OUP user menu

fMRI reveals distinct CNS processing during symptomatic and recovered complex regional pain syndrome in children

A. Lebel, L. Becerra, D. Wallin, E. A. Moulton, S. Morris, G. Pendse, J. Jasciewicz, M. Stein, M. Aiello-Lammens, E. Grant, C. Berde, D. Borsook
DOI: http://dx.doi.org/10.1093/brain/awn123 1854-1879 First published online: 20 June 2008

Summary

Complex regional pain syndrome (CRPS) in paediatric patients is clinically distinct from the adult condition in which there is often complete resolution of its signs and symptoms within several months to a few years. The ability to compare the symptomatic and asymptomatic condition in the same individuals makes this population interesting for the investigation of mechanisms underlying pain and other symptoms of CRPS. We used fMRI to evaluate CNS activation in paediatric patients (9–18 years) with CRPS affecting the lower extremity. Each patient underwent two scanning sessions: once during an active period of pain (CRPS+), and once after symptomatic recovery (CRPS). In each session, mechanical (brush) and thermal (cold) stimuli were applied to the affected region of the involved limb and the corresponding mirror region of the unaffected limb. Two fundamental fMRI analyses were performed: (i) within-group analysis for CRPS+ state and CRPS state for brush and cold for the affected and unaffected limbs in each case; (ii) between-group (contrast) analysis for activations in affected and unaffected limbs in CRPS or post-CRPS states. We found: (i) in the CRPS+ state, stimuli that evoked mechanical or cold allodynia produced patterns of CNS activation similar to those reported in adult CRPS; (ii) in the CRPS+ state, stimuli that evoked mechanical or cold allodynia produced significant decreases in BOLD signal, suggesting pain-induced activation of endogenous pain modulatory systems; (iii) cold- or brush-induced activations in regions such as the basal ganglia and parietal lobe may explain some CNS-related symptoms in CRPS, including movement disorders and hemineglect/inattention; (iv) in the CRPS state, significant activation differences persisted despite nearly complete elimination of evoked pain; (v) although non-noxious stimuli to the unaffected limb were perceived as equivalent in CRPS+ and CRPS states, the same stimulus produced different patterns of activation in the two states, suggesting that the ‘CRPS brain’ responds differently to normal stimuli applied to unaffected regions. Our results suggest significant changes in CNS circuitry in patients with CRPS.

  • children
  • pain
  • fMRI
  • plasticity
  • reflex sympathetic dystrophy

Introduction

In adults, complex regional pain syndrome (CRPS) is a clinical syndrome featuring severe pain, hypersensitivity to innocuous (allodynia) and noxious (hyperalgesia) somatosensory stimuli, autonomic/neurovascular signs (coldness, poor circulation, abnormal sweating, swelling and skin discolouration) and trophic signs (abnormal hair and nail growth, muscle atrophy, joint contractures) (Wasner et al., 1998; Sieweke et al., 1999; Birklein et al., 2000). Though the pathophysiology of CRPS has not been clearly defined, it has been suggested to be a form of painful small fibre sensory neuropathy (Santiago et al., 2000). CRPS types I and II correspond approximately to the older terms ‘reflex sympathetic dystrophy’ (RSD) and ‘causalgia’, respectively (Grabow et al., 2004), but may in fact be similar diseases (Oaklander et al., 2006). CRPS, as a chronic neuropathic pain syndrome, likely involves peripheral and central sensitization of neuronal function (Janig and Baron, 2002). At the level of the spinal cord and brainstem, these disturbances are thought to be related to the altered neuroplasticity that leads to abnormal central pain processing (Porreca et al., 2002; Dubner, 2004). Current thinking is that CRPS is a disease of the CNS (Janig and Baron, 2002) with a phenotype that includes alterations in autonomic function (Birklein et al., 1998; Meier et al., 2006); sensory systems (Drummond and Finch, 2006; Maihofner et al., 2006b), as indicated by pain that progresses from its initial site proximally up a limb and even to the contralateral side (Maleki et al., 2000); and motor systems (Verdugo and Ochoa, 2000), as indicated by associated dystonia and movement disorders. Higher level functions such as visuospatial perception are also involved (Sumitani et al., 2007), as indicated by neglect-like symptoms (Galer et al., 1995; Galer and Jensen, 1999; Frettloh et al., 2006; Maihofner and DeCol, 2007).

Recent studies investigated alterations in brain function in adult patients with CRPS (Maihofner et al., 2003, 2005, 2006a; Pleger et al., 2006). Von–Frey stimulation of the affected limb evoked pinprick hyperalgesia and produced greater contralateral activation than identical stimulation of the unaffected limb in primary (S1) and secondary (S2) sensory cortex, insula, anterior cingulate cortex and frontal cortices (Maihofner et al., 2005). Mechanical allodynia evoked by brushing the affected limb was reported to correspond with activation of motor (M1) and cognitive regions (frontal regions), areas involved in emotional processing (e.g. anterior and posterior cingulate cortex, temporal lobe), parietal association cortices, as well as pain sensory regions (e.g. S1, insula) (Maihofner et al., 2006a). Of note was the significant negative activation in visual, posterior insular and temporal cortices in response to brushing that evoked allodynia. Other types of chronic pain, such as phantom limb pain, have been shown to produce significant cortical reorganization (Flor et al., 1995; Karl et al., 2001). In a recent study using magnetic source imaging, cortical reorganization was reported in the contralateral SI cortex in patients with CRPS (Maihofner et al., 2004). The reorganization involved parts of the body (lips and fingers) that did not have pain, but exchanged representations following recovery from CRPS. An fMRI study also demonstrated cortical reorganization in CRPS: BOLD signals in contralateral S1 and S2 induced by innocuous electrical stimulation were reduced in the affected limb in comparison with the unaffected limb (Pleger et al., 2006).

Studies in adults have evaluated the fixed disease, presumably because it is difficult to recruit subjects with CRPS who respond to therapy in a short time period. In contrast, CRPS symptoms in paediatric patients frequently fluctuate and often clinically resolve within several months to 2 years (Low et al., 2007). While a specific and accepted definition of CRPS in children does not exist, this population affords several advantages in the study of CRPS: (i) unlike adults with CRPS, paediatric patients are generally free of additional complicating neuropathic pain conditions; (ii) children are more likely to demonstrate robust neuronal plasticity in an otherwise normal brain and (iii) paediatric imaging provides insights into specific brain regions involved in dysfunction (e.g. precursors of autonomic change, movement disorders, etc). Although these differences exist, the basic CNS changes in paediatric CRPS may be similar to those in adults. Thus, imaging pain-related responses in paediatric patients with CRPS may facilitate identification of CNS changes associated with this challenging chronic pain disorder.

Here we report on pain-related BOLD responses in a group of paediatric patients studied during fulminating CRPS and again following their clinical recovery. By studying the patterns of CNS activation evoked by mechanical and thermal stimuli to affected and unaffected regions in children with CRPS, we hoped to identify changes in the CNS response that correlate with the pain experienced during stimulation of the affected limb and thus gain insight into the mechanisms that underlie pain and other symptoms of CPRS. We also sought to determine whether the changes in activation associated with the CRPS+ state disappear or persist upon resolution of CRPS pain.

Methods

Subjects

Human experimentation in children

The experimental procedure was approved by the institutional review boards (IRB) at both the McLean Hospital (for brain imaging) and at the Children's; Hospital Boston (for patient recruitment). Because this was a study involving pain in children, special procedures were adopted. One such safeguard was to halt the pain stimulus if the subjects reported a pain visual analog score (VAS) of >8/10. In addition to parental consent, parents were present during all steps of the study. A post-scan evaluation questionnaire was completed by subjects to document their experience in the scanner and the painful stimuli they had received. In addition, as part of the IRB oversight, a report was sent to the IRB upon completion of each scanning session. The following nomenclature is used throughout the report for disease or pain state = CRPS+ or C; for post-CRPS/recovered state = CRPS or P; for affected limb = A, for unaffected limb = U, for brush = brush and for cold = cold. For example, CAbrush would indicate CRPS+ state, affected limb, for brush and PUbrush would represent post-CRPS, unaffected limb for brush.

Inclusion criteria

Twelve patients aged 9–18 years (13.5 ± 1.63 years, mean ± SEM) with CRPS affecting the lower extremity unilaterally were recruited from the clinical caseload of the Pain Management Center at Children's; Hospital Boston. For functional magnetic resonance imaging during an attack, patients needed to have (i) refrained from using symptomatic pro re nata (when necessary) analgesic drugs at least 4 h prior to the examination; (ii) experienced a moderate to severe pain (i.e. pain intensity greater than 5 on a visual analog scale) and (iii) experienced unilateral limb pain (Fig. 1A).

Fig. 1

(A) Maps of region affected in lower extremity for each subject. The spatial extent of spontaneous (yellow) and evoked (red) pain for each subject as determined during the initial clinical evaluation. (B) Paradigm. The experimental paradigm for functional imaging. Subjects underwent two scanning sessions: Scan 1 with the pain (CRPS+ state) and Scan 2 without the pain (CRPS state). For each scanning session, the area most sensitive to presented stimuli was identified prior to the scans. During each functional scan, cold or brush stimuli were applied to these areas on the affected (red) and unaffected mirror region (green) limbs. Stimuli were presented in a semi-randomized fashion.

