Subcortical brain systems play an important role in the regulation of a variety of cognitive and emotional processes.¹ The basal ganglia and thalamus, along with the surrounding limbic structures, comprise MacLean's paleomammalian brain,² a region that serves to integrate appetitive drives with approach or avoidance to environmental stimuli. In humans, frontal—subcortical circuits comprise one of the major organizing neural networks in the brain. Discrete pathways connect prefrontal neurons sequentially to caudate, globus pallidus, dorsomedial thalamus, and finally back to the original prefrontal locus.³ These loop circuits process information supporting executive cognitive functions such as organization, motivation, and social comportment.⁴ This region of the brain receives its blood supply from relatively long, small-diameter, penetrating branches of the anterior and middle cerebral arteries. This unique circulatory arrangement is a consequence of the rapid evolutionary expansion of the neocortex and leaves these deep white and gray matter tissues vulnerable to ischemic injury and resulting neuropsychiatric dysfunction.

Changes in subcortical white matter and deep gray matter nuclei are often noted in the brains of elderly people. Typically, these changes appear as foci of increased signal (hyperintensity) on T₂-weighted magnetic resonance (MR) imaging or as areas of low density (attenuation) on computed tomography (CT) (F1, F1, F1). Such changes were previously interpreted to reflect degeneration of the white matter typical of Binswanger's disease. More recently, similar changes have been reported in patients with Alzheimer's disease, in nondemented patients with cerebrovascular disease or psychiatric illness, and in apparently asymptomatic individuals. As a result, important questions are being raised about the frequency, pathophysiology, and potential clinical significance of these subcortical changes.

Given the frequent use of brain MR imaging in neuropsychiatry, clinicians and neuroscientists are likely to encounter patients with changes in the subcortical white and gray matter. The purpose of this article is to review relevant data on the occurrence, etiology, and correlates (both clinical and neurobiological) of these subcortical changes. The potential implications of these changes for the pathophysiology of neuropsychiatric disorders will be discussed, with a particular focus on the relationship between subcortical brain changes and aging, cognitive function, and mood in older adults.

Pathologic changes in the subcortical white matter were noted as early as 1854 by Durand-Fardel, who described "L'atrophie interstitielle du cerveau" (brain interstitial atrophy) on macroscopic examination of the brain.⁵ Binswanger was the first to suggest that these white matter changes might be associated with clinical symptoms. In 1894, he reported 8 patients with a form of dementia characterized pathologically by subcortical white matter lesions and ventricular enlargement, but with strikingly normal-appearing cerebral cortex.⁶ Although no histology was described, he suggested that the white matter atrophy resulted from reduced perfusion caused by arteriosclerosis. Alzheimer, in 1902,⁷ provided histological confirmation of the cortical sparing and relatively focal white matter involvement and named the condition after his colleague, Binswanger. He also noted severe atherosclerosis of the medullary arteries as well as the presence of lacunar infarctions in the white matter, internal capsule, basal ganglia, thalamus, and pons. In a major review, Olszewski, in 1965,⁸ translated the publications of Binswanger and Alzheimer, described two additional neuropathologic cases of Binswanger's disease, and introduced the term subcortical atherosclerotic encephalopathy. Since then numerous reviews and case studies have provided helpful clinical and neuropathological data on patients with Binswanger's disease.^9—11

More recently, emphasis has shifted toward an antemortem recognition of subcortical white and gray matter changes on brain MR imaging or CT. The observation that these changes may occur in clinically nondemented individuals has resulted in considerable confusion in terminology. Although the term Binswanger's has frequently been used as if it were synonymous with these MR imaging or CT changes, this usage is technically incorrect, and the diagnosis of Binswanger's disease should be reserved for those cases in which it is justified by clinical features such as a slowly developing dementia appearing typically in middle age.^12,13

Newer terms have been introduced that refer to these imaging findings, including white matter lucency, incidental MR lesions, periventricular hyperintensity, patchy white matter lesions, leukoencephalopathy, and unidentified bright objects.^14—17 Hachinski et al.¹⁸ proposed the term leukoaraiosis to refer to rarefaction ("araiosis") or diminution of density of the white ("leuko") matter. These terms can be criticized because they imply that the imaging changes are limited to the subcortical white matter, despite the common occurrence of deep gray matter and brainstem involvement.^15,18—20 We prefer Awad and colleagues' term subcortical encephalomalacia ("malacia" meaning morbid softening or thinning), which emphasizes the anatomic distribution of the imaging changes without restricting them to white matter.¹⁹ As will be seen below, changes in subcortical white and gray matter may be produced by a variety of different pathologic processes that may be indistinguishable by radiologic findings alone.

