Neural Indicators of Fatigue in Chronic Diseases: A Systematic Review of MRI Studies

While fatigue is prevalent in chronic diseases, the neural mechanisms underlying this symptom remain unknown. Magnetic resonance imaging (MRI) has the potential to enable us to characterize this symptom. The aim of this review was to gather and appraise the current literature on MRI studies of fatigue in chronic diseases. We systematically searched the following databases: MedLine, PsycInfo, Embase and Scopus (inception to April 2016). We selected studies according to a predefined inclusion and exclusion criteria. We assessed the quality of the studies and conducted descriptive statistical analyses. We identified 26 studies of varying design and quality. Structural and functional MRI, alongside diffusion tensor imaging (DTI) and functional connectivity (FC) studies, identified significant brain indicators of fatigue. The most common regions were the frontal lobe, parietal lobe, limbic system and basal ganglia. Longitudinal studies offered more precise and reliable analysis. Brain structures found to be related to fatigue were highly heterogeneous, not only between diseases, but also for different studies of the same disease. Given the different designs, methodologies and variable results, we conclude that there are currently no well-defined brain indicators of fatigue in chronic diseases.


Introduction
The perception of fatigue is subjective, due to the interchangeable use of "tiredness" and the clinically pertinent experience of fatigue. There is no general consensus on the definition of fatigue. Nevertheless, for clinical purposes, it has been defined as an overwhelming feeling of physical and/or mental tiredness, along with a lack of energy which constrains the daily activities of the patient [1].
As opposed to weakness related to primary muscle disorders, fatigue has been conceived as a central nervous system event, considering that it remains after resting or sleeping. It may involve lack of attention, decline of executive and cognitive functions, difficulties in information processing or loss of productivity [2,3].
This symptom is often reported as one of the most burdensome, and the main cause of decrease in quality of life within chronic diseases, such as Rheumatoid Arthritis (RA) [4,5], Systemic Lupus Erythematosus (SLE) [6], primary biliary cirrhosis [7], or Human Immunodeficiency Virus (HIV) [8] among others.

Methods
To ensure high quality reporting, the current systematic review adheres to the recommendation for systematic reviews of the PRISMA statement [41]. The literature search was based on a predefined list of search terms, inclusion, exclusion and quality criteria. The approaches used for data collection, extraction of characteristics and analysis of results are fully described.

Information Sources
A comprehensive search strategy was developed for studying fatigue within chronic diseases through neuroimaging techniques. Experimental and observational studies, with prospective or retrospective data collection, cross-sectional and longitudinal design, cohort-, case control-and randomized-designed studies were included in this review.
The review was carried out by searching electronic databases and through consultation with experts in the field and a medical librarian specialist. Studies were identified through Medline (Ovid Medline (R) 1946 to 22 April 2016), Embase (1974 to 27 April 2016), PsycInfo (1806 to 22 April 2016) and Scopus. All studies were included, without any language or date restriction. The last search was run on 28 April 2016.

Search
The following search terms were used for all databases: (fatigue OR lethargy OR weakness OR weariness OR debility OR enervation OR exhaustion OR faintness OR feebleness OR heaviness OR languor OR lassitude OR listlessness OR burnout OR fatigation OR tiredness OR overtired* OR asthenia) [title] AND ("magnetic resonance" OR MRI OR fMRI OR neuroimaging OR "diffusion tensor imaging" OR "voxel-based morphometry" OR "voxel based morphometry") [title/abstract] AND (brain OR cerebr* OR neur*) [title/abstract] NOT ("multiple sclerosis" OR MS) [title].

Study Selection
In the search of neural changes related to fatigue, the SR was focused on the study of mental or cognitive fatigue, discarding physical fatigue due to muscular disorders. The non-invasive and non-ionizing nature, high resolution, great contrast between tissues and ability to detect subtle changes, makes the MRI one of the favourite tools in current clinical and research practice [42]. This SR focuses on MRI studies, excluding other neuroimaging techniques such as Computed Tomography (CT) or Positron Emission Tomography (PET). Another fact to take into account is that primary central nervous system (CNS) diseases generally induce gross structural abnormalities on MRI. Such abnormalities may mask some of the more subtle neural correlates (measured by more advanced MR protocols), which are expected to provide insight in the context of fatigue. Therefore, CNS diseases such as Parkinson's disease, post-stroke syndrome, multiple sclerosis or amyotrophic lateral sclerosis among others were excluded.
As opposed to epidemiological studies, most of the neuroimaging studies make use of small sample sizes. However, good statistical power can still be achieved with sample sizes between 10-15 subjects [43,44]. In this review, studies with a minimum of 10 subjects were included.
Therefore, studies were included if they met the following criteria: (1) observational or interventional study design; (2) investigation of mental or cognitive fatigue; (3) application of structural or functional magnetic resonance imaging and (4) study of chronic disease. Studies were excluded if they: (1) study muscular or physical fatigue; (2) use non MR techniques (i.e., Positron Emission Tomography (PET), Computed Tomography (CT), etc.); (3) study primary CNS diseases (as Multiple Sclerosis (MS), Amyotrophic Lateral Sclerosis (ALS), stroke, etc.); (4) had sample sizes with less than 10 participants and (5) were single cases. After deletion of duplicates, the titles and abstracts of all records were reviewed.

