Neural underpinnings of behavioural strategies that prioritize either cognitive task performance or pain

Ò

PAIN 154 (2013) 2060–2071

www.elsevier.com/locate/pain

Neural underpinnings of behavioural strategies that prioritize either cognitive task performance or pain Nathalie Erpelding a, Karen D. Davis a,b,c,⇑ a

Division of Brain, Imaging, and Behaviour—Systems Neuroscience, Toronto Western Research Institute, University Health Network, Toronto, Ontario, Canada Institute of Medical Science, University of Toronto, Toronto, Ontario, Canada c Department of Surgery, University of Toronto, Toronto, Ontario, Canada b

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

a r t i c l e

i n f o

Article history: Received 24 April 2013 Received in revised form 3 June 2013 Accepted 17 June 2013

Keywords: Attention Cognitive interference CTA fMRI Pain Resting-state connectivity sMRI TBSS VBM

a b s t r a c t We previously discovered that when faced with a challenging cognitive task in the context of pain, some people prioritize task performance, while in others, pain results in poorer performance. These behaviours, designated respectively as A- and P-types (for attention dominates vs pain dominates), may reflect pain coping strategies, resilience or vulnerabilities to develop chronic pain, or predict the efficacy of treatments such as cognitive behavioural therapy. Here, we used a cognitive interference task and pain stimulation in 80 subjects to interrogate psychophysical, psychological, brain structure and function that distinguish these behavioural strategies. During concurrent pain, the A group exhibited faster task reaction times (RTs) compared to nonpain trials, whereas the P group had slower RTs during pain compared to nonpain trials, with the A group being 143 ms faster than the P group. Brain imaging revealed structural and functional brain features that characterized these behavioural strategies. Compared to the performance-oriented A group, the P group had (1) more gray matter in regions implicated in pain and salience (anterior insula, anterior midcingulate cortex, supplementary motor area, orbitofrontal cortex, thalamus, caudate), (2) greater functional connectivity in sensorimotor and salience resting-state networks, (3) less white matter integrity in the internal and external capsule, anterior thalamic radiation and corticospinal tract, but (4) were indistinguishable based on sex, pain sensitivity, neuroticism, and pain catastrophizing. These data may represent neural underpinnings of how task performance vs pain is prioritized and provide a framework for developing personalized pain therapy approaches that are based on behaviour–structure–function organization. Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

1. Introduction Intuitively, pain should interfere with the ability to sustain a high level of performance during an attention-demanding task. However, we discovered that some individuals improve cognitive task performance in the context of pain (ie, attention dominates, designated A-type), whereas others show decline of cognitive task performance during pain (ie, pain dominates, designated P-type) [76]. These behavioural strategies may lie at the core of understanding individual variability in pain coping strategies and the effectiveness of multidimensional treatment approaches for pain such as cognitive behavioural therapy. ⇑ Corresponding author. Address: Division of Brain, Imaging and Behaviour— Systems Neuroscience, Toronto Western Research Institute, Toronto Western Hospital, 399 Bathurst St, Room MP14-306, Toronto, Ontario M5T 2S8, Canada. Tel.: +1 (416) 603 5662; fax: +1 (416) 603 5745. E-mail address: [email protected] (K.D. Davis).

It is not known why in some individuals (P-type) pain disrupts cognitive abilities while in others cognitive performance is improved in the context of pain (A-type). Our previous study provided evidence that the brain reflects these behavioural strategies in that the A group, but not the P group, exhibited attenuation of pain-evoked functional MRI (fMRI) responses in primary and secondary somatosensory cortices (S1, S2), and the anterior insula (aIns) during task performance [76]. However, factors contributing to pain coping strategies (eg, individual characteristics, personality, sensory sensitivity or complexities of brain structure and functional network connectivity) remain unknown. Pain is of biological importance for survival and thus requires attention [33,60]. Some studies report that pain captures attention and disrupts working memory by reducing performance in cognitive-attentional tasks [9,13,20,21,31,49,53]. Other studies suggest that difficult cognitive tasks reduce pain perception [3,17,30,40, 46,47,52,56,64,67,70,71,73,85–87,93]. Accordingly, the A and P characterization helps explain inconsistencies in previous studies.

0304-3959/$36.00 Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.pain.2013.06.030

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

Pain and cognitive performance likely share but also compete for mental resources, the outcome of which underlies attentional switch abilities in A- and P-type behaviour. Here, we determined whether these differential effects relate to psychophysical sensitivities to pain, personality factors, brain structure and function. Subjects were assigned to groups on the basis of whether their reaction times (RTs) were faster (A-type) or slower (P-type) in a modified version of the numerical interference task during pain compared to nonpain trials. We predicted that the subject groups are distinguished by their gray and white matter, and connectivity in brain regions and networks involved in pain, attention and salience. To test this, we used voxel-based morphology and cortical thickness analysis to measure gray matter and probabilistic tractography to assess white matter connectivity between gray matter regions that showed group differences. We then used tract-based spatial statistics (TBSS) to investigate white matter integrity group differences in these tracts. Finally, we used resting-state fMRI to determine whether individuals could be distinguished by functional connectivity of sensorimotor and salience resting-state networks.

