Lesion location

The relationship between lesion location and PSF remains controversial.9 18 48 51–53 53–55 The consensus is that lesion location is not determinant of development of fatigue9 11 30 41 49 56; however, some studies suggest there may be a higher incidence of fatigue in subcortical strokes when compared with cortical strokes.18 Subcortical strokes may be further broadly classified as basal ganglia and cerebellar strokes, with basal ganglia strokes more likely to give rise to fatigue possibly due to disturbances in the limbic–motor integration networks.48 57 There is not enough information available about detailed distribution of cortical lesions and the incidence of fatigue; however, any lesion to attention networks could result in fatigue, as poor attention may be a key element of high effort, a feature of fatigue, as described later in the review.58 Posterior cerebral artery strokes resulting in thalamic and brainstem lesions have previously been associated with high levels of fatigue. High fatigue in posterior artery strokes may be a result of poor attention mediated by lesions in attentional networks including the ascending reticular activating system and the lenticular, hypothalamic and thalamic nuclei.18 50 Some authors investigated the difference in incidence of fatigue based on type of lesion. Patients with large vessel stroke experienced greater fatigue than small vessel involvement,59 while other studies report an association with cerebral microbleeds.60 Those with stroke reported greater fatigue than transient ischaemic attack (TIA) despite minimal impairment.14 An association was seen between fatigue and white matter lesion.61 These findings suggest that the presence of a brain lesion rather than poststroke disability or vascular risk factors might be important in the aetiology of PSF. Furthermore, small vessel disease and development of PSF remains controversial. Whether lesion location significantly influences the development of PSF is still an open question, and future studies will benefit from a systematic anatomical correlation of lesion location with fatigue incidence and severity.

Proinflammatory response in the acute phase and subsequent fatigue

In a systematic review published in 2012 and based on examination of the literature until the last trimester of 2010, Kutlubaev and colleagues51 conclude that biological factors probably play a major role in PSF. The mechanisms remain uncertain when looking at possible involvement of stroke-induced alterations in hypothalamic–pitutary–adrenal axis and in neurotransmitter systems. In a more recent systematic review, Ponchel et al 62 propose that inflammatory factors appear to play a role, although it still needs to be confirmed. Inflammation is well known to be associated with fatigue. Several mechanisms have been proposed for this association (64). Inflammation is mediated by the production and release of proinflammatory cytokines that take place in the brain during stroke because of the peripheral blood mononuclear cells and is amplified by resident macrophages, known as microglia.63 In addition, inflammation can propagate from the brain to the periphery and vice versa by various communication mechanisms.64 Brain proinflammatory cytokines can act on neurotransmitters such as serotonin and dopamine by several mechanisms including alterations in synthesis, packaging in microvesicles, release and reuptake. These mechanisms have been mainly studied in the context of inflammation-induced depression.65 Inflammation can also induce oxidative stress which, if anti-inflammatory and antioxidant processes are deficient, can lead to neurodegeneration that affects primarily dopaminergic neurons in the mesostriatal and mesolimbic pathways.66 Inflammation can also activate the kynurenine (KYN) pathway that, in the context of activated microglia and ongoing neuroinflammation, favours the formation of neurotoxic KYN metabolites to the detriment of neuroprotective KYN metabolites and therefore enhances further the risk for neurodegeneration. In this section we will examine the evidence supporting the involvement of each of these mechanisms in stroke-associated fatigue.

Inflammation and fatigue

There is cumulative evidence for an inflammatory reaction in acute ischaemic stroke (AIS), indicating important interactions between the nervous and immune systems.67 68 Cytokines are upregulated in the brain after stroke and are expressed in invading immune cells and in glial cells and neurons.69 Brain inflammation caused by the stroke episode does not remain local. As mentioned in the previous paragraph, it propagates to the periphery. Peripheral inflammation preceding and following a stroke episode has potent modulatory effects on the pathology of stroke. As inflammation measured by peripheral concentrations of C reactive protein and cytokines in serum plasma depends on both pre-existing inflammatory clinical status related to, for example, atherosclerosis and cardiovascular disease and propagation of brain inflammation to the periphery, the results of clinical studies on the relationship between the inflammatory response following AIS and infarct volume70–78 and stroke subtype73 are likely to be very variable. The role of the cytokines involved thus remain unclear, and whether postischaemic inflammatory responses are deleterious or beneficial to brain recovery is still a matter of discussion.79 80 A study, in which the serum levels of 13 cytokines were evaluated in blood samples taken from 45 acute stroke patients (within 72 hours of stroke onset) and 40 healthy controls,81 showed for instance significantly higher levels of interleukin (IL)-1ra, IL-6, IL-8, IL-9, IL-10, IL-12, IL-18 and CXCL-1 in stroke patients. The authors concluded that this evidence suggested an early proinflammatory response and an early activation of endogenous immunosuppressive mechanism following stroke.

