Interhemispheric and Corticothalamic White-Matter Dysfunction Underlies Affective Morbidity and Impaired Pain Modulation in Chronic Pain

Abstract

BACKGROUND:

Although patients with chronic pain show behavioral signs of impaired endogenous pain modulation, responsible cerebral networks have yet to be anatomically delineated. We used diffusion tensor imaging (DTI) to examine the white-matter alterations in patients with chronic pain compared with healthy subjects. We further measured thermal pain modulatory responses using the offset analgesia (OA) paradigm. We tested whether the white-matter indices be associated with psychophysical parameters reflecting morbidity and modulatory responses of pain in patients, and whether they could serve as diagnostic biomarkers of chronic pain.

METHODS:

Twenty-six patients with chronic pain and 18 age- and gender-matched healthy controls were enrolled. After completing psychophysical questionnaires, they underwent OA measurement and whole-brain DTI in a 3 Tesla magnetic resonance imaging scanner. Fractional anisotropy (FA) and radial diffusivity (RD) of the white-matter were computed and compared between the groups with tract-based spatial statistics using the FMRIB Software Library (FSL) software. Correlations were sought among white-matter indices, thermal pain responses, and psychophysical parameters. The white-matter indices and OA-related parameters were tested whether they distinguish patients from controls by receiver operating characteristic analysis.

RESULTS:

During OA, patients showed a shorter latency to the maximum (maximum visual analog scale [VAS] latency, 16.0 ± 3.7 vs 18.9 ± 3.1 second [mean ± standard deviation, SD]; P = .032) but a longer latency to the minimum pain (OA latency, 15.6 ± 3.5 vs 11.1 ± 4.2 seconds; P = .004) than controls. They showed a smaller mean FA (0.44 ± 0.12 vs 0.45 ± 0.11; P = .012) and a larger mean RD of the global white-matter (0.00057 ± 0.00002 vs 0.00056 ± 0.00002; P = .038) than controls, at specific areas including the corpus callosum, anterior thalamic radiation, and forceps major. FA of the splenium of the corpus callosum was associated with maximum VAS latency (r = 0.493) and OA latency (r = −0.552). The Pain Catastrophizing Scale scores showed strong negative correlations with FA across those specific areas (r = −0.405). Those latencies during OA and white-matter metrics distinguished patients from controls (P < .05).

CONCLUSIONS:

Patients with chronic pain showed dysfunction of the white matter concerned with interhemispheric communication of sensorimotor information as well as descending corticothalamic modulation of pain in association with affective morbidity and altered temporal dynamics of pain perception. We suggest that an impaired interhemispheric modulation of pain, through the corpus callosum, might be a novel cerebral mechanism in chronification of pain.

KEY POINTS.

Question: What is the functional relevance of white matter dysfunction in patients with chronic pain?

Findings: Patients with chronic pain showed extensive white-matter disintegrity specifically over the anterior thalamic radiation and corpus callosum, in association with altered pain modulatory responses by offset analgesia and emotional morbidity of pain.

Meaning: In addition to the descending pain modulatory system, the interhemispheric inhibition of pain might be a novel pain modulatory circuit that turns dysfunctional in chronic pain conditions.

Chronic pain is a major health problem worldwide, and is estimated to prevail up to 19% of the general population.¹ Because of its heterogeneous etiologies and intractable nature even with healed or absent peripheral pathology, evidence for central sensitization of pain has been postulated and sought for in the brain. While human neuroimaging studies have unveiled cerebral plastic changes in pain-evoked activation,² functional connectivity,^3,4 and gray matter volume,^3,5 anatomical changes in neural circuitry remain less explored. To investigate such anatomical biomarkers for chronic pain, we used diffusion tensor imaging (DTI), which visualizes water molecule diffusion in magnitude and direction and quantifies the microstructural integrity of the white-matter.⁶ Briefly, water molecules diffuse along the direction of nerve fibers, and the degree of diffusion is calculated as fractional anisotropy (FA), radial diffusivity (RD), etc. Those metrics enable the estimation of neuronal pathological changes such as demyelination, edema, and inflammation of axonal fibers.⁶ Earlier DTI studies,^7–14 however, have yet to delineate the functional, pathophysiological significance of white-matter abnormalities in chronic pain.

