The Vascular Depression Hypothesis: Mechanisms Linking Vascular Disease with Depression

Abstract

The ‘Vascular Depression’ hypothesis posits that cerebrovascular disease may predispose, precipitate, or perpetuate some geriatric depressive syndromes. This hypothesis stimulated much research that has improved our understanding of the complex relationships between late-life depression (LLD), vascular risk factors, and cognition. Succinctly, there are well-established relationships between late-life depression, vascular risk factors, and cerebral hyperintensities, the radiological hallmark of vascular depression. Cognitive dysfunction is common in late-life depression, particularly executive dysfunction, a finding predictive of poor antidepressant response. Over time, progression of hyperintensities and cognitive deficits predicts a poor course of depression and may reflect underlying worsening of vascular disease. This work laid the foundation for examining the mechanisms by which vascular disease influences brain circuits and influences the development and course of depression. We review data testing the vascular depression hypothesis with a focus on identifying potential underlying vascular mechanisms. We propose a disconnection hypothesis, wherein focal vascular damage and white matter lesion location is a crucial factor influencing neural connectivity that contributes to clinical symptomatology. We also propose inflammatory and hypoperfusion hypotheses, concepts that link underlying vascular processes with adverse effects on brain function that influence the development of depression. Testing such hypotheses will not only inform the relationship between vascular disease and depression but also provide guidance on the potential repurposing of pharmacological agents that may improve late-life depression outcomes.

There is substantial heterogeneity in the biological factors that influence the risk of developing depression across the lifespan. Genetic and epigenetic factors influence gene expression or protein function across a variety of molecular systems, resulting in a vulnerability to developing depression in context of psychosocial adversity.¹ This heterogeneity is even more apparent in depression occurring in older adults, or late-life depression (LLD), as aging-related changes across multiple organ systems also contribute to depression. Aging-related changes in the neurologic, immune, and endocrine systems have all been implicated in depression, but one of the best studied in LLD is the role of vascular disease.

The ‘Vascular Depression’ Hypothesis proposed that “cerebrovascular disease may predispose, precipitate, or perpetuate some geriatric depressive syndromes”.² The hypothesis was accompanied by two proposals on the core characteristics of and how to define a vascular depression subtype. First, Alexopoulos and colleagues proposed a working definition based on the presence of vascular risk factors. The clinical presentation of “vascular depression” was characterized by cognitive deficits, psychomotor retardation, lack of insight, and disability disproportional to the depression severity.^{2, 3} Investigators subsequently focused on the cognitive dysfunction occurring in LLD and its relationship to response to antidepressants.⁴ Second, Krishnan and colleagues proposed a MRI-based definition ⁵ requiring evidence of vascular changes on neuroimaging, referred to as MRI hyperintensities. Both definitions are important as subsequent work demonstrated the internal validity of a vascular depression diagnostic subtype characterized by MRI findings and executive dysfunction.⁶

White Matter Lesions, White Matter Hyperintensities, and MRI-Defined Vascular Depression

The hallmark of MRI-defined vascular depression is the presence of white matter lesions (WMLs) identified as white matter hyperintensities (WMH) on T2-weighted or fluid attenuated inversion recovery (FLAIR) MRI. Throughout this review we refer to the radiological findings as WMHs, while reserving the term WML when discussing the underlying changes in the cerebral white matter.

WMHs are associated with advanced age ⁷ and cerebrovascular risk factors including diabetes, cardiac disease, and hypertension.^8-12 Vascular dysregulation contributes to WMH development as white matter is sensitive to transient ischemia ¹³ and many larger WMHs are ischemic in origin.^{14, 15} Hypertension and blood pressure variability are associated with LLD ^16-18 and also contribute to WMH development,^{19, 20} particularly when accompanied by impaired cerebral vasomotor reactivity and altered autoregulatory processes.^{21, 22} Such deficits reduce cerebral blood flow (CBF) and may lead to WMHs.^{23, 24}

WMHs in Late-Life Depression

LLD is consistently associated with greater WMH severity ^25-28 and greater measured WMH volumes.^{10, 29-31} These observations guided proposals for MRI-defined vascular depression diagnostic criteria that include a lesion severity criterion.^{32, 33} In such conceptualizations, age at the initial depressive episode is an important consideration. Compared with early-onset depression, individuals with a later onset (e.g., after 50 years) exhibit greater WMH severity ^{5, 18, 31, 34-38} and cognitive impairment.^{39, 40} However, “vascular depression” as a potential diagnostic entity may not be limited to late-onset patients. Individuals with an earlier onset are at increased vascular risk as precedent depression is associated with increased risk of vascular disease and stroke.^41-45 Depression in early and mid-life may promote inflammation ^46-48 or epigenetic modifications of genes related to vascular homeostasis.⁴⁹ Thus some individuals with early onset depression may be prone to “vascular depression” later in life.