On CT, the white matter changes of subcortical encephalomalacia appear as areas of hypodensity that involve the periventricular and deep white matter of the centrum semiovale, as well as the deep gray matter nuclei (F1, F1, F1). The changes are frequently symmetric, may extend for varying distances into the deep white matter, and do not enhance with contrast.

Brain MR imaging is more sensitive than CT for detecting periventricular and deep white matter changes (F1, F1, F1). The changes usually appear as areas of increased signal (hyperintensity) on both proton density— and T₂-weighted images; on T₁-weighted images these lesions may appear hypointense; some authors take this finding as evidence of frank infarction. Fluid-attenuated inversion-recovery (FLAIR) imaging sequences are especially sensitive to these white matter changes and allow for heavily T₂-weighted images without the obscuring of adjacent brain by the high CSF signal. The white matter changes on MR imaging may appear in a variety of patterns. Independent punctate or multipunctate/diffuse foci are common. Additionally, a thin, continuous or discontinuous hyperintense lining of periventricular white matter known as a "rim" is often seen. Rounded hyperintense foci at the angles of the frontal or occipital horns are referred to as "caps." Finally, a "halo" of periventricular hyperintensity of a variable thickness and patchy distribution can extend irregularly into deep white matter. The arcuate U fibers are typically spared.

Punctate foci of hypodensity on CT or hyperintensity on proton density— and T₂-weighted MR imaging images may also be seen in deep gray matter nuclei such as the basal ganglia and thalamus. The pons may also be involved.²⁰ MR imaging is more sensitive than CT for detecting these changes. Additional neuroanatomic findings may include ventricular dilatation and cortical atrophy.^15,17,21

Subcortical encephalomalacia on brain imaging should be distinguished from similar-appearing changes produced by other disease processes. The differential diagnosis includes white matter edema from tumor, stroke, or trauma; demyelination from diseases such as multiple sclerosis or following infection, radiation, anoxia, or toxins; and the late-onset leukodystrophies.²² Typically, these conditions show diffuse involvement of the white matter, including the arcuate fibers, and the foci of increased T₂ signal enhance with contrast. A variety of dementing conditions may be associated with subcortical changes, including Pick's disease, progressive subcortical gliosis, presenile glial dystrophy, vascular dementia, and normal-pressure hydrocephalus (NPH). NPH can mimic Binswanger's disease since it includes ventriculomegaly and the clinical triad of dementia, gait apraxia, and incontinence. Isotope cisternography, dynamic MR imaging, and lumbar puncture may assist in the diagnosis of NPH. Similar subcortical white and gray matter changes are occasionally noted in patients with probable Alzheimer's disease, where recent speculation has considered cerebral amyloid angiopathy as a possible cause of the subcortical lesions in some patients.²³ CADASIL (cerebral autosomal dominant arteriosclerotic encephalopathy and leukomalacia) is a familial disorder characterized by diffuse cerebral amyloid angiopathy and a progressive dementia. MR imaging studies reveal diffuse subcortical hyperintensity on T₂-weighted images.²⁴

For most elderly subjects, these disease processes are readily excluded and the subcortical encephalomalacia appears as an incidental or unexpected finding on MR or CT imaging. In these cases, the precise etiology of the subcortical changes remains unclear because of limited cliniconeuropathologic data. Periventricular hyperintensities on proton density— or T₂-weighted MR imaging in the form of caps or rims are quite common in normal samples and do not appear to constitute a disease state. Histological studies suggest that the caps and rims (F1, F1, F1) reflect increased water content from a variety of structural conditions, including loosely arranged fine-fiber tracts with low myelin, a patchy loss of ependyma with astrocytic gliosis ("ependymitis granularis"), and the normal convergence of flow of interstitial fluid within the periventricular region.^24,25 Periventricular hyperintensity in the form of a smooth halo (F1, F1, F1) has been associated histologically with disruption of the ependymal lining, with subependymal gliosis and concomitant loss of myelin.²⁵

