Complex regional pain syndrome: The matter of white matter?

Abstract Introduction Many central pathophysiological aspects of complex regional pain syndrome (CRPS) are still unknown. Although brain‐imaging studies are increasingly supporting the contribution of the central nervous system to the generation and maintenance of the CRPS pain, the brain's white‐matter alterations are seldom investigated. Methods In this study, we used diffusion tensor imaging to explore white‐matter changes in twelve CRPS‐type‐1 female patients suffering from chronic right upper‐limb pain compared with twelve healthy control subjects. Results Tract‐based spatial‐statistics analysis revealed significantly higher mean diffusivity, axial diffusivity, and radial diffusivity in the CRPS patients, suggesting that the structural connectivity is altered in CRPS. All these measures were altered in the genu, body, and splenium of corpus callosum, as well as in the left anterior and posterior and the right superior parts of the corona radiata. Axial diffusivity was significantly correlated with clinical motor symptoms at whole‐brain level, supporting the physiological significance of the observed white‐matter abnormalities. Conclusions Altogether, our findings further corroborate the involvement of the central nervous system in CRPS.


| INTRODUCTION
Although the complete pathophysiology of complex regional pain syndrome (CRPS) is still poorly understood, previous brain-imaging studies have suggested the involvement of central nervous system in the maintenance and progression of the pain (Marinus et al., 2011). For example, the CRPS patients can have altered hand representation as a sign of cortical reorganization (Di Pietro, Stanton, Moseley, Lotze, & McAuley, 2015;Juottonen et al., 2002), and some of the observed brain alterations are associated with clinical characteristics of CRPS, such as motor dysfunction (Maihöfner et al., 2007).
Two studies have previously analyzed DTI in CRPS, both focusing on fractional anisotropy (FA). In whole-brain analysis, van Velzen, Rombouts, van Buchem, Marinus, and van Hilten (2016) found no FA abnormalities, whereas Geha et al. (2008) showed FA to be lowered in the left cingulum; the subsequent region-of-interest analysis revealed decreased axial diffusivity (AD) and increased radial diffusivity (RD) in the same area. However, because FA is sometimes insensitive to changes in other DTI measures (Acosta-Cabronero, Williams, Pengas, & Nestor, 2010;Hasan, 2006), it would be important to examine each of these measures in a whole-brain manner.
In the present study, we performed whole-brain tract-based spatial-statistics (TBSS) analysis of FA, mean diffusivity (MD), AD, and RD to further examine changes in white-matter integrity in CRPS type-1 patients. To address the functional relevance of these changes, we also explored the associations between the motor dysfunction and disease duration, and the DTI measures.

| Participants
Twelve female patients with chronic type-1 CRPS affecting rightdominant upper-limb (mean ± SD age 46 ± 6 years, median 46, range 35-57) participated in the study. As the control group, we selected twelve optimally age-and scanner-matched (see Image acquisition) subjects (44 ± 10 years, median 45, range 25-60) from sixteen successfully scanned healthy female participants. The groups did not differ in age (p > .05, Wilcoxon rank-sum test). All subjects were right-handed by the Edinburg Handedness Inventory and self-report.
Before participation, all subjects gave informed written consent in accordance with the Declaration of Helsinki. The study protocol was ap-  (Harden, Bruehl, Stanton-Hicks, & Wilson, 2007) and our criteria stipulating (1) severe pain (pain intensity greater than four on 11-point numeric scale); (2) right-handedness; (3) no other major psychiatric or neurological diseases, or drug/alcohol addiction; (4) no contraindication for MRI.
In clinical examination, the diagnostic symptoms and signs were evaluated by an experienced neurologist (author H.H.) with a subspecialty in pain medicine. The disease duration was a mean ± SD of 5.8 ± 4.5 years. Table 1 summarizes details of clinical examination and demographic data.

