|Home | About | Journals | Submit | Contact Us | Français|
One third of all stroke survivors develop post-stroke depression (PSD). Depressive symptoms adversely affect rehabilitation and significantly increase risk of death in the post-stroke period. One of the theoretical views on the determinants of PSD focuses on psychosocial factors like disability and social support. Others emphasize biologic mechanisms such as disruption of biogenic amine neurotransmission and release of proinflammatory cytokines. The “lesion location” perspective attempts to establish a relationship between localization of stroke and occurrence of depression, but empirical results remain contradictory. These divergences are partly related to the fact that neuroimaging methods, unlike neuropathology, are not able to assess precisely the full extent of stroke-affected areas and do not specify the different types of vascular lesions. We provide here an overview of the known phenomenological profile and current pathogenic hypotheses of PSD and present neuropathological data challenging the classic “single-stroke”-based neuroanatomical model of PSD. We suggest that vascular burden due to the chronic accumulation of small macrovascular and microvascular lesions may be a crucial determinant of the development and evolution of PSD.
More than a hundred years ago, Meyer  postulated that depression may be the consequence of the combined effect of brain injury, affecting mainly the left frontal lobe as well as other lobar convexities, and psychosocial vulnerability such as past psychiatric history. Although the association of depression with stroke has been recognized clinically for several decades, only in the past 25 years systematic studies on this subject have been conducted (for review see ). According to the World Health Organization , stroke is the most common severe neurological disorder and the third leading cause of mortality among adults, with 15 million people worldwide suffering stroke annually. A recent review reported a markedly stable frequency of depressive symptoms in 33% of all stroke survivors at any time during the follow-up . In addition, stroke survivors have more than six fold higher risk of developing clinically overt depression even two or more years after index stroke compared to age-matched controls . Depression is significantly more likely after stroke than other illnesses with comparable disability and adversely affects both short-term recovery and long-term rehabilitation . Moreover, death risk is three-fold higher over the next decade in depressed stroke survivors compared to those without depression .
The possibility that the risk of depression after stroke is related to lesion location was developed in the mid-seventies at John Hopkins University and originated the concept of post-stroke depression (PSD) . The first animal data in this field showed an association between the laterality of brain damage after middle cerebral artery occlusion and both brain catecholamine concentration and behavioral patterns in rats suggesting that the ischemic damage may induce depression via the disruption of biogenic amine pathways . Later on, the first associations between the localization of stroke and occurrence of depression were reported by the same group in humans . Several subsequent studies suggested a link between PSD and macrovascular lesions in various brain regions such as the left frontal lobe, bilateral prefrontal cortex, left anterior and right occipital cortex (for review see ). Despite these research efforts, the relationships between vascular lesions and the occurrence of depression remain a matter of debate [12–14]. Equally intriguing are the recent contradictory results on the location of brain lesions in PSD [8,11,15,16]. Methodological issues such as differences in sampling, assessment interval from stroke and type of lesions considered have been evoked to explain the discrepant results. After providing a brief overview of the clinical, epidemiological and treatment issues related to PSD, the present review focuses on the pathogenetic models of this condition and critically discusses its neuroanatomical substrates with particular reference to the contribution of neuropathological data in this field.
The peak prevalence of major depression occurs 3–6 months after the stroke with a range of 9–34% during this time-frame . The greatest increase in the prevalence of PSD is observed between the initial and 1-month repeated assessments despite an overall decrease of disability . However, the risk appears to remain high even 1–3 years after stroke so that if one considers all depressive episodes occurring both in the acute and later post-stroke periods, PSD prevalence would be even higher reaching 25% to 79% . Thus far, no precise data are available regarding the impact of PSD on community healthcare resources but it is thought that PSD patients are more often in institutional care, are more frequent users of health services and show poorer functional recovery compared to non-depressed stroke survivors . Only physical disability, stroke severity and cognitive impairment have been consistently identified as risk factors for PSD . The depression-stroke Italian multicenter study added female gender and previous cerebrovascular or depressive episodes as significant risk factors for this condition . More rarely cited psychosocial risk factors include positive family history of anxiety or mood disorder, high trait neuroticism on the Eysenck inventory, older age, bereavement, sleep disturbance, poor social support and recent adverse life events .