Localization of WMHs

Although many studies do not specify lesion location, distinctions in WMH location are important in understanding their role in LLD. In nondepressed samples, periventricular WMHs are more closely associated with cognitive impairment than are deep WMHs.^50-52 It is possible this is due to anatomical differences, as the periventricular region has a high density of long associating fibers with cortical-subcortical connections, while subcortical deep white matter has a high density of shorter U-fibers connecting adjacent cortical regions.^{52, 53} However, as WMH severity across regions is highly correlated, distinctions between periventricular and deep WMH may be arbitrary and reflect total WMH disease.⁵⁴ Nonetheless, others have localized WMHs associated with depression to the frontal ^55-58 and temporal lobe.⁵⁹ More recently several groups reported that LLD is associated with greater WMH severity in specific white matter fiber tracts including the cingulum bundle, uncinate fasciculus, and superior longitudinal fasciculus.^60-62

WMHs and Cerebral Function

A small number of studies suggest that greater WMH burden is associated with disrupted function in motor, cognitive, and affective circuits. During a simple motor task, high WMH burden is associated with low BOLD functional signal,⁶³ suggesting that WMH disease interferes with the neural circuit mediating the task. Other studies ^{64, 65} associated low BOLD activity with poorer performance on executive tasks, supporting the model of altered white matter integrity leading to cognitive network dysfunction. Greater WMH volume is further associated with altered default mode network connectivity at rest between the posterior cingulate cortex and medial prefrontal cortex.⁶⁶ During affective tasks, in LLD WMHs are associated with altered functional activity with an exaggeration of the functional ‘depression’ fMRI patterns: individuals with high WMH burden showed hyperactivation in limbic regions when presented with fearful faces.⁶⁷

Neuropathological Studies

Studies utilizing both neuroimaging and neuropathological techniques demonstrate that WMHs represent a wide range of pathological processes, including perivascular demyelination, arteriosclerosis, ischemia, gliosis, or partial loss of myelin and axons.^{15, 68, 69} Generally, confluent deep WMH but not periventricular WMH appear to be related to ischemic processes.⁷⁰

There are also regional differences in WMH etiology. Seminal work by Thomas and colleagues showed that deep WMH and punctate lesions of depressed older adults are most likely to have ischemic origins. In LLD ischemic lesions are also more likely to occur in the DLPFC, instead of the ACC or occipital lobe.¹⁴ Similarly, depressed elders exhibit increased expression of cellular adhesion molecules (CAMs) in the DLPFC, but not the ACC or occipital lobe.^71-73 CAMs are inflammatory markers whose expression is increased by ischemia, supporting a role for ischemia in LLD and highlighting the relationship between vascular and inflammatory processes. Although ischemic pathology is thus consistently localized to the DLPFC, other frontal regions may also be involved. Depressed elders exhibit decreased density of pyramidal neurons in the orbitofrontal cortex, which may be the result of vascular processes ⁷⁴ in arterioles and medium arteries ⁷⁵ or alterations in astrocyte-associated immune function.⁷⁶

Neuropathological studies also demonstrate that LLD can develop in the absence of significant vascular abnormalities. In a study of late-onset depressed elders, depression was not related to either lacunes or microvascular lesions.⁷⁷ Similarly, in a population-based study where depression was ascertained in a pre-mortem diagnostic interview, depression was associated neither with cerebrovascular nor Alzheimer pathology.⁷⁸ However, in contrast to prior studies,⁷⁹ depression was associated with the presence of Lewy bodies.⁷⁸ Such findings highlight the heterogeneity of pathologies that present as depression.

