1

Proposed journal section: Disorders of the Nervous System

Title: When the brain expects pain: Commonneural responses to pain anticipation are related to clinical pain and distress in fibromyalgia and osteoarthritis

Abbreviated title: Pain anticipation in fibromyalgia and osteoarthritis

Authors:

(Corresponding author) Christopher A Brown, PhD. Research Associate, Human Pain Research Group,The University of Manchester, Manchester Academic Health Science Centre,Clinical Sciences Building, Salford Royal NHS Foundation Trust, Salford, M6 8HD, United Kingdom

Email:

Tel: 0161 206 4528

Wael El-Deredy, PhD. Senior Lecturer, School of Psychological Sciences, The University of Manchester, ZochonisBuilding, Brunswick Street, Manchester, M13 9PL, United Kingdom.

Email:

Tel: 0161 275 2566

Anthony KP Jones, MD. Professor of Neuro-Rheumatology, Human Pain Research Group, University of Manchester, Clinical Sciences Building, Salford Royal NHS Foundation Trust, Salford, M6 8HD, United Kingdom

Email:

Tel: 0161 206 4266

Number of pages:33

Number of figures:2

Number of tables:3

Number of words:Abstract 239, Introduction 494, Discussion 1821, Whole manuscript 5895 (including abstract and figure legends).

Keywords:Event-related potentials; electroencephalography; expectancy; chronic pain; human.

Abstract

Supra-spinal processes in humans can exert a top-down enhancing effect on nociceptive processing in the brain and spinal cord. Studies have begun to suggest such influences occur in conditions such as fibromyalgia (FM), but it is not clear if this is unique to FM pain or common to other forms of chronic pain, such as that associated with osteoarthritis (OA). We assessed top-down processes by measuring anticipatory evoked-potentials and their estimated sources, just prior (<500ms) to laser heat painstimulation, between 16 patients with FM, 16 patients with OA and 15 healthy participants (HPs) using whole-brain Statistical Parametric Mapping. Clinical pain and psychological coping factors (pain catastrophizing, anxiety, and depression) were well matched between the patient groups such that these did not confound our comparisons between FM and OA patients. For the same level of heat pain, insula activity was significantly higher in FM than the other two groups during anticipation, and correlated with the intensity and extent of reported clinical pain. However, the same anticipatory insula activity also correlated with OA pain, and to the number of tender points across the two patient groups, suggesting common central mechanisms of tenderness. Activation in dorsolateral prefrontal cortex (DLPFC) was reduced during anticipation in the both patient groups, and was related to less effective psychological coping. Our findings suggest common neural correlates of pain and tenderness in FM and OA that are enhanced in FM but not unique to this condition.

Introduction

Fibromyalgia (FM) is characterized by chronic widespread pain and tenderness(Wolfe et al 1990a),but the mechanisms of FM painremain poorly understood. Conversely, it is often assumed that osteoarthritic (OA) pain arises solely from peripheral mechanisms in affected joints. Yet,in OA there is a poor relationship between radiographic evidence of joint damage and pain(Bedson and Croft 2008).Many patients also report pain at multiple sites irrespective of tissue damage (Natvig et al 2000) and there is evidence of central sensitization (Lee et al 2011). There may therefore be overlapping central mechanisms in FM and OA.

Investigations into FM have largely focused on possible centralcontributions to pain (Yunus 1992;Clauw 2009). Abnormalities have been identifiedin spinal mechanisms (Staud and Smitherman 2002;Price et al 2002;Staud 2002) and in diffuse noxious inhibitory controls(Lautenbacher and Rollman 1997;Staud et al 2003), which rely on spinal and supraspinal pathways. Brain-imaging has shown augmented responses to experimental pain stimuli compared with pain-free controls(for a review, see Gracely and Ambrose 2011) consistent with either central augmentation or lack of inhibition.

FM is associatedwith psychological co-morbidities such as anxiety and depression (Thieme et al 2004), pain catastrophizing (Hassett et al 2000), and cognitive impairments(Baumstark et al 1993), which are, to an extent, shared in patients with OA (Edwards et al 2011). Pain catastrophizing has been associated with increased activity in brain areas related to anticipation of pain(Gracely et al 2004).Both catastrophizing (Sullivan et al 2001;Seminowicz and Davis 2006) and anticipation (Koyama et al 2005;Brown et al 2008a;Brown et al 2008b) have been shown to augment pain and its neural processes. However, it is not clear whether supraspinal processing abnormalities are unique to patients with FM or are just related to psychological factors that can also occur in conditions such as OA.

