What lies beneath: White matter microstructure in pediatric myalgic encephalomyelitis/chronic fatigue syndrome using diffusion MRI

Recent studies in adults with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) suggest that changes in brain white matter microstructural organization may correlate with core ME/CFS symptoms, and represent a potential biomarker of disease. However, this has yet to be investigated in the pediatric ME/CFS population. We examined group differences in macrostructural and microstructural white matter properties, and their relationship with clinical measures, between adolescents recently diagnosed with ME/CFS and healthy controls. Forty‐eight adolescents (25 ME/CFS, 23 controls, mean age 16 years) underwent brain diffusion MRI, and a robust multi‐analytic approach was used to evaluate white and gray matter volume, regional brain volume, cortical thickness, fractional anisotropy, mean/axial/radial diffusivity, neurite dispersion and density, fiber density, and fiber cross section. From a clinical perspective, adolescents with ME/CFS showed greater fatigue and pain, poorer sleep quality, and poorer performance on cognitive measures of processing speed and sustained attention compared with controls. However, no significant group differences in white matter properties were observed, with the exception of greater white matter fiber cross section of the left inferior longitudinal fasciculus in the ME/CFS group compared with controls, which did not survive correction for intracranial volume. Overall, our findings suggest that white matter abnormalities may not be predominant in pediatric ME/CFS in the early stages following diagnosis. The discrepancy between our null findings and white matter abnormalities identified in the adult ME/CFS literature could suggest that older age and/or longer illness duration influence changes in brain structure and brain–behavior relationships that are not yet established in adolescence.

With uncertain etiology and lacking a diagnostic biomarker, the diagnosis is characterized by fatigue persisting for at least 3 months, substantial reduction in daily activities, and dysfunction in various symptom domains. The latter includes neurological symptoms such as pain, sleep, cognition, and post-exertional malaise, as well as autonomic, neuroendocrine, and immune system dysfunction (Jason et al., 2006(Jason et al., , 2010. Symptom severity may improve over time with clinical management, yet a large majority of adolescents continue to meet diagnostic criteria for ME/CFS at least 2 years following initial diagnosis, with a particular persistence of symptoms in the domains of fatigue, pain, and quality of life (Josev et al., 2021). The impact on academic, social, and emotional development of affected adolescents is cause for concern among clinicians, educational staff, and families (Crawley et al., 2011;Crawley & Sterne, 2009), and there are clear financial consequences for patients' families and wider society (Jason et al., 2008;Missen et al., 2012;Velleman et al., 2016).
Delineating the pathophysiological mechanisms of pediatric ME/ CFS remains an essential goal for our understanding, management, and treatment of this debilitating condition.
Inflammation of the central nervous system may underlie many of the core neurological ME/CFS symptoms, as identified through morphological changes or abnormalities of brain white matter (Nakatomi et al., 2014;VanElzakker et al., 2018). Indeed, increased inflammatory-related expression of activated microglia and astrocytes has been observed in various cortical and subcortical brain regions in adults with ME/CFS (Nakatomi et al., 2014). It has been proposed that with illness chronicity, the central nervous system experiences a prolonged state of low-grade neurological and systemic inflammation, resulting in structural and functional dysfunction over time, and loss of normal homeostatic processes (Nacul et al., 2020).
Although studies are limited in number, different magnetic resonance imaging (MRI) methods have been used to investigate the macrostructural and microstructural properties of white matter in adults with ME/CFS, including voxel-based morphometry (VBM), diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), neurite orientation dispersion and density imaging (NODDI), and T1-weighted/T2-weighted ratio mapping. Previous reviews examining brain differences in adults with ME/CFS (Almutairi et al., 2020; Morris et al., 2018;Shan et al., 2020) have identified significantly increased white matter hyperintensities in ME/CFS compared with various control groups (i.e., head injury controls, Natelson et al., 1993;healthy controls, Lange et al., 2005; ME/CFS controls with comorbid psychopathology, Lange et al., 2005). Reduced white matter volumes have also been observed in adults with ME/ CFS compared with controls in the occipital lobe (Puri et al., 2012); frontal cortex (Barnden et al., 2011); frontal occipital lobe and fasciculus (Shan et al., 2016); prefrontal cortex, ventrolateral thalamus, and internal capsule (Barnden et al., 2015). Of the few adult studies examining white matter at a microstructural level, ME/CFS samples have been small (n ≤ 20), effect sizes for diffusion parameters have been unavailable, and results have been heterogeneous with respect to the pattern and distribution of affected brain regions. For example, Zeineh et al. (2015) noted increased fractional anisotropy in the right anterior arcuate fasciculus (correlating with illness severity), while Kimura et al. (2019) noted decreased fractional anisotropy in the genu of the corpus callosum and right internal capsule. Thapaliya et al. (2021) observed decreased axial and mean diffusivity in the descending corticocerebellar tract (midbrain and pons) and increased transverse diffusivity in the medulla, while Rayhan et al. (2013) observed increased axial diffusivity in the right inferior fronto-occipital fasciculus. Kimura et al. (2019) also noted decreases in mean kurtosis and decreased neurite density index in the right superior longitudinal fasciculus of adult ME/CFS patients. In some cases, these brain changes correlate with ME/CFS symptoms and illness characteristics such as fatigue, pain perception, cognitive difficulties, symptom severity, and disease duration (Barnden et al., 2011(Barnden et al., , 2015Rayhan et al., 2013;Shan et al., 2016;Zeineh et al., 2015).
While white matter abnormalities have been identified in adults with ME/CFS, such heterogeneity in results has made it difficult to hypothesize the mechanisms underlying white matter change. For instance, white matter volume loss in ME/CFS has been argued to represent demyelination and disturbed integrity of cortico-cortical or cortico-subcortical connections, while observations of increases in white matter volume in similar ME/CFS samples have been argued to reflect compensatory myelination mechanisms in response to neuroinflammation (Barnden et al., 2011(Barnden et al., , 2015Nakatomi et al., 2014). The heterogeneity of reported results may be explained by variation in imaging techniques, protocols, diagnostic criteria, and research inclusion criteria used. The variation may also relate to the heterogeneity of ME/CFS, with findings potentially