Exclusion criteria

These included: (i) claustrophobia; (ii) significant medical problems such as uncontrolled asthma or seizure disorder, acute cardiac disease, psychiatric problems and other (non-CRPS) neurological disease; (iii) pregnancy; (iv) magnetic implants of any type and (v) weight >285 lbs. Informed consent and patient assent were obtained from all subjects and their parents.

Experimental plan

The experimental procedure is outlined in Fig. 1B. Patients meeting experimental criteria were studied twice during separate sessions [Session I—painful state (CRPS+) and Session II—non-painful state (CRPS)]. The subject's response to mechanical (brush) stimuli and thermal pain thresholds were measured in random order in several cutaneous areas on both lower limbs during each session (QST I and II). Following the completion of the quantitative sensory testing, patients underwent fMRI scanning during the application of mechanical and thermal stimuli to their affected region and mirror side. Patients were able to stop the study at any point. All experiments were performed using a 3T (Siemens Allegra) open body magnet.

Quantitative sensory testing (QST) and in scanner psychophysical measures

QST

Patients were tested in a quiet, temperature-regulated (25°C) room at the Brain Imaging Center at McLean Hospital. During the pre-scan testing, cold thresholds and responses to mechanical stimuli (pain intensity and defining the spatial extent of mechanical allodynia) were measured in random order in several cutaneous areas as appropriate in the painful and non-painful state: (i) the pain region within the ipsilateral-affected skin (Fig. 1A) and (ii) the corresponding contralateral (mirror) region. To determine cold pain thresholds, the skin was cooled down linearly at a slow rate (1°C/s) until pain sensation was perceived, at which time the subject stopped the stimulus by pressing a button on a patient response unit (method of limits). To determine pain evoked by mechanical allodynia, the skin was brushed with a hand-held soft bristle brush. The use of two different modalities (cold and mechanical) allowed detection of sensitization within the brain during the painful (sensitized) and non-painful clinical state.

In scanner psychophysical measures

Pain ratings (VAS 0–10) for the stimuli were obtained within the scanner using a turn-dial and visualized screen prompt. In addition, subjects were asked to complete a Post-Study Questionnaire (Supplementary Data) following each study. The questionnaire asked questions regarding the subjects’ level of pain (e.g. ‘How painful did you find the sensory testing part?’), and questions regarding potential future participation (see Web B for details on the questionnaire).

Functional imaging

Upon the completion of the QST and the determination of thermal pain threshold in the lower extremities, subjects were placed in the magnet for functional imaging. After standard anatomical scans, functional scans were obtained in a semi-random sequence for brush and cold stimulation of the lower extremities. Two sets of four functional scans were collected for each side of the body, with two scans for brush and two scans for 1°C below cold pain threshold on the affected side. Each brush scan was comprised of two stimuli (25 s duration, 1 Hz, 30 s inter-stimulus interval). For the cold scans, two pulses of cold stimuli (cold pain threshold −1°C; ramp: 4°C/s; duration 25 s stimulus interval: 30 s inter-stimulus interval) were applied to the same skin areas during both visits. Baseline temperature in each case was 32°C. Thermal stimuli were applied using a 3.0 × 3.0 cm2 Peltier thermode. These devices for use in the fMRI environment were developed at the NMR Center at the Massachusetts General Hospital with Medoc, Haifa, Israel.

fMRI data acquisition and psychophysics data acquisition

Subjects were scanned on a 3.0 T Trio (Siemens) using a quadrature Siemens head coil. We used specially defined imaging parameters that have been established for the cortex and brainstem. Anatomical images were acquired using a magnetization prepared rapid gradient echo (MPRAGE) sequence [128 1.33-mm-thick slices with an in-plane resolution of 1 mm (256 × 256)]. Magnitude and phase images were then acquired to unwarp functional scans. Slice location, number and thickness were the same as the ones used in the functional scans.

Functional image datasets were processed and analysed using FSL (FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl) (Smith et al., 2004). Processing steps included spatial filtering (FWHM = 6 mm), high-pass temporal filtering (σ = 70.0 s), motion correction, geometric unwarping of EPI images, removal of all non-brain areas in images and mean-based intensity normalization. In addition, motion of >3 mm resulted in subjects being excluded; all data was motion corrected. First-level fMRI analysis was performed on individual subject's data using FMRI Expert Analysis Tool (FEAT) using FMRIB's Improved Linear Model (FILM) with local autocorrelation correction (Woolrich et al., 2001). For the brush stimuli, stimulus application timings were used as explanatory variables (EVs). For the cold stimuli, recorded temperature profiles were used as EVs, in addition to their temporal derivatives. The EVs were convolved with a gamma function with a 3 s SD and a 6 s haemodynamic lag. The resulting statistical parametric maps from first-level analysis were registered to the MNI 152 Brain (Montreal Neurological Institute) using FMRIBs Linear Image Registration Tool (FLIRT www.fmrib.ox.ac.uk/fsl). The registered statistical parametric maps were used in the group-level analysis. In addition to comparisons, independent analysis of CRPS+ and CRPS states for brush and cold were performed. For subjects that had their right leg affected, brains were flipped as we have previously described (Moulton et al., 2007) to allow for inter-subject comparisons.

Biostatistical analysis of functional scans

All experiments involved a scan in the pathological state (i.e. pain, CRPS+) and another in the normal (i.e. pain-free, CRPS) state. The analysis to test each hypothesis consisted of paired t-tests of main effects as determined by random effects analysis.

Group-level analysis was carried out using a two-sample paired t-test using FEAT. In total, eight comparisons were completed—four comparisons between pre- and post-CRPS recovery and four comparisons between the affected and unaffected CRPS sites. As shown in Fig. 2A and summarized in Table 1, we performed the following analyses on the processed data: Analysis 1—Averaged data for activation patterns for each stimuli applied to the affected and unaffected limbs during the CRPS+ (C) or CRPS (P) states. Here we wished to evaluate the de novo effects of brush or cold on brain activation when stimulating the most painful region affected, or the mirror location on the unaffected limbs. Analysis 2—Contrast analyses were performed to compare activation during the CRPS+ (C) and CRPS (P) state for each limb and between limbs. Paired t-tests were used to determine differences in these analyses. Specifically, the following contrast analyses were performed: (i) contrast of affected versus unaffected limb activation in the CRPS+ state; (ii) contrast of affected limb activation in CRPS+ versus CRPS state; (iii) contrast of unaffected limb activation in CRPS+ versus CRPS state and (iv) contrast of affected versus unaffected limb activation in the CRPS state.

Fig. 2

(A) Analysis approach. Flow chart of the analyses performed for the CRPS+ and CRPS conditions, including generation of average maps for affected and unaffected states (for brush and cold) and contrast analyses. These contrast analyses allowed for the determination of effects of limb on CRPS state and effects of CRPS state on limb. (B) Contrast analyses. The six (I–VI) outcomes for comparison of Condition A (affected) to Condition B (control). The Condition C (outcome) is defined as noted under ‘Terminology’. Each outcome falls under one of the following categories: Increased response, decreased response or valence change (see text).

View this table:
Table 1

Matrix for data analysis

AnalysisPain stateData setTable
AverageCRPS+ stateCAbrushA Web
CAcoldB Web
CUbrushC Web
CUcoldD Web
CRPS statePAbrushE Web
PAcoldF Web
PUbrushG Web
PUcoldH Web
ContrastAffected CRPS+ versus affected CRPSCAbrush versus PAbrush CAcold versus PAcold3 4
Affected CRPS+ versus unaffected CRPS+CAbrush versus CUbrush CAcold versus CUcold5 6
Affected CRPS versus unaffected CRPSPAbrush versus PUbrush PAcold versus PUcold7 8
Unaffected CRPS+ versus unaffected CRPSCUbrush versus PUbrush CUcold versus PUcold9 10
  • CRPS sate: C = CRPS+ state (pain state); P = CRPS state (no pain state). Limb: A = affected; U = unaffected. Stimulus: brush, cold.

Modification of contrast analysis to reflect differences in activation valence

Contrast analysis requires further refinement to account for positive and negative activations in the averaged group data. Specifically, given that the analysis is carried out on a voxel-by-voxel fashion, it is possible to compare a positive activation in one condition with a negative one in another. For proper comparisons (contrasts), we have grouped the potential six situations into three outcomes described in Fig. 2B. The outcomes are: (i) increased activation of the test condition (A) versus control condition (B); (ii) decreased activation of test versus control and (iii) valence change, meaning that activation in condition A had a different valence (sign) compared to condition B (e.g. positive activation in A and negative activation in B). An increase in activation of (A) versus (B) could result from a larger positive activation in A compared to a smaller positive activation in B, or from a more negative activation in A compared to a less negative activation in B (Fig. 2B). In the former case the contrast is positive while in the latter is negative; however, by tracking the sign of the conditions A and B we can properly label both as increased activations. Valence changes reflect potentially significant changes in brain activation from one condition to another. The potential for one area to be positively activated in condition A and negatively activated in condition B indicates, more than a large contrast difference, a change in brain response from probable excitation to inhibition. Hence, those changes cannot be discussed by simply observing the contrast difference. In order to perform these analyses, group statistical maps for each condition (A and B) are used to generate masks for the different situations: positive activation in conditions A and B (posA, posB) and negative activation in conditions A and B (negA, negB). The contrast results (A versus B, meaning A minus B) are assessed with the use of these masks; the overlap mask of posA and posB applied to the contrast map will reveal areas for which A is larger than B (positive contrast posC) and areas for which A is smaller than B (negative contrast negC) (Fig. 2B). The other situations are generated in a similar fashion as depicted in Fig. 2B.