2. Methods 2.1. Subjects Eighty healthy right-handed subjects (40 women and 40 men; age range 19 to 36 years, mean ± SD age 24.5 ± 4.9 years) were recruited for the study and provided informed written consent to experimental protocols approved by the University Health Network research ethics board. Each subject underwent 2 experimental sessions. Session 1 included questionnaires, psychophysical tests to determine individuals’ thermal and pain sensitivity, temporal summation (TS) of heat pain and a cognitive interference task to categorize subjects into A and P groups on the basis of their behavioural responses. In session 2, structural and functional MRI data were acquired for each subject. The 2 sessions were held between 2 and 12 days apart.

2061

A modified version of the tonic heat pain model (THPM) introduced by Lautenbacher et al. [57] was used to induce painful stimulation during the cognitive interference task. The THPM consists of repeated pulsating heat stimuli that reach temperatures of 1 °C above the subject’s HP threshold. The use of THPM is advantageous because it produces a stable and reliable pain sensation without inducing sensitization or habituation effects [57]. Additionally, stimulation can be repeatedly applied over a prolonged period of time without reaching pain tolerance limits [57]. Here, heat stimuli were applied to the left volar forearm in 60 s blocks and corresponded to the length of 1 block of the cognitive interference task. Each stimulation block started from a baseline of 32 °C. Temperature was then increased from the baseline temperature to the target temperature (ie, 1 °C above the subject’s HP threshold) at 7 to 10 °C/s. The temperature was held at this level for 1 s, then decreased to 0.3 °C below the HP threshold and was kept at this temperature for 1 s; rates of 2 °C/s were used to pulsate stimuli between these 2 target temperatures. At the end of each stimulation block, temperature returned to the baseline temperature of 32 °C with rates of 7 to 10 °C/s. We purposely avoided collecting pain intensity ratings during and after the cognitive task to prevent subjects from diverting their attention towards the pain stimulus, which would have biased their behavioural during the task. Next, TS of heat pain was assessed on the right volar forearm. The baseline temperature was set to 32 °C, and then 10 consecutive 48 °C heat pulses were delivered with an interstimulus temperature of 40 °C at 0.5 Hz and fixed ramp rates of 10 °C/s. Subjects were instructed to rate their pain intensity after each heat pulse on a verbal numerical rating scale that ranged from 0 to 100 (0 = no pain, 100 = worst pain imaginable). For each subject, TS of heat pain was evaluated in 4 consecutive blocks that were separated by 60-s intervals. For TS analysis, the first run was considered as practice run and was discarded from further analysis. To investigate subjective pain intensity increases over the course of the 10 delivered suprathreshold heat pulses, the percentage change of the last heat pulse compared to the first heat pulse was calculated for each run separately, and percentage changes were then averaged over the last 3 stimulation blocks. Between-group effects of TS of heat pain were statistically evaluated by ANOVA.

2.2. Questionnaires 2.4. Cognitive interference task Subjects completed the NEO-Five Factor Inventory (NEO-FFI) [19] and the Pain Catastrophizing Scale (PCS) [83]. 2.3. Pain sensitivity and tonic heat pain Heat stimuli were applied to subjects’ volar forearm with a 30  30 mm Peltier thermode (TSA-II NeuroSensory Analyzer, Medoc Ltd, Israel) to determine thermal thresholds and to evaluate TS of heat pain. Details of the testing of cool detection (CD), warm detection (WD), cold pain (CP) and heat pain (HP) thresholds have been previously described [35]. Briefly, 3 consecutive stimulus trials were used for each detection threshold measurement and pain thresholds were measured in 5 consecutive trials on the left volar forearm. For each modality, the baseline temperature was 32 °C. The ramp rates (ie, ascending and descending) for CD and WD were 1 °C/s and consisted of 1.5 °C/s (ascending) and 10 °C/s (descending) for HP and CP. CD and WD had interstimulus intervals of 6 s, and CP and HP intervals were set at 10 s. The order of measurement was kept the same for each subject and consisted of CD, WD CP and then HP. CD and WD thresholds were determined by averaging the last 2 out of the 3 repetitions. CP and HP thresholds were based on the average of the 3 last measures of the 5 trials. A and P group differences for CD, WD, CP and HP were analyzed by univariate analysis of variance (ANOVA).

In our previous study [76], the A/P classification was based on the Stroop interference task. Because this task typically produces small RT differences (ie, a few milliseconds), here we wished to develop a more robust task in which larger RT differences could be used for a clearer classification into A-and P-types. Therefore, in the present study, a modified version of the numerical interference task [32,92] was used to separate participants into the P- and Atype groups. Our preliminary results for the numerical interference task revealed large RT differences between task conditions that were based on task congruency (ie, >100 ms RT difference between task conditions). The numerical interference task was modified to increase difficulty and thus amplify RT differences between task conditions, allowing for easier subject classification. Subjects viewed a screen that displayed 3 vertically aligned boxes, each containing digits between 1 and 9 (Fig. 1A). Two task conditions varied in difficulty based on congruency. In the easier Value (V) Task, subjects had to determine the highest value of digits across the 3 boxes (dominant information; correct response was ‘‘4’’ in the example shown in Fig. 1A, top). In the more difficult Number (N) Task, subjects were instructed to determine the greatest number of digits across the 3 boxes (nondominant information; correct response was ‘‘8’’ in the example shown in Fig. 1A, top). Subjects entered their responses using a numeric keypad with their right hand and were instructed to respond as quickly as possible but

