Association of white matter hyperintensities and gray matter volume with cognition in older individuals without cognitive impairment

White matter hyperintensities (WMH) are brain white matter lesions with high signal on T ₂-weighted fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI). These lesions are believed to represent underlying pathologic changes which are variable in nature and severity, including alterations in myelin and axon structure, gliosis, and small vessel disease (Gouw et al. 2011). Clinical experience and mounting research data suggest that WMH are very common in aging and associated with a range of vascular diseases including stroke (Breteler et al. 1994; Raz et al. 2012; Burton et al. 2004). The clinical significance of WMH has been well studied, and associations have been demonstrated with impaired neurologic function including cognitive and motor impairment, as well as with death (Kerber et al. 2006; Silbert et al. 2008; Debette and Markus 2010). Yet, despite better recognition of WMH with increased availability and use of sophisticated neuroimaging technology, and the consensus that WMH should no longer be considered an “incidental” finding on neuroimaging, little information is available on the relationship of WMH with neurologic function in individuals without overt neurologic impairment. The data to date suggest that WMH are likely associated with subtle changes in neurologic function, including lower cognitive function, even among healthy older individuals without overt cognitive impairment [e.g., without Alzheimer’s disease or mild cognitive impairment (MCI)], and these results have been supported by meta-analyses (DeCarli et al. 1995, Gunning-Dixon and Raz 2000; Nordahl et al. 2006; Ishikawa et al. 2012). However, few studies have systematically examined the relationship of WMH with a range of cognitive domains among older individuals without neurologic conditions or diseases, and results of these few studies are mixed, with data suggesting associations with some domains (e.g., executive function) but not others (Vannorsdall et al. 2009; Murray et al. 2010; Hedden et al. 2012).

While a body of literature supports a relationship of gray matter to cognition in older persons, including among persons without known cognitive impairment (Jack et al. 2000; Kramer et al. 2007; Fleischman et al. 2013), few reports are available that take the effects on cognition of both WMH and gray matter volumes into account simultaneously. Indeed, among studies examining white matter together with gray matter, many had small samples of cognitively normal subjects, were restricted to MRI volumes in select brain regions, or did not measure WMH lesions (Hedden et al. 2012; Mueller et al. 2010; Taki et al. 2011; Fletcher et al. 2013; Royle et al. 2013; Wirth et al. 2013; Papp et al. 2014). No study that we know of has directly tested independence of both total WMH and total gray matter volume effects on cognition among a large group of older individuals without cognitive impairment. Studies that consider multiple brain measures simultaneously are needed to improve understanding of the interactions of neuropathologic events and the pathophysiologic cascade. Further, studies of cognitively-normal individuals are critical to shed light on the earliest factors that may play a role in neurologic decline and, as such, are important for identifying plausible strategies for prevention of cognitive impairment, and, ultimately, for improving public health.

This study examines the relation of total WMH volume to cognitive function in different cognitive domains, and tests whether the relation is independent of total gray matter volume. We used brain MRI and neuropsychologic data from more than 200 well-characterized, older community-dwelling women and men without dementia or MCI, participating in an epidemiologic study of aging, the Rush Memory and Aging Project. In the first set of analyses, we examined the relation of total WMH volume on brain MRI scans to level of cognitive function in composite measures of five separate cognitive domains. Secondary analyses determined the relative contribution of WMH to cognition, compared to that of demographic variables, and addressed whether findings were modified by age. In the second set of analyses, we examined for independence of effects of total WMH volume and total gray matter volume on cognition.

In this study of more than 200 older individuals without dementia or mild cognitive impairment, we found that WMH on brain MRI were common, associated with older age, and related to cognition. Specifically, larger total WMH volumes were associated with lower levels of perceptual speed, but not with memory or other cognitive functions. The effect of total WMH volume on perceptual speed was nearly half that of the demographic variables combined, and the association of total WMH volume with perceptual speed was independent of age. Notably, when considering total WMH and gray matter simultaneously, we found that the association of larger total WMH volume with lower perceptual speed in older individuals without overt cognitive impairment was independent of total gray matter volume.

It is now well recognized that WMH are commonly observed in the brains of older persons, and that the burden of WMH increases with age (Breteler et al. 1994; Ylikoski et al. 1995; Gouw et al. 2008; Raz et al. 2012). This burden may be, in part, due to increased neurologic diseases in aging, including cerebrovascular disease (Burton et al. 2004; Rost et al. 2010). Yet, WMH are also common in middle-aged healthy adults (with one study reporting 51 % of 428 healthy subjects in their 40’s having WMH), and in older adults without known neurologic diseases or conditions (Wen et al. 2009; Silbert et al. 2008). Findings from our study are consistent with this literature. With a mean age of about 80 years and no clinically recognized cognitive impairment, all of the subjects in this study had some WMH, mostly in relatively small volumes, as would be expected in a healthy volunteer cohort.