The histopathologic correlates of punctate and multipunctate/diffuse subcortical hyperintensity (SH) in the deep white matter are varied, but most evidence points to changes in perivascular tissue.²⁵ Punctate SH usually does not represent infarction, but rather tissue changes induced by perivascular edema, impaired diffusion of nutrients through thickened vessel walls, état criblé (a state of dilated perivascular—Virchow-Robin—spaces together with sclerosis and ectasia of the medullary arteries and arterioles), or mechanical damage to the adjacent tissue induced by a water-hammer effect of pulsating arterioles. The multipunctate/diffuse SH lesions are linked to more severe tissue changes, including gliosis, edema, reduced myelination, and axonal destruction and infarction.²⁶ The relative proportions of these various tissue changes are not readily separable with conventional MR imaging techniques. Furthermore, although the tissue changes associated with the various types of SH changes represent a continuum of severity, it has not been established whether these tissue changes truly evolve through such stages of severity.

To summarize, the imaging changes of SH (excluding mild periventricular changes such as caps and rims) appear to reflect a combination of edema, vascular ectasia, dilated perivascular spaces, demyelination, and occasionally lacunar infarctions. The pathogenesis of these structural changes is likely related to the unique anatomy of the vascular supply to the subcortical white and gray matter regions, which renders these areas vulnerable to ischemia.^9,10,19,27 As noted above, these areas are perfused by long, penetrating medullary arteries that are relatively small in diameter, originate at right angles from the parent vessel, and have few or no collaterals (F2). These anatomic features predispose the medullary vessels to arteriosclerotic damage, which results in decreased perfusion of the end zone or watershed areas that they supply.²⁵ Clinicopathologic changes might then evolve that are indistinguishable by MR imaging alone, along a spectrum from asymptomatic état criblé in the earliest stages, to edema and white matter tract degeneration, to the accumulation of lacunar and nonlacunar infarctions and clinically symptomatic cerebrovascular disease such as Binswanger's disease and vascular dementia. Contributing to this process would be a variety of other conditions that might affect perfusion of the subcortical regions, including hypertensive arteriosclerosis, chronic recurrent hypotension, cerebral amyloid angiopathy, senile arteriolar hyalinosis, age-related meningeal thickening, and diminished autoregulation of cerebral circulation associated with aging.

Thus, as was conceptualized by Awad et al.,¹⁹ SH may actually reflect "wear and tear" of brain parenchyma that accompanies aging and chronic cerebrovascular disease. It should be noted, however, that progressive vasculopathy per se does not account for all visible subcortical MR imaging lesions. A wide variety of other lesions can mimic the MR imaging appearance of SH, including telangiectasias, isolated white matter infarcts, diverticula of the trigone of the lateral ventricle, and even asymptomatic zones of multiple sclerosis.²⁸ Furthermore, in a few cases of severe vascular disease in the deep subcortical regions, postmortem MR imaging has been normal.^19,28 Clearly, additional neuropathologic studies are needed to resolve these issues.

Subcortical Hyperintensity in Aging

Numerous MR imaging studies have demonstrated that aging is associated with an increased occurrence and severity of SH.^29—33 The precise frequency and distribution of SH in usual or normal aging are difficult to determine, however, given differences between studies in subject sample (age range, health status, etc.), imaging technology, and lesion definition and measurement. In our study of healthy adults with no lifetime history of neurologic or psychiatric illness, SH was present in the deep white matter in 48 subjects (64.0%), in the periventricular white matter in 9 (12.0%), in the basal ganglia in 9 (12.0%), in the thalamus in 4 (5.3%), and in the pons in 16 (21.3%).³⁴ We found that the age-specific odds of subcortical hyperintensity increased by 5% to 9% per year of age, depending on the anatomical region involved (F3). The literature consistently demonstrates that patients with risk factors for vascular disease (e.g., smoking, diabetes mellitus, hypertension) and those with a high medical burden have a greater prevalence and severity of SH. Indeed, evidence on MR imaging of old cerebral "microbleeds" is reported in approximately 6% of neuropsychiatrically normal community volunteers and is associated with more extensive SH.³⁵ Recent data in twins suggest that genetic factors may also contribute to individual differences in volume of subcortical hyperintensity.³⁶