2062

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

substantial habituation in the later blocks. Consequently, in order to minimize potential learning and habituation effects, we classified subjects into A and P groups on the basis of the data from blocks 4 (pain) and 5 (no pain), which showed the most stable RTs and percentage ACs. Each block contained 24 trials that were each displayed for 2.5 s, which were averaged for each block to ensure reliable RTs per block. The interblock interval was 60 s and V Task and N Task were presented consecutively (ie, 8 V Task blocks and then 8 N Task blocks). The order of V Task and N Task was randomized across subjects (ie, 40 subjects performed the V Task and then the N Task, and 40 subjects performed the N Task and then the V Task). Concurrent painful thermal stimuli were delivered throughout the second, fourth, sixth and eighth block for every subject. E-Prime (Psychology Software Tools Inc) was used to display stimuli and record subjects’ RTs and ACs. Subjects were divided into A and P groups on the basis of whether their RTs during the N Task increased (P group) or decreased (A group) during concurrent pain. The N Task requires a high amount of cognitive-attentional resources because subjects must compare information from 2 different sources and process the nondominant stimulus information (ie, number of digits) over the dominant information (ie, value of digits) [32]. 2.5. MRI acquisition

Fig. 1. Cognitive interference task design, subject RTs and task-by-group interaction. (A) Task design. A trial had 3 vertically aligned boxes that contained 1 to 9 digits each. In the Value (V) Task, subjects were instructed to determine the highest value of digits. In the Number (N) Task, subjects were instructed to determine the largest number of digits. Tasks consisted of eight 60 s experimental blocks with 24 trials that were presented for 2.5 s. Interblock intervals were 60 s. (B) RTs for the N Task for ‘‘No Pain’’ (block 4) and ‘‘Pain’’ (block 5). Lines indicate data for individual subjects and the bars indicate group averages. All subjects in the A group had decreased RTs with pain (left panel), and all subjects in the P group had increased RTs with pain (right panel). (C) Task-by-group interaction. There was a significant task-by-group interaction (F(1, 68) = 135. 83, P < .001). Post hoc tests revealed that A- and P-type groups had significant RT differences in the ‘‘Pain’’ (block 5) condition (P < .01), but not in the ‘‘No Pain’’ (block 4) condition (P = .53).

to maintain accuracy (AC). Subjects were not specifically instructed on how to cope with the pain (for example, by trying to ignore or focus). For each subject, 8 V Task and 8 N Task 60-s blocks were presented. The first 3 blocks were considered practice blocks because our preliminary analyses revealed significant learning effects in the first 3 blocks. Additionally, preliminary analyses revealed

Subjects underwent MRI on a 3 T MRI system (GE Medical Systems, Milwaukee, WI) fitted with an 8-channel phased-array head coil. For each participant, we collected (1) a high-resolution T1weighted anatomical scan, (2) 2 diffusion-weighted imaging scans, and (3) a resting-state fMRI scan using an echo-planar pulse imaging (EPI) sequence. Subjects were instructed to remain still during the MRI acquisition to obtain good image quality. A high-resolution anatomical whole-brain scan (180 axial slices; 256  256 matrix; 25.6 cm field of view; 1  1  1 mm voxels) was acquired using a T1-weighted inversion recovery prepped, 3-D fast spoiled gradient echo (IR-FSGPR) sequence (flip angle 15°; TE 3 ms; TR 7.8 ms; TI 450 ms). Two separate diffusion-weighted imaging scans (64 slices; TR 17000 ms; TE 83.3 ms; 96  96 matrix; 23 cm field of view; 2.4  2.4  2.4 mm voxels) were acquired for each subject using 60 noncolinear, isotropic directions (b = 1000 s/mm2) and 10 non-diffusion-weighted images (b = 0 s/mm2). A 5-min resting-state fMRI was collected using T2⁄-weighted EPI scan (40 slices; 64  64 matrix; 20 cm field of view; 3.125  3.125  4 mm voxels; TE = 30 ms; TR = 2000 ms). Subjects were instructed to relax, keep their eyes closed and think of nothing in particular [24,44]. As a result of technical issues with the scanner, one subject had a missing EPI scan, so that the restingstate analysis was performed on 69 subjects. 2.6. Gray matter data preprocessing and analysis 2.6.1. Cortical thickness To assess A and P group differences in cortical gray matter thickness, we used FreeSurfer software (http://surfer.nmr.mgh.harvard.edu/). These methods have been described in detail elsewhere [23,38,39] and also have been previously used by our lab [11,12,35,62,84]. Briefly, preprocessing steps involved intensity normalization, skull stripping, Talairach transformation, hemispheric separation and tissue segmentation. Next, the white matter/gray matter border (ie, white surface) and gray matter/ cerebrospinal fluid border (ie, pial surface) were identified and transformed into surfaces. We then calculated the distance between the 2 surfaces for every point and each hemisphere separately. Each subject’s cortical surface was anatomically parcellated; sulci and gyri were labeled and aligned to the FreeSurfer’s average