The Depression-Executive Dysfunction (DED) Syndrome and Cognitive Deficits in LLD

A “depression-executive dysfunction syndrome” ⁴ was conceptualized as the clinical expression of frontal network impairment caused by vascular and other aging related factors. Accordingly, this syndrome describes depressed patients with vascular disease and evidence of impairment in networks related to mood and executive function.⁸⁰

Executive function is a frontally mediated domain ^81-83 that encompasses cognitive processes including selective attention, response inhibition, and performance monitoring. It is clinically expressed as difficulty with planning, sequencing, organizing, and abstracting. These deficits are common in LLD,^84-87 particularly in late-onset depression.^{40, 86, 88} Efforts to characterize the presentation of the DED Syndrome have dichotomized samples by their performance on executive function tasks. When compared with depressed elders without executive dysfunction, patients with DED exhibit reduced fluency, impaired visual naming, suspiciousness, anhedonia, psychomotor retardation, and significant disability.^89-91 As discussed later, studies across eight different samples identified executive dysfunction as a predictor of poor antidepressant response.^92-99

Broader Cognitive Deficits in LLD: Relationship with Vascular Risk Factors

Individuals with LLD exhibit executive dysfunction and deficits across other cognitive domains including episodic memory, working memory, visuospatial ability, and processing speed.^{84, 87, 100-105} Processing speed deficits may influence other cognitive deficits ^103-105 and mediate in part executive task performance.¹⁰⁶ Although cognitive performance improves with successful antidepressant treatment, such deficits can persist and performance may not be restored to a normal level.^{100, 107-109}

Cognitive deficits are associated with vascular dysfunction. In nondepressed cohorts, there are associations between episodic memory, psychomotor speed, and executive function, and measures of vascular dysfunction, such as hypertension, heart rate variability, and orthostasis.^110-112 In LLD, Framingham Vascular Risk scores ¹¹³ are associated with deficits in processing speed, executive function, and episodic memory.^{98, 104} Greater WMH burden is associated with executive dysfunction, perseveration, and slowed processing speed ^{12, 61, 114-117} but also episodic and visuospatial memory deficits.^{102, 110, 118, 119} These cross-sectional observations are supported by longitudinal studies demonstrating that progression of WMH severity parallels cognitive decline.^120-122

WMHs localized to specific fiber tracts are associated with specific cognitive deficits. Executive function deficits are associated with WMHs in the uncinate fasciculus, superior longitudinal fasciculus, and the cingulum bundle.¹²³ Similarly, episodic memory deficits are associated with WMHs in the cingulum bundle and inferior longitudinal fasciculus.^{123, 124}

Vascular Depression and Course of Depression

Cognitive Dysfunction and Antidepressant Outcomes

In LLD, the relationship between treatment outcome and executive dysfunction is well established. Multiple studies demonstrate that executive dysfunction predicts poor acute response of LLD to antidepressants.^{92, 94, 96, 97, 99, 125} Executive dysfunction is also associated with high relapse and recurrence rates,¹²⁶ although not all studies found this association.¹²⁷ Much of this work used the Mattis Dementia Rating Scale Initiation-Perseveration (I/P) subtest, which yields a composite score of several executive skills.¹²⁸ Notably, a recent meta-analysis examined the relationship between antidepressant response and performance across multiple executive tests, including the I/P subtest, the Wisconsin Card Sort Test, the Stroop Color-Word test, and others. The authors concluded that the I/P subtest was the only test providing reliable discrimination between antidepressant responders and nonresponders.¹²⁹ Further analysis of I/P subtest components showed that semantic strategy (semantic clustering) while performing the I/P verbal fluency task explained most of the variance in predicting remission.¹³⁰ Effective semantic strategy appears to be associated with treatment response regardless of the probing task; effective semantic strategy during verbal learning task was also associated with high LLD remission rate.¹³¹ These observations suggest that preserved semantic organization, rather than fluency or verbal learning, is critical for remission during antidepressant treatment. However, impaired response inhibition is another aspect of executive dysfunction that influences antidepressant response.^{92, 95, 125, 132}

Corroborating the role of the cognitive control network in LLD are reports examining resting state functional connectivity (FC). Low resting FC in the cognitive control network, but not in the default network, predict poor acute antidepressant remission rates along with persistence of depression, apathy and dysexecutive behavior.¹³³

Other cognitive deficits can influence outcomes. Poorer performance on tests of episodic memory, language processing, and processing speed are associated with poorer acute antidepressant response ⁹⁸ and may predict poorer long-term course of depression.¹³⁴ Further analysis is needed to identify a superordinate cognitive dysfunction that accounts for the relationship of these tests to antidepressant response.

