Background
Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a complex and debilitating chronic illness. Patients with ME/CFS experience overwhelming fatigue that severely impacts their quality of life, with 25% of ME/CFS patients chronically bedridden or house-bound. However, there is no known underlying disease process, no biologically based treatment, and no objective diagnostic criteria for ME/CFS.
Although the aetiology of ME/CFS remains unclear, the well-documented neurological symptoms, sleep and autonomic dysfunction, abnormalities in cognition, and altered sensory and pain perception, suggest that abnormal brain function assumes a crucial role in the underlying disease process [
1]. Given the aforementioned constellation of symptoms, ME/CFS has been classified as a neurological disease (ICD-10 G93.3; ICD-11 8E49) by the World Health Organization.
Researchers have used multiple neuroimaging techniques to explore structural, neurochemical, and functional brain changes in patients with ME/CFS since the 1990s. However, the results have been mixed for various reasons, such as small sample sizes and ill-defined disease classification. In short, it is recognised that abnormal brain function plays a critical role in ME/CFS. However, little is known about what abnormalities are consistently observed across multicentre studies and to what extent the observed abnormalities are reproducible. Therefore, there is a pressing need to summarise the consistency of observed abnormalities across studies in this field so as to inform future directions of neuroimaging research in ME/CFS.
This study systematically reviewed the neuroimaging studies of ME/CFS from January 1988 to July 2018 to specifically address the following three questions. (i) What structural or functional differences were consistently (i.e. by two or more research teams) observed in radiological brain imaging studies of ME/CFS? (ii) What brain measures were consistently associated with ME/CFS symptom phenotypes? (iii) What differences or symptoms and associated brain measures were inconsistently reported? The aim of this study was to facilitate a better-informed hypothesis of ME/CFS aetiology based on consistent findings, to reconcile some inconsistent findings, and to identify a future research focus.
Methods
The objectives and analyses methods of this retrospective systematic review were specified in advance, documented and registered in the PROSPERO database [
2]. The registration was submitted after the initial search but before the start of data extraction (examining contents of the selected papers).
Data Source and Searches
An investigator (ZS) searched electronic databases, including Ovid MEDLINE, PubMed (NCBI), and Scopus, to identify relevant articles published between January 1988 (year of the first ME/CFS case definition) and July 2018. Cross-referencing supplemented the search during reviewing of full-text articles.
The search field included the title, abstract, and keywords with any combination of a population condition and an exposure condition as the search condition. The population condition included chronic fatigue syndrome, myalgic encephalomyelitis, systemic exertion intolerance disease, and their abbreviations. The exposure condition included neuroimage, neuroimaging, magnetic resonance imaging (MRI), functional magnetic resonance imaging, diffusion tensor imaging, magnetic resonance spectroscopy, arterial spin labelling, diffusion-weighted imaging, computed tomography, position emission tomography (PET), single-photon emission computed tomography (SPECT), ultrasound, and their abbreviations.
Study Selection
We included the following studies: (i) peer-reviewed English language scientific journal articles; (ii) ME/CFS definitions including systemic exertion intolerance disease proposed by Institute of Medicine [
3], International Consensus Criteria [
4], Paediatric ME/CFS definition [
5], Canadian consensus criteria [
6], Centers for Disease Control and Prevention (CDC) Holmes criteria [
7], London criteria [
8], Reeves criteria [
9], CDC Fukuda criteria [
10], Oxford criteria [
11]; iii) neuroimaging techniques of MRI, PET, SPECT, computed tomography (CT), ultrasound.
We excluded the following studies: (i) non-peer reviewed articles; (ii) articles not written in English; (iii) review, hypothesis without experimental data, editorial, and opinion articles, case reports and protocols, since this study is to summarise consistent observations; (iv) the modalities of electroencephalography, magnetoencephalography, and functional near-infrared spectroscopy because these techniques do not produce images and the scope of this study was limited to radiological brain imaging modalities.
Two investigators (ZS and LB) independently reviewed the abstract of each study to determine eligibility. Any disagreement was resolved by discussion to consensus.
Risk of bias evaluation
The risk of bias for each study was evaluated using a tool adapted from QUADAS-2 [
12], a quality assessment tool of diagnostic accuracy studies. This study reviewed observational studies; thus, the QUADAS-2 tool was adapted in 4 domains: (i) patient selection, (ii) index test, (iii) control standard, and (iv) control of confounding factors.
The signalling questions for the domain of patient selection included: (i) whether a consecutive or random sample of patients was enrolled; and (ii) whether the study avoided inappropriate exclusion.
The signalling questions for the domain of index test included: (i) whether a full brain imaging protocol was reported; and whether any potential bias existed in the protocol; ii) whether a significance threshold was pre-specified and whether the multiple comparison correction was appropriately handled.
The signalling questions for the domain of control standard included: (i) whether controls have similar age as patients or were controlled for age; (ii) whether the control group had the same gender ratio as the patient group or were controlled for gender ratio and (iii) whether controls were selected or recruited from the same or a similar population.
The signalling questions for the domain of control of confounding factors included: (i) whether psychiatric disorders including anxiety and depression were screened or controlled for; (ii) whether obesity was screened for, or body mass index was controlled for and (iii) whether neurologically active medication was screened.
Each domain of a study was summarised as low risk (no signalling question violated), moderate risk (one signalling question violated), or high risk (more than one signalling questions violated). If a study had one or more domain with high risks, the study results were interpreted with caution and relevant to the violations.
Synthesis of results
The following contents were extracted from each article: author details, affiliations, country, ME/CFS diagnostic criteria, number of participants, demographics, aim of paper, image modality, analysis method, and observations.