Not surprisingly, a consensual definition of PSD is difficult to achieve . Typically, DSM-IV diagnostic criteria classify PSD within the group of “mood disorders due to stroke, with depressive features” or “with major depressive-like episode” (DSM-IV, American Psychiatric Association, 1994). Whyte and Mulsant  defined PSD as “depression occurring in the context of a clinically apparent stroke (as opposed to silent vascular disease)” emphasizing the sequential nature of the two temporally related events – stroke followed by depression – although an etiopathogenetic relationship is not assumed. The clinical patterns of PSD remain quite vague. Paradiso et al. indicated that early-onset PSD is characterized by anxiety, loss of libido and feelings of guilt, whereas later-onset PSD is more frequently associated with diurnal variation of mood and social isolation . In an attempt to delineate PSD from major depression, Gainotti et al. proposed that catastrophic reactions, hyperemotionality and diurnal mood variation are more consistently associated with PSD, suicidal thoughts and anhedonia being more frequent in endogenous depression . However, catastrophic reactions and altered emotional regulation have been described in several neurological conditions with or without depression raising doubts about the specificity of this cluster for defining PSD .
The course of PSD seems to be dependent on the timing of onset. Early onset depressive episodes tend to have a shorter course with more probable spontaneous remission in comparison to later-onset depressive episodes. Astrom et al. showed that, at 1 year, 60% of the patients with early depression (0 to 3 months) had recovered and that those who had not recovered at this follow-up had a high risk of developing chronic depression . The results of Andersen et al. confirmed this notion showing that stroke survivors who experience onset of depression seven weeks or later after their stroke have a lower rate of spontaneous recovery . There may be different biological or psychological processes working in acute PSD in comparison to subacute or chronic mood disorder. Early PSD may represent a transient reaction to the stroke with remission based on neural regeneration or biochemical mechanisms (changes in receptor regulation, rate of transmitter or enzyme synthesis). In contrast, late PSD may reflect neuropathological changes secondary to the stroke with a delayed but enduring impact on mood .
Several placebo-controlled, randomized double-blind studies support the efficacy of antidepressant medication in PSD. Citalopram , nortriptyline , and fluoxetine  were all reported to be significantly more efficacious than placebo in the treatment of depressive symptoms after stroke. The two latter agents also improved activities of daily living  and reduced mortality . Similarly, sertraline showed to be superior to placebo in preventing the occurrence of PSD . Besides its direct effect on depressive symptoms, antidepressant treatment in PSD may also exert outcome-improving effects on the brain in at least two ways . An action on monoaminergic nuclei modulating cortico-striato-pallido-thalamo-cortical (CSPTC) pathways is one of the options. Five CSPTC pathways referred to as the oculomotor, dorsolateral prefrontal, lateral orbital frontal and anterior cingulated pathways form closed loops, which originate from the prefrontal cortex and project to the globus pallidus, substantia nigra and thalamus . The last three pathways are involved in both executive functions and affective regulation. Their functionality is thought to be closely modulated by the activity of monoaminergic nuclei (namely, raphe, locus coeruleus and ventral tegmental area) which are preferential sites of action of antidepressant medication . Alternatively, antidepressants may promote the secretion of neurotrophins favoring brain recovery. One of these neurotrophins, brain-derived neurotrophic factor (BDNF), and its receptor trkB, are increased by antidepressant treatment and it has been shown that BDNF-dependent signaling, affecting intracellular signal transduction and synaptic connectivity, is required for antidepressants to induce mood recovery . One should, however, keep in mind that the positive impact of antidepressants in PSD is not generally accepted. For instance, a recent systematic review of pharmacological treatments in PSD was unable to find any strong evidence on the effectiveness of pharmacotherapy in either producing a remission or preventing the onset of depression after stroke .