WMHs as a Predictor of Antidepressant Outcomes

Although greater WMH severity is often associated with poorer acute antidepressant response,^{36, 98, 99, 135-138} results are inconsistent.^{97, 139-142} Importantly, these studies assess total cerebral WMH severity, while the effect of WMLs on the antidepressant response may depend on lesion location.¹³⁸ Unfortunately, methodological differences in assessing WMH severity make it difficult to reconcile discrepant findings across studies.^{143, 144}

Most studies used cross-sectional WMH measures to predict outcome. However, WMHs are a progressive rather than a static process ^{11, 145} and rate of change may be a more important predictor. The few studies in LLD cohorts examining this question found that greater increases in WMH volume over two- and four-year intervals are associated with nonremission or relapse.^{146, 147}

Some have examined both WMH severity and cognitive dysfunction as predictors of outcome. In these studies cognitive measures are generally a better predictor of antidepressant response than are WMH measures. The largest published study examining a 12-week response to sertraline found that poorer response was independently associated with greater WMH severity and poorer performance in cognitive domains of episodic memory, language processing, processing speed and executive function. However, in models controlling for baseline depression severity, WMH severity was no longer associated with response.⁹⁸ Smaller studies have found similar patterns, wherein initial associations between WMH severity and response were not statistically significant in models incorporating cognitive measures.^{97, 99}

Other measures of white matter integrity are associated with antidepressant response. In voxel-wide analysis, microstructural white matter abnormalities (low fractional anisotropy, FA) in multiple frontolimbic brain areas, including the rostral and dorsal anterior cingulate, dorsolateral prefrontal cortex, genu of the corpus callosum, hippocampus, and posterior cingulate, were associated with non-remission.¹⁴⁸ In region of interest analyses, low FA in the corpus callosum, left superior corona radiata, and right inferior longitudinal fasciculum was associated with lower remission rates.¹⁴⁹ In contrast, others found that higher prefrontal white matter FA was associated with a failure to remit to antidepressants,¹⁵⁰ a finding that parallels reports in depression of increased functional connectivity to a dorsal nexus.¹⁵¹ Notably, depressed elderly 5-HTTLPR polymorphism short allele carriers had lower FA than long allele homozygotes in frontolimbic areas, including the anterior cingulate, posterior cingulate, dorsolateral and medial prefrontal regions.¹⁵² Such differences in FA can be caused by several pathologies, not all of which are vascular in nature.

Vascular Risk Genes in Depression

There is a wide literature examining genetic influences in LLD, including reports examining polymorphisms in BDNF, APOE, and serotonin transporter (5-HTTLPR) genes, amongst others (see recent reviews by Lotrich ¹⁵³ and Benjamin and Taylor ¹⁵⁴). Within this field, there are an increasing number of reports associating LLD with genetic polymorphisms increasing vascular risk.

One example of this literature includes studies examining genetic variation of the renin-angiotensin system (RAS). The RAS regulates fluid homeostasis and blood pressure and is a target for two classes of antihypertensive drugs. Beyond influencing vascular function, RAS polymorphisms may also increase the risk of depression via modulation of monoamines ¹⁵⁵ or contributing to hypothalamic-pituitary-adrenal (HPA) axis dysregulation.¹⁵⁶ Results from studies examining the common angiotensin converting enzyme (ACE) insertion/deletion (I/D) polymorphism are mixed in adult depression ^157-162 and suicidality,^162-164 although other ACE polymorphisms may be important.^{161, 165, 166} Despite its key role in the RAS, the gene encoding for the angiotensin II type 1 receptor (AGTR1) is less well studied. The CC genotype of the A1166C (rs5186) polymorphism is associated with increased responsiveness to Ang-II.¹⁶⁷ However, despite one study associating the CC genotype with increased risk of MDD,¹⁵⁹ this was not replicated in a elderly sample.¹⁶⁸

