteine (tHcy) level is associated with an increased risk of symptomatic vascular disease, with initial studies^1-3 also suggesting associations with clinically silent findings on cranial magnetic resonance imaging (MRI) thought to represent small-vessel disease of the brain. If such associations were causal, treatments aimed at lowering tHcy level, such as certain vitamins, could lessen the burden of vascular disease on the brain. Levels of tHcy were measured in a subset of participants in the Cardiovascular Health Study (CHS). We examined the association between these levels and findings on cranial MRI to confirm the results from previous studies^1,2 and to examine how other factors related to tHcy levels and not previously examined—creatinine, folate, and vitamins B₆ and B₁₂ and methylenetetrahydrofolate reductase (MTHFR) genotype—affected the associations.

Members of the CHS cohort were recruited from a random sample of people on the Health Care Financing Administration Medicare eligibility lists in 4 US communities. Participants had to be 65 years or older and could not be institutionalized, wheelchair bound in the home, or undergoing active treatment for cancer. Among those who were eligible, 57.3% were enrolled. Details about the study design and characteristics of the original 5201 participants are published elsewhere.^4,5 Participants underwent an extensive baseline examination between 1989 and 1990, including questionnaires, a physical examination, and laboratory testing. All participants were invited to undergo cranial MRI, and 3660 underwent scanning in a standard fashion between 1991 and 1994. Magnetic resonance imaging included sagittal T1-weighted localizer images and axial T1-weighted, spin density, and T2-weighted images.⁶ All axial images had a 5-mm thickness and no interslice gaps. Without knowledge of any clinical information, neuroradiologists at the reading center estimated the white matter, ventricular, and sulcal grades using a 10-point system, from 0, representing no abnormality, to 9, the most abnormal, as detailed previously.⁷ A brain infarct was defined as an area of abnormal signal intensity, at least 3-mm large, in a vascular distribution, and without mass effect.⁶ Because of small numbers, the upper grades for white matter, ventricles, and sulci were collapsed into a single category, leaving 5 levels for each of these MRI findings.

As part of an ancillary study, fasting blood specimens collected at the baseline examination of the original cohort were used to determine plasma levels of tHcy, folate, and vitamins B₆ and B₁₂ in 1300 participants (25.0% of the cohort): all those with an incident myocardial infarction (n = 358), all those with an incident stroke (n = 226), and 716 persons randomly selected from the remaining CHS participants. Without knowledge of the characteristics of the CHS participants, the plasma tHcy level was measured by high-pressure liquid chromatography and electrochemical detection, as previously described.⁸ Plasma vitamin concentrations were measured using radioimmunoassay. One system (Quantaphase II Assay System; Bio-Rad Laboratories, Hercules, Calif) was used for folate and vitamin B₁₂, and a kit (3H-REA; ALPCO Diagnostics, Windham, NH) was used for pyridoxal phosphate, the bioactive form of vitamin B₆. The DNA extracted from peripheral leukocytes was used to determine MTHFR C677T genotype, as previously described.⁹ Plasma assays were performed on blood samples that had not been previously thawed.

For these analyses, 58 of the 3660 participants who underwent MRI were excluded because they had a missing value for 1 or more of the MRI variables defined herein. Of the remaining participants, 722 had also been selected for the ancillary study and 2880 had not. Because 22 participants had missing values for tHcy, 700 remained with results for the MRI scan and tHcy level. Before MRI, 20 participants had a transient ischemic attack; 53, a stroke; and 5, both. The 78 participants with a history of transient ischemic attacks, strokes, or both and the 622 without such a history were more likely to be older white men than the 2880 who had undergone MRI but did not have tHcy level determined. Table 1 details these differences and others among the 3 groups. In general, the 78 participants had more prevalent disease and the 622 had less than the 2880. The 3660 participants who underwent MRI were healthier than the 1541 who did not.⁷ For the 622 participants, the interval from when blood specimens were collected to MRI being performed was a mean (SD) of 3.20 (0.53) years and a median of 3.13 years.