Research into the pathogenesis of PSD is characterized by the striking opposition between two theoretical models: that of the psychosocial vulnerability and that of the biological determinism. This polarity of thought may have impeded the development of a comprehensive approach to PSD prevention and treatment. In this section we will attempt to explore both the relative impact and conceptual limits of each hypothesis.
The notion that depression in stroke patients involves the psychosocial response to disability and loss has been present from the pioneer reports on emotional disorders following stroke. Considering the brain as “the most cherished organ of humanity”, Fisher considered depression an understandable reaction to cerebrovascular disease . The relationship between PSD and disability is complex since bidirectional causality has been sustained. Disability after stroke may trigger PSD which in turn led to reduced functional outcome [19,29,40]. Astrom et al. showed that several psychosocial factors play a role in the development of both early and late-onset PSD . The absence of family support is a strong determinant of acute depressive reactions following stroke. Dependence for the activities of daily living (ADL) is not associated with acute depressive reactions but it becomes the most important predictor of depression after the first three months. This observation may support the notion that functional impairment does not determine the onset of depression but rather interacts with it to limit the results of long-term rehabilitation. One year post-stroke, the persistence of few social contacts is the main psychosocial factor associated with PSD. In the same line, during the 3 years following stroke depressed patients have a marked paucity of social contacts compared to non-depressed patients. Although these data point to the influence of psychosocial determinants in the development of PSD, one should acknowledge that the distinction between dependent and independent variables is quite ambiguous in this context (i.e. reduced social contacts can be a cause as well as a result of depression).
The biogenic amine and cytokine hypotheses represent the two main theories surrounding a biological origin of PSD . Robinson and Bloom postulated that ischemic lesions may interrupt the biogenic amine-containing axons ascending from the brainstem to the cerebral cortex leading to a decreased production of serotonin (5HT) and norepinephrine (NE) in limbic structures of frontal and temporal lobes as well as basal ganglia . The deficit of biogenic amines in these areas or the failure to up-regulate 5HT receptors after stroke may trigger PSD . Supporting this hypothesis, significantly lower CSF concentrations of the serotonin metabolite 5-hydroxy-indoleacetic acid were measured in PSD patients compared to non-depressed stroke survivors . Equally in line are the recent PET-scan findings of Moller et al.  on 5HT1A receptor availability after stroke suggesting that changes of serotonin neurotransmission may occur in the early phase of stroke and be modulated by treatment with serotonin reuptake inhibitors (SSRI). In another study , growth hormone (GH) response to desipramine was significantly blunted in PSD patients compared to non-depressed post-stroke patients, suggesting diminished α2 adrenoreceptor function as a possible marker of PSD.
Alternatively, the proinflammatory cytokines IL-1, tumor necrosis factor α (TNF-α), IL-6, IL-8 and IL-18 that are involved in the initiation and amplification of the inflammatory response to acute cerebral ischemia may also have depressiogenic properties. In fact, the participation of proinflammatory cytokines in the etiology of depression has been explored by several authors [47,48]. Interleukine-1β (IL-1β) and TNF-α can induce a depressive-like behavior in mice resembling the somatic syndrome of depression in humans. It is known that autoimmune disorders and HIV, primarily affecting the immunological system, are associated with depression. Additionally, increased brain levels of IL-1β, TNF-α, IL-6 have been found in patients suffering from mood disorders . Cytokine activation may induce depressive symptoms by increasing the activity of the hypothalamus-pituitary-adrenal (HPA) axis and altering biogenic amine neurotransmission in several hypothalamic nuclei as well as limbic sites. Some cytokines are also known to up-regulate the tryptophane-catabolizing enzyme indoleamine 2,3-deoxygenase (IDO) leading to decreased synthesis of 5HT. Although attractive, this hypothesis shows two main conceptual weaknesses. First, no relationship was established between the level of proinflammatory cytokine induction and PSD in humans. Most importantly, molecular cascades after stroke also include a marked induction of anti-inflammatory cytokines which may counterbalance the depressiogenic effect of proinflammatory cytokines [50,51].