RAS variation is further related to differences in brain structure and function. Several studies in older adults have now associated variants in ACE and AGTR1 with altered cerebral frontotemporal structure ^169-172 and abnormal default mode network activity.¹⁷³AGTR1 variants are associated with greater WMH progression, particularly in men,¹⁶⁸ and this effect may be localized to specific tracts.⁶⁰

RAS polymorphisms may also influence antidepressant outcomes. In a general adult population, the ACE D variant and the AGTR1 C1166 allele are associated with better acute antidepressant response.^{174, 175} However, this contrasts with a study of LLD, wherein AGTR1 C1166 allele homozygous individuals exhibited a poorer antidepressant response over a longer period.¹⁷⁶ Such age differences could be important, reflecting stress reactivity in earlier life and allostatic effects of an overly active RAS later in life.

Mechanistic Hypotheses

According to a model by Alexopoulos,¹ the clinical expression of LLD is mediated by altered brain activity in cognitive and affective circuits, characterized as hypometabolism of dorsal cortical regions and hypermetabolism of ventral limbic structures. Age-, disease-, and stress-related neurobiological changes serve as etiological factors contributing to alterations in circuits regulating mood and cognition. The impact of these neurobiological contributors can be increased by vulnerability factors originating from preexisting differences in mood circuitry.

Arguably, such etiological factors may lead to depression only after crossing a certain threshold above which they acquire a dose effect relationship (Figure 1). In this conceptualization, single or multiple potential “etiological factors” may reach an initial severity threshold and contribute to LLD “vulnerability,” due to frontolimbic compromise evident through imaging or abnormal performance on cognitive or affective tasks. When etiological factors further increase in severity and cross a second threshold, they may have a greater and direct effect on mood circuit function, leading to affective and cognitive symptomatology. Of course, instead of threshold effects such relationships could be cumulative following linear or curvilinear patterns

Figure 1.

Proposed progression of etiological processes affecting frontolimbic function. After crossing an initial threshold, vascular or other processes may serve as one of many factors that increase vulnerability to depression through compromise of affective and cognitive brain circuits. As the underlying disease process progresses, its adverse effects on the affective and cognitive circuits would likewise increase, resulting in the circuit dysfunctions that characterize depression.

Such a threshold model can account for multifactorial contributions to LLD. It also explains why our best clinical and biological measures predict only a modest part of the variance in distinguishing LLD from controls and in predicting treatment response as any individual measure may itself not cross a crucial threshold. This model allows for mixed syndromes of multiple etiological contributors if each disturbance alone or in combination is strong enough to cross the threshold required to induce neural circuit changes mediating LLD.

Given the central role of vascular disease and WMHs in much of the literature, it is worth examining their role within this model. WMHs can be conceptualized as visible markers of ischemic damage, although vascular compromise may result in insults to white and/or gray matter that are not visible on conventional neuroimaging.¹⁷⁷ In this context, greater WMH severity represents a state of vascular compromise and ischemic injury in specific regions. Depending on the severity of the underlying WMLs and how strongly they affect affective and cognitive circuitry, such damage may itself predispose individuals to depressive episodes. However, WMHs may also be an index of the severity of underlying vascular disease affecting neural circuits. In this case, the direct effect of WMHs on neural circuits may be modest compared with that of the underlying vascular process.

The Disconnection Hypothesis

Building on the concept of “disconnection syndromes,” ischemia and WMLs may contribute to depression by disrupting neural connections among regions regulating mood and cognition.¹⁷⁸ In this model, global cerebral WMH severity is less germane to LLD than is focal damage to specific fiber tracts and neural circuits. Such focal damage would adversely affect the tract's structural and functional connectivity. In turn this state adversely affects the function of connected regions at rest and during cognitive tasks and contribute to neural circuitry alterations that mediate clinical symptoms and influence antidepressant response.

