White Matter Lesions and Brain Atrophy: More than Shared Risk Factors? A Systematic Review

In healthy individuals, global and regional brain volume decline starts at a slow rate in early adulthood and accelerates in older age [1, 2, 3]. The rate of this decline varies considerably among individuals [3]. This is of clinical importance since the extent and rate of progression of global and regional brain atrophy are associated with future cognitive deterioration and conversion to dementia [4, 5, 6].

Although brain atrophy may be caused by specific neurological diseases, it is often found on MRI in the elderly without apparent neurological symptoms. Evidence has accumulated that vascular factors play an important role in the etiology of brain atrophy and development of dementia. These factors include hypertension [7, 8], hyperlipidemia [8], diabetes mellitus [9], obesity [10], large amounts of alcohol [11], and cigarette smoking [8, 9].

In addition to brain atrophy, the presence and severity of white matter lesions (WML) increase with age [12, 13, 14]. They too are associated with future cognitive decline and dementia [15, 16, 17, 18]. Patients with subcortical vascular ischemic disease, which is characterized by substantial WML, show progressive cognitive impairment in specific domains and have a considerable risk of developing dementia [19, 20]. Furthermore, vascular risk factors, including hypertension [12, 14, 21, 22, 23, 24, 25], diabetes mellitus [26], obesity [27, 28], and smoking [14, 23], are also associated with WML.

More than 100 years ago, Alois Alzheimer and Otto Binswanger had already described the concomitant presence of subcortical vascular pathology and pronounced atrophy of the white matter with enormously enlarged ventricles in postmortem studies [29]. More recently, concomitant brain atrophy and WML have frequently been observed in elderly people on magnetic resonance imaging (MRI). However, it is unknown which factors may explain the co-occurrence of WML and brain atrophy. One possibility is that their coexistence may be explained by shared vascular risk factors or other shared factors associated with aging. Another possibility is that the disturbances of white matter integrity can contribute to the pathogenesis of brain atrophy by causing ischemic damage to axons, oligodendrocytes, and other glial cells [30].

In this study, we systematically reviewed available studies that investigated the relation between WML and global or regional brain atrophy on MRI to assess whether there is sufficient evidence to conclude that WML and brain atrophy are associated with each other independently of shared risk factors. Finally, we discussed the proposed mechanisms that could underlie the relation between WML and brain atrophy.

We searched PubMed for studies published in English between 1980 and October 2007. Since several synonyms exist for WML and brain atrophy, we made search terms for WML and brain atrophy and entered them as follows: (‘search term for WML’ AND ‘search term for brain atrophy’). We combined this search with limits for language and publication date (fig. 1). The last search was performed on October 31st, 2007. Additional literature was obtained from reference lists of relevant articles.

One investigator (A.P.A.A.) screened all titles and abstracts, and all full-text articles were evaluated by 2 investigators independently of each other (A.P.A.A. and L.G.E.). A consensus meeting was held for cases of disagreement. Since we used a broad search term, we expected a large number of irrelevant articles. Therefore, we defined the following exclusion criteria on the basis of which abstracts could be discarded: case reports and reviews, studies performed in a pediatric study population, and studies in which only computed tomography (CT) was used for imaging. Since we were interested in WML of possible ischemic origin, we also excluded studies that investigated hyperintensities on MRI with other known causes (hematological disorders, metabolic or toxic causes, non-infectious inflammatory or autoimmune diseases, infectious causes, genetic disorders, radiotherapy or chemotherapy, and head trauma).

For all potentially relevant articles, we retrieved the full-text version. Data was extracted from the articles if they met the following criteria: (1) the relation between WML and brain atrophy was studied, (2) the report was published in English, (3) it was an original study, (4) MRI was used for imaging, (5) assessment methods for WML and brain atrophy were specified, and (6) a sample size of at least 20 participants per study group was defined.

For all relevant studies, we recorded the source population, the design (cross-sectional or longitudinal), the sample size, and the mean age of the participants. Next, we described the types of WML and measures of brain atrophy that were investigated and the methods that were used to assess WML and atrophy. Due to heterogeneity in the assessment of WML and brain atrophy, and due to variation in the units of measurements, it proved difficult to compare the measures of effect across studies quantitatively. Therefore, we decided to summarize the results qualitatively and recorded whether the relation between WML and measures of atrophy was statistically significant (p < 0.05). If different subtypes of WML or brain atrophy were investigated within 1 study, we recorded the results separately. Finally, we checked whether adjustments were made for potential shared risk factors in the analyses, and recorded the variables for which adjustments were made.