A large body of literature shows that WMH have detrimental effects on neurologic function, including cognition (Silbert et al. 2008, 2012; Debette and Markus 2010; Wakefield et al. 2010; Kloppenborg et al. 2014). WMH have been associated with cognition among persons with a range of cognitive function including dementia, and clinical diagnoses including Alzheimer’s disease (Aggarwal et al. 2010; Smith et al. 2011; Lo et al. 2012; Carmichael et al. 2012; Maillard et al. 2012). Associations with cognition among persons with mild cognitive impairment have also been reported (Yoshita et al. 2006; Debette et al. 2007; Carmichael et al. 2010; Lo et al. 2012). However, little is known about WMH and cognition among persons without dementia (Aggarwal et al. 2010), and even less so among persons without dementia or mild cognitive impairment. Several factors may contribute to the limited knowledge in this area. Some studies have evaluated cognition in seemingly healthy older individuals, but have not excluded persons with mild cognitive dysfunction (Au et al. 2006; Wright et al. 2008; Ishikawa et al. 2012). Other studies identified cognitive impairment but employed a non-specific screening test (e.g., MMSE < 24), which does not exclude all cases with mild dementia or mild cognitive impairment (Söderlund et al. 2006; van den Heuvel et al. 2006; Wakefield et al. 2010). Finally, some studies used only a single measure of cognition or had a small sample (e.g., <50 subjects) (van der Flier et al. 2005; Silbert et al. 2009; Raji et al. 2012).

A few larger studies of WMH and cognition in the elderly, specifically excluding persons with dementia and mild cognitive impairment, have examined the association of WMH with different cognitive domains. In one study, WMH burden was related to performance on timed tasks (Vannorsdall et al. 2009). Two subsequent studies showed an association of WMH with lower executive function but not memory (Murray et al. 2010; Hedden et al. 2012). Here, we found that larger total WMH volume was related to lower scores on a composite measure of perceptual speed. Our finding contributes to the mounting evidence showing that greater WMH is associated with lower cognition, and lower perceptual speed in particular, in individuals without overt cognitive impairment. Further, while the effect of WMH on cognition may be comparable to that of age as suggested by one study (Hedden et al. 2012), we found that WMH volume accounted for nearly half of the variance in perceptual speed explained by demographic variables including age, underscoring the strength of the effect of WMH on cognition in older individuals without overt cognitive impairment.

Factors accounting for the relation of WMH to cognition remain to be elucidated. As increasing age is associated with WMH and also with cognitive impairment, we considered the role of age in our study. All analyses controlled for age, yet we found an association of larger WMH volume with lower cognition, suggesting that age is not a crucial confounder in this relationship. Furthermore, we found no evidence of modification by age of the association between WMH and cognition, suggesting that the relation of WMH to cognition is independent of age. Additional research is needed that examines other factors which may modify the relationship of WMH with cognition among overtly healthy older individuals. Plausible factors include genetic factors, vascular risk factors and diseases (such as hypertension and diabetes), inflammation, and oxidative stress (Novak et al. 2006; Wright et al. 2009; Xu et al. 2010; Satizabal et al. 2012; Raz et al. 2012).

Few studies have examined effects of both WMH and gray matter volumes on different cognitive domains. We are aware of only one other study which considered both total WMH and total gray matter volumes that included persons with normal cognition (He et al. 2012). In this clinic-based study of persons with a range of cognitive function (47 with dementia, 65 with MCI, and 97 with normal cognition), WMH and gray matter volume were separately associated with episodic memory, but in analyses taking both MRI measures into account simultaneously, only gray matter remained associated (He et al. 2012). The authors concluded that gray matter is more closely related to cognition than WMH. A more recent study included exclusively persons without dementia (n = 81, all with MMSE ≥ 24), and considered volume in a specific gray matter region (the hippocampus), in addition to WMH (Papp et al. 2014). In this study, WMH and hippocampal volumes were independently associated with processing speed when both MRI measures were considered simultaneously (Papp et al. 2014). Thus, while research has been done in persons without dementia as noted above, we are not aware of any other study in older persons that also restricted their study to those without mild cognitive impairment (MCI), and directly examined the effects simultaneously of both total WMH and total gray matter volumes on different cognitive domains. In comparison to the studies by He et al. and Papp et al., we found, in about 200 community-dwelling persons without dementia or MCI, that larger total WMH volume and smaller total gray matter volume were each separately associated with lower perceptual speed, but when taking both MRI measures into account, WMH remained related to perceptual speed while gray matter volume was no longer related. These results suggest that the association of WMH volume with perceptual speed is independent of gray matter volume. Furthermore, the results suggest that the relationship of gray matter volume to cognition may be accounted for, at least in part, by WMH volume. Also, our finding of an association of gray matter volume with episodic memory, similar to the finding by He et al. (2012) suggests that the relation of MRI data to cognition may vary depending on the particular domain of cognition being examined.