This view is supported by studies examining WMH location. Recent work demonstrates that LLD is associated with greater WMH severity in specific tracts including the cingulum bundle, uncinate fasciculus, and superior longitudinal fasciculus.^60-62 Additionally, greater WMH severity in the uncinate and superior longitudinal fasciculi is associated with executive dysfunction ^{61, 123, 124} and greater depression severity.¹⁷⁹

These macrostructural MR findings are supported by diffusion tensor imaging (DTI) studies examining white matter microstructure. DTI changes reflect various pathologies leading to decreased myelin integrity, including demyelination secondary to cerebrovascular and inflammatory changes. WMLs occurring in fiber tracts are also associated with DTI changes.^{177, 180} LLD is associated with differences on DTI measures of the uncinate fasciculus, cingulum bundle, anterior thalamic radiation, and superior longitudinal fasciculus.^{181, 182} The uncinate fasciculus and cingulum bundle are particularly relevant to LLD ^183-188 because of their role in cognition and emotion processing,^{188, 189} although other tracts are also associated with antidepressant response in LLD.¹⁴⁹ Importantly, differences in the uncinate fasciculus and cingulum bundle are also observed in younger depressed adults.^{190, 191} Although those younger adult studies do not necessarily suggest a vascular cause in that population, they support that structural alterations in these fiber tracts are associated with depression.

Structural differences have functional effects. Fiber tract structural connectivity is positively correlated with resting state FC of connected regions.^192-196 These data warrant a comprehensive examination of the relationships among ischemic focal WML injury, damage to tract microstructure, regional connectivity, and circuit function. Examination of such a “lesion model” can identify neural circuit deficits contributing to depression and predictive of poor outcomes.

The Inflammation Hypothesis

Aging- and disease-related processes promote proinflammatory states in older individuals.^{197, 198} Further, immune activation can be a characteristic of depression ^{48, 199} and precipitate depressive symptoms.²⁰⁰ Alexopoulos and Morimoto recently proposed that immune dysregulation may promote the development of affective and cognitive symptoms in LLD.²⁰¹ This view is supported by studies in midlife adult populations. Administration of cytokines or induction of peripheral inflammation results in an inflammatory response, which in turn is correlated with fatigue, slowed reaction time,²⁰² and mood reduction.^{200, 203} Even without medical illness, depressed individuals exhibit increased levels of proinflammatory cytokines ^{204, 205} and reduced anti-inflammatory cytokine levels.^{206, 207}

Several mechanisms may explain these relationships. Proinflammatory cytokines affect monoamine neurotransmitter pathways,²⁰⁸ including indoleamine 2,3-dioxygenase upregulation and kynurenine pathway activation.^209-211 This results in decreased tryptophan and serotonin and increased synthesis of detrimental tryptophan catabolites that promote hippocampal damage and apoptosis.^{211, 212} Cytokines, including IL-1β, also reduce extracellular serotonin levels by activating the serotonin transporter.²¹³ Additionally, proinflammatory cytokines disrupt glucocorticoid receptor function ²¹⁴ and reduce neurotrophic support.²¹⁵

Peripheral inflammation is associated with altered brain function in areas mediating cognition and mood. In healthy individuals, typhoid-induced IL-6 levels correlate with lower mood, increased subgenual anterior cingulate cortex activity during affective processing, and altered subgenual cingulate connectivity.²⁰³ Laboratory-administered social stressors increase inflammatory maker levels that in turn are associated with activation of the dorsal anterior cingulate cortex and anterior insula.²¹⁶ In bereaved women undergoing grief elicitation, inflammatory markers were positively associated with subgenual cingulate and orbitofrontal cortex activation.²¹⁷

Supporting this work is a growing literature demonstrating that genetic polymorphisms promoting pro-inflammatory processes may contribute to depression ^{218, 219} and in LLD may be related to age of onset.²²⁰ Such relationships are likely moderated through an effect on cytokine levels that in turn are associated with sad mood ²²¹ and altered emotion processing.²²²

Regarding LLD, the aging process disrupts immune function,²²³ increasing peripheral immune activity and shifting the CNS into a proinflammatory state.¹⁹⁸ Elevated peripheral cytokine levels are associated with depressive symptoms in older adults, with the most consistent finding being for IL-6, but also implicating IL-1β, IL-8 and TNFα.^224-228 Proinflammatory states in older adults are associated with cognitive deficits, including poorer executive function, poorer memory performance, worse global cognition, and steeper decline in cognition.^229-231 Finally, greater IL-6 and C-reactive protein levels are associated with greater WMH burden.^232-235