Significance
Little is known about the neuroinflammatory effects of myalgic encephalomyelitis/chronic fatigue syndrome (ME/ CFS) on brain white matter in the context of adolescence.
Such knowledge could be pivotal in fast-tracking treatment and prevention of this and other post-viral fatigue syndromes like "long covid." This case-controlled study represents the first to investigate white matter microstructural organization in pediatric ME/CFS, and its relationship with clinical measures. The findings suggest that white matter abnormalities may not be predominant in pediatric ME/ CFS in the early stages following diagnosis. Longitudinal studies are needed to determine whether ME/CFS-specific white matter changes become apparent at later stages of this debilitating pediatric illness. reflecting varying phenotypes of the condition with different underlying pathophysiological mechanisms.
Further complicating the matter is the lack of studies investigating the ME/CFS neuroinflammatory hypothesis in a developmental context. Compared with adulthood, childhood and adolescence are characterized by significant brain maturation, which may lead to variation in the manifestation of (or susceptibility to) microstructural change as a result of neuronal damage or inflammation (Figaji, 2017;Morris et al., 2018). With the lack of brain imaging studies in pediatric ME/CFS to date, it is unclear whether white matter abnormalities are present in children and/or adolescents with ME/CFS, or if present, whether they compare with the white matter abnormalities identified in adults. The presence of white matter changes in adolescents newly diagnosed with ME/CFS would suggest acute neuroinflammatory effects of the illness (i.e., prodromal or early stages of the illness), while abnormalities identified only later in their illness course may be more suggestive of neuroinflammation secondary to illness chronicity.
To investigate neuroinflammation in pediatric ME/CFS, the current study employed diffusion-weighted MRI to examine the presence and degree of pathological changes in brain white matter microstructure on a whole-brain and regional level. To the authors' knowledge, this case-controlled study represents the first to investigate brain white matter microstructure in children and adolescents with ME/CFS. Given the lack of background research in this field, we used an exploratory (rather hypothesis-driven) approach to this study, and to ensure a robust analysis, five methodological approaches were used and explained in the Aims below. Our team previously showed that functional connectivity assessed via restingstate fMRI did not differentiate pediatric ME/CFS patients and healthy controls, even following a cognitively effortful task, and was not associated with clinical variables such as fatigue level or cognitive ability (Josev et al., 2020). It was unclear whether these findings (a) represented a genuine lack of difference between patients and controls in neural connectivity, (b) related to patients being early in their illness course, prior to the potential chronic effects of neuroinflammation on connectivity, or (c) whether resting-state fMRI was an imprecise tool for identifying pediatric ME/CFS diagnostic biomarkers, which may instead be detected at a deeper, more subtle level. Indeed, diffusion MRI may prove a useful tool in identifying potential diagnostic biomarkers in pediatric ME/CFS given that it is more sensitive to changes in neural tissue microstructure (and to some degree, macrostructure) than structural or functional MRI (Le Bihan & Iima, 2015).