Little is known about the natural history of SH in usual aging. Lesion progression has been documented in longitudinal imaging studies with follow-up of from 2 to 5 years.^37—39 In about half of the subjects the worsening was minor, whereas in the remainder more extensive changes were observed.³⁷ Reported risk factors for progression are diastolic hypertension and lesion severity.^37,38

An emerging literature suggests that SH may have functional and clinical significance in normal elders. Quantitative EEG studies in normal aging show a relationship between SH and EEG abnormalities, most notably decreased coherence (a measure of shared electrical activity between brain regions) in the frontal and visual networks of connections.⁴⁰ In contrast, studies of evoked related potentials found no relationship between SH and P300 latency in normal elderly subjects.⁴¹ Metabolic brain imaging studies have reported reductions in cerebral blood flow and metabolism among healthy elderly volunteers with relatively large volumes of SH.^42—44 These metabolic changes may be especially prominent in the frontal lobes,⁴² where the SH lesions predominate. MR imaging has recently been used to examine the functional correlates of SH.⁴⁵ On MR spectroscopy, SH has been associated with a decreased ratio of adenosine triphosphate/inorganic phosphate, but not with any changes in phosphocreatine, phosphomonoester, or phosphodiester,⁴⁶ and with an increased ratio of choline/N-acetyl aspartate (NAA), but not with any changes in creatine/NAA or lactate/NAA.⁴⁷

Most of the research on the clinical significance of SH has focused on vascular events, equilibrium, and cognition. Given its presumed vascular etiology, it is not surprising that in patients with clinical vascular disease, SH is predictive of vascular events,⁴⁸ ischemic or hemorrhagic stroke,^49,50 and vascular death.⁴⁸ In patients with transient ischemic attack or minor stroke within the past 6 months, SH also predicts risk of hemorrhagic complications from oral anticoagulation with phenprocoumon.⁵⁰ Less is known about the impact of SH upon risk of overt cerebrovascular disease in normal elders.

Fibers essential for proper gait and balance course through the periventricular region. Several studies have found relationships between the extent of periventricular SH and impairment in equilibrium among elderly patients with complaints of disturbed balance^51—53 as well as in asymptomatic elderly volunteers.^54—56

Although a growing literature exists on the relation between cognition and changes in subcortical white and gray matter in normal aging (t1), these results are conflicting, and the studies are difficult to compare given methodological differences. Most studies find no relationship between SH and scores on dementia rating scales,^57—62 but such measures may be relatively insensitive to subtle changes in cognition. Positive findings have generally been more common among studies using more sensitive neuropsychological assessment batteries, with measures of frontal lobe function demonstrating the closest relationship with SH.^{36,37,42,57,63—67,69} Most^57,70 but not all³⁹ of the small number of longitudinal studies find associations between SH and cognitive decline, but the correlations are generally small and thus account for a relatively small proportion of the variance in test scores (t1).

In an early study from our laboratory,⁷¹ we examined relationships between SH and two specific neuropsychological instruments: the Benton Facial Recognition Test⁷² and the WAIS-R Digit Symbol. The former was chosen because it had yielded the highest level of significance of any test reported to be associated with SH; the latter, because it had been reported to be related to SH by two independent groups (t1). In addition, both the Benton Facial Recognition Test and the Digit Symbol subtest of the Wechsler Adult Intelligence Scale—Revised (WAIS-R)⁷³ were favored a priori because subcortical pathology might be expected to disrupt visuospatial perception and visuomotor execution, respectively. We found that performance on both tests was highly related to age and education, but not to the presence of SH. The majority of our subjects had relatively mild findings of SH, however, and it thus remains possible that more severe changes that reach a certain "threshold" volume might be associated with cognitive dysfunction in otherwise healthy adults.^{42,61,62,64,65} Indeed, extensive damage to subcortical white matter tracks would disrupt frontal-subcortical circuitry and possibly provide an anatomical substrate for the mental slowing and disturbed executive functioning seen with aging. Lesion location may also be important, as evidenced by the work of DeGroot et al.,⁶⁵ who found relationships between cognitive dysfunction and periventricular SH but not deep white matter SH, when each was analyzed conditional upon the other.

