The major findings of this study are: 1. WML predicted frontal cortical volume, particularly the medial orbitofrontal cortical volume among all groups after controlling TIV, age, gender, years of education, and the presence of lacune; 2. WML predicted depressive symptom severity directly and this relationship was not mediated by regional frontal cortical volume; 3. The relationship between WML and cognitive function was indirect and mediated by regional frontal cortex. These findings confirm the previous result from this collaborative project reporting that WML are related to depression and gray matter volume is related to cognition . Our study has important implications for understanding the potential mechanisms underlying these dissociations. Each of these findings will be discussed below.
The first major finding of this study is that all lobar WML are associated with frontal cortical volume irrespective of groups. This association is strongest with medial orbitofrontal cortex. Our findings are consistent with one previous functional MRI study showed that WML were associated with decreased frontal activation  in healthy elders and one previous FDG PET study showed that WML predicted decreased frontal glucose metabolism in all cognitive groups , however, these studies did not explore specific associations with orbitofrontal cortex.
WML can be related to the frontal cortical atrophy because WML and frontal cortical atrophy both are related to aging, ischemic damage, and cortical degenerative changes. In frontal lobes, lateral and orbitofrontal regions in frontal cortex have been reported to be prominently affected by cerebral aging among elders [27, 28]. Thus, the selective volume loss of medial orbitofrontal cortex associated with WML could be explained by cerebral aging. However, we found that WML volume was related to regional frontal cortical volume after controlling for age, and we thus conclude that this relationship is not explained solely by normal aging. It is not clear that the ischemic damage to the axons of subcortical white matter causes retrograde degeneration of neuronal bodies in frontal cortex or degenerative neuronal cell death of frontal cortex causes anterograde Wallerian degeneration of the axons of subcortical white matter. Longitudinal study will be needed to find the causal relationship between subcortical WML and frontal cortical atrophy.
The second major finding of this study is that WML, especially frontal and temporal WML, are directly related to depressive symptom severity and frontal cortex do not mediate this association.
Previous studies have also reported that geriatric depression is associated with WML [6–8]. Depressive symptoms in elders consist of diverse cognitive, emotive, and physiological symptoms. Therefore, geriatric depression can be caused by abnormalities of brain connections rather than a single brain area abnormality and WML association with decreased anisotropy and reduced myelin integrity of white matter tracts in frontal lobes  indicates the possibility of disruption of white matter tracked by WML.
In this study, depressive symptom severity is not associated with regional frontal cortices. However, some other studies have reported that geriatric depressive disorder is associated with orbitofrontal cortical atrophy in elders [30, 31], but these studies did not consider WML effect on depression. Our findings may be in part due to our sample criteria which exclude severe depression because prefrontal cortical atrophy was reported to be associated with severe depression and not to mild depression .
The third major finding of this study is that WML are related to impaired cognitive function but the relationship between WML and impaired cognitive function was indirect and mediated by volume of frontal cortex. Orbitofrontal cortex is commonly related to memory encoding  and to apathy which can influence the executive performance negatively . Therefore, orbitofrontal cortical atrophy can cause cognitive impairment including memory and executive dysfunction.
Previous studies [9, 10, 35] which reported the relationship of WML to cognitive function usually did not consider the role of cortical atrophy. In several studies there was no direct association between WML and cognitive function once global cortical atrophy was accounted for [3, 11, 36, 37]. These findings add to the view that cognitive dysfunction related to WML may be result of the frontal cortical atrophy secondary to WML rather than the direct result of WML.
Several features of this study differed from previous reports. First, we measured WML volume and regional cortical volumes instead of ordinal scoring system. There were numerous visual WML rating scales and these subjective scales displayed ceiling effects and showed poor sensitivities . Therefore, volumetric measurement of WML and other cortical areas may yield more accurate results. Second, we used SEM to explore the mediation effect of frontal cortex. SEM has several advantages. We can have multiple indicators for each construct  and multiple indicators do not cause collinearity problems differently to linear regression model .
We have several limitations. These results are from a cross-sectional data and therefore causality between WML and atrophy cannot be confirmed. We do not divide WML into periventricular WML and deep WML, so we cannot analyze the effect of WML sites on cognition and depression. We excluded severely depressed patients who had recurrent suicidal ideas or severe impairments in social function. Therefore, we cannot rule out the possibility of the association between frontal lobes and severe depression. Due to relative small sample size, we cannot control vascular risk factors and drug effects.