Successful antidepressant treatment may reduce proinflammatory markers.⁴⁸ This action may be a direct effect as in vitro studies demonstrate that antidepressants reduce proinflammatory markers while increasing levels of anti-inflammatory cytokines.^236-240 Results in clinical populations are inconsistent, but a recent meta-analysis concluded that antidepressants, particularly SSRIs, reduce IL-6, IL-1β, and TNFα.²⁴¹ Additional evidence supports that some anti-inflammatory drugs and may have antidepressant properties,^242-244 particularly in individuals with elevated pre-treatment proinflammatory markers.²⁴⁵

Importantly, pro-inflammatory processes also contribute to neurodegenerative processes. Increased peripheral inflammatory marker levels are associated with increased risk for all-cause dementia ²⁴⁶ and central inflammatory processes significantly contribute to the pathogenesis of Alzheimer's disease.²⁴⁷ However, nonsteroidal anti-inflammatory drugs have not consistently demonstrated clinical benefit in Alzheimer's disease ²⁴⁷ and in a recent consensus statement were considered but not recommended as a priority for further investigation.²⁴⁸

The Hypoperfusion Hypothesis

Vascular dysregulation is common in LLD ^249-254 and CBF reductions can impair regional brain function, contributing to affective and cognitive symptoms. Regional cerebral metabolic activity is tightly correlated with CBF, which is regulated by local interactions between neurons, glia, and the vasculature.^{255, 256} Blood flow to the brain is influenced by systemic hemodynamics and cerebrovascular autoregulation, with cerebral arteries contracting or dilating as arterial pressure changes. These processes interact to maintain stable perfusion. However, these processes are impaired in the context of vascular disease:²⁵⁶ hypertension, diabetes, and atherosclerosis lead to vascular wall hypertrophy, reduced arterial lumen diameter, reduced arterial distensibility, and endothelial cell dysfunction.^256-258 Such vascular changes, including increased intima media thickness, increased arterial stiffness, and endothelial dysfunction are pronounced in LLD populations ^249-254 and endothelial function may be particularly poor in antidepressant nonresponders.²⁵⁹ Such vascular pathology results in reduced blood flow velocities and decreased vasomotor reactivity,^{260, 261} adversely affecting CBF.

Perfusion deficits do not need to cause ischemia in order to influence brain function. Reduced CBF impairs protein synthesis ²⁶² crucial for cognitive processing ^{263, 264} and for maintaining the integrity of cortical functional maps.²⁶⁵ Thus mild CBF reduction may impair cognitive and affective processes, while greater CBF reduction in the context of autoregulatory deficits may cause ischemic injury. The subcortical white matter is particularly sensitive to these changes because it is supplied by terminal arterioles with limited collateral flow ²⁶⁶ and so susceptible to infarction due to impaired autoregulation.²⁶⁷ Greater WMH severity may be a marker of broader deficits in perfusion and autoregulation as individuals with greater WMH severity exhibit reduced CBF in both white matter and gray matter regions.^268-270 Depressed elders with greater WMH severity also exhibit decreased perfusion in the cingulate gyrus,²⁷¹ a region involved in cognitive and affective processing.²⁷²

PET, SPECT, and MRI arterial spin labeling (ASL) have identified CBF differences in midlife adult depressed populations.^273-276 Advanced age is itself associated with decreased frontotemporal CBF,²⁷⁷ an effect mediated by vascular risk factors.²⁷⁸ In LLD, perfusion deficits are more severe, particularly in the medial and lateral prefrontal cortex (PFC), subcortical, and temporal structures.^{269, 279-283} Deficits in the dorsolateral PFC may improve with antidepressant response, while other perfusion deficits may not.²⁸¹ Others report no change in CBF in antidepressant nonresponders,^284-288 suggesting that persistently reduced regional CBF may be a biomarker of nonresponse.

Cerebral hypoperfusion is associated with Alzheimer's disease and mild cognitive impairment,^289-291 but also with abnormal performance on specific cognitive tasks.^291-294 For example, resting hippocampal blood flow is positively correlated with spatial memory performance.²⁹² Although not specifically a measure of perfusion, reduced arterial CBF predicts slower processing speed,²⁹⁵ a domain that may mediate other deficits in LLD.¹⁰³ Perfusion deficits may thus contribute to the affective and cognitive symptoms observed in LLD. Intriguingly, these data support investigating the utility of agents that may improve perfusion or restore endothelial function.^{296, 297}

Role of Vascular Risk Factors in Late-Life Depression

Fifteen years after the formulation of the “vascular depression” hypothesis,² a confluence of findings suggests that:

1. Vascular risk factors and radiological findings are often associated with LLD.

2. Cognitive deficits across multiple domains are common in LLD and associated with vascular risk factors and greater WMH severity. Such deficits, particularly executive dysfunction, predict poor response to antidepressants.