| Aims
The main aim for this study was to investigate the integrity of brain white matter in adolescents newly diagnosed with ME/CFS compared with healthy adolescent controls, from both a macrostructural and microstructural perspective. First, at a macrostructural level, we aimed to determine whether there were group differences in (a) white matter volume, gray matter volume, and cerebrospinal fluid, using a volumetric tissue analysis and VBM approach and (b) group differences in regional brain volume and cortical thickness, using an anatomical surface-based parcellation approach. Second, at a microstructural level, we aimed to investigate whether there were group differences in (i) fractional anisotropy and mean/axial/ radial diffusivity, using a diffusion-weighted imaging (DWI) model approach, (ii) group differences in neurite dispersion and density, using a NODDI model approach, and (iii) group differences in fiber density and fiber cross section, using a fixel-based analysis (FBA) approach. Finally, we aimed to explore brain-behavior relationships between group differences observed in these aforementioned white matter measures and clinical measures of cognitive ability and general intellectual functioning, fatigue, mood (anxiety and depression), sleep quality, pain, and fatigue-related quality of life.

| Participants
Forty-eight female and male adolescents (25 with ME/CFS and 23 healthy controls) participated in the original MRI study (Josev et al., 2020) and all were included in the analysis of the structural MRI data. For the DWI analysis, 45 participants were included in the final dataset for analysis (24 ME/CFS, 21 healthy controls); three participants had to be excluded due to extensive motion artifact or signal loss due to metallic dental implants. In brief, ME/ CFS participants aged 13-18 years were recruited through the hospital's ME/CFS Clinic and diagnosed by pediatricians specializing in ME/CFS using the Canadian Criteria adapted for pediatrics (Jason et al., 2006). Healthy adolescent controls aged 13-18 years were recruited through advertising posters around the hospital clinics and were defined as healthy if they had no history of ME/CFS or other chronic illnesses. Exclusion criteria for both groups were (a) insufficient English, (b) presence of major depression and/or anxiety disorder, (c) history of psychosis or bipolar disorder, (d) pre-existing developmental disability or brain injury, and (e) current use of any medication that may affect brain function. The inclusion and exclusion criteria were determined by the treating pediatrician for the ME/CFS group and by the study coordinators for the healthy control group (i.e., practicing clinical neuropsychologists with expertise in pediatric ME/CFS).

| Design, procedure, and measures
Participants attended a study visit at the hospital which comprised of a brain MRI scan, as well as a clinical assessment of fatigue, general intellectual functioning, cognitive ability, and academic ability (note, only measures used for the current study are described here, see Josev et al., 2020 for full detailed procedure). Participants also completed a battery of standardized questionnaires prior to the study visit using REDCap (version 5.10.2, Vanderbilt University, Tennessee, USA, 2014) (Harris et al., 2009), comprising assessment of mood (anxiety and depression), sleep quality and hygiene, pain, fatigue-related quality of life, and other illness characteristics including school attendance, time since symptom onset at diagnosis, and illness trigger. Table 1 shows the specific descriptive and clinical measures used in the study.

| T 1 -weighted structural processing
Volumetric tissue analysis was first performed to assess wholebrain group differences in total gray matter, white matter, and cerebrospinal fluid (CSF); each T 1 -weighted image was first segmented into gray matter, white matter, and CSF in native space using SPM12, using the same algorithm described by Ashburner and Friston (2003). Voxel-by-voxel analysis with VBM was also performed using a standard workflow beginning with tissue segmentation (gray matter, white matter, CSF) using the default preprocessing parameters in SPM12 (Ashburner & Friston, 2005, 2000.