Subsequent analyses focused on the 622 participants who lacked a history of transient ischemic attacks, strokes, or both and who had tHcy measurements available. Results were similar and conclusions the same whether we considered tHcy level as a continuous variable or as quintiles, with exceptions as noted. Following the approach used in the Rotterdam Scan Study,² we used quintiles, which were based on all 1300 participants in the ancillary study. We examined the association between quintiles of tHcy and various factors collected at baseline using linear regression with tHcy level as the dependent variable. Given the strong relation between tHcy level and age and sex, these 2 variables were included in all models as independent variables. To evaluate associations with individual MRI findings, we used linear regression with the grades (5 levels) as the dependent variable and logistic regression with infarcts (present or absent) as the dependent variable. Initial models were with and without adjustments for age and sex. Again, following the approach used in the Rotterdam Scan Study,² we next adjusted for age, sex, systolic blood pressure, use of antihypertensive drugs, diabetes mellitus, and smoking. We then further adjusted for markers of atherosclerotic disease, including the presence of 50% or more carotid stenosis, the intima-media thickness of the common carotid artery, and the ankle-arm index. Finally, we examined how associations were affected by controlling for levels of creatinine, folate, and vitamins B₆ and B₁₂ and MTHFR genotype. Results of the analyses described herein were similar, and conclusions the same, if we used analysis of covariance instead of linear regression.

In a previous CHS report,¹⁰ cluster analysis was used to define patterns of MRI findings. We repeated the logistic regression described in this section using the 5 clusters: normal (n = 146), atrophy (n = 188), leukoaraiosis (n = 98), simple infarct (n = 99), and complex infarct (n = 91). The normal cluster, which lacked infarcts and was low on all grades, always served as the comparison group. The atrophy and leukoaraiosis clusters lacked infarcts, differing mostly by the latter having higher white matter grades than the former. The simple and complex infarct clusters had infarcts, differing mostly by the latter having higher white matter grades than the former. Results were similar when a cluster of interest was compared with all other clusters combined. Also in a previous report,¹¹ incident MRI-defined infarcts were identified in a group of CHS participants who underwent 2 MRI scans separated by about 5 years.¹¹ Logistic regression was used to assess the association between tHcy level and incident MRI-defined infarcts (n = 223). Evidence for effect modification was sought by adding interaction terms to the models described herein. A commercially available software program (SPSS, version 11.0.1; SPSS Inc, Chicago, Ill) was used for these analyses, which were based on the updated CHS database incorporating minor corrections through April 2002.

Several risk factors and biomarkers were significantly associated with tHcy level after controlling for age and sex, which were strongly and independently associated with tHcy level (Table 2). The strongest associations were for the biomarkers other than MTHFR genotype, which was not significantly associated with tHcy level.

Table 3 lists the associations between quintiles of tHcy level and MRI findings. Considering the individual MRI findings, ventricular and sulcal grades were significantly associated with tHcy level, but the significance was lost after adjusting for age and sex. Neither white matter grade nor infarcts were significantly associated with tHcy level before or after adjustment for age and sex.

The results concerning the MRI patterns defined in a previous cluster analysis¹⁰ are also given in Table 3. The linear trend across the quintiles of tHcy was significant only for the complex infarct cluster after adjustment for age and sex. Significance remained after adjustment for other risk factors (P = .03) and, in addition, for atherosclerotic markers but was slightly diminished after further adjustment for creatinine level, vitamin levels, and MTHFR genotype. In these analyses, when tHcy level was included as a continuous variable rather than as quintiles, all of the P values for the complex infarct cluster were smaller than those listed in the table for linear trend. For example, the odds ratio for the complex infarct cluster when adjusted for risk factors was 1.15 (95% confidence interval [CI], 1.04-1.27) (P = .008), and was largely unchanged when the model was, in addition, adjusted for the atherosclerotic markers. With additional adjustment for creatinine level, vitamin levels, and MTHFR genotype, the odds ratio increased to 1.16 (95% CI, 1.01-1.33) (P = .04). None of the other clusters had significant linear trends for tHcy level after adjustment for age and sex.