In summary, the occurrence of SH increases with age, and its severity increases in the presence of risk factors for vascular disease or high medical burden. Pathologic correlates of SH vary but appear to reflect the effects of brain hypoperfusion. The presence and severity of SH are associated with changes in brain electrophysiology, blood flow, and metabolism, as well as with dysequilibrium and alterations in cognitive (especially frontal lobe) functioning in normal elderly subjects.

Subcortical Hyperintensity in Patients With Dementia

The precise frequency, severity, and distribution of SH in patients with dementia is difficult to determine because of methodological differences between studies in their subject samples (e.g., inclusion of patients with underlying vascular disease), imaging technology, and definitions of SH (t2). In general, the literature suggests that patients with vascular dementia have a greater frequency and severity of SH (both periventricular and deep white matter) than do patients with Alzheimer's disease or control subjects,^74—83 consistent with the presumed vascular etiology of SH (t2). Thus, SH has been identified in 33% to 97% of patients with vascular dementia^75—78 compared with 5% to 56% of patients with Alzheimer's disease.^{75,76,79—83}

The data are less clear in studies of Alzheimer's disease, again in part because reports differ in the inclusion of patients with risk factors for vascular disease and in the definition of lesions^{58,60,84—97} (t2). Most studies examining SH in the deep white matter do not find a greater prevalence of SH in patients in patients with Alzheimer's disease than control subjects (t2). Conflicting findings have been reported for periventricular SH, perhaps due in part to the methodological differences described above (t2).

In the one published study of dementia with Lewy bodies, Barber et al.⁹⁴ found a higher prevalence of SH (both periventricular and deep white matter) in patients than control subjects.

The pathophysiology of SH in patients with Alzheimer's disease is not known but is presumed to be similar to that in control samples, namely a subcortical microvasculopathy.^15,98,99 No relationship has been found between apolipoprotein E E4 allele and the occurrence of SH in patients with probable Alzheimer's disease, vascular dementia, or dementia with Lewy bodies.^100,101

Few studies have addressed brain anatomic correlates of SH in patients with Alzheimer's disease.^90,91 Periventricular SH appears to be associated with ventricular dilation (t2). Fazekas et al.⁸⁵ have proposed that ventricular enlargement caused by the loss of periventricular fibers in Alzheimer's disease may disrupt the integrity of the ventricular lining. This disruption may result in increased transependymal flow of CSF into the periventricular white matter with resulting imaging changes of SH.

Data conflict regarding the relationship between SH and cognitive function in patients with Alzheimer's disease (t2). Studies using global screening instruments such as the Mini-Mental State Examination (MMSE) or Cognitive Capacity Screening Examination generally show no effect of SH on global cognitive functioning.^{87,88,90,93,102—104} One exception is Bondareff et al.,¹⁰¹ who noted a correlation between a volumetric measurement of SH (the "lesion:brain ratio") and lower MMSE scores. Positive findings are more common among studies that employ a more extensive neuropsychological test battery,^{103,105—107} although negative findings have also been reported.¹⁰⁸ Cognitive domains most commonly affected include attention,¹⁰⁷ tactile identification of objects,¹⁰³ and frontal lobe function (t2), although the amount of variance in cognition accounted for by the SH is typically quite small.¹⁰⁵

The various functional and clinical consequences of SH in dementia have been examined in several studies. Deep white matter/basal ganglia SH has been associated with reductions in global cerebral perfusion;¹⁰⁹ frontal perfusion;¹¹⁰ bilateral perfusion in frontal lobe, basal ganglia, and thalamus;¹⁰⁸ and reduced hippocampal perfusion⁹⁰ in studies using single-photon emission computed tomography (SPECT). Other SPECT studies, however, show no correlation between SH and cerebral perfusion in patients with Alzheimer's disease.¹¹¹ Yamaji et al.,¹¹² using positron emission tomography (PET), noted significantly lower global cerebral blood flow in their cohort of patients with Alzheimer's disease and SH versus patients without SH. It is of interest that oxygen extraction was increased in involved areas, suggesting adequate compensation for the reduction in blood flow. This finding may account for the number of studies that show little or no cognitive impairment in the setting of SH.