To fill such a knowledge gap, anatomical signatures of chronic pain should be sought for correlations with dysfunction of endogenous analgesia. We assessed endogenous analgesia by offset analgesia (OA), a phenomenon in which a slight decrease in noxious thermal stimulus leads to a disproportionately large decrease in pain perception.¹⁵ OA was shown attenuated in patients with neuropathic pain,¹⁶ fibromyalgia,¹⁷ and various chronic pain conditions.¹⁸ By functional magnetic resonance imaging (fMRI) during OA we further found suppressed blood oxygenation level-dependent activities in the descending pain modulatory and reward systems in chronic pain.¹⁹ Questions remain as to the anatomical substrates for those altered cerebral responsiveness. Plastic changes in neural circuits should be visualized in association with derangement of pain modulatory mechanisms as well as psychophysical properties in chronic pain.

In this study, we used DTI of the whole brain to investigate differences in anatomical indices of the white-matter between patients with chronic pain and age- and sex-matched healthy controls. Patients were recruited from a variety of disease types, resulting in a more generalized group of patients with chronic pain. We hypothesized that those patients and healthy subjects might have different metrics of white-matter microstructure in pain-modulatory regions, such as descending pain modulatory and reward systems, and that these metrics would be related to various psychophysical parameters associated with chronic pain. Furthermore, we tested whether the white-matter changes be associated with altered temporal dynamics of pain perception, measured by the OA paradigm,^18,19 particularly latencies of pain perception that might potentially reflect pain modulatory interactions in the brain.

METHODS

Subjects

All protocols of this study were approved by the local Institutional Review Boards of Tokyo Medical and Dental University (No. 1525) and Jikei University (No. 10032), and registered at the Japanese clinical trials registry (UMIN-CTR, http://www.umin.ac.jp/ctr/, No. UMIN000011253; principal investigator: J. Kurata, date of registration: July 23, 2013). We followed the ethical guidelines of the Declaration of Helsinki. Written informed consent was obtained from all subjects participating in the study, allowing withdrawal from the study at any time.

From February 2014 to September 2018, 26 patients with chronic pain and 18 age- and sex-matched healthy subjects were recruited at Tokyo Medical and Dental University Hospital through posters and word of mouth. All subjects were aged between 20 and 75 years; right-handed; of Asian ethnicity; confirmed free from the exclusion criteria such as inability to obtain consent; any contraindications to MRI such as pregnancy, claustrophobia, in-body metal objects; dementia or other psychiatric disorders; and obesity (body mass index >35 kg/m²). Inclusion criteria for patients with chronic pain were persistent pain for longer than 3 months despite medication or interventional therapy; and pain intensity ≥15 by a visual analog scale (VAS; 0 for no pain, 100 for maximum pain imaginable). Patients were allowed to use oral medications to treat or alleviate symptoms until just before the experiment. Healthy controls were matched in age and sex to the patients and had no history of chronic pain. Exclusion criteria were the same as for patients with chronic pain.

Experimental Design

All subjects were instructed to refrain from caffeine and alcohol consumption for 24 hours before the experiment. All subjects were asked to provide their age and gender on the day of the experiment, and the patients with chronic pain were also asked to provide information about VAS indicating current pain intensity, duration of pain, and illnesses. Medication history of the patients was obtained from their medical records. All subjects then completed a series of psychophysical questionnaires. Of all subjects, 14 of 26 patients in the chronic pain patient group and 15 of 18 healthy subjects group underwent thermal pain stimulation including OA stimulation paradigm. All subjects then underwent multimodal MRI consisting of DTI, fMRI during OA, resting-state fMRI, and 3-dimensional high-resolution anatomical imaging. Part of those results were reported elsewhere^3,19 except for those of DTI.

Psychophysical Questionnaires

Before imaging, all subjects were asked to answer the following questionnaires: the Pain Catastrophizing Scale (PCS)^20,21 to measure catastrophic thoughts about pain; the short-form McGill Pain Questionnaire (MPQ)²² to assess daily pain symptoms: the painDETECT questionnaire (PD-Q)²³ to assess neuropathic pain; and the Beck Depression Inventory (BDI)²⁴ to assess depressive symptoms.

Thermal Pain Stimulation and Definition of OA-Related Parameters

Part of subjects (14 patients and 15 healthy subjects) underwent an off-line thermal pain experiment before MRI. We used a Peltier-type thermal stimulator (PATHWAY: Medoc) and a 30 mm-radius thermode attached to the left volar forearm for thermal stimulation. Subjects were asked to continuously record VAS from 0 mm (no pain) to 100 mm (most intense pain imaginable) using a computerized VAS recorder (CoVAS: Medoc) during thermal stimulation at a sampling rate of 9 Hz. Subjects underwent 2 series of thermal pain stimuli, with intervening 5-minute rest, each including 3 OA blocks (a series of 32°C-baseline, 45°C-T1 [10-second duration], and 46°C-T2 [5, 10, or 15 seconds], 45°C-T3 [20 seconds], followed by 32°C-baseline; Figure 1) in a pseudorandom order at ramp rates of +56°C/s and −6°C/s, totaling in 6 stimulus blocks. The thermode was moved to a different adjacent position between the series to avoid adaptation or sensitization. During the experiment, subjects were asked to attend to pain stimulus and VAS responses.