This was followed by spatial normalisation in standard Montreal
Neurological Institute (MNI) space via registration with modulation to ensure voxel-wise correspondence across participants and preservation of tissue volume. The normalised tissue segments were smoothed with an 8mm full-width at half-maximum Gaussian kernel. Automatic surface-based brain parcellation, based on the Destrieux atlas (Destrieux et al., 2010), was also performed on the T 1 -weighted images using FreeSurfer 6.0. (Fischl, 2012) in order to assess group differences in regional brain volumes and cortical thickness.

| Preprocessing
MRtrix3 was used to perform the preprocessing image analysis.
The b = 1000 s/mm 2 and b = 3000 s/mm 2 shells were preprocessed separately using a combination of the MRtrix3 (version 3.0 RC3, http://www.mrtrix.org/) and FSL software packages. The DWI data were preprocessed sequentially to correct for data thermal noise (Veraart et al., 2016), Gibbs-ringing artifacts (Kellner et al., 2016), eddy current and motion-induced distortions (Andersson et al., 2003;Andersson & Sotiropoulos, 2016), EPI susceptibilityinduced geometric distortions (Bhushan et al., 2012), and b1bias field-induced inhomogeneity (Tustison et al., 2010). Finally, the b = 3000 s/mm 2 shell was registered to the b = 1000 s/mm 2 shell using the Linear Image Registration Tool (FLIRT) within FSL (Jenkinson et al., 2002). In addition, to account for the different TE and TR between the two shells, each shell was normalized by the b = 0 s/mm 2 image which involved dividing every voxel in each shell by the average b = 0 s/mm 2 measurement for that shell (Owen et al., 2014). The aligned b = 3000 s/mm 2 shell was then merged with the b = 1000 s/mm 2 shell.

| Diffusion tensor model
The DTI model was then fitted on preprocessed b = 1000 s/mm 2 to generate the DTI images and metrics of FA, MD, AD, and RD for between-group analysis, using the dtifit tool (Basser et al., 1994) in FSL.

TA B L E 1
Descriptive and clinical measures used in analysis of the ME/CFS and healthy control groups.

Clinical measures used in analysis
Depression Hospital Anxiety and Depression Scale (HADS) (Zigmond & Snaith, 1983) Self-report instrument to evaluate frequency of anxiety and depression (14 items total, 7 items per subscale).
Questions are scored 0-3 points, with higher scores indicating greater levels of depression and anxiety. For the current study, only the depression total score was used, as the SCAS provides a more comprehensive measure of anxiety  (Varni et al., 1987) Self-rated 100 mm scale to measure intensity of present pain, from "not hurting" or "no pain" (0) to "hurting a whole lot" or "severe pain" (100) Fatigue-related quality of life Using this method, the NODDI model generated neurite dispersion and density images.

| Registration and template creation
Eigenvalue and eigenvector maps were generated for each participant, and converted to a single volume in DTI-TK format.
DTI-TK (Zhang et al., 2006(Zhang et al., , 2010 was used to create an unbiased population-based template that uses both the average diffusion features (e.g., diffusivities and anisotropy) as well as the anatomical shape features (such as tract size) in the population. Each participant's DTI model matrices and NODDI model matrices were registered to the population-based tensor template using diffeomorphic alignments.

| Fixel-based analysis
A relatively recent analysis framework for DWI data, fixel-based analysis (FBA), was employed to measure biological changes occurring in white matter fibers within each voxel (Dhollander et al., 2021). for changes to both within-voxel fiber density and macroscopic atrophy, thereby providing an estimate of total intra-axonal volume or overall "carrying capacity" of the white matter fiber bundle (Haykal et al., 2020;Raffelt et al., 2017).
FBA was performed in accordance with a recommended pipeline (Raffelt et al., 2017). The preprocessed two-shell DWI images were upsampled to isotropic 1.3 mm 3 voxels to increase the anatomical contrast, as recommended by Dhollander et al. (2021).
The response function of each subject was estimated and the group average response function was calculated. The fiber orientation distribution estimation (FODs) was estimated for each participant using the group average response function. An unbiased study-specific FOD template was employed, using 21 samples to ensure parity from each group. All subject FOD images were finally registered to the study-specific FOD template and transformed to the template space. Whole brain tractography was performed on the FOD template where 20 million tracts were generated, and spherical-deconvolution informed filtering of tractograms (SIFT) was implemented to filter this down to 2 million tracts; the density of which corresponds to the fiber density present in the data (Smith et al., 2013). Tractography was performed in MRtrix3 using the probabilistic tractography algorithm with default parameters