Extracted findings were classified as frequent, consistent, single, and inconsistent observations. The frequent, consistent, and single observations were defined as same or similar findings reported by five or more research teams, by two to four teams, and one research team without inconsistent finding reported, respectively. The inconsistent observations were defined as findings reported by one or more research teams with contradictory findings reported. Studies that had one common author and from the same institution were considered as studies from the same research team.
Discussion
The synthesis of the reviewed studies generated two frequent observations (larger recruitment of brain regions during cognitive tasks and abnormalities in the brain stem), three consistent observations (sluggish fMRI signal response to tasks, reduced serotonin transporters, and regional hypometabolism), and five single observations (regional T1 and T2 spin-echo intensity differences and abnormal clinical correlations, white matter microstructural changes, regional metabolite abnormalities, associations between intracranial compliance and severity of orthostatic intolerance symptoms, and associations between cerebral vascular control and skeletal muscle pH).
Larger recruitments of brain areas could be potentially explained by the sluggish BOLD response to tasks, which was observed consistently in two fMRI studies [
29,
40]. The BOLD response is determined by neurovascular coupling (NVC) which consists of an initial feedforward mechanism of glutamate activation of a Ca
2+ dependent signalling pathway in both neurons and astrocytes to release vasoactive factors to increase local blood flow, and secondary feedback driven by metabolism [
77]. Several recent studies have demonstrated that Ca
2+ mobilisation is impaired in ME/CFS from genetic [
78,
79], molecular biological [
80], and electrophysiological aspects [
81]. These observations would suggest, it is plausible that glutamate-Ca
2+ NVC pathways in patients with ME/CFS may be impaired because of reduced TRPM3 activity and Ca
2+ mobilisation.
Impaired NVC could have multiple detrimental effects on the brain. Firstly, deficits in matching local CBF to neuronal activity may lead to hypoxia. Secondly, inadequate energy supply to activated neurons may potentially lead to energy failure of ionic pumps. Thirdly, oxidative stress that impairs blood vessel endothelial cell function may cause blood–brain barrier (BBB) breakdown and neuroinflammation. These potential detrimental effects are generally diffuse and multifocal in the brain, leading to reduced cognitive efficiency in ME/CFS and manifesting as more extensive functional recruitment during cognitive task performance. A similar manifestation has been observed in patients with other neuropathologies that result in diffuse damage such as traumatic brain injury [
82]. Reduced cognitive efficiency may be the neurological underpinning of the observation that patients with ME/CFS often report subjective changes when performing cognitive tasks despite normal objective performance. Impaired NVC in ME/CFS is further supported by structural MRI observations of regional white matter loss in the left inferior fronto-occipital fasciculus [
47] and the brain stem [
30‐
34], both being brain regions sensitive to hypoxia [
83,
84].
Dysfunction of NVC is closely related to neuroinflammation, which is another neurological disease process proposed to underly ME/CFS [
85,
86]. Dysfunction of NVC may trigger neuroinflammation. The typical NVC dysfunction features of inflammatory and angiogenic activation triggered by hypoxia may jointly result in a leaky BBB, brain oedema, and neuronal dysfunction and damage [
87]. Vice versa, neuroinflammation and infection may result in NVC dysfunction, whereby, activation of microglia, especially the formation of the M2 phenotype, exacerbates damage to the BBB. Neuroinflammation molecular patterns can interact with neurotransmitters, increase procoagulant activity and thrombosis, and cause endothelial injury and damage, resulting in NVC dysfunction [
88]. Although there is no direct evidence to support either dysfunction of NVC or neuroinflammation as the primary aetiology factor of ME/CFS, we posit that dysfunction of NVC constitutes a critical underlying disease process in ME/CFS.
The synthesis of articles in this review also generated three consistent observations: sluggish BOLD response to task [
29,
40], reduced serotonin transporter, and regional hypometabolism. The sluggish BOLD response is consistent with NVC dysfunction. Two studies observed reduced serotonin transporter in ME/CFS [
41,
42]. Serotonergic imbalance is a prominent feature in major depression [
89] and a large percentage of patients with ME/CFS experience depression. These observations may help explain the well documented depressive symptoms in ME/CFS. The evidence of regional hypometabolism [
37,
43] is aligned with an NVC dysfunction. The hypoxia, oxidative stress, and energy deficiency caused by NVC dysfunction may lead to reduced glucose uptake and result in a vicious cycle [
90]. The inconsistent observations of volumetric differences in ME/CFS are consistent with the notion that the detrimental effects of NVC dysfunctions are diffuse and multifocal in the brain. Different studies may generate positive or negative findings depending on the disease severity and patient selection.
This review noted ME/CFS as an under-recognised disease and that brain imaging studies of ME/CFS suffered from small sample size. A small sample size not only reduces the chance of detecting a true effect but also reduces the likelihood that a statistically significant result reflects a true effect [
91]. As such, we argue that there is a pressing need to establish a collaborative neuroimaging databank for ME/CFS in addition to a proposed ME/CFS biobank [
92]. Notably, more than 80% of reviewed studies did not control for lifestyle differences. Physical activity may impact brain structures and functions [
93,
94]. Given that patients with ME/CFS have to reduce their daily activities because of fatigue, objective measurement of physical activity in both patients and controls is essential to differentiate between brain differences that are unique in ME/CFS and those due to reduced physical activity.
This review was unable to perform a formal meta-analysis because: (i) There were few studies on neuroimaging of ME/CFS. (ii) Each study used an ad hoc analysis method. (iii) Different aspects were investigated in different studies. Nevertheless, this systematic review of neuroimaging in ME/CFS provides a snapshot of ME/CFS brain imaging studies to date and identifies deficiencies that need to be accounted for in future studies.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.