3. WMHs, although associated with cognitive dysfunction, do not consistently predict poor antidepressant response. However, progressive WMH disease is associated with poorer depression course.

The link between progressive vascular disease and the development and persistence of LLD invites study of the underlying mechanisms by which vascular disease alters brain function in such a way to mediate the development of clinical symptoms. The ultimate goal of this line of study is to develop novel treatments or repurpose existent agents to improve LLD outcomes.

We propose three mechanistic paths to “vascular depression.” Although the processes of these pathways could proceed independently, they are complementary and interconnected and so may all contribute to LLD (Figure 2). For example, inflammation participates in vascular remodeling, accelerates vascular damage ²⁹⁸ and can contribute to perfusion deficits. In turn, regional perfusion deficits adversely affect gray matter function and contribute to the development of WMLs that disrupt regional structural and functional connectivity.

Figure 2.

This model links the disconnection, inflammation, and hypoperfusion processes proposed in late-life depression. Vascular disease may contribute to altered brain function characteristic of depression (dorsal hypometabolism, ventral hypermetabolism) either through structural damage adversely affecting connectivity, through perfusion deficits altering regional function, or both. Proinflammatory processes increase vascular risk but may also affect brain function through independent processes.

In this model, vascular disease is an important and central contributor to LLD. Not only is vascular disease common and perhaps unavoidable in later life, but it also interacts with other pathological processes related to depression. However, vascular disease is not the only contributor to LLD. Other biological and environmental factors can increase the risk for depression by alterating neural circuit structure and function. Similarly, genetic influences may increase risk of depression by acting at multiple points, such altering function of neural circuits, changing the biological response to stress, adding to vascular risk, or predisposing to proinflammatory states.

We are not proposing a single mechanism underlying LLD. Non-vascular factors clearly contribute to LLD. The same genetic, epigenetic, and environmental factors that contribute to depression in younger adults continue to confer vulnerability to depression in later life. Moreover, although alterations in immune and endocrine regulation affect vascular risk, they may also increase the risk of depression through independent mechanisms.

Just as there are non-vascular factors contributing to LLD, there may be additional mechanisms linking vascular disease with LLD. Altered homocysteine metabolism is associated with LLD,²⁹⁹ with hyperhomocysteinuria potentially contributing to vascular risk by increasing atherosclerosis and endothelial dysfunction.³⁰⁰ Another example is altered sympathoadrenomedullary system and hypothalamic-pituitary-adnenal (HPA) axis function ³⁰¹ that, among other activities, affects inflammation.³⁰² These and other mechanisms may contribute to the development or perpetuation of LLD both through vascular and non-vascular pathways and require further study. Such studies would benefit from the incorporation of genomic approaches along with measures of systemic function, thus linking molecular mechanisms with clinically salient measures and outcomes.

Demonstrating that such mechanisms affect neural circuit function opens the door for intervention studies examining whether modulation of these mechanisms can improve depression outcomes. There is already preliminary support for using anti-inflammatory drugs in depression. Etanercept, a soluble TNFα receptor, and celecoxib, a cyclo-oxygenase-2 inhibitor, may reduce depressive symptoms in patients with inflammatory diseases ^{242, 243} and infliximab may improve depression in patients with greater pre-treatment inflammation.²⁴⁵ Similarly, omega-3 fatty acids are anti-inflammatory and also affect monoamine neurotransmission,³⁰³ suggesting an antidepressant role for omega-3 supplementation. However, recent meta-analyses have not provided strong support for omega-3 supplementation for depression, although larger well designed trials may be warranted.^{304, 305} Finally, drugs that modulate the renin-angiotensin system improve cerebral perfusion ²⁹⁷ and may improve cognition,^306-308 although they have not been well studied in depression. We suggest that study of these agents in LLD may produce a novel, targeted pharmacology for the large group of LLD patients who are unresponsive to conventional antidepressants.