Right versus left brain analysis

In order to compare the affected and unaffected sides, each functional image was flipped along the y-axis (anterior–posterior axis) before being registered to the standard brain (see later). Thus, the analysis determined which regions were modulated relative to the side of stimulation. The use of flipped brains in fMRI analysis is a well-described procedure in clinical pain studies (Maihofner et al., 2006a; Pleger et al., 2006; Schweinhardt et al., 2006).

fMRI data thresholding

For each group-level comparison, activated and deactivated voxels were determined by using a Gaussian mixture modelling (GMM) approach. Regions having similar patterns of activity can be expected to have differently scaled and shifted Gaussian distributions. This variability was accounted for by using GMM, a multiple comparisons-based analysis generally used for unsupervised classification of data into multiple categories (Pendse et al., 2006). A minimum cluster criterion of seven voxels in original space was also implemented.

Clusters of activation

The zstat map can be considered as a 3D function f (x, y, z) with peaks (maxima) and valleys (minima). The goal of the clustering procedure was to discern peaks by growing connected clusters around each maximum in the first step. In the next step, all voxels above the chosen threshold were assigned to clusters from step 1 using a minimum cluster distance. After active (deactive) voxels were associated with a particular cluster of activation, a region of interest (ROI) was assigned to a cluster based on the coordinate of its maxima as determined using the WFU_PickAtlas tool.

Brain region nomenclature

The nomenclature of the Pick Atlas (Lancaster et al., 2000; Maldjian et al., 2003) is used for all references to brain regions.

Results

Patients

Twelve paediatric patients with clinical criteria meeting CRPS (Meier et al., 2006) were recruited to the study. Four patients fell out of the study for the following reasons: three did not return following their first session for non-specific reasons unrelated to the experience of the first scan (incomplete resolution of pain, involvement in other activities, travel distance), and one felt claustrophobic during her first scan. Only those subjects that completed the protocol and provided usable fMRI datasets are included in this report (n = 8).

A summary of patient demographics is presented in Table 2. The mean age was 13.5 ± 1.63 (SEM) years old, and the average duration of their disease (pain) was 13.3 ± 2.35 months (mean ± SEM). Most patients had developed CRPS subsequent to a trauma. All eight were girls with unilateral pain in the lower extremity; six were affected on the left side. In all cases, the subjective distribution of pain was confined to the affected limb (Fig. 1A). Representative clinical profiles of two patients are provided in the supplementary information (Supplementary DataPatient Medical Histories).

View this table:
Table 2

Patient demographics

SubjectSexAgeEtiologySpontaneous pain (VAS)Laterality locationCRPS onsetCRPS offsetDuration (month)
004F14Neuralgia of peroneal nerve8.0L knee7/20037/200412
005F17Twisted ankle1.0L ankle9/20037/200510
006F16Unknown8.0L foot10/20047/20059
007F11Injury to peroneal nerve4.0L foot6/20048/22/0512
008F10Ankle sprain4.0L foot5/20049/200516
011F14Knee injury Arthroscopic surgery4.0L knee6/20059/200515
012F16Dislocated patella Injury to peroneal Nerve8.0R knee3/20056/200615
014F10Unknown4.0R leg1/20044/200627
Mean±13.505.1313.30
SEM1.631.612.35

Psychophysical results document reversal of pain symptoms

Ratings outside the magnet

The average spontaneous pain rating prior to the first scan was 5.1 ± 1.6 (mean ± SEM) on a VAS scale of 0–10 (Table 2) and no spontaneous pain at the time of their second scan. Prior to each scanning session, subjects underwent psychophysical evaluations for thermal stimuli (cold pain threshold) and for mechanical allodynia (brush), once during a time when they had pain (CRPS+), and again when they were asymptomatic for spontaneous pain (CRPS). VAS scores for pain elicited by brush prior to scanning were 4.8 ± 0.4 (CRPS+) and 1.0 ± 0.3 (CRPS) (Fig. 3A). The average cold pain threshold in the CRPS+ state was 5.9 ± 0.2°C (mean ± SEM) and 2.1 ± 0.3 (mean ± SEM) in the CRPS states (Fig. 3A). These differences in pain intensity in the CRPS+ versus CRPS states were significant (P < 0.01; Student's t-test).

Fig. 3

Psychophysical measures. (A) Pre-scanning group VAS pain ratings: VAS ratings (0–10; ± SEM) for cold stimuli and brush stimuli collected prior to scanning, during the CRPS+ (Visit 1) and CRPS state (Visit 2). (B) Group VAS pain ratings during the functional scans: Average (± SEM) maximal pain ratings by the eight subjects for brush and cold for affected versus unaffected limb for Visit 1 and Visit 2.

Ratings inside the magnet

All subjects rated their pain inside the magnet using a dial. No subject terminated scanning due to excessive pain, as their VAS levels never exceeded 8/10. Average peak ratings for cold and brush for each scanning session (CRPS+ and CRPS) were determined (Fig. 3B), revealing a large difference in ratings for the affected (A) versus unaffected (U) limb for each visit. The more clinically significant changes (P < 0.01) are those observed between visit 1 (CRPS+) and visit 2 (CRPS) for both brush and cold. All subjects initially presented with dynamic mechanical and cold allodynia with associated spontaneous pain, and showed dramatic reduction of both in the CRPS state (P < 0.01 for both, Fig. 3B). There was no difference between visits in VAS scores for brush applied to the unaffected limb. Unexpectedly, cold applied to the unaffected limb elicited a slightly greater VAS score in the CRPS state (P = 0.06).

Post-fMRI ratings

Subjects completed a questionnaire that documented their experience with the stimuli and the overall experiment (Supplementary Data—Post Scan Questionnaire). Verbal reporting of pain levels following cold and brush decreased significantly between visit 1 (CRPS+) and visit 2 (CRPS), as did reported levels of anxiety (Fig. A, Supplementary). Responses to the question ‘How painful did you find the fMRI/brain picture part?’ were very low and did not change between visits. These results are consistent with expectation given that patients had both spontaneous and evoked pain at the time of visit 1, and only very minimal evoked pain during visit 2.

fMRI results

The fMRI results were compared as shown in Table 2 and reported for each condition or comparison in Tables A–H (Web) and Tables 3–6. For the two subjects with affected right limbs, brain images were flipped to allow comparison, thus laterality (Lat) in the tables refers to the corrected brain side where left (L) is ipsilateral to the affected limb and right (R) is contralateral to the affected limb. Foci of activation are listed based on their location within structures in the three main regions of the brain (cortical, subcortical and brainstem/cerebellum). Tables show the coordinates (mm), statistical significance (z-stat) and the volume of activation for each activation cluster.

Analysis 1: averaged data

Brain BOLD responses to mechanical and cold stimulation during active paediatric CRPS and following recovery

Brush or cold applied to the affected (painful) region or to the mirror region elicited significant changes in BOLD signal in many brain regions (complete results presented in Tables A–H, Supplementary Material). Some of the most salient are summarized later.

CNS response to stimulation of the affected limb during active CRPS (CRPS+)

Widespread cortical changes in BOLD following brush or cold stimuli are similar to the pattern of BOLD response seen in adult CRPS patients with cold allodynia

Brush and cold stimuli elicited activation in many areas of the cortex. Cortical regions showing increased activation following brush stimulation included the parietal lobe, contralateral and ipsilateral SI, the anterior and middle cingulate, and a number of regions in the anterior insula (Supplementary Table A). These areas are involved in primary sensation, sensory association and emotional processing (Apkarian et al., 2005). An even greater number of regions showed brush-induced decreases in BOLD signal, including regions in the frontal lobe (which is involved in cognitive and emotional processing) and the parietal lobe. Additional foci of decreased BOLD signal were observed in the middle cingulate, middle temporal lobe and in the parahippocampus and hippocampus. Although VAS ratings for cold and brush to the affected limb were similar (∼6/10 for cold and ∼5/10 for brush), cold applied to the affected area produced many fewer changes in BOLD signal than brush; all but one of these changes were decreases (Supplementary Table B). Some of these decreased activations occurred in the same regions affected by brush (e.g. middle temporal region), but some were specific to cold. For example, cold resulted in decreased activation in the nucleus accumbens (NAC), which our earlier studies suggest may be an aversive readout for thermal pain (Becerra et al., 2001; Becerra and Borsook, 2008). These changes in cortical BOLD patterns are similar to the pattern of BOLD response seen for cold allodynia in adult CRPS patients (Seifert and Maihofner, 2007). The relative lack of sub-cortical activations is somewhat surprising, since allodynia in chronic pain is associated with activation in these structures (Becerra et al., 2006).