| Structural image statistics
For the structural image analysis, group differences in gray and white matter tissue volume, CSF, FreeSurfer regional brain volumes, and cortical thickness were investigated using the statistical package R. A generalized linear regression design was used, with the MRI measure as the dependent variable (i.e., gray matter volume, white matter volume, CSF volume, etc.); group as the independent variable; and age, sex, and intracranial volume (ICV) used as covariates. To account for multiple comparisons, false discovery rate (FDR)- Firstly, groupwise (ME/CFS vs. healthy control group) comparisons of diffusion measures across the skeleton were conducted, using twosample unpaired (independent groups) t-test designs, with age and sex included as covariates. In all statistical tests, 5000 permutations were performed. All results were reported at p < .05 after thresholdfree cluster enhancement (TFCE) (Nichols & Holmes, 2002) and familywise error rate correction.

| FBA statistical analyses
General linear models (GLMs) were computed for permutation-based testing of FBA metrics. We compared FD, log-FC, and FDC between the groups using connectivity-based fixel enhancement (CFE) in MRtrix3 (Raffelt et al., 2015) with age, sex, and ICV used as covariates. The FBA results reported here were generated using 5000 permutations and reached familywise error (FWE)-corrected statistical significance at pFWE < .05. The JHU-ICBM FA template (Smith et al., 2004) was registered to the FOD template, and the transformation was applied to the JHU white matter labels atlas. These JHU white matter labels were overlaid onto the output scale image from the fixel statistical analyses to identify the regions that showed group differences, with the FD, log-FC, and FDC extracted from any significant clusters identified.

| Sample characteristics
Characteristics of the sample are shown in Table 2  (3-6 months, 7-12 months, 13-24 months, and >24 months). Most of the control group did not miss any school in the last term (69%, n = 16), whereas 78% (n = 18) of the ME/CFS group missed between 50% and 100% of the school term.

| Structural MRI analysis (SPM)
Volumetric tissue analysis revealed no significant group differences in whole-brain gray matter volume (p = .53), white matter brain volume (p = .53), or CSF (p = .91) between the ME/CFS and control groups, after adjustment for age, sex, ICV, and FDR correction for multiple comparisons. ICV was also similar between the groups, adjusted for age, sex, and FDR correction (p FDR = .54). VBM also revealed no significant differences between the groups in terms of gray and white matter volume, after adjustment for age, sex, ICV, and FWE correction (p FWE < .05). FreeSurfer regional gray matter volumes and cortical thickness were also not significantly different between groups, adjusted for age, sex, ICV, and FDR correction (p FDR < .05).

TA B L E 2
Participant characteristics for the ME/CFS and control groups.

| DTI model and NODDI analysis
Tract-based spatial statistics of DTI matrices (FA, MD, AD, and RD) and NODDI matrices (odi and ficvf) revealed no significant group differences between the ME/CFS and healthy control groups, after adjustment for age and sex, and after TFCE (threshold-free cluster enhancement) correction (p TFCE < .05).

| Fixel-based analysis
FBA results using CFE showed no significant group differences between the ME/CFS and healthy control group in terms of fiber density (FD) or the combined measure of fiber density and cross section (FDC), adjusted for age, sex, and ICV. A significant difference was found between groups, whereby the ME/CFS group showed greater Note: N.B.: One ME/CFS participant did not complete the questionnaires (HADS, SCAS, ASWS, ASHS, and SEIFA index), and another ME/CFS participant did not complete the cognitive ability measures due to excessive fatigue. There was also missing data for parent report of perceived illness trigger (n = 2), % school missed (n = 3), and academic measures (n = 1).