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

surface map and smoothed with a 5 mm full-width half-maximum (FWHM) Gaussian smoothing kernel. Statistical significance of group differences was set at P < .01, corrected for multiple comparisons based on Monte Carlo permutations with 5000 iterations using AlphaSim (http://afni.nimh.nih.gov/afni/doc/manual/AlphaSim) as previously used by our group [11,12,35,62,84]. Simulations in AlphaSim revealed that using an image-wide threshold of P < .01, corrected clusters with 122 contiguous vertices were required to reach significance. Analysis was restricted to a mask that contained S1 and S2, anterior cingulate cortex (ACC), midcingulate cortex (MCC), posterior cingulate cortex (PCC), insula (Ins) and prefrontal cortex (PFC). This mask was created using FreeSurfer’s parcellation atlas (aparc.a2009s) [27]. 2.6.2. Voxel-based morphometry We used FSL-VBM to analyze A and P group differences in subcortical gray matter. FSL-VBM constitutes a voxel-based morphometry style analysis [2,42]. Preprocessing steps included brain extraction using BET [77], tissue-type segmentation using FAST [94] and alignment of gray matter partial volume maps to MNI152 standard space using FLIRT [50,51]. A study-specific template was built using 30 subjects of the A group and a random subset of 30 subjects of the P group. These 60 gray matter images were nonlinearly aligned to the ICBM-152 gray matter template. This procedure was necessary to create an unbiased study-specific template, as different subject group sizes would have resulted in higher template accuracy for the larger subject group during the registration step. Next, the full set of 70 gray matter images was nonlinearly normalized onto this study-specific template. The registered partial volume images were then modulated to correct for local expansion or contraction by dividing them by the Jacobian of the warp field. Finally, the modulated registered gray matter volume images were smoothed with an isotropic Gaussian kernel with a sigma of 2 mm (4.6 mm FWHM). To identify group differences in both whole brain and subcortical gray matter, we used permutation-based nonparametric testing implemented in FSL (randomise v 2.0). Subcortical gray matter analysis was restricted to a subcortical mask that contained thalamus, caudate, putamen, hippocampus, amygdala, and periaqueductal gray (PAG). This mask was created using the Harvard– Oxford Subcortical Structural Atlas in FSL (http:// www.cma.mgh.harvard.edu/). We corrected for multiple comparisons using Monte Carlo simulations (5000 iterations) in AlphaSim. We calculated that with image-wide threshold of P < .01, a corrected cluster (P < .01) of 347 contiguous voxels for the whole brain gray matter and a cluster of 109 contiguous voxels for the subcortical gray matter mask was required to be significant. 2.7. White matter data preprocessing and analysis To investigate white matter connectivity and integrity differences between subject groups, we used a data-driven approach by combining probabilistic tractography and TBSS. Probabilistic tractography and TBSS are 2 DTI methods that evaluate white matter properties and organization in the brain based on water diffusion. Probabilistic tractography quantifies the likelihood of connectivity between brain structures. TBSS is a method to investigate white matter integrity in specific white matter tracts. Using probabilistic tractography, we first estimated structural connectivity between gray matter regions that differed between A and P groups. In the next step, we used TBSS to quantify fractional anisotropy (FA) differences between subject groups in the white matter tracts that derived from the probabilistic tractography analysis. DTI preprocessing steps included visual inspection of the raw DTI images, eddy current distortion correction using affine registration to a reference b0 (ie, first collected b0) and noise reduction

2063

by averaging each subject’s 2 DTI runs to a single volume. Then a b0-derived brain mask was applied to each diffusion-weighted image, and individual brain masks were created using BET [77]. These preprocessed DTI images were then used to run probabilistic tractography and TBSS. 2.8. Probabilistic tractography Probabilistic tractography was performed using FMRIB’s diffusion toolbox (FDT) [7]. Tract estimations between gray matter regions that resulted from cortical thickness analysis and voxelbased morphometry were conducted using probabilistic fiber tracking [7] with 5000 tract-following samples and a curvature threshold of 0.2. A dual-fiber approach was chosen to account for crossing fibers, thus yielding more reliable tract estimation results [8]. Then affine transformation registration matrices were created between individual high-resolution T1, diffusion tensor images, and standard space using FLIRT [50,51]. Seeds for tractography were based on significant gray matter regions that resulted from our analysis of group differences using cortical thickness analysis and voxel-based morphometry. To do this, masks of our significant cortical clusters were transformed from each subject’s individual FreeSurfer cortical surface space to T1 standard space in FSL. Subcortical findings from voxel-based morphometry using FSL-VBM already resulted in T1 standard space. Masks were then converted to standard diffusion space, and used for probabilistic tractography. Given that we used probabilistic tractography to assess structural connectivity between gray matter regions that were significantly different between subject groups resulting from gray matter analysis, we based our tractography analysis previously documented structural connections between gray matter regions of interest. Because we performed probabilistic tractography between significant gray matter regions, within the same hemisphere, we investigated, for the left hemisphere, structural connectivity between (1) thalamus and aIns, (2) thalamus and aMCC, (3) thalamus and caudate and (4) aIns and aMCC. For the right hemisphere, we assessed structural connectivity between (1) thalamus and aIns, (2) thalamus and SMA, (3) thalamus and OFC and (4) aIns and OFC. To prevent the tractography algorithm from considering tracts crossing to the opposite hemisphere, we entered an exclusion mask that consisted of a single slice along the midline in the sagittal plane in the model. Furthermore, exclusion masks posterior to the thalamus seeds were used to further exclude indirect fibers tracts from the thalamus. We computed structural connectivity entering each of those gray matter regions as a seed region and as a target region in separate connectivity analyses, and then averaged these tract images to obtain mean structural connectivity for each gray matter seed combination. Furthermore, we included all the tracts from the seed that reached the target region and terminated the tracts as soon as they hit the target region; hence we entered the target region as both inclusion mask (ie, waypoints mask) and termination mask. Consistent with previous work, we excluded tracts that had <10 voxels (ie, 0.02%) hitting the target region, and subsequent group maps were thresholded such that at least 25% of the subjects showed common tracts between seed and target [6]. On the basis of these criteria, tracts between (1) the left aIns and the left aMCC, (2) the right thalamus and the right OFC and (3) the right aIns and the right OFC had to be excluded from further analysis. Thus, white matter tracts between the left thalamus and the left aIns, the left thalamus and the left aMCC, the left thalamus and left caudate, the right thalamus and the right aIns, and the right thalamus and the right SMA were added and binarized to create a single binarized group map that was used as mask to perform TBSS.