Unaffected limb—CRPS+ state

During symptomatic CRPS, innocuous bush or cold stimulation of the unaffected limb produces greater CNS BOLD activation than the same stimuli to the affected limb

None of the subjects experienced spontaneous pain in the unaffected limb and brush and cold stimuli did not elicit pain prior to or during the first scanning session (CRPS+). Brush to the unaffected limb produced almost twice as many foci of increased BOLD signal in cortical regions as stimulation of the affected limb (Supplementary Table E). Activation was seen in the anterior and middle ACC as well as in the anterior and posterior insula. There were prominent activations in the thalamus and the basal ganglia, which were not seen after brushing the affected limb. These regions appear to be involved in processing noxious thermal stimuli in healthy subjects (Bingel et al., 2004) and patients with neuropathic pain (Becerra et al., 2006). There were relatively few regions showing decreased BOLD signal. Cold stimulation of the unaffected limb in the CRPS+ state produced a similar pattern of response, with many more activations than seen for stimulation of the affected limb (Supplementary Table F).

Affected limb—CRPS state

Some of the CNS BOLD responses associated with allodynia in the painful state persist following clinical recovery

In the CRPS state, subjects had no spontaneous pain and evoked pain to brush and cold applied to the affected limb during the scanning was low (i.e. <2/10 for brush and <2.5/10 for cold; see Fig. 3B). However, brush stimulation of the affected region after recovery still induced many foci of activation, primarily areas of decreased BOLD signal within frontal, parietal and temporal cortex, parahippocampus and hippocampus (Supplementary Table C). Some of these changes were very similar to those elicited by brush to the affected limb in the CRPS+ state, including foci of activation in frontal lobe regions, middle and lingual regions of the temporal lobe, the parahippocampus and hippocampus. These similarities suggest that altered processing of stimuli to the affected region persists even though the spontaneous pain has resolved and evoked pain is minimal. Interestingly, the BOLD pattern elicited by brushing the affected region in the recovered state included a greater number of activation foci in the parahippocampal regions than seen in the active CRPS state. Cold stimuli applied to the affected limb in the CRPS state also resulted in many activations in the frontal, cingulate and insular regions (Supplementary Table D). As seen for brush, decreases in signal predominated. The total number of activations following cold applied to the affected limb in the CRPS state was more than double that observed during the CRPS+ state.

Unaffected limb—CRPS state

Following recovery, stimuli to the unaffected limb induces many fewer activations than observed during the painful state

Upon recovery from CRPS, brush to the unaffected limb resulted in very few activations: all eight clusters were cortical and showed increased signal (Supplementary Table G). This is in stark contrast to the many foci of activation observed following brush to the same location on the unaffected limb during the painful CRPS+ state (Supplementary Table E). Similarly, cold applied to the unaffected limb elicited far fewer regions of activation during the CRSP state (Supplementary Table H) than during the CRPS+ state (Supplementary Table F). For cold, this finding was less dramatic, with many regions of decreased BOLD signal still induced. In addition, areas of decreased signal were seen in the caudate nucleus, which does not show changes following cold stimuli in the CRPS+ state.

Percentage of total brain activated

In the CRPS+ state, stimulation of the affected limb predominantly induces decreases in BOLD, but increases in BOLD predominate following stimulation of the unaffected limb

To get a sense for overall changes in CNS activation following these thermal and mechanical stimuli, we calculated the number of voxels activated by each stimulus as a percentage of total pixels imaged. Either mechanical or thermal stimuli to the affected limb in both the CRPS+ and CRPS states induced a greater percentage of voxels with negative changes in BOLD signal than positive changes (Fig. 4, histograms). As noted in the figure, each group of histograms refers to the percentage of positive (red) and negative (blue) voxels showing significant activation relative to the brain for the four conditions (A = affected CRPS+; B = affected CRPS; C = unaffected CRPS+; and D = unaffected CRPS). For brush, negative bold signals exceeded positive bold signals for the affected limb in both states (Brush, Histograms, A, B). In contrast, brush stimuli to the unaffected limb in CRPS+ elicited a greater percentage of voxels with positive (20%) than negative changes (5%) (Brush, Histogram, C); this appeared to be true in the CRPS state as well, but far fewer voxels were activated (Brush, Histogram D). Overall, cold stimuli applied to the unaffected and affected limbs in the CRPS+ and CRPS states, consistently produced more or equal negative BOLD signal (Cold, Histograms A–D, Fig. 4).

Fig. 4

Average activation maps. Serial coronal sections showing significantly increased (red) and decreased (blue) BOLD responses for brush and cold applied to the affected or unaffected limb during the CRPS+ or CRPS state. Conditions (e.g. affected limb CRPS+) are arranged in rows, with columns for Brush (AD) or Cold (EH). S = superior, I = inferior, L = ipsilateral to stimulation (left hemisphere in unflipped brains), R = contralateral to stimulation (right hemisphere in unflipped brains). See ‘Methods’ section for details. The histograms below show the percentage of significant positive and negative voxels relative to the brain. The letters on the x-axis correspond with each of the figures above.

Analysis 2: contrast analyses

We performed contrast analysis for four comparisons: affected CRPS+ versus affected CRPS; affected CRPS+ versus unaffected CRPS+; affected CRPS versus unaffected CRPS; and unaffected CRPS+ versus unaffected CRPS for brush and cold (Supplementary Table 1A). Representative examples from a traditional contrast analysis for the same data are shown in Fig. 5. Comparison is complicated by the fact that the same region may show positive signal change under one condition and negative signal change in another. We interpret this reversal of polarity, or valence change, as indicative of a significant process in the brain, such as a change from excitation to inhibition. In a typical contrast analysis, valence changes may artificially enhance differences. We have therefore segmented our contrast analysis to distinguish valence changes from standard differences in positive BOLD activation. The results of these analyses are presented in Tables 3–10, in Fig. 6 for brush and Fig. 7 for cold. Each contrast is presented as Increased Response, Decreased Response or Valence Change (Fig. 2B) both in the tables and in the text.

Fig. 5

Contrast activation maps (standard method). Serial coronal sections through the brain for each contrast shown on the left for brush (A–D) and cold (E–H). S = superior, I = inferior, L = ipsilateral to stimulation (left hemisphere in un-flipped brains), R = contralateral to stimulation (right hemisphere in un-flipped brains). See ‘Methods’ section for details. The histograms below show total activation for positive and negative voxels activated as a percentage of total brain activation. The letters on the x-axis correspond with each of the figures above. Compare with contrast maps (Figs 6 and 7) that take into account ‘increased response’, ‘decreased response’ and ‘valence change’.

Fig. 6

Contrast maps for brush: increased and decreased responses, and valence change. Contrast analyses (arranged in rows) for increased activation (left panels), decreased activation (middle panels) and valence change (right panels). The statistical thresholds are shown below each figure for condition A < condition B (blue-light blue) and for condition A > condition B (red-yellow). Note that each bar has different z values that were defined for threshold of activation using GMM (see text). S = superior, I = inferior, L = ipsilateral to stimulation (left hemisphere in un-flipped brains), R = contralateral to stimulation (right hemisphere in un-flipped brains). See ‘Methods’ section for details.

Fig. 7

Contrast maps for cold: increased and decreased responses and valence change. Contrast analyses (arranged in rows) for increased activation (left panels), decreased activation (middle panels) and valence change (right panels). The statistical thresholds are shown below each figure for condition A < condition B (blue-light blue) and for condition A > condition B (red-yellow). Note that each bar has different z values that were defined for threshold of activation using GMM (see text). S = superior, I = inferior, L = ipsilateral to stimulation (left hemisphere in un-flipped brains), R = contralateral to stimulation (right hemisphere in un-flipped brains). See ‘Methods’ section for details.

Affected limb—CRPS+ versus CRPS state

Many cortical regions have larger decreases in BOLD signal following stimulation to the affected region in the painful state than in the recovered state

Brush (Table 3): The CRPS+ state showed increased response over the CRPS state in many cortical regions. Only one cluster, in the supplemental motor area, had a positive activation that increased further in the painful condition. In many other cortical regions—predominantly the middle frontal lobe, fusiform and precuneal regions of the parietal lobe, middle temporal lobe, cingulate and parahippocampus—there was a greater negative activation in the painful condition. Other regions exhibited decreased responses in the CRPS+ state (i.e. changed from positive in CRPS to less positive in CRPS+ or from negative to less negative), including the inferior frontal lobe, angular cortex and also subcortical regions of sublenticular extended amygdala (SLEA) and ventral tegmentum (VT). The SLEA has previously been reported to show activation in experimental (Becerra et al., 2001) and pathological pain (Becerra et al., 2006). Valence changes were observed in a few regions including the cingulate, which went from negative to positive signal change and angular gyrus of the parietal lobe, which changed from positive to negative. The finding that it is predominantly cortical regions which show increased responses indicates that they are more affected in the painful CRPS+ state versus the non-painful CRPS state.