Considering fatigue as a dependent variable in the relationship between stroke and inflammation still adds to the complexity. Despite these difficulties, Ormstad and colleagues investigated the association between PSF, PSD, stroke type, infarct volume, laterality and the levels of various cytokines.82 83 PSF and PSD were measured using the FSS and Beck Depression Inventory, respectively, at 6, 12 and 18 months after stroke onset. The results indicated a role for the poststroke proinflammatory response in the appearance of PSF. The finding that IL-1β seems to be a predictor of PSF82 suggests fatigue after stroke could be part of what has been described as inflammation-induced sickness behaviour.84 Animal models of stroke support this possibility. For instance, Kunze et al 85 showed that experimental stroke in Lewis rats resulted in behaviour consistent with depression, while Wistar and Sprague-Dawley rats exhibited sickness-like behaviour including fatigue-like behaviour. The role of genetic factors has also been taken into consideration in clinical studies. For instance, Becker and colleagues86 showed that single neucleotide polymorphisms in the gene coding for the IL-1 receptor antagonist and the gene coding for Toll Like Receptor 4 (TLR4) were associated with high and low fatigue, respectively. However, the small size of the sample (n=39) precludes any definitive conclusion.

A possible neurochemical mechanism for cytokine-mediated fatigue, as discussed earlier, could be due to reduced capability for 5-hydroxytryptamine (5-HT) synthesis. However, the ineffectiveness of serotonin reuptake inhibitors on PSF87 may rule out serotonergic pathways as a source of fatigue. Hypodopaminergic activity induced by inflammatory cytokines could potentially be the source of cytokine-induced fatigue.88 89 An association between serum cytokines and PSD was not found, despite numerous reports of an association between PSD and PSF.83 Recent hypotheses suggest that glutamate might be the source of affective symptoms related to depression90 91 but may not mediate fatigue unrelated to depression.

Inflammation-mediated activation of the KYN pathway and fatigue

Inflammation can induce neurotoxicity by upregulating the indoleamine 2,3-dioxygenase (IDO) enzyme (which catalyses the rate-limiting step in the synthesis of KYN from tryptophan (TRP)) in multiple central and peripheral cell types.92 93 Activation of IDO can be measured by increased plasma/serum KYN over TRP ratio. KYN by itself acts as a ligand of the aryl hydrocarbon receptor, which generates regulatory T cells. The evidence in favour of such a mechanism in PSF is still weak. One study in particular reported reduced percentage of circulating regulatory T cells combined with increased systemic Th17 and proinflammatory cytokines and reduced anti-inflammatory cytokines.94 In addition, interferon-gamma, the cytokine primarily responsible for upregulating IDO, was not higher in stroke81 when compared with healthy controls. KYN produced in the brain in response to injury or transported from the periphery is further metabolised into the neuroprotective KYN metabolite kynurenic acid (KA) by KYN aminotransferase and the neurotoxic KYN metabolites 3-hydroxykynurenine and quinolinic acid by KYN monooxygenase. Activation of microglia, the main cellular source of proinflammatory cytokines in the brain during neuroinflammation, favours the neurotoxic pathway at the same time as it results in the extracellular release of glutamate. Whether microglial activation and the formation of neurotoxic KYN metabolites is involved in PSF is not clear. A few studies based on variations in plasma levels of KYN metabolites indicate activation of the KYN pathway could play a role in poststroke sequelae.95–98 Darlington et al 96 showed increased TRP catabolism after stroke and suggest that oxidative TRP metabolism contributes to oxidative stress and brain damage. Brouns et al 95 showed a correlation between KYN/TRP ratio and stroke severity and long-term stroke outcome that did not correlate with KA/3-hydroxyanthranillic acid ratio. Mo et al 98 showed an upregulation of IDO activation in ischaemic stroke. Ormstad et al 97 indicated an increase in TRP oxidation and reduced capability for 5-HT synthesis in the brain following AIS.