TA B L E 2 (Continued)
F I G U R E 1 Fixels with significantly higher log FC in the left inferior longitudinal fasciculus in the ME/CFS group compared with the control group, adjusted for age and sex only (p FWE < .001). This result was no longer significant once additionally corrected for ICV (p FWE > .05). Row 1: Significant fixels converted to a voxel map (hot) for visualization, overlaid on the population-based FOD directionally encoded color (DEC) map (Dhollander et al., 2015) (red = left-right, green = front-back, blue = up-down). Row 2: Magnification of color-coded significant fixels (p FWE < .05) to view fiber direction in the significant region, overlaid on the FOD-based DEC map. fiber-bundle cross section (log of FC) in a cluster representing the left inferior longitudinal fasciculus compared with the healthy control group, after adjustment for age and sex (p FWE < .001) (Figure 1 below).
However, this cluster did not survive correction for ICV (p FWE > .05).

| Association between clinical factors and group differences in white matter macrostructure and microstructure
No significant group differences were observed in any of the structural MRI measures investigated in this study, once corrected for age, sex, ICV, and multiple comparisons. As such, the planned multivariate regression analyses to determine relationships between group differences in the MRI measures and clinical measures were not considered meaningful for this analysis and were therefore not conducted.

| DISCUSS ION
To our knowledge, this is the first case-controlled study to examine white matter microstructural properties in children and adolescents with ME/CFS. It is also the first to employ a fixel-based analysis approach in estimating axonal properties in ME/CFS. We used a robust approach to analysis that employed volumetric tissue analysis, VBM, DTI, NODDI, and FBA.
After correction for age, sex, ICV, and multiple comparisons, results from our study revealed that the ME/CFS and control groups were similarly matched in macrostructural brain properties in terms of whole-brain and regional volumes of white and gray matter, CSF, and cortical thickness. Similarly, no significant group differences were observed in any of the microstructural properties including the diffusion-weighted matrices, NODDI matrices, or within the FBA analysis. Of note, the ME/CFS group did show significantly greater fiberbundle cross section of the left inferior longitudinal fasciculus (ILF) compared with the healthy control group, after adjustment for age, sex, and correction for multiple comparisons; however, this did not survive correction for ICV. Compared with controls, and as previously shown (Josev et al., 2020), the ME/CFS group also reported greater subjective fatigue, fatigue-related quality of life, pain, and poorer sleep quality. The ME/CFS group also performed lower than controls on objective cognitive tests of processing speed and sustained attention. Overall, our analysis suggested that despite significant group differences in clinical measures, macrostructural and microstructural white matter properties are comparable between newly diagnosed patients with pediatric ME/CFS and their healthy peers.
While the left ILF finding in the current study appeared to be related to intracranial volume rather than being specific to ME/CFS, it is interesting to note that two previous diffusion studies have also identified the ILF (among other white matter tracts) as a pathway with greater white matter microstructural integrity in adults with ME/CFS relative to healthy controls. Thapaliya et al. (2020) observed higher mean T1w/T2w ratio in the right ILF of ME/CFS patients, which was suggestive of greater myelin (or iron) content. This may have arisen as a compensatory mechanism in response to compromised microstructure in temporal-occipital or anterior temporal regions. Zeineh et al. (2015) also observed increased fractional anisotropy in the anterior region of the right ILF in right-handed ME/CFS patients, which was interpreted as reflecting either strengthening of fibers or weakening of crossing fibers in this region. However, in contrast to our own findings, both these studies identified the contralateral ILF, and in both studies, the ILF findings were unrelated to clinical parameters.
Furthermore, the greater left ILF fiber cross section noted in the ME/CFS group in our study (prior to ICV correction) occurred in isolation, without an accompanying increase in the measures of fiber density (FD) and combined fiber density and cross section (FDC).
The ILF finding is therefore unlikely to represent a morphological change to macroscopic fiber arrangement (i.e., a greater intra-axonal volume where the fiber bundle occupies a greater number of voxels) because the volume is relative to the fiber orientation, with the combined FDC measure providing the most sensitive evaluation of fixelwise effects (Raffelt et al., 2017). On balance, our isolated left ILF finding is unlikely to represent an early pathophysiological change or brain biomarker of early onset ME/CFS.
