Severity of white matter lesions (WML) is divided into 3 categories based on the distribution of periventricular WML scores. The risk of dementia is expressed as age- and sex-adjusted hazard ratios. The number of dementia cases and total number of participants, respectively, in consecutive periventricular WML severity categories were 16 and 749 (grade 0-3), 18 and 261 (grade >3-6), 11 and 66 (grade >6-9), and in consecutive subcortical WML severity categories 24 and 763 (0-1 mL), 10 and 244 (>1-6 mL), and 10 and 65 (>6-29.5 mL).
Bars represent age- and sex-adjusted mean white matter lesion severity (standard error) at baseline for participants without (n = 1032) and participants with (n = 45) incident dementia during follow-up.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Prins ND, van Dijk EJ, den Heijer T, et al. Cerebral White Matter Lesions and the Risk of Dementia. Arch Neurol. 2004;61(10):1531–1534. doi:10.1001/archneur.61.10.1531
To study the association between white matter lesions (WML) in specific locations and the risk of dementia.
The Rotterdam Scan Study, a prospective population-based cohort study. We scored periventricular and subcortical WML on magnetic resonance imaging and observed participants until January 2002 for incident dementia.
We included 1077 people aged 60 to 90 years who did not have dementia at baseline.
Main Outcome Measure
Incident dementia by Diagnostic and Statistical Manual of Mental Disorders, Third Edition (DSM III-R) criteria.
During a mean follow-up of 5.2 years, 45 participants developed dementia. Higher severity of periventricular WML increased the risk of dementia, whereas the association between subcortical WML and dementia was less prominent. The adjusted hazard ratio of dementia for each standard deviation increase in periventricular WML severity was 1.67 (95% confidence interval, 1.25-2.24). This increased risk was independent of other risk factors for dementia and partly independent of other structural brain changes on magnetic resonance imaging.
White matter lesions, especially in the periventricular region, increase the risk of dementia in elderly people.
Cerebral white matter lesions (WML) in elderly people are thought to result from small-vessel disease and are considered to be a risk factor for dementia.1 Evidence relating WML to dementia is mainly derived from studies in patients with stroke and fromcross-sectional studies in patients with dementia. White matter lesions increase the risk of poststroke dementia and, together with lacunar infarcts, are considered the primary type of brain lesions in subcortical ischemic vascular dementia.1,2 Small-vessel disease may also contribute to the development of Alzheimer disease (AD), because patients with AD were found to have more WML than controls.3
White matter lesions are also frequently seen on magnetic resonance (MR) imaging of elderly patients without dementia, but only a few studies investigated the extent to which WML increase the risk of dementia in the general population.4,5 We investigated the risk of dementia for WML in specific locations in the Rotterdam Scan Study. Furthermore, we assessed whether the association between WML and dementia is independent of other risk factors for dementia and other structural brain changes on MR imaging.
The Rotterdam Scan Study is a prospective, population-based cohort study designed to study causes and consequences of age-related brain changes in elderly people. The characteristics of the 1077 participants have been described previously.6 All participants were free of dementia at baseline.5 Baseline examination from 1995 to 1996 comprised a structured interview, neuropsychological tests, physical examination, and blood sampling; all participants underwent MR imaging of the brain. From 1999 to 2000, 787 of the 973 participants who were alive and eligible were reexamined at the research center similar to the baseline examination (response rate, 81%). All participants were continually monitored for mortality, dementia, and stroke until January 1, 2002.