Figure 1.

Schematic diagram of the offset analgesia paradigm and pain responses. Gray lines show temperature and blue lines show temporal changes in VAS. The baseline temperature started at 32°C, followed by thermal stimulation at 45°C for 10 s (T1), then 46°C thermal stimulation for 5, 10, or 15 s, respectively (T2), followed by 20 s of 45°C thermal stimulation (T3), and finally back to baseline temperature. Maximum VAS value during the T2 and the first 10 s of T3 was defined as T_max. Minimum VAS at T3 during the 10 s after the VAS reached T_max was defined as OA_min. The magnitude of OA (ΔOA) was defined as ΔOA = T_max − OA_min, and the ratio of ∆OA to T_max (ΔOA/T_max) was defined as ΔOA%. Furthermore, maximum VAS latency was defined as the time until the maximum VAS (T_max) was reached by thermal stimulation, and OA latency was defined as the time from T_max to OA_min. OA indicates offset analgesia; VAS, visual analog scale.

The OA-related parameters were defined as shown in Figure 1. T_max was defined as the maximum VAS value during T2 and the first half of T3, and OA_min the minimum VAS during T3 within 10 seconds after T_max. The T_max − OA_min difference was defined as the magnitude of OA (ΔOA) and the ratio of ∆OA to T_max was defined as ΔOA%. Furthermore, maximum VAS latency was defined as the time from the beginning of T1 to T_max, and OA latency as the time from T_max to OA_min. These parameters were calculated as an average of 2 measures each, either at T2 = 5, 10, and 15 seconds.

MRI Acquisitions

Multimodal MRI was performed using a 3T MRI scanner (Signa HDxt3.0T Optima Edition, GE Healthcare) and an 8-channel birdcage head coil. An MRI-compatible headphone was used to minimize MRI noise and to communicate with research staff. A high-resolution, T1-weighted, 3-dimensional anatomical image of the whole brain was obtained with the following parameters: TR, 7.8 milliseconds; TE, 3 milliseconds; inversion time, 400 milliseconds; flip angle, 12°; FOV, 256 × 205 mm²; slice thickness, 1 mm; 137 contiguous sagittal slices, resulting in a voxel size of 1 × 1 × 1 mm³. DTI was obtained with the following parameters: number of applied axes, 30; b-value, 1000 sec/mm² with 1 b0 image; TR, 11,000 milliseconds; TE, 72 milliseconds; matrix, 128 × 128; FOV, 24 cm; number of slices, 46; slice thickness, 3 mm. The other imaging modalities included thermal stimulus-evoked and resting-state fMRI, whose results were reported elsewhere.^3,19

Data Preprocessing of Diffusion Tensor Images

The analysis was performed using the FMRIB Software Library (FSL) version 5.0.9 (www.fmrib.ox.ac.uk/fsl) and its toolboxes. First, all images were manually verified for the absence of significant artifacts or anatomical abnormalities. Eddy current correction and head motion correction were then performed using FMRIB’s Diffusion Toolbox (FDT).²⁵ Brain extraction toolbox (BET) was used to create an image from which nonbrain regions were excluded.²⁶ DTIFIT reconstruction of the FDT was used to obtain FA, which indicated the degree of diffusion anisotropy within a voxel and takes values between 0 and 1, with larger values indicating greater directional dependence of the Brownian motion of water molecules.⁶ Mean diffusivity (MD) was also calculated by fitting a tensor model at each voxel of diffusion data. Axial diffusivity (AD) and RD maps were calculated from each eigenvalue using the fslmaths command line. Specifically, the AD map was eigenvalue 1 and the RD map was calculated from the average of eigenvalue 2 and eigenvalue 3.