Other reported clinical correlates of SH in patients with Alzheimer's disease include a higher risk of subsequent stroke,¹¹³ higher apathy scores, motor incoordination and gait disturbance,¹⁰⁸ urinary incontinence, and purposeless activity such as wandering.¹¹⁴

Barber et al.⁹⁴ reported that frontal white matter hyperintensities were associated with depressive symptomatology in patients with Alzheimer's disease, dementia with Lewy bodies, or vascular dementia. They also observed that occipital white matter hyperintensities were associated with the absence of visual hallucinations.

In a one-year longitudinal study, Lopez et al.¹¹⁵ found no difference in cognitive change between patients with Alzheimer's disease and periventricular SH versus those patients without SH. In addition, the frequency of major depression, delusions, and hallucinations did not differ between the two groups. However, the presence of periventricular SH did predict the development of clinically significant cerebrovascular disease.

In summary, both periventricular and deep white matter/basal ganglia hyperintensities appear to be more common and generally more severe in patients with vascular dementia than in those with probable Alzheimer's disease or Lewy body dementia, and all three groups may have more extensive changes than healthy age-matched control subjects. In patients with dementia, more extensive or severe SH may be associated with brain structural (especially ventricular enlargement) and metabolic changes; with impairment in cognition, gait, and motor coordination; with depressive symptoms (especially for frontal SH); and with subsequent risk of stroke.

Subcortical Hyperintensity in Patients With Mood Disorders

It has been known for some time that patients with Binswanger's disease may present not only with dementia but also with a variety of other neuropsychiatric symptoms, including delirium, behavioral disturbances, mania, psychosis, social withdrawal, abulia, and depression.^{9,11,14,15,17}

Early clinical reports suggested an increased rate of SH in elderly patients with severe depression referred for ECT^14,15,17,116 or in patients with bipolar illness.¹¹⁷ Subsequently a large number of controlled investigations demonstrated that SH is indeed more common or severe in patients with major depression or bipolar disorder (t3^118—147). The precise frequency of SH in these populations is difficult to determine, however, given differences between studies in subject samples (e.g., in age, sex, and health status including vascular risk factors or disease), in imaging technology, and in lesion definition and measurement (t3). A recent meta-analytic study¹⁴⁸ of the extant literature reported a common odds ratio (Mantel-Haenszel technique) of 3.2 (95% confidence intervals of 2.11 and 4.82) for SH in patients with major depression, and 3.29 (95% CI of 2.14 and 5.07) in patients with bipolar disorder. The SH lesions in patients with mood disorders are typically localized to the frontal white matter and the subcortical gray nuclei (thalamus, basal ganglia; t3). As noted above for nonpatient samples, SH in patients with mood disorders is associated with age and overall medical burden, including risk factors for vascular disease (t3). As such, SH in patients with mood disorders is also associated with other age- or illness-related changes in brain structure, including global or regional atrophy and ventricular enlargement (t3). Finally, SH and microinfarction are associated with depression in patients with vascular dementia.¹⁴⁹

The pathophysiology of SH in patients with mood disorders is not known but is presumed to be similar to that in control samples, that is to say a subcortical microvasculopathy.^15,98,99 Little is known about the natural history of SH in patients with mood disorders. One study of 7 relatively young adults with bipolar disorder found no changes in SH ratings at 1-year follow-up.¹³⁴ In a prospective study of the brain anatomic effects of ECT, we observed an increase in the severity of SH 6 months after a course of ECT in 5 of 35 patients with depression.¹⁵⁰ Such changes were interpreted to reflect progression of ongoing cerebrovascular disease rather than an effect of ECT.

Although studies in control samples have demonstrated functional brain correlates of SH by using quantitative EEG, PET, functional MR imaging, and MR spectroscopy as noted above, very few such data have been reported for patients with mood disorders.^151,152 Ebmeier et al.¹⁵³ found a correlation between periventricular SH and widespread periventricular and bilateral temporal perfusion defects on PET among 39 elderly patients with major depression. No relationships were observed for SH in the deep white matter. We observed a relationship between SH in the pons and shortened REM sleep latency in patients with major depression, a finding of interest given that REM sleep is felt to be mediated by systems located in the pons.¹⁵⁴