| Motor-symptom-severity index
To evaluate the overall severity of motor-symptoms in the affected hand, we created motor-symptom-severity index (MSSI) as a compound score of five subjective and objective measures: T A B L E 1 Demographic and clinical examination data of complex regional pain syndrome patients (1) movement-related pain (11-point numeric rating scale) and (2) upper-limb disability (Disabilities of Arm, Shoulder, and Hand questionnaire i.e., DASH, Institute for Work & Health [Hudak, Amadio, & Bombardier, 1996]) questionnaires, and (3) hand dexterity (ninehole peg test), (4) total active range of wrist movement, and (5) grip strength assessed by a skilled physiotherapist. We calculated MSSI as the average z-score of these five components after the grip strength and wrist AROM z-scores had been multiplied by −1, so that for all components a higher score implied more severe motor symptoms. See Supporting Information for details.

| Tract-based spatial statistics
The preprocessing of DTI data included motion and eddy-current correction, and brain extraction using FSL toolkit (http://www. fmrib.ox.ac.uk/fsl/; Smith et al., 2004). The B-matrices were also rotated (Leemans & Jones, 2009). A brain mask was created from the diffusion-weighted image with b-value of zero, using the FSL's brain extraction tool (Smith, 2002).
Tensors were estimated on the corrected data within the brain mask using FSL's dtifit and the resulting images were converted into DTI-TK (http://www.nitrc.org/projects/dtitk) format for tensor-based spatial normalization (Zhang et al., 2007). First, population-specific tensor template was bootstrapped using the IXI aging DTI template (Zhang, Yushkevich, Rueckert, & Gee, 2010 wrapped to the population-specific template in standard space. This tensor-based normalization has been shown to be superior in detecting white-matter differences compared with low-dimensional registration using scalar values, such as FA (Wang et al., 2011;Zhang et al., 2007).
The FA, MD, AD, and RD maps were reconstructed from the spatially normalized tensor images for each participant. These DTI parameters were compared voxel-by-voxel with TBSS (Smith et al., 2006), as part of FSL. Briefly, the mean FA skeleton image that represents the center of tracts, consistent across subjects, was obtained by thinning the across-subjects averaged FA image (threshold of 0.2). Then, each subject's aligned FA, MD, AD, and RD images were projected onto the mean FA skeleton.
Between-group comparisons of the resulting skeletonized data were conducted using permutation test (FSL's Randomise v2.1; 10,000 permutations), with multiple comparisons corrected using thresholdfree cluster enhancement method (Smith & Nichols, 2009). We considered a difference to be statistically significant at a corrected p < .05.
Confounding factors of age and scanner were included as covariates.
Statistically significant tracts were labeled according to the whitematter atlas from Johns Hopkins University (JHU ICBM-DTI-81 whitematter labels).
Whole-brain skeletal FA, MD, AD, and RD were calculated by averaging over the FA skeleton. Between-group comparisons in these values were performed using unpaired two-tailed two-sample t tests.
Age and scanner were regressed out using multiple linear regression before the comparison analysis.
To evaluate the bias caused by head motion, the motion-correction parameters (absolute displacement and mean absolute translations) were also compared between the two groups using unpaired twotailed two-sample t test. The mean translation values were calculated across all three axes.

| Correlation between DTI parameters and clinical characteristics
Pearson coefficients were computed between the whole-brain skeletal DTI and clinical characteristics including the disease duration and MSSI (Matlab, Mathworks Inc.). Voxelwise correlation analyzes between DTI measures and clinical characteristics were performed using FSL's Randomise (v2.1, 10,000 permutations) and multiple correlations were corrected using the treshold-free cluster enhancement method. Age and scanner were included as covariates.