In this study we compared central processing of pain in groups of patients with OA and FM (comparable in clinical pain levels and catastrophizing) relative to pain-free controls. We used electroencephalography (EEG) with source estimation to measure the neural generators of anticipatory and pain-evoked responses to experimental acute laser pain. The high temporal resolution of EEG provides an advantage over types of neuroimaging that are reliant on slow haemodynamic responses or blood flow. Previously, we showed in a healthy population (Brown et al 2008a) and in patients with musculoskeletal pain (Brown and Jones 2013)that ‘late’ anticipatory responses, within half a second prior to pain onset, can be reliably localized to pain processing regions. These localized responses correlate with expectancy and pain ratings (Brown et al 2008b) and provide unique picture of the brain state in preparation for pain. In the present study we hypothesized that FM and OA pain would be associated with common abnormalities in anticipatory neural networks, suggesting shared top-down influences on pain, in brain regions known to activate during pain anticipation and be modified by a psychosocial intervention (insular, mid-cingulate and dorsolateral prefrontal cortices) (Brown and Jones 2013).

Materials and Methods

The research study was approved by Salford LocalResearch Ethics Committee in the United Kingdom. The study conforms with the Code of Ethics of the World Medical Association (Declaration of Helsinki). 47right-handed participants were recruited.All subjects gave written informed consent to take part in the study. 16 patients had a diagnosis of fibromyalgia (FM, age 48.6 ± 8.6 (mean ± SD)), 16 of osteoarthritis (OA, age 54.3 ± 9.8), and 15 were pain-free healthy participants (HP, age 46.3 ± 7.3). The gender of the participants was well matched, with only one male in each of the FM and HP groups and two in the OA group. Age was significantly different between patient groups, and age was therefore used as a nuisance variable for all statistical analyses.

FMand OA patients fulfilled the American College of Rheumatology (ACR) criteria for the diagnosis of FM (Wolfe et al 1990b) and OA (Altman et al 1986;Arnett et al 1988). While new criteria for FM currently exist (Wolfe et al 2011), data collection for this study began prior the new criteria being available and so these were not used. Participants were excluded from the study if their medical records showed a history of neurological disorder, morbid psychiatric disorder (including major depression and anxiety-related disorders confirmed by a psychiatrist) or cardiovascular disease. It was expected that patients would be recruited with sub-clinical levels of anxiety and depression that are normal for chronic pain populations. All participantswere non-medicating at the time of the study, having been requested to withdraw from any analgesic medication for the purpose of the study.

Experimental protocol

Questionnaires were posted to participants to fill out two weeks prior to the experimental session. The Hospital Anxiety and Depression Scale (HADS,(Zigmond and Snaith 1983)) was used to assess mood symptoms. It is composed of statements relating to anxiety or depression. The Pain Catastrophising Scale (PCS,(Sullivan et al 1995)) was used as a measure of pain-related psychological coping.

On the day of the experiment, tender point (TP) examinations were first carried out by a trained research nurse, and scored using the Manual Tender Point Survey according to a published methodology (Okifuji et al 1997). Due to the lack of availability of a research nurse on the day, tender points were examined in only 8 out of the 15 healthy volunteers and 14 out of the 16 OA patients, while all 16 FM patients were examined.

Clinical pain levels (“In general, how severe is your pain?”)was measured using a 0 – 10 visual analogue scale (VAS) ranging from “no pain” to “very severe”. Pain interference (“How much does the pain interfere with your life?”) was also assessed using a VAS ranging from “not at all” to “completely”. Healthy volunteers completed these scales as well as patients.

Neural responses to acute pain

Acute pain was induced using aCO2laser that specifically activates nociceptors in the skin (Meyer et al 1976). Heat stimuli of 150ms duration and a beam diameter of 15mm were applied to the dorsal surface of the subjects’ right forearm. Between stimuli, the laser was moved randomly over an area 3cm x 5cm to avoid habituation, sensitization or skin damage. Subjects wore protective laser safety goggles during the experiment.