Global White-Matter Comparisons and Tract-Based Spatial Statistics Analysis

Global white-matter metrics (FA, MD, AD, and RD) were computed in individual subjects and compared between the groups using an unpaired Student t test. Then, whole-brain voxel-wise statistical analysis was performed on the volumes of each metric using tract-based spatial statistics (TBSS) by FSL.²⁷ Each subject’s FA data were nonlinearly registered to the FMRIB58 FA standard space template and aligned to the Montreal Neurological Institute space (MNI152). The mean image of all aligned FA images was then created and thinned to create a skeletonized mean FA image representing the centers of all tracts. It was thresholded at an FA value of 0.2 to exclude gray matter and cerebrospinal fluid. MD, AD, and RD images were also aligned to the MNI space and projected onto the mean FA skeleton. Nonparametric permutation methods (Randomise tool in FSL) were used to find differences in contrast between patients with chronic pain and healthy subjects (n = 5000, P < .05 with age and gender as covariates). For multiple comparisons, we used the threshold-free cluster enhancement²⁸ with family-wise error corrections.

Region-of-Interest Analysis

Significant white-matter clusters were localized on the MRI Atlas of Human White-Matter included in the FSL (ICBM-DTI-81 white-matter labels atlas, and JHU white-matter tractography atlas).^29,30 The regions-of-interest (ROIs) were created from the overlaps between the clusters and the atlas masks, and were used to extract average metric values from each individual image.

Statistical Analysis

The number of subjects was determined conveniently based on sample sizes in other similar studies.^8,11,13 We used the SPSS software (version 23, IBM) for statistical analysis with P < .05 considered significant. Demographic data such as age and the VAS, indicating current pain intensity, and psychophysical questionnaires from the PCS, MPQ, PD-Q, and BDI were compared between the groups by an unpaired Student t test, after confirming normality with the Shapiro-Wilk test; gender using the χ² test. OA-related parameters and DTI metrics were compared between the groups by an unpaired Student t test or Mann-Whitney’s U test, after confirming normality with the Shapiro-Wilk test. A receiver-operating characteristic curve (ROC) analysis was performed for parameters that might help discriminating between patients and healthy subjects. Cutoff values, specificity, and sensitivity were calculated by Youden’s J statistic.

Correlation Analysis

Individual DTI metrics derived from the significant ROIs were sought for correlation with psychophysical and OA-related parameters, using Pearson’s analysis with false discovery rate (FDR) correction for multiple comparisons and significance threshold at P < .05.

RESULTS

Subject Demographics

Twenty-six patients with chronic pain and 18 healthy controls were recruited as shown in Table 1A. The patients showed significantly higher values in all the psychophysical parameters indicating morbidity of chronic pain. Etiologies of chronic pain consisted of fibromyalgia (n = 6), limb pain (6), low back pain (4), migraine (2), complex regional pain syndrome (2), phantom limb pain (1), cervical spondylosis (1), chronic widespread pain (1), postherpetic neuralgia (1), rheumatoid arthritis (1), and trigeminal neuralgia (1).

Table 1.

Demographics of Subjects and Pain Perception Profiles During OA

A. All the Subjects
	CP (n = 26)	HC (n = 18)	P
Number (male/female)	9/17	8/10	.545
Age (y)	48.8 ± 13.9	43.2 ± 12.2	.175
PD-Q	15.8 ± 8.2	1.9 ± 3.8	<.0001
MPQ	15.0 ± 9.1	0.17 ± 0.51	<.0001
PCS	31.4 ± 12.5	4.4 ± 7.0	<.0001
BDI	16.4 ± 12.2	3.3 ± 4.6	<.0001
VAS	50.7 ± 27.1	0.83 ± 2.57	<.0001
Pain duration (mo)	102.2 ± 84.1	n/a
B. The Subjects With Thermal Stimulation
	CP (n = 14)	HC (n = 15)	P
Number (male/female)	4/10	8/7	.264
Age (y)	43.1 ± 12.7	43.3 ± 12.1	.967
PD-Q	16.6 ± 8.4	2.2 ± 4.1	<.0001
MPQ	13.9 ± 9.5	0.07 ± 0.26	<.0001
PCS	29.2 ± 14.0	4.4 ± 7.4	<.0001
BDI	17.4 ± 13.5	3.1 ± 4.9	.001
VAS	46.4 ± 24.4	0.67 ± 2.58	<.0001
Pain duration (mo)	124.6 ± 104.0	n/a
C. Pain Perception Profiles During Thermal Pain Stimulation
	T2 (s)	CP (n = 14)	HC (n = 15)	P
	5	40.0 ± 30.8	47.7 ± 29.8	.446
∆OA	10	55.2 ± 26.7	49.1 ± 30.3	.572
	15	64.8 ± 27.8	53.4 ± 26.8	.274
	5	0.49 ± 0.35	0.70 ± 0.32	.093
∆OA%	10	0.68 ± 0.32	0.72 ± 0.31	.715
	15	0.76 ± 0.26	0.81 ± 0.22	.591
	5	79.3 ± 17.3	69.4 ± 29.3	.683
T max	10	82.3 ± 13.9	69.6 ± 27.0	.129
	15	85.5 ± 15.2	66.5 ± 28.7	.041
	5	13.9 ± 1.9	14.5 ± 1.5	.371
Maximum VAS latency	10	15.5 ± 3.1	16.0 ± 3.0	.681
	15	16.0 ± 3.7	18.9 ± 3.1	.032
	5	7.7 ± 2.3	6.8 ± 3.2	.439
OA latency	10	10.6 ± 2.9	9.6 ± 4.0	.486
	15	15.6 ± 3.5	11.1 ± 4.2	.004