View this table:
Table 3

Contrast analysis affected limb for brush: CRPS+ versus CRPS state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
 Cortical
  Frontal
            Supplemental_Motor_AreaR5.66568−14660.232
Negative (negAnegBnegC)
 Cortical
  Frontal
            MiddleR−1.8867250−141.096
            MiddleR−1.7009650−80.568
            Inferior_OrbitalL−2.938−3236−122.616
            MiddleL−1.7098−3220320.264
  Parietal
            FusiformL−1.752−30−24−320.832
            FusiformR−2.591422−38−180.864
            PrecuneusL−1.6224−4−50420.648
            PrecuneusR−1.901316−58220.760
            AngularR−2.560646−62322.008
  Occipital
            CalcarineR−1.85916−60180.312
  Temporal
            InferiorL−1.8387−362−380.264
            MiddleR−2.6457560−261.424
            MiddleL−1.5748−58−6−180.616
            MiddleL−1.5552−50−12−200.456
            MiddleL−1.78−62−14−100.976
            MiddleL−1.6895−64−24−60.288
            MiddleL−1.7201−62−28−40.376
            MiddleL−1.8744−62−42−40.440
            LingualL−3.8783−22−56−632.672
  Cingulum
            AnteriorL−1.6797−840−41.088
            AnteriorR−1.7613630−60.816
            MiddleL−2.2299−8−36400.616
  Parahippocampus
            ParaHippocampalL−1.4965−22−24−200.272
            ParaHippocampalL−1.9545−30−38−100.304
            ParaHippocampalL−2.0323−32−42−40.400
  Sub-cortical
            HippocampusL−3.1183−26−16−121.184
Decreased response
Positive (negAnegBposC)
Negative (posAposBnegC)
 Cortical
  Frontal
            Inferior_OrbitalR−2.61483638−82.104
  Parietal
            AngularL−3.0961−44−56381.856
            AngularL−2.6039−46−58260.744
  Cingulum
            AnteriorR−1.56291046161.272
  Sub-cortical
            SLEA−1.9483−10−8−60.440
            VT−1.83912−18−220.296
  Brainstem/cerebellum
            Cerebellum_Crus1R−1.896450−72−240.664
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            Paracentral_LobuleR6.44578−26660.768
  Cingulum
            MiddleL5.9182022360.280
  Brainstem/cerebellum
            Cerebellum_8L6.1476−34−64−520.552
Negative (negAposBnegC)
 Cortical
  Frontal
            Inferior_OrbitalR−2.61483638−82.104
  Parietal
            AngularL−3.0961−44−56381.856
            AngularL−2.6039−46−58260.744
  Cingulum
            AnteriorR−1.56291046161.272
  Brainstem/cerebellum
            Cerebellum_Crus1R−1.896450−72−240.664

Cold (Table 4): For cold, contrast analysis of CRPS+ versus CRPS showed few increased or decreased responses. Valence changes were observed manly in the frontal lobe. The relative lack of changes is perhaps counterintuitive since there is a significant difference in pain VAS responses for these two conditions. This may be because the cold stimuli were not suprathreshold (i.e. temperatures used not low enough). These frontal lobe changes may reflect cognitive and emotional processing of the stimulus in the absence of perceived sensory changes.

View this table:
Table 4

Contrast analysis affected limb for cold: CRPS+ versus CRPS state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
Negative (negAnegBnegC)
Decreased response
Positive (negAnegBposC)
Negative (posAposBnegC)
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            PrecentralR4.663344−14480.240
            PrecentralR5.509736−24563.264
  Brainstem/cerebellum
            Cerebellum_6L5.0768−16−62−160.504
Negative (negAposBnegC)
 Cortical
  Frontal
            Middle_OrbitalR−5.07313048120.656
            Inferior_   OperculumR−5.11235618100.592

Affected limb CRPS+ versus unaffected limb CRPS+

Many cortical regions have larger decreases in BOLD signal following stimulation to the affected region in the painful state than in the recovered state

This analysis is similar to the previously reported comparisons between the affected and unaffected limbs during the painful condition in adult CRPS (Maihofner et al., 2006a) except that we take into account the valence of activation in our comparison. Contrast analysis results for brush and cold are noted later.

Brush (Table 5): Contrast analysis comparing the response to brush applied to the affected limb versus the mirror area of the non-painful limb during the CRPS+ state revealed significant differences in the parietal, occipital and temporal lobes. These included positive and negative increases and decreases in signal. In addition there were many regions that exhibited valence change. Most of these were within the same cortical areas noted for increased or decreased responses (frontal, parietal and temporal regions). However, brainstem regions [including the periaqueductal gray (PAG) and pons] and cerebellar regions also showed valence changes. For example, the valence of the response in the PAG changed from negative to positive, suggesting that its inhibitory role may be diminished (Becerra et al., 2001). The extensive differences in the BOLD response presented here and in the figure are consistent with large differences in VAS pain scores for the affected versus unaffected region.

View this table:
Table 5

Contrast analysis affected versus unaffected limb for brush: CRPS+ state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
 Cortical
  Frontal
            Supplemental_Motor_AreaR4.581712−4646.048
            SuperiorL3.7717−20−4642.024
  Parietal
            PostcentralL4.5204−46−20280.960
            SupraMarginalR4.904444−38240.568
            SupraMarginalR3.253754−44362.680
  Occipital
            Rolandic_OperculumR4.469548−18142.120
            Rolandic_OperculumR4.889146−32221.056
            MiddleL3.3233−44−6620.224
  Temporal
            HeschlR6.133836−24100.592
            SuperiorL3.5799−58−36220.568
            MiddleR4.778746−6641.368
  Cingulum
            MiddleL4.5552222363.832
  Insula
            AnteriorR4.094144684.328
Negative (negAnegBnegC)
 Cortical
  Frontal
            Superior_MedialL−2.0363−258200.680
            Superior_MedialL−2.0904−145000.416
            Middle_OrbitalL−2.409−250−125.992
            SuperiorL−2.5624−2036400.432
  Parietal
            FusiformL−2.9101−34−22−320.488
  Temporal
            Pole_MiddleR−1.9591428−360.472
            MiddleL−2.1687−562−220.528
            InferiorR−2.3258560−320.848
            MiddleL−2.1171−58−4−180.984
            InferiorR−2.193654−8−420.896
            SuperiorL−1.7756−64−1020.672
            InferiorR−3.21352−18−202.368
            LingualL−2.0812−22−56−80.264
            MiddleR−2.831264−5860.544
Decreased response
Positive (negAnegBposC)
 Cortical
  Frontal
            PrecentralR3.344854−6481.104
            PrecentralR3.559342−10500.848
            Supplemental_Motor_AreaR5.25468−245411.768
            PrecentralR4.19426−24602.712
  Parietal
            PrecuneusL4.4738−2−46563.520
            PrecuneusL3.418−18−50101.840
            FusiformL4.4458−34−56−122.656
            PrecuneusL3.5304−2−58261.992
  Temporal
            SuperiorR2.684750−8−100.720
            MiddleR3.099746−58180.792
            RingualR3.945626−6023.352
  Parahippocampus
            ParaHippocampalR5.656422−28−141.736
            ParaHippocampalL3.4796−24−28−142.744
            ParaHippocampalR3.687134−36−101.488
  Sub-cortical
            HippocampusR4.512124−22−120.584
            HippocampusL4.3257−36−30−100.688
  Brainstem/cerebellum
            Cerebellum_4_5R3.613814−42−200.272
            Cerebellum_6L4.5406−30−46−281.112
            Cerebellum_4_5R3.443410−46−61.440
Negative (posAposBnegC)
 Cortical
  Frontal
            MiddleL−2.7826−3256202.120
            Middle_OrbitalR−1.94533054200.480
            MiddleL−1.8831−2646180.304
            Inferior_TriangularR−1.66964624300.376
            Inferior_TriangularL−1.6938−461640.304
  Cingulum
            AnteriorL−2.3796−1026280.232
  Insula
            AnteriorL−1.9222−2824−20.528
            AnteriorL−2.9546−3618−101.024
            PosteriorL−2.1342−34−22200.232
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            Superior_OrbitalR5.423328−2623.312
            Superior_OrbitalR5.313726−4661.984
            PrecentralR4.743424−16580.520
            PrecentralL5.1058−28−20584.272
  Parietal
            SupraMarginalR4.600158−40241.096
            PostcentralR8.927114−407026.032
            InferiorL3.3723−52−40360.360
            InferiorR3.11242−42440.304
            PrecuneusR4.805218−50143.208
            PrecuneusL4.3855−14−54501.352
  Occipital
            Rolandic_OperculumR3.249658−4162.080
            SuperiorR4.567226−64263.032
  Temporal
            SuperiorR5.226252−2242.272
            SuperiorL5.0409−52−2641.704
            SuperiorR4.487250−32102.008
            SuperiorL4.8838−50−3280.440
            MiddleL5.2774−46−4082.368
            MiddleR2.675760−62−20.600
  Insula
            PosteriorR5.887136−181412.496
  Sub-cortical
            CaudateR4.45042262011.352
  Brainstem/cerebellum
            Pag3.31928−28−100.320
            Brainstem4.786312−34−280.920
            Brainstem/Pons5.68614−38−404.288
            Brainstem4.48678−38−300.328
            Pons4.5877−16−40−401.856
            Cerebellum5.240226−36−340.888
            Cerebellum_4_5R4.625620−38−260.248
            Cerebellum_4_5L5.6889−18−42−286.792
            Cerebellum_9R4.701214−48−481.304
            Cerebellum_6R5.344730−48−305.224
            Cerebellum_4_5L4.6109−6−52−61.808
            Cerebellum_8R4.619628−56−483.280
            Cerebellum_6R4.611140−56−262.440
            Cerebellum_6L4.6583−34−56−261.504
            Cerebellum_6L4.5938−30−56−281.136
            Cerebellum_8L5.6512−36−60−502.184
            Cerebellum_6L4.3286−32−60−202.568
            Cerebellum_4_5L5.2659−8−60−120.856
            Cerebellum_8L5.39−30−62−501.848
            Cerebellum_6L5.2731−4−62−205.976
            Vermis_65.5998−2−64−104.240
            Cerebellum_8L4.4258−6−70−400.960
            Cerebellum_Crus1L4.7841−12−72−280.792
            Cerebellum_7bL4.6123−34−72−521.392
            Cerebellum_8R4.735422−74−483.304
Negative (negAposBnegC)
 Cortical
  Frontal
            MiddleL−3.3395−325427.136
            Superior_MedialL−2.8436−1042180.640
            Superior_MedialL−1.9593240460.240
            Inferior_OrbitalL−3.1602−2228−121.520
            Inferior_OrbitalR−2.89292826−165.048
            Inferior_OrbitalL−1.4921−4624−60.216
            Inferior_OperculumR−2.10533824300.368
            Inferior_TriangularR−2.5689562200.424
            Inferior_TriangularL−1.7479−4622200.304
  Parietal
            PostcentralL−2.4616−62−16200.392
  Occipital
            Rolandic_OperculumL−1.9092−38−30180.320
  Temporal
            Pole_SuperiorL−1.8565−4222−220.248
            Pole_SuperiorL−2.6915−5018−160.448
            Pole_SuperiorL−2.4098−364−200.304
            MiddleL−3.6851−66−4680.936
  Sub-cortical
            CaudateL−1.9177−166140.216
            ThalamusL−2.0278−10−28−20.448
            SLEA−2.3536−8−10−80.888