Ormstad et al 99, also investigated the mechanisms involved in PSD and PSF by studying the relationship between KYN pathway activity in the acute ischaemic phase and subsequent PSF and PSD in the stroke sample referred in their previous publication. Compared with the other neutral amino acids that compete with TRP for entry into the brain, plasma levels of TRP index were significantly lower in patients with an FSS score of ≥4 than in those with an FSS score of <4 at 12 months. However, in contradiction with the neurotoxic hypothesis, the serum levels of KA were significantly higher in patients with an FSS of score ≥4 than in those with an FSS score of <4 at 18 months. This indicates stroke patients with PSF might have a lower bioavailability of TRP in the acute stroke phase. In contrast to PSF, no predictors of PSD were found. These findings suggest that the immune response and IDO activation that follows AIS can predict PSF but not PSD. Interestingly, the TRP index did not correlate with fatigue at earlier time points (6 months) and similarly another study100 also showed no correlation between KYN/TRP ratio and fatigue in the very early stages poststroke (1 month). Emerging evidence appears to suggest that early activation of the KYN pathway does not manifest immediately as fatigue and depression but may have long-term consequences that has led to the proposal of a biopsychosocial model for PSF and depression.101

Neurophysiological and behavioural perturbations in chronic PSF

A particularly challenging aspect of investigating persistent fatigue is the difficulty in differentiating the causes from the effects of fatigue. Persistent symptoms such as fatigue result in significant behavioural and neurophysiological changes, which in turn may cause further fatigue resulting in a vicious cycle.101 Thus, identifying the mechanism that first establishes persistent fatigue is not straightforward. To date, investigations addressing behavioural and neurophysiological underpinnings of PSF have all been correlational studies, which for reasons stated below, do provide minimal insights into the direction of causality between behaviour, fatigue and physiology. However, causality remains to be confirmed by interventional studies aimed at alleviating fatigue.

Physical deconditioning is a common sequela after stroke and is believed to trigger PSF. Physical deconditioning results in fatigue and, subsequently, avoidance of physical activity. This further deteriorates deconditioning and can lead to more fatigue.102 A recent systematic review however failed to find an association between fatigue and any measures of physical activity or fitness.103 Both motor and cognitive deficits are seen in PSF. In the first year after stroke, in the absence of any obvious motor deficit, PSF was associated with poor attention and executive function as measured using phasic attention test and modified Stroop task.104 While fatigue correlated with attentional and executive function, neither correlated with lesion location. Another independent investigation105 confirmed poor executive and memory functions in high PSF in the first 6 months after stroke. They also showed that side and size of lesion did not correlate with presence of PSF. Despite subtle differences between the two investigations in terms of PSF definition and tests used to measure cognitive function, both investigations suggest that at 6 months poststroke, cognitive impairment correlates with fatigue with no obvious link to the lesion itself.

Several studies show that PSF is not correlated with motor deficits. However, a closer look at these investigations show that tests used to identify motor deficits are normally questionnaire-based scores of activities of daily living such as Barthel index, Rankin scale and National Institutes of Health Stroke Scale.25 28 34 46 56 86 104 105 Some studies use laboratory-based tasks to measure motor function such as action research arm test, nine-hole peg test (NHPT), grip strength106 and scales such as Fugl-Meyer.107 These measures, although useful to identify gross motor limitations relevant to daily life, are not accurate measures of motor impairment. A recent investigation showed that in evenly matched, minimal motor functional deficit high and low PSF groups, there was a significant reduction in ballistic movement speeds in the affected limb of the high fatigue group,108 while no difference in simple reaction times, attention and information processing speed was observed. The significantly slower ballistic speeds did not reflect on laboratory test scores, such as the NHPT times, possibly because tasks such as NHPT capture movement velocity and other features such as dexterity. Does this mean that, after all, PSF is related to motor impairment? The above investigation did not determine if those with slower movement speeds were capable of achieving higher speeds. It could be that in the task, one chose to move slower than their maximum speed. Those with slower movement speeds also reported heaviness of the affected limb,109 possibly a central sensory processing problem, which in turn may have led to choosing slower movement speeds. Those with slower movement speeds also exhibited low motor cortex excitability.106 It is unclear if these two findings have a direct relationship as it has previously been shown that motor cortex excitability does not encode movement speeds,110however, motor cortex excitability seems to be crucial in motor learning involving movement speeds.111 Low motor cortex excitability was also significantly correlated with high levels of PSF, but interestingly, voluntary activation, a measure of excitability of structures upstream of motor cortex such as secondary motor areas, was not significantly related to fatigue. Thus, PSF appears to be associated with behavioural108 and perceptual109 sensorimotor deficits with some underlying neurophysiological106 perturbations in the sensorimotor pathway.