Also, tHcy level was not significantly associated with incident MRI-defined infarcts before or after controlling for age and sex (Table 3), although the number of participants included in these analyses was small (n = 223). The percentage with MRI-defined infarcts in each quintile was 12.7% in quintile 1, 16.0% in quintile 2, 19.6% in quintile 3, 18.8% in quintile 4, and 11.4% in quintile 5.

For the results presented in Table 3, we sought evidence and found none for significant effect modification (statistical interaction) by sex, baseline myocardial infarction status, the interval between obtaining blood sample and MRI, and MTHFR genotype (data not shown). Also, none of the associations of MRI findings and tHcy level became significant after adjusting for other risk factors, atherosclerotic markers, creatinine level, vitamin levels, and MTHFR genotype, as individual groups of variables or in combinations (data not shown). Results of the analyses summarized in Table 3 were not substantively changed by controlling for the interval between obtaining blood sample and MRI, by focusing on participants with a tHcy level at the 90th percentile and above, or by including the 78 participants with a history of transient ischemic attack and/or stroke (data not shown).

Among these participants in the CHS, plasma tHcy level was not associated with individual MRI findings of white matter grade or infarcts, although a significant association was seen for an MRI pattern combining infarcts and high white matter grade. Plasma tHcy level was significantly related to ventricular and sulcal enlargement on MRI, but these associations lost their significance and the pattern of increasing grades for these MRI findings with increasing tHcy level was eliminated after adjustment for age and sex, both of which were strongly and independently associated with tHcy level. Finally, tHcy level was also not related to incident MRI-defined infarcts in participants who underwent 2 MRI scans separated by about 5 years, although small numbers (n = 223) may have compromised these results. The lack of significant associations was not altered when multivariable models included, as groups or in combinations, other cerebrovascular risk factors, markers of atherosclerosis, or biomarkers, including creatinine, folate, vitamins B₆ and B₁₂, and MTHFR genotype. Results were similar regardless of whether the 78 participants with a history of transient ischemic attack, stroke, or both were excluded.

The only significant associations with tHcy level in the current analyses were for an MRI pattern combining infarcts and high white matter grade, the complex infarct cluster.¹⁰ In these analyses, the normal cluster, which lacked infarcts and had low white matter grades, served as the reference group. The linear trend for the association remained significant (P = .04) after adjustment for other risk factors and atherosclerotic markers. The odds ratio was 3.3 (95% CI, 0.96-11.20) when comparing the top quintile of tHcy level with the bottom quintile. Further adjustment for creatinine level, vitamin levels, and MTHFR genotype slightly weakened the association, with a comparable odds ratio of 3.2 (95% CI, 0.81-13.10). When tHcy level was considered as a continuous variable rather than as quintiles, adjustment for creatinine level, vitamin levels, and MTHFR genotype increased the strength of the significant association. The odds ratio for tHcy level in the fully adjusted model was 1.16 (95% CI, 1.01-1.33) (P = .04), which could be interpreted as the risk of an MRI pattern of complex infarct increasing by 16% for each 0.135 milligram (1.0 micromole) per liter increase in tHcy level. If associations were mediated through 1 or more of creatinine level, vitamin levels, and MTHFR genotype, we would have expected a greater weakening of the association than was observed.