Synthesis of Results
Brain correlates of fatigue were extracted for each of the studies. The outcomes were analysed according to: (1) illness and (2) neuroimaging techniques. Fatigue inducing tasks were reported for functional neuroimaging studies.

Quality Assessment
Those studies accomplishing the inclusion and exclusion criteria were evaluated for quality assessment using an established 10-item quality score [45]. The list of questions are: (1) Does the study have a clearly defined research objective? (2) Does the study adequately describe the inclusion criteria? (3) Does the study adequately describe the exclusion criteria? (4) Does the study report on the population parameters/demographics? (5) Does the study report details on assessment of fatigue? (6) Does the study provide details of imaging protocol? (7) Does the study provide a proper control group? (8) Does the study apply proper statistical analysis? Correction for multiple comparisons? (9) Does the study adequately report on the strength of the results? and (10) Do the authors report on the limitations of their study?

Study Selection
The search of Medline, Embase, PsycInfo and Scopus databases provided a total of 1202 citations. After deleting duplicates, 727 papers remained. Of these, a total of 655 studies were discarded after reviewing titles and abstract, as they did not meet the inclusion or exclusion criteria. Two hundred and forty-seven were discarded for reporting single cases. Ninety-five of the studies were reviews, book chapters, did not have an abstract or were theoretical reports. a further 169 studies were excluded for being focused on muscular or physical fatigue. From the remaining papers, 38 more were discarded, as they studied fatigue in healthy participants only, and another 59 studies because they investigated non related illnesses (brain cancer, ALS, MS, infarction, stroke, clinically isolated syndrome, stress-related exhaustion, cocaine addiction or scleroderma). Twenty-two more studies were excluded for not studying fatigue, 7 for not using neuroimaging techniques, 17 more for using non related imaging techniques (PET, CT, etc.), and 1 because it focused on animal models. Finally, 72 remaining papers which met the inclusion and exclusion criteria were selected for full-text reading.
Finally, 26 studies remained [39,40,. Figure 1 shows the process followed in this review.  Table 1 summarizes the study characteristics pertinent to our research: demographic features, study design, duration of follow-up, task during the fMRI, fatigue assessment, imaging modality and statistic methods.

Synthesis of Results
Brain correlates of fatigue were extracted for each of the studies, and they are summarized in Table 3.

[40] CFS
Negative correlation between fatigue and fC between salience network and posterior cingulate cortex. Negative correlation between fatigue and fC between resting state network and anterior midcingulate cortex.
100 [106] CFS Positive correlation between fatigue and connectivity between posterior cingulate cortex and dorsal anterior cingulate cortex. 100 [107] CFS No significant findings. 100 [108] CFS Negative correlation between fatigue and activation in the right globus pallidus. 100 [109] CFS Negative correlation between fatigue duration and WM volume in the midbrain. 90   Neural correlates of fatigue in ankylosing spondylitis were mainly found in the parietal lobe, specifically in the inferior parietal sulcus [93] and postcentral gyrus [93,97], and the basal ganglia (caudate nucleus and putamen [97]). Other regions such as superior temporal polysensory area and insula were correlated with fatigue reduction after anti TNF-α therapy. Fatigue scores [97] were negatively correlated with GM volume in the dorsal and ventral attention networks, and positively correlated in the executive control network. In the case of granulomatosis with polyangiitis (GPA) [96], higher activation during a fatigue task in the medial frontal gyrus, paracentral lobule, globus pallidus and thalamus was found, while Basu et al. suggested fornix and cingulum play an important role in GPA-related fatigue [102]. With regard to chronic fatigue syndrome (CFS), results between studies were heterogeneous. Most of the significant correlates of fatigue were found in the frontal lobe (dorsolateral prefrontal cortex [94,110], dorsomedial prefrontal cortex [110], medial prefrontal cortex [111], lateral prefrontal cortex [91], medial superior frontal gyrus [106], and bilateral supplemental and premotor region [112]) and the limbic system (parahippocampal gyrus [39,110], anterior cingulate [39,40,106,111], posterior cingulate [39,40,110], and thalamus and hippocampus [39]). Other areas related to fatigue were found in the parietal lobe (supramarginal gyrus and postcentral gyrus [39], and superior parietal region [112]), midbrain [109] and globus pallidus [108]. Cope et al. found white matter lesions in a minority of all groups [114], while Zeineh et al. and De Lange et al. reported no significant findings [107,113]. Studies in other diseases reported varying results. Regarding cancer, dorsomedial prefrontal cortex [98] and executive network [99] were found to be related with fatigue. In [100] no significant findings were reported. Other correlates with cognitive fatigue were found in globus pallidus in the case of cirrhosis [101], ventral striatum in Hepatitis C [104] and inferior fronto-occipital fasciculus in Gulf War Illness [103], whereas in postpoliomyelitis (PPS), putamen, reticular formation and medial leminiscus was related with fatigue [92], while Trojan et al. did not find significant results [95]. Finally, basal ganglia was related to fatigue in HIV patients [105], although the quality of this study was judged to be only fair.