Our search resulted in 689 articles. As expected, the majority of these articles were not relevant for this review because we used a broad search term. After screening of the titles and abstracts, we retrieved the full-text version of 145 articles. Of these, 41 met the inclusion criteria and 7 additional relevant articles were identified through reference lists of the included articles. The most important reason why articles were excluded was because WML and brain atrophy were both studied as the determinant or as the outcome of interest, but the relation between them was not investigated.

The source populations were very heterogeneous. However, the study populations in the individual studies were more homogeneous. As both WML and brain atrophy are associated with vascular disease and cognitive impairment, the majority of the original studies limited their study populations to either patients from one of these categories or to individuals from the general population. We classified the 48 included articles into 3 main categories to preserve the categorization made in the original studies: general population, subjects with cognitive impairment, and patients with vascular disease.

The general population category consisted of studies that investigated a community-dwelling population or studies that were performed in healthy controls or in patients without neurological or psychiatric disease. In the category of patients with cognitive impairment, study populations were included with (probable) Alzheimer’s disease, mixed dementia or cognitive impairment without dementia. In 1 study, patients with Lewy body dementia were also included [31]. Studies investigating patients with vascular disease were further divided into a group of patients with vascular risk factors only and a group with clinically manifest cerebrovascular disease.

We also distinguished studies that investigated indicators of global brain atrophy from studies that investigated indicators of regional brain atrophy. Measures of global brain atrophy included total brain atrophy, cortical gray matter atrophy, white matter atrophy, and ventricular enlargement. The main measure of regional atrophy was medial temporal lobe atrophy (MTA), including atrophy of the hippocampus, amygdala, and entorhinal cortex.

Studies in which several study populations were investigated and studies reporting on the relation between WML and both global and regional brain atrophy were described in all relevant categories.

The characteristics and results of the 20 general population studies are summarized in table 1. There were 17 cross-sectional studies [9, 13, 22, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45], 2 longitudinal studies [46, 47] and 1 study is included twice in table 1, since both cross-sectional and longitudinal results were reported [48].

Thirteen studies involved a community-based study sample [9, 13, 22, 32, 33, 35, 37, 38, 41, 43, 45, 46, 47], while 5 studies involved healthy controls or hospital personnel [34, 36, 40, 44, 48]. Two studies included neurologically healthy patients with different amounts of WML [39, 42].

Of the 13 cross-sectional studies that investigated global atrophy, 10 found a significant relation between more WML and more global brain atrophy [9, 13, 32, 33, 34, 35, 36, 37, 38, 39]. This relation was also observed in the 2 longitudinal studies investigating global atrophy [46, 47]. Of the 5 cross-sectional studies that examined regional brain atrophy, 3 studies found that WML were associated with more medial temporal lobe atrophy and with smaller hippocampal and amygdala volumes [40, 41, 42]. However, in the 2 other cross-sectional studies there was no association between WML and hippocampal atrophy [43], or hippocampal and entorhinal volumes [48]. Also, WML volume at baseline was not associated with the atrophy rates of the entorhinal cortex and hippocampus during follow-up [48].

In 7 of the 20 general population studies, no adjustments were made for age [22, 32, 36, 37, 40, 42, 45]. Shared risk factors were only considered in 4 out of 20 studies [9, 38, 41, 47], all of which observed an association between WML and measures of brain atrophy.

Nineteen cross-sectional studies [31, 36, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65] and 2 longitudinal studies [66, 67] were performed in patients with cognitive impairment. Characteristics and results of these studies are presented in table 2. In 5 studies [49, 51, 52, 58, 60], both global and regional brain atrophy were investigated, and therefore these studies are presented twice in table 2.

Study populations consisted of patients referred for evaluation of cognitive impairment or dementia [31, 36, 49, 50, 54, 55, 56, 57, 59, 62, 63] or of patients who were recruited at specialized dementia centers [52, 60, 64, 65, 67]. Two studies were performed in MCI patients participating in a clinical trial [53, 66], one study was part of a large genetic study in AD patients and their non-demented siblings [58], and in the other study participants were selected because they met the criteria for probable AD and did or did not show severe WML on MRI [61]. Finally, 1 study was performed in a heterogeneous group of patients with lacunes and a spectrum of cognitive impairment, in patients with probable AD without lacunes, and a control group of individuals without cognitive impairment and without lacunes [51].

The relation between WML and global brain atrophy was assessed in 13 cross-sectional studies. In 9 of these, WML were associated with more global brain atrophy and with smaller cortical gray matter volumes [31, 49, 50, 51, 52, 58, 60, 63, 64]. In the other 4 studies, these relations were not found [36, 55, 61, 62]. The relation between WML and measures of regional brain atrophy was investigated in 11 cross-sectional studies and in 2 longitudinal studies. A significant association between WML and atrophy of the hippocampus or medial temporal lobe was found in 7 cross-sectional studies [52, 53, 54, 56, 57, 58, 59], although in 1 of these studies only periventricular WML, and not deep WML, were associated with medial temporal lobe atrophy [54]. In another cross-sectional study [65], a relation was found between WML in the frontal and parieto-occipital regions and more hippocampal atrophy. In 3 cross-sectional studies [49, 51, 60], no relation was found between total WML volume or WML ratings and volumes of the hippocampus or entorhinal cortex. In 1 longitudinal study, presence and progression of periventricular WML, but not deep WML, were associated with progression of medial temporal lobe atrophy [67]. In another longitudinal study, total WML volume at baseline was not associated with the rate of hippocampal atrophy [66].