Do any of the above findings suggest a causal role for motor and cognitive, behavioural and neurophysiological deficits in development of PSF? The attention and executive impairments seen in PSF are global. Structural and physiological changes remote to the site of lesion resulting in global attentional and executive impairment is a well-documented phenomenon after stroke. However, without further investigations, the correlations between fatigue and attention/executive impairment cannot be interpreted as being causal to fatigue; fatigue-related sensorimotor behavioural, physiological and perceptual findings are confined to the affected hemisphere. Should they be a consequence of fatigue, one would expect a more global effect on behaviour, hence there may be a causal role for sensorimotor alterations in development of fatigue. A small pilot interventional study reported significant benefits to using cognitive behavioural therapy and graded exercise112 for PSF. However, the effect size was not great enough to suggest that the intervention had targeted the underlying causal neural perturbations, but rather may have succeeded in compensating for some of the fatigue resultant behavioural changes. Moreover, the length of follow-up was not sufficient to determine fully the long-term effects of the intervention.

Active inference-based theoretical model of PSF20 113–115

Poor attention, slower processing speed, diminished memory, reduced movement speed and perceived limb heaviness: how might the above deficits give rise to fatigue, when no seemingly direct cause for fatigue such as sustained exertion exist? To answer this, we must first understand how sustained exertion might give rise to fatigue. With sustained activity, performance in the task drops and, importantly, subjects report fatigue, or more precisely, the need for higher effort to maintain task performance and at task failure, the inability to exert required effort to perform the task.116 It is also well established that task performance and report of fatigue are not correlated.117 Hence, the notion of higher effort is inextricably linked to fatigue, while drop in task performance itself may be seen as behavioural consequence of fatigue. A dissociation between task performance and fatigue, with a closer link between effort or ‘estimated action cost’ and fatigue, is indicative of a possible causal link between estimated action cost and fatigue. A recent review discussed how the low motor cortex excitability, reduced movement speed and perceived limb heaviness could underpin aberrant ‘estimated action cost’ or effort. Aberrant effort results in higher than normally required effort for everyday tasks.20 Persistent high effort for simple tasks gives rise to fatigue. This fits well with the clinical symptoms of PSF, such as fatigue without prior activity, fatigue that does not respond to rest and fatigue that limits everyday activities.

Effort is a perceptual inference that has its origins in intention (efferent information) and is modulated by feedback (afferent information). The active inference theory of sensorimotor control integrates efferent and afferent inputs to explain movement initiation and motor control113 and provides a framework within which effort, as defined above, may arise.118 The active inference theory of sensorimotor control postulates that the (efferent) output from corticomotor system is in the form of sensory (proprioceptive) predictions and (afferent) input from the somatosensory systems is in the form of sensory (prediction) errors. For a movement to be initiated, that is, sensory predictions to be fulfilled, ascending sensory errors must not be attended to, and to maintain status quo, sensory errors must be attended to. This property of altering precision of sensory errors is known as ‘sensory attenuation’. It has been postulated118 that in PSF, inference of high effort could be the result of poor sensory attenuation. In the context of muscle contractions, the inability to suppress ascending prediction errors is inferred by the brain as needing more than the estimated effort to perform the contraction. The motor cortex encodes prediction errors as shown by a suppression of sensory attenuation when motor cortex excitability is artificially reduced using non-invasive brain stimulation.119 The observed reduction in motor cortex excitability in stroke survivors with high fatigue106 further suggests poor sensory attenuation may be the mechanism by which fatigue arises. Evidence from other pathological conditions implicating dopaminergic systems in poor sensory attenuation115 combined with PSF-related molecular disturbances and preclinical work in muscle metabolism poststroke120 121 lend support to both central and peripheral mechanisms playing a role in poor sensory attenuation.