Cold (Table 6): The contrast analysis for cold was less dramatic than for brush, despite the significant difference in VAS ratings (∼6/10 in CRPS+ state and 2/10 in CRSP state; see Fig. 3B). A few regions showed increased or decreased responses. Positive valence changes were observed in frontal regions, primary somatosensory region, cingulate and parahippocampal areas.

View this table:
Table 6

Contrast analysis affected versus unaffected limb for cold: CRPS+ state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
Negative (negAnegBnegC)
Decreased response
Positive (negAnegBposC)
 Cortical
  Frontal
            Middle_OrbitalL3.261−1056−101.192
            SuperiorR2.338825440.864
            Middle_OrbitalL3.2762−1052−80.400
  Brainstem/cerebellum
            Cerebellum_4_5R3.353912−46−60.600
Negative (posAposBnegC)
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            PrecentralR3.513550−10400.224
            PrecentralR3.115342−14480.456
            PrecentralR4.10732−22561.664
            PostcentralR2.782320−26580.456
  Parietal
            PostcentralR3.361822−42703.856
  Cingulum
            AnteriorR3.5258438−20.376
            AnteriorL3.432−63800.360
  Parahippocampus
            ParaHippocampalR3.873422−22−160.488
            ParaHippocampalR3.577132−26−140.336
  Sub-cortical
            HippocampusL4.0691−18−8−120.848
  Brainstem/cerebellum
            Cerebellum_9R3.550114−54−560.704
            Cerebellum_9L3.7321−16−54−520.672
            Cerebellum_4_5R2.5588−54−60.280
            Vermis_4_52.8657−2−58−160.272
            Cerebellum_8R3.040818−62−540.288
            Cerebellum_8R3.049414−66−500.920
            Cerebellum_8L4.0653−16−66−502.040
            Cerebellum_8L3.5465−4−66−401.448
            Cerebellum_6L5.5357−14−66−1433.896
Negative (negAposBnegC)
 Cortical
  Frontal
            Inferior_TriangularL−7.2979−4444100.552

Affected limb CRPS versus unaffected CRPS

Overall, few responses to evoked pain for increased and decreased responses to brush and cold and valence changes more prominent following cold stimuli

Brush (Table 7): At the second scanning session, after symptomatic recovery, CNS responses to brush to the affected and unaffected sides were very similar; contrast analysis revealed only a few regions of increased or decreased activation and 15 regions with positive or negative valence change. This is consistent with the significant, but very small, difference in VAS responses to brushing the affected and unaffected regions after recovery.

View this table:
Table 7

Contrast analysis affected versus unaffected limb for brush: CRPS state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
 Cortical
  Occipital
            Rolandic_OperculumR5.394244−32220.360
Negative (negAnegBnegC)
 Cortical
  Frontal
            Inferior_OrbitalR−2.73332626−181.112
Decreased response
Positive (negAnegBposC)
 Cortical
  Frontal
            Middle_OrbitalR3.56043054220.640
            SuperiorR4.66731452400.768
            SuperiorR4.5421030600.584
            Superior_OrbitalR4.67882414660.264
  Parietal
            PrecuneusL5.5102−12−404611.160
Negative (posAposBnegC)
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            Superior_MedialL5.9997−240560.640
  Parietal
            FusiformL4.6975−26−68−160.680
  Temporal
            MiddleR4.016142−60140.216
  Insula
            PosteriorR6.070434−22221.664
  Sub-cortical
            SLEA5.311710−12−103.600
  Brainstem/cerebellum
            Cerebellum_Crus1R4.508850−62−281.280
Negative (negAposBnegC)
 Cortical
  Frontal
            Inferior_TriangularL−3.1967−343800.224
            Inferior_TriangularL−3.4541−522820.608
        Inferior_OperculumL−3.3209−4812220.544
            PrecentralR−4.867542−16362.064
  Parietal
            PostcentralR−4.194458−4241.576
            PostcentralL−4.4329−62−6340.840
            PostcentralL−3.4335−50−8260.600
            PostcentralL−4.6002−50−12381.992
  Brainstem/cerebellum
            Cerebellum_8R−3.489922−72−580.568

Cold (Table 8): In contrast to the results for brush, this analysis revealed many more increases and decreases in BOLD response to stimulation of the affected versus unaffected limb in the CRPS state. In addition, a relatively large number of cortical regions showed valence changes; in particular, there were positive valence changes in frontal and parietal regions, and negative valence changes in temporal regions.

View this table:
Table 8

Contrast analysis affected versus unaffected limb for cold: CRPS state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
Negative (negAnegBnegC)
 Cortical
  Parietal
            FusiformL−3.7562−34−44−241.016
  Temporal
            MiddleL−3.4128−54−18−60.272
            SuperiorL−3.7621−58−2681.488
            MiddleL−3.1719−64−4260.528
  Parahippocampus
            ParaHippocampalL−3.4913−18−8−260.240
Decreased response
Positive (negAnegBposC)
 Cortical
  Frontal
            Middle_OrbitalR2.55773628480.280
            Superior_OrbitalR2.81212220640.240
            Superior_OrbitalR2.66472820600.240
            SuperiorL2.5427−1412700.696
            Middle_OrbitalR3.3306428561.304
  Parietal
            InferiorL2.4303−60−24460.224
  Occipital
            SuperiorR2.970124−64400.496
Negative (posAposBnegC)
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            SuperiorR3.6858452421.264
            SuperiorR3.62191452180.600
            Superior_OrbitalR2.52692252400.240
            MiddleL3.2239−4050140.568
            Middle_OrbitalR2.71373246401.448
            MiddleL3.2102−4044200.416
            MiddleL3.5368−2842140.656
            Inferior_TriangularR3.8602524244.224
            Middle_OrbitalR3.39253418361.000
            Inferior_OperculumR2.91175618120.320
            Superior_OrbitalR4.38422410581.632
            PrecentralR2.7295568320.272
            SuperiorL2.5299−168680.224
            PrecentralR2.9159484300.264
            PrecentralR2.7709380420.240
  Parietal
            InferiorR3.323158−44460.624
            SuperiorR3.51136−44421.088
            InferiorR2.873844−44460.248
            AngularR2.986456−50340.408
            InferiorR4.235356−54461.584
            AngularR3.228844−54360.632
            AngularR3.465552−56380.504
            AngularR3.197340−58420.496
            AngularR2.904936−60360.232
            PrecuneusR2.49518−62420.376
  Insula
            AnteriorR3.0596302880.216
  Brainstem/cerebellum
            Cerebellum_Crus2R3.711350−66−400.424
Negative (negAposBnegC)
 Cortical
  Frontal
            Inferior_OrbitalL−3.8853−5024−102.272
            Inferior_TriangularL−4.1379−4622162.040
            MiddleL−3.521−3418340.328
  Parietal
            AngularL−3.8265−52−70240.456
  Occipital
            Rolandic_OperculumL−4.3307−42−22161.744
            SuperiorL−4.0061−16−70280.944
            CalcarineL−5.1526−12−70106.416
  Temporal
            Pole_SuperiorL−3.751−3614−200.280
            Pole_SuperiorL−3.1469−5012−100.768
            MiddleL−4.5803−54−34−61.040
            LingualL−4.4159−14−48−40.912
            MiddleL−4.5812−60−50123.240
            RingualR−4.0658−5400.432
            LingualL−3.9407−18−54−120.360
            RingualR−3.75668−6441.352
  Sub-cortical
            HippocampusR−3.729922−32−80.336
  Brainstem/cerebellum
            Cerebellum_6L−3.3931−30−48−300.360

Unaffected limb CRPS+ versus unaffected limb CRPS

Predominant changes observed for valence change, suggesting possible reversal of pain effect (spontaneous) during CRPS+ state on CNS processing

As noted in the averaged data, both brush and cold to the unaffected limb produced a very large number of activations in the CRPS+ state (Supplementary Tables E and F), in contrast to the same stimuli applied to the limb in the CRPS state (Supplementary Tables G and H). These were also reflected in the contrast analysis.