2064

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

2.9. Tract-based spatial statistics TBSS allows voxel-by-voxel comparisons of white matter integrity changes and/or differences in the brain using various DTI metrics like FA. FA in white matter tracts was analyzed on the whole brain level and within the connectivity mask that was obtained from probabilistic tractography. Subjects’ FA maps were processed with TBSS [78,79]. First, FA images were created by fitting a tensor model to the raw diffusion data using DTIFIT in FDT v 2.0. FA maps were then nonlinearly registered to a 1  1  1 mm FA map (FMRIB58_FA). Next, a mean FA image was created and thinned to generate a mean FA skeleton that included white matter tracts that were common to all subjects. Each subject’s peak FA value was then projected onto this skeleton. The WM skeleton was thresholded at 0.2 and used for voxelwise cross-subject statistics. We analyzed group differences in FA on the whole brain and in the white matter skeleton. To do this, we extracted mean FA values in every voxel of the brain, and mean FA values in every voxel of the white matter skeleton. We tested for group differences by univariate ANOVAs. Additionally, we examined FA group differences within the structural connectivity mask that was derived from probabilistic tractography. To do this, we ran voxelwise permutation-based nonparametric testing in FSL that was restricted to this mask. Statistical significance (P < .01) was corrected for multiple comparisons using AlphaSim (image-wide threshold at P < .01), and resulted in a cluster threshold of 4 contiguous voxels. Significant TBSS results were ‘‘thickened’’ with ‘‘tbss_fill’’ in FSL for visualization purposes. This command thickens significant findings along the white matter tract (FA threshold >0.2) using a 3-mm Gaussian smoothing kernel. 2.10. Resting-state functional connectivity data preprocessing and analysis Resting-state fMRI preprocessing steps were carried out using FMRI Expert Analysis Tool (FEAT). Resting-state fMRI is used to quantify functional interactions between spatially distinct brain regions in task-free conditions. In this study, we investigated whether A-type and P-type subjects exhibit different patterns of resting-state connectivity in brain networks that are involved in pain and attention. Thus, we limited functional connectivity analysis to sensorimotor and salience resting-state networks (RSNs). For each scan, the first 4 volumes were discarded to allow for stabilization of the magnetic field. The images were then motion corrected, nonbrain tissue was removed using BET [77], spatially smoothed with a 5-mm kernel, temporally filtered with a 0.01 Hz high-pass filter and registered to the high-resolution T1 and MNI152 2-mm standard space using FLIRT [50,51]. Next, RSNs were extracted with standard group independent component analysis (ICA) using probabilistic ICA (pICA) [5] implemented in FSL MELODIC. ICA is a technique that separates a set of signals into independent (ie, uncorrelated and non-Gaussian) spatiotemporal components [5]. For pICA preprocessing, scans were subjected to variance normalization and were then temporally concatenated to create a single 4-D image. Next, the data set was decomposed into 25 independent components (ICs). Finally, the component output images were transformed to MNI152 2-mm standard space using FLIRT [50,51]. Before statistical inference, 13 out of the 25 ICs were identified as anatomically and functionally relevant RSNs. The 12 remaining components were classified as reflecting distinct artifacts due to motion, fluctuation in cerebrospinal fluid and physiological or scanner noise. Criteria for identifying relevant ICs were: (1) signal within a low frequency range of 0.1 to 0.01 Hz [18,58], (2)