A growing literature has focused on the clinical significance of SH in patients with mood disorders. Onset of major depression at a late age (>60 years old) has been associated with SH in most^{15,17,123,155—160} but not all^{119,120,125,126} studies of patients with unipolar disease, as described below. Most studies in patients with bipolar disorder find no relationship between SH and age at onset.^134,144,145 Most reports find no relationship between SH and psychotic symptoms^{128,134,138,145} or severity of depression.^{17,58,129,134,138,156,161} Apathy, anhedonia, and functional disability have been suggested as possible correlates of SH in elderly patients with major depression.¹⁶² No relationship between SH and a family history of mood disorder has been found for patients with major depression,¹⁵⁵ but conflicting findings have been reported for patients with bipolar disorder.^{134,140,145,147} Although SH is associated with dysequilibrium in elderly control samples as mentioned above, no study has examined such relationships in elderly patients with mood disorders.

Several studies have found a relationship between SH and cognitive impairment in patients with major depression.^{17,58,155,156,161,163} The few studies that have reported negative findings typically used screening measures rather than more sensitive neuropsychological test batteries^119,129 (t3). Consistent with the effects seen in control samples as described above, the cognitive disturbance is typically manifest as cognitive slowing, memory deficits, and frontal lobe/executive dysfunction (t3). Conflicting findings have been reported for bipolar disorder.^134,147

Little is known about the impact of SH on treatment outcomes in patients with mood disorders. In a small (N=39) study of older inpatients with severe (mainly psychotic) depression, Hickie et al.¹⁵⁵ found that SH was associated with a poorer response to either pharmacotherapy or ECT. While our experience is consistent with the former finding, we observe a good response to ECT even in depressed patients with extensive SH.^14,15,17,150 We and others have found, however, that the presence or severity of SH may be predictive of treatment side effects, including delirium from ECT¹¹⁶ or antidepressant medications¹⁶⁴ as well as neuroleptic-induced parkinsonism.¹⁶⁵ Similar observations were reported by Fujikawa et al.¹⁶⁶ in a study of patients with major depression and silent cerebral infarction on brain MR imaging. They found that those patients with silent cerebral infarction (SH or cortical infarct) had a greater frequency of "adverse central nervous system reactions" to antidepressant drugs; these reactions included delirium, parkinsonism, dyskinesia, and akathisia.

With regard to the long-term prognosis of patients with SH and mood disorders, one group found a correlation between SH and number of prior hospitalizations but not duration of current bipolar illness.^134,140 In a longitudinal investigation of mainly older patients with depression, Hickie et al.¹⁶⁷ found that cognitive decline and functional impairment were associated with the severity of SH on MR imaging when SH were noted 6 months to 2 years previously.

The association of SH and mood disorders has led to interesting speculation about the direction of causation. Clearly, the occurrence of SH will be greater in patients with overt vascular disease, but as noted in t3, the relationship between SH and mood disorders remains significant in some studies even after adjustment for medical burden and traditional vascular risk factors. The association between SH and late-age-onset mood disorders has also led to the suggestion that SH or the vascular disease with which it is associated may be causative of the depression, at least in some patients.^162,168,169 Consistent with this view is the finding of a lack of family history of mood disorder in patients with SH and depression.¹⁵⁵ Indeed, depression is a common symptom of subcortical brain disease (e.g., Parkinson's disease, Huntington's disease, multiple sclerosis) and the frontal predominance of the SH lesions in patients with mood disorders positions such lesions to disrupt (disconnect?) frontal-subcortical circuits that appear to be important in regulation of emotional behavior.^17,170,171 At a neurochemical level, one report suggests that SH in geriatric depression is associated with reduced platelet [³H]-imipramine binding, a potential marker for presynaptic serotonergic function.¹⁷² Another potential causative mechanism is suggested by the association of SH and cognitive impairment; such impairment might result in limited coping strategies with which to respond to stress, leading to depression.¹⁶⁰

There are other lines of evidence to suggest, however, that depression may be causative of cerebrovascular disease. Depression is associated with a variety of pathophysiologic alterations that may increase vulnerability to vascular disease,¹⁷³ including increased platelet activation.¹⁷⁴ Indeed, recent prospective data demonstrate an association between baseline depressive symptoms and subsequent stroke as well as stroke mortality,^175—177 even after adjustment for established stroke risk factors. Treatment of depression with medications that induce hypotension could also theoretically cause or worsen SH and cerebrovascular disease,¹¹⁸ but no such relationships have been found.^{134,135,138,140,145} There is no evidence that SH in patients with mood disorders is causally related to treatment with ECT.^{17,120,129,145,150}