Details of the MR imaging examinations in the Rotterdam Scan Study have been published.6 We considered WML to be in the periventricular region if they were directly adjacent to the ventricle; otherwise, we considered them subcortical. Periventricular WML were scored semiquantitatively for locations at the frontal and occipital horns, and the lateral walls of the ventricles, in order to obtain a total periventricular score (range, 0-9). For subcortical WML, a total volume as appearing on hardcopy was approximated based on the number and size of lesions in the frontal, parietal, temporal, and occipital lobes (range, 0-29.5 mL).6 We rated cortical atrophy on a semiquantitive scale (range, 0-15) and assessed subcortical atrophy by the ventricle to brain ratio (range, 0.21-0.45). Cerebral infarcts were defined as focal hyperintensities on T2-weighted images, 3 mm or larger, and with a corresponding prominent hypointensity on T1-weighted images if located in the white matter.5
Participants with dementia were carefully excluded at baseline.5 We screened all participants for dementia at follow-up with the Mini-Mental State Examination (MMSE),7 and the Geriatric Mental State Schedule8 Screen positives were subsequently evaluated using the Cambridge Mental Disorders of the Elderly Examination.9 Participants who were then thought to have dementia were examined by a neurologist and underwent extensive neuropsychological testing. In addition, we continually monitored the medical records of all participants at their general practitioners’ offices and at the Regional Institute for Outpatient Mental Health Care to obtain information on newly diagnosed dementia until January 1, 2002.5 A panel that reviewed all available information diagnosed dementia and its subtypes according to standardized criteria.10-12 We defined the onset of dementia as the date on which the clinical symptoms first allowed the diagnosis of dementia to be made.
The following variables assessed at baseline were used as possible confounders: age, sex, educational status,13 hypertension, diabetes mellitus, smoking, APOE genotype,14 history of stroke, and incident stroke.5
We assessed the association between WML and measures of generalized brain atrophy with Pearson correlation coefficient, and the association between WML and the presence of cerebral infarcts with linear regression analysis. To examine the relationship between WML and the risk of dementia and AD, we used Cox proportional hazards regression models. We analyzed periventricular and subcortical WML in categories of severity to analyze the shape of the relationship and as a continuous variable (per standard deviation). Adjustments were made for age and sex, and analyses were repeated with possible confounders and measures of other structural brain changes added to the models. Additionally, we excluded participants with a history of stroke at baseline, and participants with a baseline MMSE score of 25 or lower. We examined possible effect modification by APOE genotype through stratified analysis.
Characteristics of the participants are presented in Table 1. Periventricular and subcortical WML were positively correlated with cortical brain atrophy (Pearson correlation coefficient 0.40, P<.01 and 0.25, P<.01) and subcortical brain atrophy (Pearson correlation coefficient 0.21, P<.01 and 0.14, P<.01). Presence of cerebral infarcts was associated with a higher severity of periventricular and subcortical WML (age- and sex-adjusted mean difference in periventricular WML severity 1.6 points, 95% confidence interval [CI], 1.3-1.9 points; in subcortical WML severity 2.0 mL; 95% CI, 1.6-2.4 mL).
During 5572 person-years of follow-up (mean per person, 5.2 years), 45 participants developed dementia (incidence rate, 8.1/1000 person-years). Alzheimer disease was diagnosed in 34 patients (76%), vascular dementia in 6 (13%), and another 5 (11%) were diagnosed as having other types of dementia (Parkinson disease dementia , multiple system atrophy , and unspecified dementia ). One hundred seventy-four participants died. The risk of dementia increased linearly with severity of periventricular WML (Figure 1 and Table 2). Increasing severity of subcortical WML tended to increase the risk of dementia, but this association was less strong (Figure 1 and Table 2). The hazard ratio for dementia per standard deviation increment in periventricular WML score remained largely the same after exclusion of participants with a history of stroke (n = 58) (hazard ratio for dementia 1.66; 95% CI, 1.22-2.27) and after adjustment for incident stroke (hazard ratio for dementia, 1.63; 95% CI, 1.21-2.19). After exclusion of participants with a baseline MMSE score of 25 or lower (n = 173), the association remained (hazard ratio for dementia, 1.50; 95% CI, 1.04-2.16).
Participants who developed dementia during follow-up had on average more severe WML at baseline in all locations within the periventricular and subcortical region (Figure 2). Periventricular WML also increased the risk of AD (hazard ratio for AD, 1.41; 95% CI, 1.01-1.98). The association of periventricular WML and AD was similar for those with and without an APOE ε4 allele (data not shown).
Higher severity of periventricular WML increased the risk of dementia, whereas the association between subcortical WML and dementia was less prominent. The association between periventricular WML and dementia was independent of possible confounders and partly independent of other structural brain changes on MR imaging. The strengths of this study are the large number of participating elderly people, its population-based design, and the fact that we had a complete follow-up for dementia through our monitoring system. Another important feature is the distinction between WML in the periventricular region and WML in the subcortical region.