Brush (Table 9): Most changes were valence changes with frontal and temporal regions predominating in both positive and negative valence change and the post-central parietal region exhibited many regions of negative valence change as well. Cold (Table 10): Results of this contrast analysis were similar to those for brush with predominant changes to valence in cortical areas. In addition, signal changes in the thalamus and hypothalamus switched from positive valence in the CRPS+ state to negative in CRPS.

View this table:
Table 9

Contrast analysis unaffected limb for brush: CRPS+ versus CRPS state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
Negative (negAnegBnegC)
 Cortical
  Frontal
            PrecentralL−4.2268−26−12680.624
            Paracentral_LobuleR−3.98156−38601.584
Decreased response
Positive (negAnegBposC)
Negative (posAposBnegC)
 Cortical
  Frontal
            Inferior_OperculumR−4.1786468220.664
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            Middle_OrbitalR6.3153054201.392
            Superior_MedialL5.5618−240540.368
            SuperiorR5.2416638560.248
  Temporal
            MiddleR7.005554−34−80.504
Negative (negAposBnegC)
 Cortical
  Frontal
            PrecentralL−4.1322−14−18680.344
            Paracentral_LobuleL−4.546−18−24640.592
  Parietal
            PostcentralL−4.3566−620160.224
            PostcentralL−5.0013−56−4180.792
            PostcentralR−6.023858−6204.536
            PostcentralL−4.2107−46−10381.064
            PostcentralR−4.682920−42600.536
  Occipital
            CalcarineR−4.631726−5040.784
  Temporal
            InferiorL−4.7274−40−4−440.392
            InferiorR−4.902850−54−100.656
            MiddleR−4.01246−58−40.464
            RingualR−4.342126−7000.632
View this table:
Table 10

Contrast analysis unaffected limb for cold: CRPS+ versus CRPS state

Brain regionLat.Z-statCoordinates (mm)Volume cm3
xyz
Increased response
Positive (posAposBposC)
 Cortical
  Frontal
            Middle_OrbitalR4.13433254161.408
  Insula
            AnteriorR4.00784020−100.224
Negative (negAnegBnegC)
 Cortical
  Frontal
            PrecentralR−2.642126−28700.512
  Parietal
            FusiformR−4.377530−42−1849.848
            AngularR−3.567946−58281.856
            PrecuneusR−2.404118−60220.336
Decreased response
Positive (negAnegBposC)
Negative (posAposBnegC)
Valence change
Positive (posAnegBposC)
 Cortical
  Frontal
            Middle_OrbitalR3.90362648320.224
            Middle_OrbitalR6.426546461613.760
            MiddleL5.954−42402413.680
            Superior_MedialL4.5692−426380.288
            Supplemental_Motor_AreaR4.258816560.352
            Inferior_OperculumR4.52475816100.584
            Supplemental_Motor_AreaR4.47631012680.216
            Supplemental_Motor_AreaR3.964610580.288
            Supplemental_Motor_AreaL5.2641−1010541.408
            PrecentralR4.04275210460.352
            Superior_OrbitalR3.92232210580.256
            PrecentralR4.5529504320.464
            SuperiorL4.7914−14−2721.512
  Parietal
            PostcentralR4.495666−14220.264
            SupraMarginalR5.59856−40424.456
            InferiorL4.0325−38−44520.392
  Temporal
            Pole_SuperiorR4.2648524−120.304
   Cingulum
            AnteriorL3.8301−634200.216
  Insula
            AnteriorR4.4283422100.344
Negative (negAposBnegC)
 Cortical
  Frontal
            MiddleL−3.0345−2220460.528
  Parietal
            SuperiorR−2.660116−46620.376
            PrecuneusR−2.405616−48480.328
            SuperiorR−2.96120−52580.576
            AngularL−3.138−42−62301.664
  Occipital
            CalcarineL−6.5005−12−66833.512
  Temporal
            MiddleL−2.7892−48−50220.280
  Cingulum
            PosteriorR−2.55562−40180.224
  Sub-cortical
            ThalamusL−2.6235−22−3020.312
            HypothalamusL−4.4212−4−2−60.328
  Brainstem/cerebellum
            Vermis_10−3.04646−46−320.792
            Cerebellum_9L−3.1003−16−52−541.752
            Cerebellum_8R−2.974418−54−560.776
            Vermis_6−2.72362−56−220.528
            Vermis_7−3.16574−74−280.928
            Vermis_7−3.03730−74−220.376

Figure 5 shows examples of the regions that were significantly activated in the contrast analyses noted earlier. Several significant changes described earlier can be seen here, including increased activation in many regions in response to brush applied to the affected limb versus the unaffected limb in the CPRS+ state (5B). Changes in brainstem signal occur in substantia nigra (5B), where signal is increased in the affected versus the unaffected side in the CRPS+ state. In the recovered state (CRPS), another brainstem region, the red nucleus, shows greater activation in the affected versus unaffected side (5C), along with activations in the basal ganglia (caudate nucleus and nucleus accumbens), and medial prefrontal and parietal cortices. No significant activations were observed in the thalamus, a major sensory processing site.

Discussion

Our results show distinct patterns of CNS activation following mechanical and thermal stimuli to the affected and unaffected regions in CRPS+ and CRPS states in children. To our knowledge, this is the first fMRI study of pain in children, and our results offer insight into central pain processing in paediatric CRPS that may also apply to the adult condition.

Reversal of pain symptoms: psychophysical measures

All subjects had spontaneous pain during their first scan and no spontaneous pain at the time of their second scan. Prior to recovery, our CRPS patients demonstrated both mechanical and cold allodynia on the affected side. These symptoms are consistent with a recent study of cutaneous abnormalities in children and adolescents with CRPS (Sethna et al., 2007). We observed that, both outside and inside the magnet, normally innocuous brush and cold stimuli applied to the affected area in patients in the CRPS+ state produced pain rated as >5/10. Following recovery, patients’ VAS ratings for evoked pain decreased dramatically for brush and for cold. Activation patterns for brush and cold applied to the affected limb during the CRPS+ and the CRPS state showed overall differences in CNS responses indicating widespread neuroplastic changes with recovery from CRPS.

CNS responses associated with mechanical and cold allodynia in children are similar to those seen in adults

During active CRPS, stimuli to the affected limb produced a greater level of positive BOLD activations than stimuli to the unaffected limb. The increased activation was particularly clear for brush stimulation, which induces mechanical allodynia in CPRS. This is consistent with the previous fMRI studies in adult CRPS patients (Maihofner et al., 2006a). Specifically, dynamic mechanical allodynia in adult CRPS subjects resulted in increased activation in the contralateral SI and MI, bilaterally in the insula and SII, parietal cortex, frontal cortex and cingulate cortex, and decreased activation in the temporal cortices. Our results for brush (Table A, Supplementary Data) show greater activation following affected limb stimulation in contralateral SI, anterior and middle cingulate, and bilaterally in the anterior insula. We also observed decreased activation in the temporal lobes as well as other regions (frontal and parietal cortices and parahippocampus). The CNS responses to cold stimulation of the affected limb in paediatric CRPS patients that we describe here also parallel data reported for adults with cold allodynia (Seifert and Maihofner, 2007). These similarities between the BOLD phenotype in adult and childlike CRPS are noteworthy because paediatric CRPS state has been considered, somewhat arbitrarily, to be clinically different from the adult condition.

In addition to spontaneous pain (77%), adult CRPS can include evoked hyperalgesia (94%), autonomic changes (98%), cognitive changes (Grabow et al., 2004; Apkarian et al., 2004) and motor dysfunction (97%) (Birklein et al., 2000). With the exception of autonomic dysfunction, these manifestations are not common in paediatric patients (Meier et al., 2006), perhaps because of the relatively rapid recovery experienced by most (within 1–2 years). In our paediatric population, we observed BOLD changes in regions that are also implicated in adult CRPS (Wasner et al., 1998; Birklein et al., 2000; van de Beek et al., 2002; Janig and Baron, 2003). Although we did not measure or observe non-pain symptoms in our subjects, it is interesting to note that many of the BOLD activations we report here are consistent with altered processing in regions that presumably contribute to non-pain CRPS symptoms in the adult. For example, changes in the parietal lobe might be related to hemi-inattention (Pleger et al., 2005), and changes in the frontal lobe might reflect other aspects of altered cognition (Grabow et al., 2004; Apkarian et al., 2004). Prominent activation was observed in most analyses in the temporal lobe, similar to those described in adult patients with CRPS (Maihofner et al., 2006b) and may relate to complex processing of fear and anxiety (Charney, 2003). The ongoing activation we describe in the basal ganglia (Supplementary Tables B, D, E, F, H and 6) might eventually cause neurochemical changes or neuronal loss, leading to CRPS-related movement disorders such as dystonia. Similarly, the changes we observe in hypothalamus may indicate altered function that eventually causes the autonomic symptoms of CRPS.