Values are mean ± SD.

Abbreviations: ∆OA, magnitude of OA; ΔOA%, ΔOA/T_max; ΔOA/T_max; maximum VAS latency, time from the start of 45°C stimulus to T_max; BDI, Beck Depression Inventory; CP, patients with chronic pain; HC, healthy controls; MPQ, short-form McGill Pain Questionnaire; OA, offset analgesia; OA latency, time from T_max to the minimum VAS during T3; PCS, Pain Catastrophizing Scale; PD-Q, painDETECT Questionnaire; T_max, maximum VAS during T2 and the first half of T3; VAS, visual analog scale.

Offset Analgesia Responses and Latencies Differed Between Patients and Controls

Out of all the subjects, 14 of 26 patients and 15 of 18 controls underwent thermal pain stimulation before MRI. Those patients and controls did not differ significantly either in age or sex (Table 1B). Patients showed a significantly larger T_max at T2 = 15 seconds (P = .041) than controls (Table 1C). Patients tended to have smaller ΔOA% at T2 = 5 seconds than controls (P = .093), but showed no differences either in ∆OA and ∆OA% compared with controls at any T2 (Table 1C).

On the other hand, patients showed a significantly shorter maximum VAS latency (P = .032) and a longer OA latency (P = .004) than controls at T2 = 15 seconds (Table 1C), which distinguished patients from controls with a high degree of accuracy. For maximum VAS latency, a cutoff value of 17.3 seconds yielded a sensitivity of 0.667 and specificity of 0.857, with an area under the curve of 0.767 (95% confidence interval, 0.585–0.948, P = .015). For OA latency, a cutoff value of 13.9 seconds resulted in a sensitivity of 0.786 and specificity of 0.867, with an area under the curve of 0.819 (95% CI, 0.654–0.984, P = .003).

Global White-Matter Integrity Showed Diagnostic Capability of Chronic Pain

Mean global FA of the white-matter was significantly smaller in patients than in controls (0.44 ± 0.12 vs 0.45 ± 0.11, respectively, P = .012). Mean global RD of the white-matter was significantly larger in patients than in controls (0.00057 ± 0.00002 vs 0.00056 ± 0.00002, P = .038). Furthermore, mean global FA distinguished patients from controls with a cutoff value of 0.443, sensitivity of 0.667, specificity of 0.731, and an area under the curve of 0.716 (95% confidence interval, 0.556–0.875, P = .016). Mean global RD distinguished patients from controls with a cutoff value of 0.000567, sensitivity of 0.615, specificity of 0.778, and an area under the curve of 0.694 (95% confidence interval, 0.535–0.854, P = .03). On the other hand, global MD and AD did not differ between the groups (data not shown).

Regional White-Matter Integrity and Its Relationships with Psychophysical and OA-Related Parameters

TBSS analysis across all the 44 subjects revealed that patients showed decreased FA over vast areas of the brain across the corticothalamic and interhemispheric fibers including the corpus callosum, internal capsule, corona radiata, external capsule, anterior thalamic radiation (ATR), forceps minor, and forceps major compared with controls (Table 2A and Figure 2, upper panel). On the other hand, the patients showed increased RD in the corpus callosum, corona radiata, anterior thalamic radiation, and forceps minor compared with controls (Table 2B and Figure 2, lower panel). There were no significant differences between the groups in AD and MD.

Table 2.