| DISCUSSION
We examined the white-matter integrity in the CRPS patients suffer-

| Potential mechanism underlying DTI changes in CRPS
Diffusion tensor imaging abnormalities have been found in a variety of chronic-pain conditions, such as migraine (Szabó et al., 2012;Yu et al., 2013), temporomandibular disorder (Moayedi et al., 2012), and F I G U R E 2 Results of tract-based spatial-statistics analysis. The red-yellow colors show voxels where complex regional pain syndrome (CRPS) patients have increased mean diffusivity (MD; top row), increased radial diffusivity (RD; 2nd row) or increased axial diffusivity (AD; 3rd row) compared with healthy control subjects (CON) (p < .05, threshold-free cluster enhancement corrected). The bottom row shows the overlap of differences. All voxels with group-differences were enlarged for better visualization using tbss_fill. The study-specific mean fractional anisotropy and the corresponding white-matter skeleton (green) were used as background images. R, right hemisphere; L, left hemisphere F I G U R E 1 Whole-brain diffusion tensor imaging (DTI) skeleton (left, shown in green) and mean skeletal DTI parameters for fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) in the complex regional pain syndrome patients (CRPS) and healthy control subjects (CON). Each symbol refers to one subject. The asterisks refer to statistically significant differences between the groups. Data were adjusted for age and scanner T A B L E 2 Tract-based spatial-statistics analysis results   The peak coordinates are given in MNI space. Only tracts with a number of voxels > 10 are shown. CRPS, complex regional pain syndrome patients; CONTROL, healthy control subjects; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; corr., corrected; L, left hemisphere; R, right hemisphere.

Region
a Could be a part of the anterior internal capsule. b The column and body of fornix.
c Includes the optic radiation.
Our current findings indicate that the white-matter integrity in the chronic CRPS is widely affected. However, the interpretations of the altered DTI parameters are not straightforward because the underlying cellular-level pathophysiology is incompletely understood (for a review, see e.g., Concha, 2014). For example, increased AD combined with increased RD has been found in several neurological diseases such as multiple sclerosis (Roosendaal et al., 2009), Alzheimer's disease (Acosta-Cabronero et al., 2010), and amyotrophic lateral sclerosis (Metwalli et al., 2010) with otherwise separate and distinctive pathophysiology.
In the current study, MD was higher in more than one-fifth of all tested voxels in the CRPS patients compared with healthy subjects. MD changes can reflect e.g., alteration of cellularity, or necrosis (Alexander et al., 2011;Concha, 2014), but MD also increases in vasogenic edema, which could be indicative of e.g., neuroinflammation (Pasternak et al., 2015) that is suspected to occur in CRPS. However, more research is needed to study the role of central inflammaltion in CRPS and its relationship to DTI abnormalities. For example, neuroinflammationspecific changes of extracellular volume are more likely to be identified with free-water imaging which is sensitive to water diffusion in the extracellular space (Pasternak et al., 2012(Pasternak et al., , 2015. The DTI changes can also be experience-related one (Scholz et al., 2009). Thus, the profound abnormalities in the CRPS patients' daily lives-such as continuous pain, altered sensory input (Rommel, Malin, Zenz, & Jänig, 2001) and reduced motor activity (Schilder et al., 2012)-could explain the large-scale pathology of the DTI alterations.
The positive correlation between motor disability (MSSI) and wholebrain skeletal AD in the patients would be in line with this.

| Caveats
Compared with FA-intensity-based registrations, the tensor-based DTI-TK registration technique, as used in the present study (see Section 2), improves the alignment of different brains and thereby the detection of group differences (Van Hecke et al., 2007;Wang et al., 2011;Zhang et al., 2007). The differences of these registration methods can affect the detection of group differences with TBSS e.g., in the cingulum bundle (Bach et al., 2014). Despite the present study's methodological advantages and the highly homogeneous patient group (dominant upper-limb CRPS), we acknowledge that the sample size is relatively small. As DTI studies on the effects of dominant-arm inactivity or chronic pain from other etiologies are not yet available, the specificity of our findings to CRPS remains to be shown. Future studies with larger number of patients are necessary to understand the spread and significance of white-matter microstructural abnormalities in the chronic CRPS patients.

| CONCLUSIONS
Several DTI parameters were altered in our chronic CRPS patients, supporting the involvement of the central nervous system in the generation and maintenance of the CRPS symptoms. The present results demonstrate specifically the presence of white-matter alterations in CRPS. The role of white-matter changes in the clinical course of CRPS should be assessed in future studies.