In 10 studies, no adjustments were made for potential shared risk factors in the relation between WML and atrophy [36, 50, 51, 52, 55, 57, 58, 59, 61, 64]. Adjustments for age or sex were made in the remaining 11 studies. Three considered hypertension or blood pressure as a possible shared risk factor [63, 65, 67]. In 1 study [54], adjustments were made for vascular risk factors, without further specification. Therefore, although in the last 4 studies associations between WML and brain atrophy were found, it is unknown whether these associations are also independent of other vascular risk factors.

The characteristics and results of the 4 studies [68, 69, 70, 71] performed in patients with vascular risk factors are presented in table 3a and the characteristics and results of the 4 studies [72, 73, 74, 75] performed in patients with symptomatic cerebrovascular disease are given in table 3b. Of these 8 studies, 5 examined the cross-sectional and 3 the longitudinal relation between WML and atrophy. Two cross-sectional studies [70, 72] examined global as well as regional atrophy. Furthermore, 1 study [74] reported the cross-sectional relation between WML and medial temporal lobe atrophy and the longitudinal relation between WML and ventricular enlargement.

The 4 studies that investigated a population with vascular risk factors recruited their patients from outpatient clinics [68, 69, 70] or from a trial on antihypertensive drugs [71]. The studies that investigated patients with manifest cerebrovascular disease included stroke survivors 3 months after discharge [72, 74], patients with a transient ischemic attack [75], and patients with a lacunar stroke [73].

In 2 cross-sectional studies performed in patients with vascular risk factors, associations were found between larger volumes of periventricular WML and more global brain atrophy and both sulcal and ventricular enlargement [68, 69]. However, in 2 other cross-sectional studies in hypertensive patients and normotensive controls, no associations between WML and total brain volume or hippocampal volume were found [70]. In a longitudinal study in hypertensive and normotensive elderly, there was no association between global atrophy rate and degree of WML at baseline, or change in WML severity during follow-up [71].

In the patients with stroke or TIA, cross-sectional associations were found between more WML and more cortical, central and medial temporal lobe atrophy [72, 73, 74, 75]. The relation between WML and ventricular enlargement was also found in the longitudinal study [74].

In 3 studies [68, 71, 75], the analyses were adjusted for age. In 1 of these studies [75] there was an association between WML and brain atrophy after adjustment for systolic and diastolic blood pressure. However, there was no association between WML and brain atrophy in another study [71] in which adjustments were made for other vascular risk factors as well. In the remaining 5 studies, no adjustments were made for age, sex or other possible shared vascular risk factors [69, 70, 72, 73, 74].

We included 48 articles that investigated the relation between WML and measures of global or regional brain atrophy. A significant relation between WML and global brain atrophy was found in 12 out of 15 general population studies, 9 out of 13 studies in patients with cognitive impairment and in 6 out of 8 studies performed in patients with vascular risk factors or symptomatic cerebrovascular disease. A significant relation between WML and medial temporal lobe atrophy – including atrophy of the hippocampus, amygdala, and entorhinal cortex – was found in 3 out of 6 general population studies, 9 out of 13 studies in patients with cognitive impairment, and in 2 out of 3 studies performed in patients with vascular risk factors or symptomatic cerebrovascular disease.

The clinical observation that WML and brain atrophy are often seen simultaneously on MRI is thus confirmed by the majority of the retrieved articles. However, most of the studies did not adjust for possible shared risk factors. Moreover, in 22 of the 48 studies no adjustments were made for age. From these studies, it is not possible to conclude that the relation between WML and brain atrophy is independent of age. When we look at all the studies that did adjust for age and other risk factors, 9 out of 10 studies found that WML were associated with measures of brain atrophy, providing at least some evidence that the relation is independent of shared risk factors [9, 38, 41, 47, 54, 63, 65, 67, 75].

Eight studies examined the longitudinal relation between WML and atrophy. Of these, 1 study did not adjust for age or other possible shared risk factors, 3 adjusted for age or age and sex, and 4 studies also adjusted for a small number of shared risk factors. Of the 7 longitudinal studies that adjusted for age or other factors, 4 found that WML were significantly associated with progression of brain atrophy, whereas 3 did not find an association. Thus, from the existing longitudinal studies, we cannot conclude that WML are a risk factor for brain atrophy.