White-Matter Tracts With Lower FA Values or Higher RD Values in Patients With Chronic Pain than in Controls

A. Lower FA
Cluster index	Cluster size (voxels)	Peak MNI coordinates	White-matter tracts	P
1	10,170	(−18, 4, 38)	Genu of corpus callosum Body of corpus callosum Splenium of corpus callosum Anterior limb of internal capsule (R, L) Posterior limb of internal capsule (R) Anterior corona radiata (R, L) Superior corona radiata (R, L) Posterior corona radiata (R) External capsule (R, L) Superior longitudinal fasciculus (R)	.012
2	1558	(22, −56, 30)	Splenium of corpus callosum Posterior corona radiata (R) Posterior thalamic radiation (R) Superior longitudinal fasciculus (R) Sagittal stratum (R)	.04
3	331	(−24, −80, 8)	Posterior thalamic radiation (L)	.035
B. Higher RD
Cluster index	Cluster size (voxels)	Peak MNI coordinates	White-matter tracts	P
1	438	(−13, 11, 29)	Genu of corpus callosum Body of corpus callosum Anterior corona radiata (L) Superior corona radiata (L)	.046
2	46	(−16, 6, 34)	Body of corpus callosum Superior corona radiata (L)	.049
3	37	(−9, 33,3)	Genu of corpus callosum	.049

Abbreviations: FA, fractional anisotropy; L, left; MNI, Montreal Neurological Institute coordinates; R, right; RD, radial diffusivity; White-matter tracts, Atlas from JHU ICBM-DTI-81 White-Matter labels.

Figure 2.

White-matter alterations in the patients with chronic pain. Upper panel, In the patients, FA decreased over vast areas of the brain across the corticothalamic and interhemispheric fibers including the corpus callosum, internal capsule, corona radiata, external capsule, anterior thalamic radiation, forceps minor, and forceps major compared with controls (blue clusters). Lower panel, RD increased in the corpus callosum, corona radiata, anterior thalamic radiation, and forceps minor compared with controls (red-yellow clusters). Nonparametric permutation methods (Rondomise tool) were used to find differences in contrast between patients with chronic pain and healthy subjects (n = 5000, P < .05 with age and gender as covariates). We used the threshold-free cluster enhancement with the family-wise error correction for multiple comparisons; FA indicates fractional anisotropy; green line, skeleton of white-matter; RD, radial diffusivity.

Across all the subjects, individual values of ROI-averaged FA and RD were examined for correlations with psychophysical parameters. FA showed significantly negative correlations with all the psychophysical parameters including VAS, BDI, PD-Q, MPQ, and PCS (P < .05, FDR-corrected) (Figure 3A), while RD showed significantly positive correlations with VAS, BDI, PD-Q, MPQ, and PCS (P < .05, FDR-corrected) (Figure 3B).

Figure 3.

Association between white-matter metrics and psychophysical parameters. A, ROI-averaged FA values were examined for correlation with psychophysical parameters in all subjects. All the psychophysical parameters were negatively correlated with FA (P < .05, FDR-corrected). B, ROI-averaged RD values were examined for correlation with psychophysical parameters in all subjects. All the psychophysical parameters were positively correlated with RD (P < .05, FDR-corrected). BDI indicates the Beck depression inventory; CP, patients with chronic pain; FA, fractional anisotropy; HC, healthy controls; MPQ, the short form McGill pain questionnaire; PCS, Pain Catastrophizing Scale; PD-Q, the painDETECT questionnaire; RD, radial diffusivity; VAS, visual analog scale.

In the patients, PCS showed particularly strong negative correlations with FA across the entire ROI (P = .04, r = −0.405), at the ATR (P = .045, r = −0.396), and at the forceps major (P = .018, r = −0.461) (Figure 4, left panel). Across all the subjects, the FA of the splenium of corpus callosum showed a positive correlation with maximum VAS latency at T2 = 15 seconds (P = .007, r = 0.493) and a negative correlation with OA latency at T2 = 15 seconds (P = .002, r = −0.552) (Figure 4, right panel). No other significant correlations were found among the other ROIs, psychophysical, and OA-related parameters.

Figure 4.

Associations of the region-specific FA with PCS and OA-related parameters. Left, In the patients, negative correlations were found especially high between (A) the FA at the anterior thalamic radiation (P = .045, r = −0.396) or (B) the FA at the forceps major (P = .018, r = −0.461) and PCS. No other significant correlations with the other ROIs and psychophysical parameters were found. Right, In all the subjects, The FA of the splenium of corpus callosum showed (C) positive correlation with maximum VAS latency (OA) at T2 = 15 s (P = .007, r = 0.493) and (D) negative correlation with OA latency at T2 = 15 s (P = .002, r = −0.552). ATR indicates anterior thalamic radiation; CC, splenium of corpus callosum; CP, patients with chronic pain; FA, fractional anisotropy; HC, healthy controls; maximum VAS latency, the time from start to T_max; OA, offset analgesia; OA latency, the time from T_max to OA_min; PCS, Pain Catastrophizing Scale.