These changes in BOLD in paediatric CRPS may reflect altered processing that is sub-clinical but could become clinically manifest if the disease does not resolve spontaneously or through treatment. Further characterization of such changes and correlation with non-pain CRPS symptoms might lead to fMRI screening and prompt earlier intervention.

Decreased BOLD signals in the painful state

Brush and cold produced more foci of decreased BOLD signal in brain regions in the CRPS+ state compared with the CRPS state. While the implications of decreased BOLD signal are not fully understood, some have suggested that it may correlate with inhibitory CNS processing (Harel et al., 2002). As such, some of these decreases occur in regions that may be interpreted as having a ‘diminished’/inhibitory function. Such a notion is made more plausible by the results showing that stimulation to the unaffected limb during the painful CRPS+ state results in widespread decreases in BOLD.

Differences in BOLD pattern induced by stimulation of the unaffected versus affected limbs in the painful state suggest central inhibition of CNS responses

During the CRPS+ state, there was greater total brain activation following stimulation of the unaffected limb mirror region than following stimulation of the affected region. While further experiments are required to assess the normal response to brush in healthy paediatric controls, this increased activation is unexpected and very different from what is seen in healthy normal adults (Davis et al., 1998; Chen et al., 2002). Significantly, the majority of changes we observed following brush or cold to the unaffected limb were decreased activations. These findings may be explained by generalized inhibition of evoked pain from the affected side or by increased sensitivity of neural circuits to stimulation on the unaffected side. Given that the subjective ratings for both brush and cold were significantly greater following stimulation of the affected side (Fig. 3A), we propose that the ongoing pain input from the affected region in active CRPS induces an overall inhibitory drive specific to the affected side. In animal models of neuropathic pain, unilateral descending facilitation is postulated to maintain hyperalgesia and central sensitization (Vera-Portocarrero et al., 2006). Although this mechanism is not well understood, especially in humans, differences in descending inhibition or facilitation (Vanegas and Schaible, 2004) may exist between the two sides of the body.

Increased inhibition may be related to several phenomena associated with CRPS—tactile impairment; hemi-neglect/hemi-inattention; and spontaneous pain and its potential deactivation during applied stimuli. Previous work has reported that there is impaired tactile discrimination in CRPS (Pleger et al., 2006), including a decrease in cortical representation in S1 and S2 of the affected hand with stimulation of the affected side. This observation may explain our finding of diminished activation for brush to the affected side compared with brush to the unaffected side in the CRPS+ state (Tables 9 and 10). Significantly, sensorimotor ‘retuning’ led to a decrease in pain and an increase in cortical map size in these areas (Pleger et al., 2004, 2005), and similar changes following recovery have been observed by others (Maihofner et al., 2004). In our study, we observed diminished activation to brush following symptomatic recovery, but increased activation with pain on the affected versus unaffected limb comparison. This may be due to the fact that brush stimulus to the two sides evokes two distinct sensory processes: (i) pain induced by mechanical allodynia on the affected side, and (ii) innocuous mechanosensation on the unaffected side. Alterations in sensation, including hemisensory deficits, may be the result of functional disturbances of noxious processing in the thalamus (Rommel et al., 2001) or other regions (Peyron et al., 2004). In our study, activation in the right parietal lobe, which is involved in hemi-inattention (Heilman and Van Den Abell, 1980; Robertson et al., 1994; Rushworth et al., 2001), was greater following brush to the affected limb than the unaffected limb in the CRPS+ state. Hemisensory deficits have been reported in patients with CRPS (Rommel et al., 1999; Galer et al., 2000), but correlation with our findings remains speculative since we did not measure or observe hemisensory deficits in our subjects. Again, we hypothesize that continued abnormal processing in these regions may lead to hemi-inattention/neglect.

CNS response to stimuli during the ‘recovered state’—evidence for persistence of altered CNS circuitry

In the asymptomatic ‘recovered state’ (CRPS), the difference in VAS pain scores following brush and cold stimuli to affected versus unaffected limbs was minimal, but there were still significant differences in the CNS responses. Thus, some alterations in the CNS response to mechanical or thermal stimuli to the affected region appear to persist in the recovered state despite the absence of allodynia. Additional support for the persistence of altered processing comes from the finding that the same stimulus applied to the affected limb in the CRPS state produced a larger number of activations than in the CRPS+ state even though stimulus-evoked VAS ratings decreased dramatically (Fig. 3B). Such ‘retuning’ has been observed in CRPS patients with shrinkage of cortical maps that parallel a decrease in pain intensity (Pleger et al., 2005). Of particular note are the regions activated in the recovered state, particularly for brush, which include regions in the frontal and parietal lobes, insula, basal ganglia, hypothalamus and PAG. Activation in these regions may correspond to cognitive and affective phenotypic manifestations of CRPS. However, we have not evaluated CNS processing later in the course of recovery, where such changes may disappear.

Valence change

The results for valence change for each contrast analysis (Tables 3–6) showed relatively larger numbers of regions showing an increase or decrease valence change suggesting that spontaneous pain may result in an alteration of baseline brain activity and a shift in the proportion of negative and positive activation. Our previous results regarding valence include the striking observation that valence changes occur in pain related structures when a pain condition is contrasted versus a non-pain one. For example, in the nucleus accumbens, there is a valence change from negative for pain onset (aversive) to positive for pain offset (rewarding) (Becerra and Borsook, 2008). In the experiments described here, we do not see large numbers of foci activated for the conditions where alterations in brain systems during the symptomatic and asymptomatic states (see ‘Discussion’ section) may produce differences but not a reversal of signal change. However, we observed valence changes for a large numbers of foci in the contrast analyses of CAbrush versus CUbrush and CAcold and CUcold, where the contrast relates to ‘opposite’ differences in state being evaluated i.e. affected versus unaffected limb (Tables 5 and 6). This also applies for the other contrasts, most notably the contrast analysis for CUbush versus PUbrush and CUcold and PUcold (Table 9 and 10). In the latter example, the responses of evoked stimuli were given during the CRPS+ (sensitized brain) and thus the opposite valence observed may reflect either normalizing of brain responses or at least different function or modulation of circuits.

Caveats

Lateralization

Most subjects were affected on their left side (of the 14 subjects originally recruited to the protocol, only two were affected on the right side). The predominant left-sided presentation of CRPS+ may be the result of right-lateralized processing of pain. Similar lateralization of experimental pain in healthy volunteers has been reported (Coghill et al., 2001; Youell et al., 2004; Symonds et al., 2006), and chronic pain is more frequently left-lateralized (Merskey and Watson, 1979). This apparent left-lateralization appears to correlate with right-hemisphere cortical dysfunctions (Coghill et al., 2001). This may have some implication on how stimulation of the unaffected side is interpreted.

A number of other caveats apply to this study including brain size and stage of development in adolescents and children, different location of stimuli applied to the lower extremity, order effects, age, gender and temporal nature of disease remission, medication, spontaneous pain during the CRPS+ state only, and the definition of CRPS in the paediatric population. These are addressed in detail in Supplementary Data—Caveats.

Conclusions

Here we present evidence that (i) child-type and adult CRPS may have similar underlying mechanisms; (ii) CRPS may result in increased overall inhibitory drive or ‘sensitized CNS’; (iii) functional abnormalities may persist after pain has resolved and (iv) persistent abnormalities in emotional circuitry may be similar to changes that cause the cognitive symptoms seen in adult CRPS. The ability to evaluate changes in neural circuitry in the paediatric population, where CRPS symptoms are usually transitory, allows us to identify specific CNS regions and circuits where anatomical and connectivity changes occur in CRPS. Our results suggest significant changes in CNS circuitry in paediatric patients with CRPS may outlast the signs and symptoms and may well be a distinctive feature of paediatric CRPS. Thus, even with a more rapid resolution of pain in children, the effect of the nerve damage and other changes that occur in CRPS at a time of developmental plasticity in brain connections, may have prolonged effects upon brain circuitry. This could impact upon pain processing in these individuals in later life.

Supplementary material

Supplementary material is available at Brain online.

Acknowledgements

The work was primarily supported by a Grant from the Mayday Foundation, New York (D.B.). Supplementary funding was provided by the Sara Page Mayo Endowment for Pediatric Pain Treatment and Research (C.B.) and Children's Hospital Boston (D.B., L.B.).

Footnotes

  • Abbreviations:
    Abbreviations:
    CRPS
    complex regional pain syndrome
    RSD
    reflex sympathetic dystrophy
    VAS
    visual analog score
    GMM
    Gaussian mixture modelling

References

View Abstract