connectivity patterns mainly located in gray matter and (3) coherent clusters of voxels within the IC map [59]. Subject-specific maps were created to test differences between A and P groups using the dual regression approach. Dual regression typically involves (1) a linear model fit (spatial regression) of the full set of group–ICA spatial maps against the separate fMRI sets, which results in time-course matrices representing each IC and subject and (2) a linear model fit (temporal regression) of these matrices against the fMRI images to estimate subjectspecific spatial maps [37]. Consequently, dual regression analysis results in separate time-course estimates for each component and subject. Inference was carried out on the subject-specific z-maps of the sensorimotor and salience RSNs. Significant differences between A and P groups were assessed using permutation-based nonparametric testing implemented in FSL (randomise v 2.0), incorporating threshold-free cluster enhancement (TFCE) with 5000 iterations [81]. Results for sensorimotor and salience RSNs were Bonferroni corrected to account for multiple comparisons; significance levels were thus set at P < .025. 2.11. Discriminant function analysis to cross validate subject classification Our subject classification was based on their performance (ie, RT differences) during a difficult cognitive task with and without pain. To ensure that our group classification was based on subjects’ coping behaviour with the given stimuli rather than random fluctuations, we performed a discriminant function analysis. The purpose of this type of analysis is to predict group membership based on a linear combination of the interval variables. Here, we used our group classification and aimed at cross validating our classification with our significant structural and functional brain findings. To do this, we calculated separate overall means for gray matter, white matter and functional resting-state connectivity data, which were then entered as separate predictors into the ‘‘leave one out’’ classification model. Our discriminant function analysis was performed on 29 A-type subjects and 40 P-type subjects (we excluded 1 subject in the A group because the EPI scan was not acquired as a result of technical difficulties with the MRI). 3. Results 3.1. Behavioural results On the basis of the N Task performance, we divided subjects into A (n = 35) or P (n = 45), depending on whether their RT decreased or increased while being exposed to painful thermal stimulation. Five subjects in each group with small RT differences (<10 ms) were eliminated, and subsequent analyses were thus based on 30 A subjects (12 women, 18 men) and 40 P subjects (24 women, 16 men). For the N Task, the A-type subjects had a mean ± SD RT decrease of 81.2 ± 55.6 ms (RTNo Pain = 1188.7 ± 217.3 ms; RTPain = 1107.5 ± 218.7 ms) during pain. In contrast, the P-type subjects had a mean RT increase of 93.7 ± 66.6 ms (RTNo Pain = 1157.4 ± 192.5 ms; RTPain = 1251.1 ± 211.3 ms) during pain (Fig. 1B). This group-specific RT difference resulted in a highly significant task-by-group interaction (F(1, 68) = 135.83, P < .001). Post hoc tests revealed that this result was due to significant RT differences between A and P subjects in the ‘‘Pain’’ condition (post hoc test, P < .01) and not due RT differences in the ‘‘No Pain’’ condition (post hoc test, P = .53) (Fig. 1C). Percentage ACs were statistically not significant differences between A subjects (ACNo Pain = 94.3 ± 6.9%; ACPain = 91.6 ± 14.4%) and P subjects (ACNo Pain = 95.5 ± 4.7%; ACPain = 94.5 ± 7.9%) (P > .05).

Ò

2065

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

For the easier V Task, RTs were relatively similar for subject groups and task condition (ie, ‘‘No Pain’’ and ‘‘Pain’’). Specifically, the A group had a RT decrease of only 3.0 ± 63.9 ms (RTNo Pain = 953.4 ± 215.1 ms; RTPain = 950.5 ± 224.2 ms) and the P group had a RT decrease of only 4.2 ± 87.2 ms (RTNo Pain = 982.7 ± 206.7 ms; RTPain = 978.5 ± 187.6 ms). Similarly, percentage AC differences for the V Task were statistically not significant (P > .05) between the A group (ACNo Pain = 97.7 ± 4.1%; ACPain = 97.8 ± 3.0%) and the P group (ACNo Pain = 98.4 ± 2.8%; ACPain = 97.2 ± 3.6%). Accordingly, MANOVA revealed that there were no significant task-by-group interactions for RTs or ACs for the V Task (all P > .05). Group demographics, mean RTs and percentage ACs, pain sensitivity variables (ie, thermal detection threshold and pain thresholds, TS of heat pain), and questionnaire scores are summarized in Table 1. There were no significant group differences in the temperature used in the THPM (46.8 ± 2.8 °C for subjects in the A group and 46.7 ± 2.8 °C for subjects in the P group). Of note is also that there were no significant group differences in neuroticism, pain catastrophizing, or any subscales of the NEO and PCS (all P > .05). Furthermore, the A and P groups did not differ in their thermal detection thresholds or their HP thresholds (all P > .05). However, the P group (13.5 ± 10.2 °C) exhibited significantly lower CP thresholds compared to the A group (8.4 ± 8.5 °C) (F(1, 69) = 5.04, P < .05) (but not significant after correcting for multiple comparisons).

Table 1 Subject demographics summarized by experimental group.a Characteristic

A-type

P-type

ANOVA

Sex (M/F) Age (years)

18/12 26.1 (5.6)

16/24 23.6 (4.2)

NSb NS

V Task RTNo Pain (ms) RTPain (ms) ACNo Pain (%) ACPain (%)

953.4 (215.1) 950.5 (224.2) 97.7 (4.1) 97.8 (3.0)

982.7 (206.7) 978.5 (187.6) 98.4 (2.8) 97.2 (3.6)

NSc

N Task RTNo Pain (ms) RTPain (ms) ACNo Pain (%) ACPain (%) CD (°C) WD (°C) CP (°C) HP (°C) TS (% change)

1188.7 (217.3) 1107.5 (218.7) 94.3 (6.9) 91.6 (14.4) 30.8 (0.7) 33.8 (0.7) 8.4 (8.5) 45.9 (2.8) 73.7 (29.5)

1157.4 (192.5) 1251.1 (211.3) 95.5 (4.7) 94.0 (7.9) 30.5 (0.9) 33.7 (0.8) 13.5 (10.2) 45.7 (2.8) 72.8 (13.5)

NS NS ⁄ NS NS

NEO-FFI Neuroticism Extraversion Openness Agreeableness Conscientiousness

51.0 55.8 62.8 47.6 46.7

50.6 58.5 59.3 50.5 45.2

(10.5) (9.0) (11.2) (13.3) (10.1)

NS NS NS NS NS

PCS Rumination Magnification Helplessness

14.3 (7.3) 5.8 (2.6) 3.3 (2.5) 4.7 (3.2)