After the pioneering work of Robinson and Szetela  who reported a significant inverse correlation between the severity of PSD and distance of the anterior border of the lesion from the frontal pole, several MRI studies attempted to explore the neuroanatomical background of PSD leading to conflicting data. A first observation in this field concerns the predominance of left hemisphere lesions in patients with PSD . This right–left difference was present both for cortical and subcortical lesions and seems to be more robust for left anterior lesions . More recent publications confirmed these findings suggesting the existence of a significant relationship between PSD severity and proximity to the left frontal pole (for review see ). However, the association between left anterior lesion location and PSD is far from being consensual. Whereas Robinson’s pooling results  suggested that left anterior cerebral lesions were associated with significantly higher depression scores than left posterior lesions (the site of the lesion explaining 50% of the variance), two systematic reviews by Singh et al.  and Carson et al.  did not support this association. The lack of uniformity in definition and measurement of depression as well as sampling differences (i.e., in-patient versus community-based series) may partly explain these discrepancies. However, the conflicting conclusions of different studies may be mostly attributed to the highly variable assessment interval from the acute post-stroke period to the chronic stroke phase . In fact, it is conceivable that lesion location is a key determinant of PSD immediately after stroke (namely by affecting biogenic amines and 5HT receptor function), while in late-onset PSD failure to adapt to the consequences of stroke can be more important (through impairments such as social isolation or visual perceptual decline and changes in arousal and attention related to right hemisphere stroke). In this context, Shimoda and Robinson  reported that, during the acute post-stroke period, PSD in inpatients was associated with left anterior lesions. By 3–6 months post-stroke, the proximity of the lesion to the frontal pole in both hemispheres rather than the hemispheric location influences the occurrence of PSD. At this period, depression was also related to the impairment in daily living activities and social functioning. At 1–2 years follow-ups, the occurrence of depression among the patients with left hemisphere stroke was mostly determined by the severity of ADL impairment. At this time period, depression in patients with right hemisphere damage was associated with lesion volume and proximity to the occipital pole.
Undoubtedly, another parameter to consider is the strategic location of ischemic damage. However, data in this field are scarce and contradictory probably reflecting the fact that unlike neuropathology, most neuroimaging methods did not assess precisely the total extent of the affected area. MRI-documented infarcts in hippocampus, basal ganglia and frontal areas have been classically associated with PSD [57–59]. Hama et al. found that the severity of affective depression was associated with left frontal lobe damage whereas apathetic depression was mostly related to basal ganglia damage . In another study, the mean frequency of infarcts in the genu of the internal capsule on the left side and bilateral pallidum as well as the mean volume of infarcts in the right occipital lobe were independent determinants of PSD .
Although difficult to perform, neuropathological analyses may be of great value in order to assess the validity of the neuroanatomical model of PSD. Unlike structural imaging, they allow for evaluating both the extent of vascular burden and different types of vascular lesions in the human brain. In fact, as for vascular cognitive impairment, one could suspect that not all vascular lesions had a deleterious impact on mood. In addition, they provide reliable estimates of the number of small vascular lesions not easily identified by MRI. We recently performed the only available neuropathological study of 95 consecutive autopsy cases of patients with stroke . Among them, there were 21 patients (9 men and 12 women) who developed PSD (mean age: 78.7±8.0 for men and 80.6±9.2 for women) and 74 (30 men and 44 women) patients without PSD (mean age: 83.5±7.2 for men and 86.5±5.4 for women). Clinical diagnosis of stroke was based on the presence of focal neurological deficits of acute onset and supportive neuroimaging. Subjects with stroke were included only if the diagnosis could be confirmed pathologically by the presence of cortical macroinfarcts and if post-stroke survival was at least one month. Cases presenting with first-onset depression more than 2 years after a stroke were excluded (4 cases). The clinical diagnosis of depression was established prospectively using the DSM-IV criteria for major and minor depression on the basis of the psychiatric examination performed by a specialist in geriatric psychiatry. Lifetime history of depression was also established on the basis of this interview. The diagnosis was confirmed in all cases after review of the medical records by the same trained psychiatrist