We were unable to confirm the findings from previous studies relating tHcy levels to individual MRI findings. Unlike other studies^1,2,12,13 of symptomatic and asymptomatic subjects with MRI-defined infarcts, we could not demonstrate a significant association between tHcy level and either prevalent or incident MRI-defined infarcts. Similarly, we could not demonstrate an association with white matter grade, as was done in the Rotterdam Scan Study² but not in other studies.^12,14,15 In the Rotterdam Scan Study, the strongest association was for those with infarcts, severe white matter lesions, or both on MRI scan, which yielded an odds ratio of 3.0 (95% CI, 1.8-5.2) (n = 1077) when comparing the top quintile of tHcy level with the bottom quintile. The trend across the quintiles of tHcy level was similar to what we observed in the complex infarct cluster, which yielded a similarly adjusted odds ratio of 3.0 (95% CI, 0.94-9.55) (n = 622). Results limited to those with infarcts and severe white matter lesions were not provided in the Rotterdam Scan Study. Comparisons with results from the Rotterdam Scan Study are difficult because of unknown baseline vitamin supplementation in both cohorts and differences between the cohorts for known baseline characteristics, methods for blood collection and tHcy determination, time between blood collection and MRI, and measurement of white matter changes. Values for tHcy quintiles used in both studies were similar.

Despite its strengths, including an extensive examination and near complete follow-up, and a large number of participants, CHS has its limitations. Prime among them is selective and incomplete participation among those invited to undergo MRI. The overall good health of the participants chosen for these analyses may explain in part the difficulty in reproducing the findings of other studies. In addition, the time between blood testing and MRI was likely longer in CHS, slightly more than 3 years, than in other studies, although stratifying by this interval or controlling for this interval in the analyses had little effect. Perhaps results may have differed if the interval were smaller or if serial tHcy levels were available. The significant findings related to complex infarcts may reflect inadequate adjustment for confounding or may be due to chance given the number of comparisons that we performed. Whether these findings apply to other racial groups cannot be addressed by this study because 96.6% of the participants were white. Finally, these issues would be better addressed by studies with a longitudinal rather than a cross-sectional design, as was used in this study.

We were unable to document an association between tHcy level and individual MRI findings, although a significant association was seen for an MRI pattern combining infarcts and high white matter grade. These findings provide some support for suggestions that an elevated tHcy level may cause subcortical vascular encephalopathy with diffuse white matter lesions and lacunes.¹⁶ Alternatively, prevalent vascular disease may predispose to an elevated tHcy level. Clinical trials^17,18 are under way to evaluate whether vitamins given to patients with symptomatic cerebrovascular disease will reduce the risk of subsequent vascular events. Whether vitamin supplements would prevent small-vessel brain disease among those without symptomatic cerebrovascular disease remains unknown and will not be resolved by ongoing clinical trials because they only include patients who have already experienced a transient ischemic attack or stroke.

Corresponding author: W. T. Longstreth, Jr, MD, MPH, Department of Neurology, Box 359775, Harborview Medical Center, 325 Ninth Ave, Seattle, WA 98104-2420.

Accepted for publication August 25, 2003.

Author contributions: Study concept and design (Drs Longstreth, Katz, Bernick, and Schwartz); acquisition of data (Drs Longstreth, Carr, Cushman, and Schwartz); analysis and interpretation of data (Drs Longstreth, Katz, Olson, Carr, Malinow, Hess, Cushman, and Schwartz); drafting of the manuscript (Drs Longstreth, Katz, Carr, and Cushman); critical revision of the manuscript for important intellectual content (Drs Longstreth, Olson, Bernick, Malinow, Hess, Cushman, and Schwartz); statistical expertise (Drs Longstreth and Katz); obtained funding (Dr Schwartz); administrative, technical, and material support (Drs Longstreth, Carr, Hess, Cushman, and Schwartz); study supervision (Drs Malinow and Schwartz).

This study was supported by contracts N01-HC-85079 through N01-HC-85086, N01-HC-35129, and N01-HC-15103 and grant R01-HL-5471 from the National Heart, Lung, and Blood Institute, Bethesda, Md.

For a full list of participating investigators and institutions in the CHS, see "About CHS: Principal Investigators and Study Sites," available at: http://chs3.chs.biostat.washington.edu/chs/.