14.5 (10.0) 5.9 (4.0) 3.1 (2.7) 5.6 (4.6)

NS NS NS NS

(10.3) (8.5) (9.0) (11.0) (11.6)

NSc

⁄c

Table 2 Gray matter regions in the P group with significantly greater cortical thickness and greater subcortical volume compared to the A group.a Area

BA

No. vertices per voxel

MNI coordinates x

L aIns L aMCC R aIns R SMA R OFC L thalamus L caudate R thalamus

24 6 11

357 208 134 134 132 114 111 110

y 30 12 32 13 16 8 18 16

T

z 15 19 9 4 30 18 6 22

2 30 11 63 24 18 18 18

3.88 2.85 3.45 4.32 2.87 3.85 2.98 3.53

aIns, anterior insula; aMCC, anterior midcingulate cortex; L, left; OFC, orbitofrontal cortex; R, right, SMA, supplementary motor area. a T scores and MNI coordinates are reported for the peak voxel. All clusters are significant at P < .01, corrected for multiple comparisons.

Both groups showed TS of heat pain from the first to the last heat pulse, but the magnitude of the summation effect did not differ between groups (A group: 73.7 ± 29.5% pain intensity increase; P group: 72.8 ± 13.5% pain intensity increase; P > .05). 3.2. Gray matter The P group had significantly greater cortical thickness than the A group in 4 cortical areas: bilateral aIns, left aMCC (BA 24), right SMA (BA 6) and right OFC (BA 11) (Table 2 and Fig. 2A). Voxelbased morphometry of subcortical areas revealed a cluster of greater gray matter in the P group compared to the A group that spanned the lateral and medial, including medial dorsal (MD), thalamus bilaterally as well as the left caudate (Table 2 and Fig. 2B). These group differences represent a 8.3% difference in regional cortical and a 8.1% difference in subcortical gray matter. There were no cortical or subcortical areas in which the A group had greater cortical thickness than the P group. Furthermore, there was no significant difference in whole brain gray matter based on voxelbased morphometry between subject groups (P > .05). 3.3. Probabilistic tractography

c

NS

AC, accuracy; CD, cool detection threshold; CP, cold pain threshold; HP, heat pain threshold; NEO-FFI, NEO-Five Factor Inventory; NS, not significant; N Task, Number Task; PCS, Pain Catastrophizing Scale; RT, reaction time; TS, temporal summation of heat pain; V Task, Value Task; WD, warm detection threshold. a Data from continuous variables are displayed as mean (SD). All comparisons were analyzed with univariate analyses of variance unless otherwise indicated. b Pearson chi-square test (2  2 table) was used to investigate sex differences between experimental groups. c Multivariate analysis of variance was used to analyze RT and AC differences between experimental groups. * Significant comparison (P < .05)

Probabilistic tractography revealed that the left thalamus and the left caudate were structurally connected through the internal capsule (IC). Similarly, white matter tracts between the left thalamus and the left aIns were found in the IC and the external capsule (EC). Structural white matter connections between the left thalamus and the left aMCC appeared in the IC, the anterior thalamic radiation (ATR) and the cingulum. Furthermore, probabilistic tractography confirmed structural connectivity between the right thalamus and the right aIns via IC and EC. The right thalamus and the right SMA were found to have white matter tracts running through the superior coronal radiation (SCR) and the corticospinal tract (CST). On the basis of predefined inclusion criteria (ie, including tracts that have >10 voxels hitting the target regions, including tracts that >25% of subjects have in common), white matter tracts between the right thalamus and the right OFC, the left aIns and the left aMCC, and the right aIns and the right OFC had to be excluded from further analysis. In summary, the final tract group map that was used for TBSS contained white matter tracts connecting the left thalamus and the left aIns, the left thalamus and the left aMCC, the left thalamus and the left caudate, the right thalamus and the right aIns, and the right thalamus and the right SMA. Accordingly, the structural connectivity mask consisted of the bilateral IC and EC, left ATR, left cingulum, right SCR and right CST.

2066

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

Fig. 2. Greater gray matter in the P group than the A group. (A) Significant cortical thickness differences between A and P groups. The P group had greater cortical thickness than the A group (blue on the brain maps) in the aIns bilaterally, left aMCC, right SMA and right OFC. The bar graphs indicate group averages ± SE. All results were significant at P < .01, corrected for multiple comparisons. (B) Significant subcortical gray matter volume differences between A and P groups. The P group had greater subcortical gray matter (blue on the axial brain slice) than the A group in the thalamus bilaterally and the left caudate. The bar graphs indicate group averages ± SE. All results were significant at P < .01, corrected for multiple comparisons. aIns, anterior insula; aMCC, anterior midcingulate cortex; OFC, orbitofrontal cortex; SMA, supplementary motor area.