Summary of Evidence
In the present article, we have systematically reviewed the available MRI studies in chronic diseases where fatigue is a common burden. In the case of ankylosing spondylitis (AS), a form of arthritis, fatigue is considered as a significant symptom [21,115]. This symptom is so prevalent in AS that it has been suggested to be considered as an independent domain from other symptoms related to the disease, such as pain or impairment [116]. This review compiled two AS studies reporting fatigue-related structures [93,97]. Some of these structures, such as insula, dorsal and ventral attention network and the executive control network are related to cognitive functions, while other structures such as the primary sensory cortex, inferior parietal sulcus and superior temporal polysensory area are involved in sensory experience. Cancer-related fatigue is the most prevalent and debilitating symptom reported by patients [117]. Approximately 90% of patients treated with radiation and 80% of patients treated with chemotherapy suffer from fatigue [118]. In this review, we gathered three studies regarding cancer-related fatigue, although just two of them, relating to breast cancer, stated significant findings. Blesch et al. reported that 99% of breast cancer patients experienced some level of fatigue [119]. Dorsomedial prefrontal cortex and the executive network were found to be related to fatigue, being both structures involved in cognitive functions. With regard to primary biliary cirrhosis, a long-term liver disease, long term fatigue affects approximately 68-85% of the patients [7,120], considered as the worst or one of the worst symptoms in around half of them [120]. Only one paper studying fatigue correlates in PBC accomplished the inclusion criteria of this review. This paper reported the globus pallidus structure, associated with the regulation of voluntary movement, to be related with fatigue. Granulomatosis with Poliangiitis (GPA) is a rare multisystem autoimmune disorder of undetermined etiology. Although fatigue prevalence in GPA has not been reported, studies within other rheumatic disorders exhibited rates between 60% and 90% [121][122][123]. In this review, two studies regarding the GPA-related fatigue were presented. Basu et al. reported several brain regions related to fatigue, such as thalamus, paracentral lobule and the medial global pallidus [96]. These areas are mainly linked to sensory and motor coordination. Same authors found the fornix and cingulum, both part of the limbic system, involved in the experience of this symptom [102]. Gulf War Illness is a chronic and multisymptomatic illness that affects military veterans of the Gulf War. This disorder is characterized by a wide range of symptoms, including significant fatigue, muscle pain and cognitive issues among others [124]. The experience of permanent fatigue is frequently described by Gulf War veterans [125][126][127][128]. Kelsall et al. reported that up to 66% of Gulf War patients suffer from mild to severe fatigue [125]. Furthermore, between 1.6 and 5.1% of these patients met the chronic fatigue syndrome criteria [126,129,130]. This review found only one study related to fatigue in Gulf War Illness [103]. The study found a correlation between fatigue and axial diffusivity in the inferior fronto-occipital fasciculus. This brain structure is related to the integration of auditory and visual association. Hepatitis C is caused by the hepatitis virus which, in turn, leads to the swollenness of the liver. a study found fatigue to be present in two thirds of patients with chronic Hepatitis C [131]. Another study reported that 53% of Hepatitis C patients suffer from fatigue. In 17% of these patients, fatigue was so severe that it led to activity impairment [132]. The only study reported in this review about Hepatitis C found the development of fatigue related to the ventral striatum [104], which mainly intercedes in reward cognition, reinforcement and motivation. HIV is a virus which attacks the immune system, which is the body's natural defence against infections and diseases. Prevalence of fatigue in HIV has varying rates according to the stage of the disease, ranging from no fatigue in the early stage of the disease, to almost 80% in AIDS patients [8,[133][134][135][136][137][138]. Just one study regarding HIV met the inclusion criteria of this review and hence, was assessed [105]. Here, basal ganglia was associated with fatigue. Patients who suffered from paralytic poliomyelitis can develop postpoliomyelitis syndrome (PPS) years or even decades later. PPS is a disorder where patients suffer from a generalized fatigue, generally reported as the most burdensome symptom of the disease [139]. Approximately 66% to 89% of PPS patients suffer from high levels of fatigue [140][141][142]. Two studies regarding PPS-related fatigue were assessed. However, just one of them found significant findings. Reticular formation, putamen and medial leminiscus were indicated as possible correlates of fatigue. These structures are thought to be involved in several functions, such as behavioural arousal and consciousness, regulation of movements and somatosensation from the skin and joints. Chronic fatigue syndrome is a complex disorder whose most common symptom is extreme fatigue. In the UK, the prevalence of CFS has been rated between 11% [143] and 15% [144] per 100,000 people. This disorder has been widely studied and it has been the most reported illness in this article. a total of 13 CFS studies related to fatigue were gathered. The majority of them found areas related to executive and cognitive functions, memory and perception. The brain structures found to be related to fatigue in the studies compared were highly heterogeneous, not only between diseases, but also for different studies within the same diseases. The most common structures were the frontal lobe, parietal lobe, limbic system and basal ganglia. These structures are associated with attention, memory, planning, integration of sensory information and learning.
Longitudinal studies, as opposed to cross-sectional studies, allow tracking of the same subjects over time, removing confounders such as cultural differences, age, etc. Therefore, they offer more precise and reliable indicators. In addition, they can give information about accumulative processes. What is more, they may allow the prediction of future changes in fatigue by assessing baseline factors. In spite of this, just seven out of 26 studies were longitudinal [91,93,99,100,104,105,114], and six involved an interventional procedure. Within these six interventional studies, one was low quality and only employed MRS [105], one only recorded structural images and did not include a proper control group [104], and another only recorded structural images [93]. The fact that longitudinal designs allow follow-up of the patient during an intervention procedure, offering more precise indicators of the population changes, and their potential as prediction models, make them the best strategy to follow in the search of neural correlates of fatigue. Therefore, longitudinal neuroimaging studies following a fatigue treatment would be of inestimable worth in the study of this symptom. sMRI, fMRI, DTI and FC studies identified significant neural correlates of fatigue, demonstrating that they are very useful techniques in the search for the processes responsible for fatigue within chronic illnesses. The integration of these techniques may be helpful in the search of biomarkers, rather than using them individually [145]. However, none of the studies in this review integrated all of them. sMRI, fMRI and DTI were integrated in [98] and sMRI, fMRI and FC in [106], both of them with no follow-up. Other studies have combined two of the techniques; sMRI and fMRI [96,108,[111][112][113], sMRI and DTI [97,102,107], or sMRI and FC [40,72]. Integration of modalities is an attractive strategy to follow in the development of a comprehensive map of the pathophysiological brain networks of fatigue. This multimodal approach has demonstrated a more comprehensive understanding of brain changes in disorders such as amyotrophic lateral sclerosis [146], schizophrenia [147][148][149][150][151][152][153][154][155][156][157], bipolar disorder [158][159][160][161], characterization of tumours [162], traumatic brain injury [163], Parkinson's disease [164], psychosis [165,166], Alzheimer's disease [167,168] and mild cognitive impairment [169][170][171]. Therefore, it should be expected that the integration of techniques may help to elucidate further brain mechanisms of fatigue.

