In this study, we demonstrate that poor executive function and episodic memory are associated with specific locations of WMH in the cerebral white matter, independent of the total volume of WMH. These observations support the hypothesis that WMH causes vascular cognitive impairment by disrupting cortical connections mediated by specific white matter tracts. We find that WMH are associated with memory impairment as well as executive dysfunction, which has also been observed in other studies.2
Most previous studies have correlated cognitive performance with global WMH volume. Relatively few studies have examined relationships between the topography of WMH frequency and the resulting cognitive deficits, and these clinical-anatomic correlation studies have generally divided the white matter into large regions of interest, such as entire lobes or periventricular vs subcortical regions.2,6,26–30
This approach does not take into account the fact that such large areas will contain multiple white matter tracts coursing to many destinations, however. Despite this limitation, some previous studies have demonstrated clinical-anatomic correlations between WMH and cognitive function. In general, periventricular WMH are more closely associated with cognitive impairment or cognitive decline, possibly suggesting a role for dysfunction of long association tracts.2,26–28
One study of stroke patients suggests that frontal WMH is associated with executive dysfunction and temporal WMH is associated with memory impairment.29
Another study found that left dorsolateral prefrontal cortex WMH is associated with decreased performance on a working memory task.6
In a study using a visual rating scale to grade severity of WMH along a white matter pathway of interest, lesions in cholinergic pathways were shown to be associated with memory impairment.30
An advantage of the present study is the use of smaller regions of interest, namely the individual MRI voxels, to perform regional clinical-anatomic correlations with higher spatial resolution. Using this technique we were able to show strong regional correlations between WMH frequency and dysfunction in specific cognitive domains. We also controlled for total WMH volume when determining regions of correlation, allowing us to determine that these regional correlations are specific to that anatomic area and are independent of total lesion volume. Because the frequency of WMH at any location is partly correlated with total lesion volume, relationships between regional WMH frequency and cognition are confounded by total lesion volume, and may not be valid unless this confounding is considered. By covarying for total lesion volume in our analyses, we can conclude with reasonable certainty that the observed relationships between regional WMH frequency and cognitive functions are equally true in those with low total WMH volume and high total WMH volume.
In this study, we were unable to directly identify the white matter tracts traveling though the regions where associations were seen between WMH frequency and cognition. We did not perform diffusion tensor imaging, which can be used to map white matter tracts. This notwithstanding, we believe it is reasonable to speculate on which tracts might be involved based on tract-tracing studies in monkeys31
and diffusion tensor imaging studies in monkeys32
that provide insights into the anatomic locations of cerebral white matter tracts. The labeled WMH regions associated with impaired executive function in may involve the following fiber tracts: the uncinate fasciculus, linking rostral temporal with orbital and medial prefrontal cortices (, label A); the inferior longitudinal fasciculus, linking rostral and caudal regions of the ventral visual steam (, labels B and C); fronto-occipital fibers (, label C); superior longitudinal fasciculus that links parietal and temporal regions with the prefrontal cortex and that is thought to play a role in spatial processing and spatial attention (, label C); cingulum bundle, relevant for motivation and behavioral control (, label D); anterior limb of the internal capsule and, on the left, genu of the internal capsule and striatal fibers coursing in the bundle of Muratoff to the caudate nucleus (, label E); and multiple types of fibers entering and leaving the cortex (, label F). The involvement of the anterior limb of the internal capsule is particularly interesting because it conveys information between prefrontal cortex and the medial dorsal and anterior thalamic nuclei, and the prefrontopontine fibers destined for the cerebellar posterior lobe—subcortical areas now known to be relevant for executive function.36,37
The labeled WMH regions associated with impaired episodic memory () may involve the following fiber tracts: fronto-occipital fibers and the inferior longitudinal fasciculus (, labels A and D); rostral and caudal limbs of the arc of the cingulum bundle (, labels B, C, and F); anterior limb and genu of the internal capsule (, label E); and bundle of Muratoff (, label G). The cingulum bundle is implicated in the motivational and emotional aspects of behavior, and it is also thought to contribute to different aspects of mnemonic processing by virtue of its connections with the hippocampus, parahippocampal regions, and retrosplenial cortex, in addition to its connections with the parietal and frontal lobes.31
The finding that WMH in the left anterior limb of the internal capsule, but not the right, was associated with worse episodic memory may reflect the predominant weighting of verbal memory, as opposed to visuospatial memory, in the factor score.
A limitation is that our findings were derived from a population that mostly consisted of subjects with mildly impaired cognition and may not be valid in other populations with different distributions of cognitive impairment. The distribution of cognitive impairment of our subjects reflects the original design of the study, with intentional enrichment of the study population for subjects with MCI.9
A sensitivity analysis showed that the cluster locations were essentially unchanged when the analyzed population was restricted to those with MCI, suggesting that the study findings are robust for this group. Further studies with larger sample sizes will be needed to confirm whether our findings are relevant to patients with overall normal cognition or dementia. Another limitation is that our analysis of the neuropsychological data did not permit investigation of components of memory such as encoding or retrieval, which may depend on different anatomic pathways; future studies will be needed to address this.
Our findings suggest that WMH location should be considered in subsequent studies that attempt to determine the relevance of WMH to cognitive function. Such studies should attempt to confirm the clinical relevance of WMH in the regions identified in our work. Future studies could also incorporate quantitative imaging markers of tissue microstructure such as measures of water proton diffusion, which seems to offer additional clinically relevant information beyond that conferred by the presence or absence of T2 hyperintensity.38
White matter diffusivity reflects microstructural changes that can be caused by processes other than WMH, and has been observed in AD (presumably as a consequence of neurodegeneration with secondary alteration of white matter integrity)39
as well as in normal aging.40
The optimal way to integrate and interpret WMH and diffusion information is a critical subject for future studies. Ultimately, a better understanding of WMH location and its significance could inform the individual clinical evaluation of patients with cognitive impairment, by allowing the discrimination of clinically relevant patterns of WMH from relatively “benign” patterns of WMH sometimes seen in cognitively normal individuals.