An initial psychophysics procedure was performed using a 0-10 numerical rating scale, which was anchored such that a level 4 indicated pain threshold, 7 indicated moderate pain, and 10 indicated unbearable pain. A ramping procedure was repeated three times to determine an intensity of laser stimulus for each subject at level 7, corresponding to a moderately painful heat level, as done previously (Brown et al 2008a).

The main experimentconsisted of the delivery of 40 moderately painful (level 7) laser pulses, occurring ten seconds apart. Each laser stimulus occurred after three preceding auditory anticipation cues spaced one second apart, to ‘count-down’ the onset of the laser stimulus so that participants could accurately predict it. The first of the auditory cues was concurrent with a visual cue indicating that the laser stimulus could be expected in three seconds time (Fig. 1), which was to act as a visual fixation point to discourage eye movements. Participants were instructed to attend to the intensity of the painand to rate it using the same 0-10 numerical scale as used in the psychophysics testing procedure as detailed above.

Electroencephalographic recordings of anticipatory and pain-evoked responses

EEG recordings were taken from 61 scalp electrodes placed according to an extended 10-20 system (Quik-Cap system, Neuroscan, Inc.). Bandpass filters were set at DC - 70Hz, with a sampling rate of 500Hz and gain of 500. A notch filter was set to 50Hz to reduce electrical interference. Electrodes were referenced to common average across all electrodes. The vertical and horizontal electro-oculograms (EOG) were measured for off-line reduction of blink and eye-movement artifacts.

Analysis of self-report measures

Differences were analyzed between groups in self-reported pain and the different measures of psychological coping. A series of univariate ANOVAs were conducted on each measure withtwo-tailed tests, with group (FM, OA, HP) as a fixed factor and age as a covariate. Pain ratings and the laser energy required to induce a level 7 pain were also compared in this way. VAS measures of clinical pain, and tender point scores, were subjected to a non-parametric test (Kruskal Wallis) due to non-normality of the data in the OA group. P-values were corrected for multiple comparisons using the False Discovery Rate (FDR) statistic set at q = 0.05 (using Matlab code by Nichols available at The resulting p-value threshold at this level was p < 0.012. For dependent variables showing significant group effects in the ANOVA, post-hoctests were performed on each group pair. For the parametric ANOVAs, we used the Scheffe test, while for the non-parametric data (VAS scores and tender points) we used the Mann-Whitney U test.

Pre-processing of EEG data

EEG data were analyzed using the EEGLAB toolbox (v4.515) running on MATLAB version 7.8. Averaged Event-Related Potentials (ERPs) covering the anticipation and pain phases of neural activity werecreated for each participant and each session, after the removal of linear trends in the data and ocular artifacts (by removing artifactual components after performing Independent Components Analysis), and filtering at 20 Hz low pass. ERPs were baseline-corrected to either the 500ms interval preceding the visual anticipation cue (for the measurement of anticipatory-evoked responses) or the 500ms preceding the laser stimulus (for measurement of the pain-evoked response).

Three 500ms temporal periods of the anticipatory brain response were extracted for analysis: a ‘baseline’ period, at -3500ms to -3000ms preceding the laser stimulus and occurring just prior to the anticipation cue (used for baseline correction of the EEG data), an ‘early’ period, at -2500ms to -2000ms preceding the laser stimulus and occurring soon after the anticipation cue, and a ‘late’ period, at –500ms to 0ms preceding the laser stimulus, as detailed and justified elsewhere (Brown and Jones 2010). The P2 peak of the Laser-Evoked Potential (LEP) was analyzed, as this was the largest amplitude potential generated post-laser stimulus and the only potential to be robustly produced across subjects with our laser stimulation parameters. For each subject and condition, P2 peak latencies were determined at the electrode for which the P2 peak showed maximum amplitude (Cz). An averaged 20ms window of LEP data was then extracted, centered on this latency.