An interesting pattern was observed when we compared the results of studies investigating global atrophy with the results of studies investigating regional brain atrophy. The majority of studies found an association between WML and global brain atrophy, which remained in the majority of the studies after adjustment for age and vascular risk factors. This association was also present in the majority of the longitudinal studies. The studies that did not find an association between WML and global atrophy generally had smaller sample sizes and used less accurate rating scales, which may have resulted in a low power to detect a significant relationship. In comparison, the results from the studies investigating the association between WML and measures of regional brain atrophy were less consistent. Interestingly, from the 12 studies that adjusted for age or other factors, all 5 studies that used the MTA score found a significant association with WML, while the 7 studies that used volumetric measurements of the hippocampus did not, except for 1 study, find an association with WML. Since the MTA score not only visually assesses volume of the medial temporal lobe, but also volume of the temporal horn of the lateral ventricle and the choroid fissure, it is plausible that concomitant ventricular enlargement accounts for the observed association between WML and MTA score [76]. Support for this explanation comes from a study that found that the MTA score correlated better with the volume of the lateral ventricle than with the volume of the medial temporal lobe itself [77].

Several mechanisms have been proposed that may explain an association between WML and brain atrophy. One possible mechanism is loss of myelin, axons, and oligodendrocytes and other glial cells in the subcortical white matter as a result of ischemic damage due to the underlying small-vessel disease [30, 78]. The association of WML with cortical gray matter atrophy may possibly be explained by ischemic damage to the axons in the subcortical white matter leading to deafferentation of cortical-subcortical connections and subsequent cortical neurodegeneration [60]. Similar mechanisms have been proposed to explain the relation between WML and atrophy of the hippocampus and amygdala, and consequently the medial temporal lobe [41, 65]. Ischemia due to the cerebral small-vessel disease may well damage the axons in the white matter, which eventually could lead to shrinkage of the hippocampus as a result of Wallerian degeneration [79]. Another explanation that has been proposed is that an impaired autoregulation due to the microangiopathy in combination with the luminal narrowing reduces the cerebral blood flow [80]. Since the hippocampus and amygdala are sensitive to hypoxia and ischemia [81, 82], this ischemia may then lead to loss of the neurons in these brain structures [83].

Other observations that may support an association between WML and brain atrophy have come from studies in other neurological conditions, such as CADASIL (cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy), multiple sclerosis [84], amyotrophic lateral sclerosis [85], and HIV [86]. Of these conditions, CADASIL is of particular interest, because the WML in that condition are also presumed to be due to vascular disease [87]. Apart from widespread WML and lacunar infarcts, brain atrophy is also a hallmark of this disease [88]. Furthermore, both hippocampal atrophy [89], subcortical atrophy [90], and global brain atrophy [91] are important predictors of cognitive performance in CADASIL patients. However, volume of WML was not associated with extent of hippocampal [89] or global brain atrophy [88, 90], although this can potentially be explained by the small sample sizes of studies in patients with CADASIL.

Given the plausible biological hypotheses, but inconsistent results, more research is warranted. Future studies should preferably assess WML as well as brain atrophy volumetrically. Visual rating scales are impeded by floor and ceiling effects and their reliability is often limited [92]. Volumetric measurements offer a more reliable, sensitive, and also objective alternative to visual rating scales [93]. However, in case volumetric assessment of WML in longitudinal studies is not feasible, at least dedicated longitudinal visual scales should be used [94]. In addition, quantitative measurements of WML and brain volumes should be normalized for head size [95]. This facilitates comparison between studies and also helps in overcoming the difficulty of interpreting the strength of the relation. Analyses on the association between WML and atrophy should at least take the effects of age and sex into account, but should also address shared vascular risk factors. Another factor that should be considered is the presence or number of lacunar infarcts. Lacunar infarcts are also considered to be caused by cerebral small-vessel disease and are also often found on MRI in the elderly [96, 97, 98]. In patients with Alzheimer’s disease and mixed dementia, lacunar infarcts may be associated with subcortical brain atrophy [60, 99]. Furthermore, in studies that examine the association between WML and atrophy of the medial temporal lobe or structures herein, global brain atrophy should be taken into account to assess whether an association is explained by global brain atrophy. Finally, to properly address the issue of the directionality of the association between WML and brain atrophy, future studies should have a longitudinal design.

In conclusion, the majority of studies found an association between WML and global brain atrophy, but it is yet uncertain if this association is independent of shared risk factors. Further studies on the relation between WML and brain atrophy are therefore warranted and should make proper adjustments in the analysis, preferably be longitudinal in design, and use volumetric assessment of WML and brain atrophy.