blind to the neuropathological findings. Both hemispheres were cut into 1 cm thick coronal slices and examined for macroscopic vascular lesions including brain infracts, cerebral hemorrhage or lacunar infarcts. Microscopic infarcts were identified and dated on two adjacent sections of each level with hematoxylin–eosin and cresyl violet stains. All cases with concurrent neuropathological evidence of Alzheimer’s disease (Braak stages IV, V, and VI or dementia with Lewy bodies were excluded. Macroscopic brain infracts, multiple lacunar infracts and substantial microvascular pathology confined to one brain region were classified as focal ischemic pathology, whereas multiple vascular lesions involving more than one cortical area were classified as multifocal ischemic pathology. Our findings showed no relationship between the development of focal and multifocal vascular pathology in a specific brain area and PSD in elderly patients without lifetime history of depression. Among demographic variables, only younger age at death was significantly related to PSD further confirming that depression after stroke is also a predisposing factor for earlier death. Although these first neuropathological data could be interpreted as an indirect support of the psychosocial model of PSD, several methodological limitations should be taken into account. First, as is the case for all neuropathological studies, autopsy-related biases cannot be excluded. Second, due to the limited number of cases with PSD, the power of the present study does not permit to identify mild size effects. Most importantly, this qualitative assessment mainly focused on macrovascular pathology and did not include a detailed assessment of the brain microvascular burden.
Considering the discrepancies of neuroimaging studies about the link between lesion location and PSD occurrence and our negative data in autopsy series, one could argue that psychological rather than neurological factors are the main determinants of PSD. Such statement may, however, be erroneous or at least premature. Paralleling the debate on the origin of PSD, the possible role of small macrovascular and microvascular lesions in triggering depressive episodes has received further support in the late 90s when several groups described a possible deleterious effect of MRI hyperintensities on affective regulation. MRI hyperintensities referred to as leucoencephalopathy, leucoaraïosis and subcortical encephalomalacia have been described both in depression and Alzheimer’s disease  but also in normal brain aging . They correspond to small vascular or microvascular pathologic changes such as demyelination, cortical microinfarcts and lacunae . Usually associated with an increased vulnerability to cognitive impairment and dementia [13,66], they gained significance in the pathogenesis of depression after the introduction of the “vascular depression hypothesis” by Alexopoulos et al. . Based on the comorbidity of depressive syndromes with cerebrovascular lesions and cardiovascular risk factors these authors postulated the existence of small macrovascular and microvascular lesions affecting brain circuits responsible for mood control. Simultaneously, Krishnan et al. described the clinical profile of vascular depression based on MRI observation of punctate hyperintensities of the subcortical gray matter, deep white matter and periventricular regions corresponding to vascularlesions . Patients classified as having vascular depression on MRI ratings were more likely to be elderly, nonpsychotic and to have a late-onset depressive disorder. As concisely stated by Newberg et al.  “Overall, the concept of vascular depression emphasizes the etiopathophysiologic relationship between cerebrovascular lesions and depression while the concept of PSD emphasizes the temporal relationship between the two”. In terms of etiopathogenesis, a progressive change of focus from macroinfarcts to small vascular lesions has been already introduced by the concept of “vascular cognitive impairment” in another highly debated field, that of the pathological basis of vascular dementia. Our recent work in this field clearly demonstrates that some microvascular lesions such as cortical microinfarcts, periventricular demyelination and thalamic and basal ganglia lacunae may have an unexpected negative impact on cognitive functions (for review see ). In a recently published article, Brodaty et al. proposed the same change of focus for PSD stating that “depression after stroke may be related to cumulative vascular brain pathology rather than side and severity of single stroke” . In this perspective, patients with chronic vascular burden could develop a PSD with a lower rate of spontaneous recovery and higher risk of chronic depression compared to an early PSD episode with spontaneous remission more probably associated with the location of macrovascular stroke lesion. Further neuropathological studies in large and clinically well characterized series are urgently needed to obtain a definite answer about the pertinence of the neuroanatomical model of PSD.
This study is supported by grants form the Jérôme Tissières Foundation (PG) and NIH grants AG02219 and AG05138 (PRH).