Of course, as noted by Videbech,¹⁴⁸ association does not prove causation, and there are other possible mechanisms to explain the relationship of SH and mood disorders. For example, the relationship may be mediated through a nonrelated confounder (e.g., medical burden), or SH could even be a result of treatment of depression (although, as noted above, the data do not appear to support this latter contention). Clearly, our understanding of the relationships between SH and mood disorders requires longitudinal studies of the development of SH and of mood disturbance, using appropriately matched control groups and standardized techniques for lesion assessment.

In summary, the prevalence and severity of SH is increased in elderly patients with major depression, in patients with bipolar disorder, and possibly in non-elderly patients with major depression. The SH in these populations is presumed to reflect subcortical microvascular disease, but the precise etiology of the vasculopathy is unknown. The SH may be associated with late-age onset of the major depression, cognitive impairment, a relatively poorer response to treatment, and an increased risk of cerebral side effects from treatment. The association of SH and mood disorders has led to interesting speculation on the mechanisms of mood disorder in patients with cerebrovascular disease.

Subcortical Hyperintensity in Patients With Other Mental Disorders

Most brain MR imaging studies of schizophrenia emphasize changes in frontotemporal and ventricular structures.¹⁷⁸ More recently, SH has been described in some patients with late-onset schizophrenia^179,180 and late-onset atypical psychosis.¹⁸¹ Sachdev and Brodaty¹⁸⁰ reported that subjects with late-onset schizophrenia had greater periventricular SH than either age-matched subjects with onset of schizophrenia before age 50 or normal control subjects. In addition, the late-onset schizophrenia group had more discrete lesions in the thalamus than the other subjects. This study did not control for the presence of cerebrovascular risk factors. The authors postulate that cerebrovascular disease may promote the onset of psychosis in vulnerable individuals. These reports are consistent with our findings of an association between SH and late-age-onset psychiatric disorder, and they warrant further study.

SH is now recognized as a common finding on brain MR imaging and CT of elderly individuals. Current clinical and neuropathologic data suggest that SH likely reflects in part abnormal water content and a spectrum of structural changes (which are difficult to distinguish by MR imaging alone), resulting from the effects of hypoperfusion of subcortical brain regions. However, because other lesions may mimic the appearance of SH and because even severe vascular disease may appear normal on postmortem MR imaging, additional neuropathologic and in vitro MR imaging studies are needed to clarify the potential etiologies of these brain imaging changes.

SH has been identified in the normal elderly population, but the lesions generally are not very extensive or severe. More data are needed to determine the effects of SH on cognitive functioning in the normal population. SH appears to be more common and generally more severe in patients with dementia, especially vascular dementia, but the severity of the cognitive impairment has not been found to be closely correlated with that of the disease process. Although little is known about the psychiatric manifestations of SH, it appears to be surprisingly common in elderly depressed patients (especially those referred for ECT), where it may serve as a risk marker for the development of late-age-onset mood disorder. The precise pathophysiologic mechanisms linking SH to clinical disorders such as dementia and late-age-onset mood disorder are unknown. The clinical implications of SH for the diagnosis, clinical phenomenology, treatment, and long-term prognosis of patients with neuropsychiatric illness require further study.

Although brain imaging may be indicated in some patients with dementia or mood disorders, current data do not support such testing solely for the purpose of "ruling out" SH in these groups. It is reasonable for the clinician to assume that SH will be present in any patient with vascular disease or risk factors or with high medical disease burden. Likewise, if SH is revealed as part of the evaluation of a patient, it may be taken as evidence of brain microvascular disease, assuming other etiologies discussed above have been excluded. In such patients, interventions aimed at the vascular disease may be appropriate. Indeed, given the potential neuropsychiatric implications of brain microvascular disease, clinicians should become proficient at helping their patients take steps to reduce the risk of vascular disease (e.g., stop smoking, maintain a healthy lifestyle, control blood pressure). In such cases, the proverbial ounce of prevention may be worth a pound of cure.