A preclinical phase of many years often precedes a diagnosis of dementia, especially in the case of AD.15 It is therefore likely that our study population contained participants with a preclinical stage of dementia that remained below detection at baseline. The association of periventricular WML and the risk of dementia did not change after exclusion of participants with a low MMSE score at baseline, which suggests that the association is not confined to participants with a preclinical stage of dementia.
Our results are in line with those from previous studies on the relationship between WML and dementia. Cross-sectional case-control studies reported positive associations of the severity of WML on MR imaging with AD and vascular dementia.3,16 In the Cardiovascular Health Study, participants with more severe WML had a 2-fold increased risk of dementia.4
Several potential mechanisms may underlie the observed associations between WML and dementia. Histopathological studies demonstrated that irregular and confluent WML correspond to ischemic tissue damage, including infarction, gliosis and rarefaction, and loss of myelin.17 This tissue damage is likely to cause disconnection of functionally related cortical and subcortical structures that are important to cognitive functioning.18 It has been suggested that periventricular WML are just an epiphenomenon of brain atrophy and are not independently related to disease.19,20 We found that the association between periventricular WML and the risk of dementia was partly independent of generalized brain atrophy. Furthermore, we found that the association between periventricular WML and incident dementia was largely independent of the presence of cerebral infarcts, of which the majority were lacunar in our study, and was not mediated by incident stroke.
Subcortical WML were not as strongly associated with dementia as periventricular WML, which is in line with previous reports.3 Several possible pathophysiologic mechanisms may explain this finding. First, WML close to the ventricles may interrupt bundles of cholinergic fibers, which extend from the nucleus basalis to the cerebral cortex, resulting in cholinergic denervation.21 Second, the white matter in the periventricular region has a high density of long association fibers, whereas subcortical white matter has a high density of U-fibers. Diffusion tensor MR imaging studies found that white matter pathologic features in patients with AD selectively involved fiber tracts connecting cortical association areas, such as the cingulate bundles and the corpus callosum.22,23 Periventricular WML may reflect vascular damage to these fiber tracts or, alternatively, represent wallerian degeneration of these tracts.
Extensive WML alone are sufficient for a diagnosis of vascular dementia,12 which leads to circularity when associations between WML and subdiagnoses of dementia are studied. However, the observed association between periventricular WML and AD suggests that WML may contribute to clinical AD. This is compatible with the view that most elderly people with dementia have mixed disease.24 Because of the small number of cases with vascular dementia, we cannot provide reliable estimates for the association between WML and the risk of vascular dementia. In conclusion, we found that higher severity of periventricular WML is independently associated with an increased risk of dementia. Longer follow-up with repeated MR imaging is needed to gain insight into whether, and to what extent, progression of WML increases the risk of dementia.
Correspondence: Monique M. B. Breteler, MD, PhD, Department of Epidemiology and Biostatistics, Erasmus Medical Center, PO Box 1738, 3000 DR Rotterdam, the Netherlands (firstname.lastname@example.org).
Accepted for Publication: November 25, 2003.
Author Contributions:Study concept and design (Drs Prins, van Dijk, Vermeer, Koudstaal, Hofman, and Breteler); acquisition of data (Drs Prins, van Dijk, den Heijer, Vermeer, and Oudkerk); analysis and interpretation of data (Drs Prins, van Dijk, den Heijer, Koudstaal, Oudkerk, Hofman, and Breteler); drafting of the manuscript (Drs Prins, van Dijk, Hofman, and Breteler); critical revision of the manuscript for important intellectual content (Drs Prins, den Heijer, Vermeer, Koudstaal, Oudkerk, Hofman, and Breteler); statistical expertise (Drs Prins, van Dijk, and Breteler); obtained funding (Drs Koudstaal, Oudkerk, and Breteler); administrative, technical, and material support (Drs Prins, van Dijk, Vermeer, and Oudkerk); study supervision (Drs Koudstaal, Oudkerk, Hofman, and Breteler).
Funding/Support: This study was supported by grants from the Netherlands Organization for Scientific Research (904.61.096), Den Haag.
Acknowledgment: We thank the Regional Institute for Ambulatory Mental Health Care, Rotterdam and Voorburg, and the general practitioners of Rotterdam and Zoetermeer for their collaboration.