Limitations
This systematic review aimed to perform a comprehensive search of all studies employing MR techniques in the investigation of neural indicators of fatigue. Although a great number of references were gathered, it is still conceivable that other related papers have been overlooked. Furthermore, the inclusion criteria set a lower limit of studies with more than 10 subjects. This limit was based on fMRI which tend to have lower sample sizes than sMRI studies. This may mean that lower power, and therefore lower quality, sMRI studies may have been included than would otherwise be the case with a larger sample size threshold. We acknowledge the diversity in the way that fatigue was measured in the studies examined. Furthermore, studies employing fMRI make use of different tasks to induce fatigue. These facts increase the risk of bias and will inevitably contribute to a disparity in results. Finally, we acknowledge that the study of brain correlates by means of neuroimaging is insufficient to offer a comprehensive explanation of the mechanisms underlying fatigue. The integration of other types of biomarkers, such as biochemical ones, may provide further understanding of the phenomenon.

Conclusions
We have found that studies searching for neural indicators of fatigue within chronic diseases are of variable design and quality. Regarding the fatigue indicators, we found that there are not well-defined neural correlates of fatigue in any chronic diseases so far. Different designs and methodologies for the same illness offered different results. From this, we conclude that there are no common correlates of fatigue across chronic diseases. According to the employed design, it seems that the best strategy in the search of neural correlates of fatigue would be to integrate different neuroimaging techniques in a longitudinal study, with a fatigue alleviating intervention. Such an approach could be a great asset to unravel the neural mechanisms of this burdensome and neglected symptom.

Conflicts of Interest:
The authors declare no conflict of interest.