DISCUSSION

We found a significant deterioration of white-matter function across the whole brain and at specific brain areas mediating interhemispheric and corticothalamic communication in patients with chronic pain compared with controls. This white-matter dysfunction correlated with morbidity of pain both in sensory and affective dimensions as well as temporal dynamics of OA. We further found that the patients showed a faster increase of pain during stimulus increase, but a slower decrease of pain after a 1˚C-decrement of stimulus than controls during the OA paradigm, in association with FA of the corpus callosum.

White-Matter Dysfunction as a Common Signature of Chronic Pain

The white-matter abnormalities were reported in specific disorders of chronic pain. For example, decreased FA was reported in the genu, body, and splenium of the corpus callosum in cervical spondylosis;⁷ in the right dorsal cingulum in irritable bowel syndrome.⁸ Other reports described similar abnormalities in chronic musculoskeletal pain (corpus callosum, corona radiata, external capsule, and posterior thalamic radiation),⁹ complex regional pain syndrome (left cingulum-callosal bundle),¹² temporomandibular disorder (anterior thalamic radiation and internal capsule),³¹ dysmenorrhea (splenium of corpus callosum, internal capsule, corona radiata, and posterior thalamic radiation),¹³ and migraine (bilateral external capsule).¹⁴

Consistent with all those earlier results, we showed that patients with chronic pain, even from heterogeneous etiologies, showed impaired white-matter integrity over vast areas of the brain including the corpus callosum, internal, external capsules, and thalamocortical pathways (Table 2 and Figure 2). We suggest those findings might constitute a common signature of white-matter dysfunction in chronic pain: FA decreased in the interhemispheric and prefrontal-thalamic fibers; while RD increased in the anterior interhemispheric fibers. Those metrics were robust enough to distinguish patients from controls, supporting feasibility of those local white-matter metrics in diagnosing chronic pain. Notably, such impaired white-matter integrity was so widely distributed over the whole brain that even mean global values of FA and RD were sensitive enough to distinguish patients from controls.

White-matter abnormalities have also been noted in many other CNS diseases. For example, a wide region of FA reductions, including corpus callosum, have been noted in diseases such as schizophrenia and major depressive disorder, and the association with chronic pain is intriguing and requires further specificity studies.

Significance of Interhemispheric White-Matter Dysfunction in Chronic Pain

FA of the white-matter is a measure for integrity of myelin sheaths, and is known to decrease by axonal damage, inflammation, and demyelination.⁶ Increased RD may also be associated with impaired myelin formation.⁶ In this study, FA decreased over vast brain areas, particularly the corpus callosum (CC), a major fiber bundle connecting the cerebral hemispheres, anatomically subdivided into the genu (gCC), body (bCC), and splenium of corpus callosum (sCC). The CC plays an important role in integrating sensory, cognitive, and motor information between the hemispheres. For example, the gCC is composed of fibers connecting the bilateral prefrontal, anterior cingulate cortices, supplementary motor, and premotor areas. The bCC also connects the bilateral sensory cortices in the precentral and postcentral gyri. The sCC is known to connect the bilateral occipital, posterior parietal, and temporal lobes, which are involved in processing visual and cognitive information.³² FA decrease, particularly in the gCC and bCC, might potentially be associated with impairments in integrating pain perception, cognition, and motor information in chronic pain. In addition, impaired bilateral interaction of the prefrontal and anterior cingulate cortices might be concerned with negative emotions in pain, because FA was negatively correlated with PCS in the majority of regions including gCC and bCC. The present findings were consistent with those of earlier reports.^7,33,34

Significance of Corticothalamic White-Matter Dysfunction in Chronic Pain

The ATR is a bundle of white-matter fibers that connects the dorsolateral prefrontal cortex (DLPFC) and thalamus. Decreased FA and increased RD of the ATR might imply dysfunction of the descending pain modulatory system.³¹ DLPFC plays a major role in activating descending pain modulation, and decrease in volume as well as the thalamus in patients with chronic low back pain,⁵ which was consistent with the ATR dysfunction as shown here.

The ATR dysfunction was associated with increased PCS, a behavioral marker for catastrophizing perception of pain.²¹ PCS and its rumination subscale were shown to predict pain disorders including chronic postoperative pain^35,36 and efficacy of pain treatment.³⁷ Taken together, deteriorated top-down, prefrontal-to-thalamic, suppression of pain through the ATR might be a potential biomarker for emotional morbidity of chronic pain.