We found regional specificity of the association of WMH with cognition, consistent with a recently published study of adults with a lower age on average and high cognitive function (mean age 60 years, and all with MMSE ≥ 25; Birdsill et al. 2014). In that study, authors exploited voxel-wise analyses and found an association of WMH, most notably in the superior corona radiata, with lower cognitive speed and flexibility (Birdsill et al. 2014). Our study extends this finding, by showing that the periventricular brain regions with WMH, including the corona radiata, and also the thalamic radiations and corpus callosum, play a role in the association of WMH with lower perceptual speed, after controlling for total gray matter volume. While the literature has previously suggested that periventricular lesions are associated with impaired perceptual speed, further studies using modern imaging and analytic techniques are needed to examine regional relationships of WMH with cognition (van den Heuvel et al. 2006).

There are several possible interpretations of our results evaluating both WMH and gray matter volumes. First, WMH volume may be a confounder in the relation of gray matter volume with cognition. Yet, the association of gray matter volume with cognition is biologically plausible and extensive literature supports such an association. Second, early pathologic processes may occur in the gray matter and lead to loss of gray matter volume (e.g., neuronal cell body dysfunction and neuronal death, resulting in gray matter atrophy), and may then be followed by changes in the white matter (e.g., changes in axons or the myelin covering the axons, and WMH), and then ultimately lead to brain dysfunction (e.g., lower cognitive function). In this case, WMH would mediate the relationship between gray matter volume and cognition. This sequence of events in the pathophysiologic cascade, in which WMH are more closely related to cognition than gray matter, contrasts with a previously published study (He et al. 2012). The discrepant results may be due to differences in source of participants (e.g., community-based individuals without overt cognitive impairment vs. persons with a range of cognitive function). Indeed, the associations of brain pathology with cognition may not be the same across a spectrum of health/disease. It is possible that once individuals develop cognitive impairment or dementia, that gray matter volumes become a more important factor in the association with cognitive function, compared to WMH. By analogy, we have previously observed differing results in the relation of two Alzheimer’s disease pathology measures (amyloid and neurofibrillary tangles) to cognition, in studies restricted to persons without cognitive impairment vs. not (Bennett et al. 2004, 2012b). In addition, discrepant results regarding the importance of WMH may also be due to differences in study design and methodology, including MRI data acquisition and image processing, neuropsychological test selection, data summary and analyses. Further study is clearly needed to definitively establish the relationship of WMH, gray matter, and cognition. Elucidation of the pathological processes that lead to lower cognitive function, and of sequences of events in that pathway (e.g., whether reduced gray matter volume occurs before WMH, or vice versa), is important to better understand mechanisms of aging and disease, and to inform on targets for intervention to prevent cognitive decline. To our knowledge, this study is the first that has tested independence of total WMH and gray matter volume effects on different cognitive domains in older individuals without overt cognitive impairment.

Weaknesses of this study should be noted. First, our aim was to examine for independent effects of total WMH and total gray matter volumes on cognition, and thus we did not examine relationships in analyses that considered both regional WMH and regional gray matter data. While our study’s novelty is enhanced by voxel-wise analyses of the association of WMH with perceptual speed, further examination of the relation of regional MRI data to cognitive domains is likely to shed additional light on underlying pathophysiologic processes and will need to be explored in subsequent studies. Second, this study presents cross-sectional analyses of the relation of WMH to cognition. Data on the evolution of WMH changes is expected to be more informative and we have plans to analyze the relation of change in WMH to cognition in the future as data become available. Third, while secondary analyses tested for effect modification by age, we did not take other factors, such as vascular factors, into account. We are collecting MRI data on brain infarcts and, as all participants were enrolled in a clinical-pathologic cohort study with a high autopsy rate, neuropathologic data on cerebral infarcts will become available for future analyses. We plan to examine this issue more directly in individuals with clinical, brain MRI, and neuropathologic data.

Strengths of the study are also worth noting. First, analyses were conducted in older community-dwelling women and men who were clinically well-characterized, and did not have dementia or mild cognitive impairment. This allowed for the examination of the relationship of WMH to cognition in a group of older individuals without overt cognitive impairment. Second, detailed neuropsychological evaluations proximate to the MRI scan were conducted, and outcome measures included five different cognitive domains. Each cognitive domain was a composite measure based on two or more individual tests, thus decreasing ceiling effects. Ceiling effects are of particular concern in this study given the high cognitive performance of individuals included in analyses. Finally, several strengths of the study are derived from the methods used to assess and analyze WMH data. Total brain WMH volume data were collected in an automated fashion, removing user bias, and were based on information from both T ₁-weighted and FLAIR images, improving robustness of segmentation and lending internal validity to the study. Furthermore, the consideration of WMH as continuous measures of volume of lesions, rather than as categories, increased the power to detect associations with cognition. Finally, we used voxel-wise analyses to examine brain regions in which presence of WMH was associated with lower cognitive function.