While our finding of comparability in white matter macrostructure and microstructure between groups is robust, we acknowledge that this study has potential limitations. Our small sample size may have underpowered our ability to detect group differences in white matter structural metrics, highlighting the need for replication and the use of larger samples. Future inclusion of an adult ME/CFS cohort for comparison would also be ideal, as it is unknown whether alterations to white matter microstructural organization observed in adult ME/CFS can be extrapolated to children and adolescents newly diagnosed with the condition. The discrepancy between the lack of white matter pathology observed in our pediatric ME/CFS cohort and the reported increases (and decreases) in white matter volume and axonal integrity in the adult ME/CFS literature may suggest that older age and/or longer illness duration could influence changes in brain structure and brain-behavior relationships that are not yet established in adolescence. Indeed, there is some evidence to suggest that white matter changes in adults with ME/CFS may be related to disease duration (Barnden et al., 2011(Barnden et al., , 2015, which could account for the lack of findings in our pediatric cohort of newly diagnosed patients. A large-scale UK study showed that children and adolescents with ME/CFS tended to have shorter illness duration at study enrollment than adults with ME/CFS (Collin et al., 2015), and the pediatric ME/CFS diagnosis itself requires a shorter fatigue duration of ≥3 months compared with ≥6 months in adults (Jason et al., 2006).
Nevertheless, our study had limited power to assess the relationship between illness duration and diffusion white matter metrics in our pediatric ME/CFS cohort, given that adolescents were newly diagnosed and there was ≤7 patients in each of the four subcategories of parent-reported time since symptom onset at diagnosis.
More longitudinal investigations have been recommended to determine whether microstructural white matter changes develop with length of illness (Almutairi et al., 2020), especially since symptoms can persist for several years after diagnosis of pediatric ME/CFS (Josev et al., 2021). In particular, increased fatigue duration may relate to reduced white matter volume in adults, particularly in the midbrain (Barnden et al., 2011), but this is yet to be investigated at the axonal level with diffusion tractography and fixel-based analysis.
Comparison of our findings with previous research was complicated by the limited number of studies investigating white matter microstructure in ME/CFS, and to date, none have been conducted in a pediatric ME/CFS population. It is clear that further studies (both exploratory and hypothesis-driven) are required to replicate previous findings and to determine whether a typical pattern of neuroinflammation exists in ME/CFS.
A key strength of our study was the inclusion of a healthy control sample which enabled us to account for the impact of brain development on chronic illness during the adolescent period. Certainly, our study highlighted the importance of considering ME/CFS in a developmental context, and this will be explored further in the future as part of a wider longitudinal study in this cohort of young patients and controls where participants underwent further clinical assessment and brain MRI approximately 2 years following their diagnosis (Josev et al., 2021). Our study was able to provide novel insights into ME/CFS in children and adolescents, an understudied group whose illness is likely to be at an earlier stage (i.e., during adolescence) and uncomplicated by factors associated with chronicity of illness. Another strength of our study was that diagnosis was made by pediatricians specializing in ME/CFS using stringent clinical consensus criteria, which allowed for a well-characterized cohort of adolescents with moderate to severe ME/CFS (see Josev et al., 2021). Our results were also consistent across the multiple analytic methods employed in this study, supporting the robustness of our findings.
Overall, while our study showed some evidence of morphological change in the left ILF in pediatric ME/CFS, our interpretation of this isolated finding was conservative, given that it was not accompanied by changes in FD or FDC, and the finding did not survive correction for intracranial volume. Furthermore, there was overall parity of macrostructural and microstructural white matter indices across all adolescents, regardless of having ME/CFS. On balance, these findings suggest that white matter abnormalities may not be predominant in pediatric ME/CFS in the early stages following diagnosis. This finding is reassuring to patients and their families, and may suggest an early window to provide interventions that promote neuroprotection and prevention of neuroinflammation.
Further longitudinal research is needed to clarify whether neuroinflammation or any ME/CFS-specific changes in brain white matter become apparent at later stages in the illness in children and adolescents.

DECLARATION OF TRANSPARENCY
The authors, reviewers and editors affirm that in accordance to the policies set by the Journal of Neuroscience Research, this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present. Program.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors do not have any conflicts of interest to declare.

PE E R R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1002/ jnr.25223.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.