Temporal Sharpening of Pain Was Evident by Latencies Rather Than Magnitude of OA

This study failed to show a significantly decreased OA magnitude in patients with chronic pain in contrast to our previous report.¹⁸ Further negative results showed intact OA in patients with chronic temporomandibular joint disorders;³⁸ reduced OA in the trigeminal region but intact OA on the forearm in patients with chronic migraine compared with healthy controls.³⁹ Based on those findings, decreased OA magnitude has not been reproducible in patients with chronic pain from heterogeneous etiologies.

On the other hand, we found that maximum VAS latency was shorter and OA latency was longer at T2 =15 seconds in patients with chronic pain than in controls. That is, patients with chronic pain felt pain faster on stimulus onset, but felt relief from pain more slowly on stimulus offset than controls. A longer T2, at 15 seconds, might have allowed pain perception to maximize during T1/T2 and to decrease to a deeper trough (larger ∆OA%), which helped visualizing differential OA latencies of pain perception between patients and controls. Remarkably, such latencies of pain perception might well have reflected the extent of temporal sharpening of pain more efficiently than OA magnitude.

Possible Impairment of Interhemispheric Pain Suppression in Chronic Pain

We found that the patients showed a shorter latency to pain maxima (maximum VAS latency) and a longer latency from pain maxima to the OA trough (OA latency) than controls. In other words, the patients perceived pain faster on stimulus onset, but relief from pain slower on stimulus decrease than controls. Those indices distinguished patients from controls, and proved associated with FA of the sCC (Figure 4, right panel). That is, subjects with smaller FA of sCC showed significantly faster pain onset on stimulus increase but slower development of OA on stimulus decrease. If we hypothesize that interhemispheric suppression of pain through the sCC might reduce the speed of pain increase (during T1/T2) and facilitate the speed of pain decrease (during T3), the sCC dysfunction might well have caused hyperalgesia on stimulus onset and retarded OA on stimulus decrease in the patients.

Therefore, we suggest a possible role for the sCC in interhemispheric modification of pain perception, dysfunction of which might be one of the cerebral mechanisms of pain chronicity. Such hypothesis was also supported by a negative association between FA of the forceps major, including sCC, and emotional morbidity of pain as measured by PCS.

Limitations

First, diverse etiologies of chronic pain might have obscured disease-specific alterations. The present findings, consistent with earlier findings from specific etiologies, should rather be interpreted as a common signature of pain chronicity. Second, although the current number of subjects was comparable with those of earlier neuroimaging studies,^8,11,13 further validation studies in larger cohorts of patients should be warranted. Differences in the white-matter metrics, nonetheless, were found robust enough to survive rigorous statistics with multiple comparisons and permutation tests, which endorsed the reliability of the present findings. Third, we did not control or cease therapeutic drugs for the benefit of patients. However, dynamics of OA were not likely to have been affected by those drugs.^16,40 Plastic alterations of the white-matter metrics might also have been free from drug-induced effects on vasculature and neurons. Fourth, cross-sectional design of this study did not allow causal assumptions. We must await future longitudinal studies to test whether OA, psychophysical, and white-matter alterations might be a cause or result of chronic pain.

CONCLUSIONS

Patients with chronic pain showed a faster increase of pain on thermal nociceptive stimulus onset but a slower decrease of pain on stimulus offset than controls during the OA paradigm. They also showed impaired white-matter integrity in the whole brain, specifically in the right anterior thalamic radiation and the corpus callosum mediating interhemispheric communication, in association with altered OA dynamics and pain catastrophizing. Those findings might imply dysfunction of both the descending pain modulatory system and interhemispheric modulation of pain, which might be postulated as a novel cerebral mechanism of pain chronification.

DISCLOSURES

Name: Jiro Kurata, MD, PhD.

Conflicts of Interest: J. Kurata is a section editor of Anesthesia & Analgesia.

This manuscript was handled by: Jianren Mao, MD, PhD.

ACKNOWLEDGMENTS

The authors cordially thank the following people: Shigehito Sawamura, MD, PhD, Department of Anesthesiology, Teikyo University School of Medicine, Tokyo, Japan, for lending PATHWAY; Hiroyuki Kobinata, MD, PhD, Eri Ikeda, MD, PhD, Shuo Zhang, MD, PhD, Takeshi Kamma, MD, Department of Anesthesiology, Tokyo Medical and Dental University, Tokyo, Japan for their help in recruiting subjects and performing experiments; Department of Radiology, Tokyo Medical and Dental University, Tokyo, Japan for their help in neuroimaging procedures.