Analysis of Event-Related Potential (ERP) data

For each temporal period (early anticipation, late anticipation, P2 peak), we performed a “scalp-region-of-interest” analysis to avoid the multiple comparisons problem of individually testing groups effects at every electrode. The voltage at nine electrodes were extracted and averaged for analysis of ERP amplitudes.The nine electrodes used included the electrode showing the maximum amplitude over the whole scalp for that time window when a grand average was created across subjects, plus the surrounding eight adjacent electrodes. This number of electrodes was regarded as reasonably broad enough to capture the peak of activity in most subjects (which wasn’t necessarily at the same electrode as in the grand average), while maintaining a reasonable degree of spatial specificity. For early anticipation the nine electrodes were centered onFCz, while for late anticipation at the P2 peak the nine electrodes were centered onCz. We used a univariate ANOVA (fixed factor: group; covariate: age) to identify group differences in the anticipatory and pain-evoked potentials. The ANOVA was performed once for each of the three time periods, and so results were judged to be statistically significant after correcting for multiple comparisons using False Discovery Rate (FDR), with a q value of 0.05 and two-tailed statistics. The resulting p value threshold was p < 0.01.

Source analysis of Event-Related Potential (ERP) data

Sources of anticipatory and pain-evoked potentials were estimated using the imaging approach to source reconstruction as implemented in SPM8 for MEG/EEG, combined withcustom MATLAB code for batch processing. For each participant, a forward model was constructed, usingan 8196 vertex template cortical mesh, coregistered to theelectrode positions of the standard 10-20 system via three fiducial markers. This produced ‘voxels’ (equivalent current dipoles) of 2mm x 2mm x 2mm. The lead-fieldof the forward model was computed using the three-shell BEMEEG head model available in SPM8. Source estimates were computed on the canonicalmesh using 256 multiple sparse priors per hemisphere including subcortical structures (Friston et al 2008), undergroup constraints (Litvak and Friston 2008). Source prior smoothness was set at 1mm. Source estimates were created based on windows of the ERP data of 500ms for pre-anticipation baseline, early and late anticipation (using the time windows described in the pre-processing section above), and 80ms for the P2 peak centered on the maximum amplitude of the peak for each individual participant. The resulting images were smoothed at 10mm FWHM, and log-transformed prior to statistical analysis to improve the normality in the distribution of the data.

Statistical analysis at the group level was performed using conventional SPM t tests. Statistical tests were performed over all voxels in the brain to enable exploratory analyses outside of our hypothesized regions of interest. To control for type I errors over the whole brain, results are reported that were significant at the voxel level after FDR correction(whole-brain) at pvoxel < 0.05, and only considering cluster sizes greater than 100 voxels. To view the images and extract clusters as volumes of interest, statistical parametricalmaps were thresholded at pvoxel < 0.001 (uncorrected). We compared the groups at each of the three time periods individually. The names of regions of activity were identified using the Automated Anatomical Labeling (AAL) system (Tzourio-Mazoyer et al 2002).

To test if there were any differences between patients in general and the HP group, analyses were conducted for each time period as follows. The two paired contrasts comparing HP patients with the FM group (“FM > HP” and “HP > FM”), were used, in addition to a further two contrasts comparing the HP group to the OA group (“HP > OA” and “OA > HP”). Age was added as a covariate for each contrast. These four contrasts were used to construct two separate conjunction analyses, for the purpose of identifying the main effect of chronic pain, i.e. common differences in both patient groups compared to the HP group. Specifically, activations were identified that were greater in the HP group than in the other two groups (i.e. [HP > FM] + [HP > OA]), and that were greater in the patient groups compared to the HP group (i.e. [FM > HP] + [OA > HP]).

To test for differences between the FM group and the other two groups, for each time period two paired contrasts were defined comparing FM patients with the HP group (“FM > HP” and “HP > FM”), and a further two contrasts comparing the FM group to the OA group (“FM > OA” and “OA > FM”),with age as a covariate. These four contrasts were then used to construct two separate conjunction analyses to test for the main effects of FM, i.e. common differences in the FM group compared to both OA and HP groups. The first test was for activations that were greater in the FM group than in the other two groups (i.e. [FM > HP] + [FM > OA]), while the second was for for activations that were lesser in the FM group (i.e. [HP > FM] + [OA > FM]).

A final analysis was performed to explain the discrepancy between the FM and OA groups in the overall size of the anticipatory-evoked potential during late anticipation. The OA group were contrasted to the FM group [OA > FM] in a single SPM t-test, with age as a covariate.