3.4. Fractional anisotropy We tested for group differences in mean FA (1) at the whole brain level, (2) along the white matter skeleton and (3) within the structural connectivity mask that resulted from probabilistic tractography. There was no statistically significant group difference for mean FA on the whole brain (A group mean ± SD: 0.20 ± 0.01; P group: 0.19 ± 0.01) or on the white matter skeleton level (A group: 0.40 ± 0.01; P group: 0.39 ± 0.02) (all P > .05). Within the structural connectivity mask, we identified several white matter tracts in which A and P groups had significant FA differences. Specifically, TBSS analysis revealed that the A group had significantly higher FA (P < .01, corrected for multiple comparisons) than the P group in the bilateral IC, bilateral EC, bilateral ATR, and right CST adjacent to our gray matter finding in the

SMA/BA6 (Table 3 and Fig. 3). There were no white matter tracts in which the P group had greater FA compared to the A group. 3.5. Resting-state fMRI Resting-state fMRI differences between A and P groups were prominent in the sensorimotor RSN and the salience RSN. In accordance with previous research [10], group ICA in our study revealed that the sensorimotor RSN comprised S1, primary motor (M1) and premotor cortices, and SMA (Fig. 4A). Interestingly, dual regression analysis showed A and P group differences in their functional connectivity between the sensorimotor RSN and the rest of the brain. Specifically, the P group had significantly greater functional connectivity (P < .025, corrected for multiple comparisons) between the sensorimotor RSN, and a large cluster encompassing

Ò

2067

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071 Table 3 White matter regions in the A group with significantly higher fractional anisotropy compared to the P group.a Region

FA

No. of voxels

A-type

L IC L ATR L IC R ATR L ATR L IC L EC L IC L EC R CST (to BA 6) R EC L ATR L IC L ATR L EC L ATR

P-type

% Increase

Mean

SD

Mean

SD

0.469 0.268 0.362 0.290 0.305 0.470 0.222 0.440 0.256 0.364 0.218 0.282 0.527 0.249 0.324 0.371

0.036 0.022 0.039 0.032 0.022 0.044 0.026 0.027 0.028 0.062 0.047 0.026 0.046 0.043 0.035 0.030

0.433 0.254 0.334 0.276 0.289 0.440 0.207 0.420 0.238 0.325 0.209 0.269 0.501 0.235 0.305 0.346

0.045 0.023 0.036 0.024 0.025 0.044 0.023 0.033 0.024 0.057 0.040 0.031 0.040 0.038 0.028 0.038

7.8 5.1 7.8 4.8 5.2 6.4 6.5 4.4 7.1 10.5 4.0 4.5 4.9 5.6 6.1 6.5

MNI coordinates x

134 50 35 24 13 9 8 7 7 6 5 5 5 4 4 4

y

21 10 20 11 8 14 34 22 34 12 29 12 11 20 32 25

T z

17 17 22 16 5 11 2 7 8 3 7 6 8 10 10 41

9 14 1 14 9 1 1 18 5 59 16 14 1 23 9 3

3.84 3.15 3.14 3.02 3.07 2.98 2.62 2.64 2.81 2.68 3.11 2.58 2.83 2.81 2.81 3.46

ATR, anterior thalamic radiation; CST, corticospinal tracts; EC, external capsule; FA, fractional anisotropy; IC, internal capsule; L, left; R, right; SD, standard deviation. a Mean and SD, T scores, % Increase, and MNI coordinates are reported for the peak voxel. All clusters are significant at P < .01, corrected for multiple comparisons.

Fig. 3. Higher white matter fractional anisotropy (FA) in the A group than the P group. Significant FA differences between A and P groups were determined using tract-based spatial statistics (TBSS) on the mean FA skeleton (blue) within a mask (green) derived from probabilistic tractography analysis using the gray matter group differences as seeds. Significantly higher FA in the A group (red) than the P group were located in the ATR, CST, EC, and IC. All results were significant at P < .01, corrected for multiple comparisons. ATR, Anterior thalamic radiation; IC, internal capsule; CST, corticospinal tract; EC, external capsule.

precuneus, S1, paracentral lobule, M1 and SMA compared to the A group (peak = 14, 18, 64; Tmax = 4.09) (Fig. 4B). Furthermore, group ICA showed that, in accordance with in previous evidence [75], the salience RSN in our study had major nodes in the dorsal ACC/aMCC, SMA, aIns, temporoparietal junction and PFC (Fig. 4C). Dual regression analysis revealed that A and P group differences in their functional connectivity between the salience RSN and the rest of the brain. The P group had significantly greater functional connectivity (P < .025, corrected for multiple comparisons) between the salience RSN, and the precuneus (peak = 6, 50, 52; Tmax = 3.99) and the PCC (peak = 6, 42, 40; Tmax = 4.12) compared to the A group (Fig. 4D). 3.6. Discriminant function analysis The discriminant function analysis revealed that 36 of 40 (90%) subjects in the P group were correctly classified into the P group, and 23 of 29 (79.3%) subjects in the A group were correctly classified into group A. In total, the canonical correlation (total of the relationship between variables) was 0.72 with an eigenvalue of 1.09, which accounts for about 52% of the shared variance between the factors. The Box’s M test showed that the variance–covariance matrices were equivalent (F(6, 23973.98) = 0.68, P = .67). Interestingly, the

analysis revealed that the gray matter data was the most important predictor in the model (.81), resting-state functional connectivity was the second most important predictor (.69), and white matter was the third most important predictor ( .55). 4. Discussion This is the first study to uncover brain structure, function and network connectivity patterns that distinguish A-type individuals who prioritize cognitive performance over pain from P-type individuals who exhibit cognitive performance declines when experiencing concurrent pain. Specifically, the P-type group was distinguished from the performance-oriented A-type group by having (1) more gray matter in areas associated with pain, attention, and salience such as the aIns, aMCC, OFC, SMA, thalamus, and caudate, (2) less white matter integrity in bundles containing tracts mediating motor functions such as the IC and EC, ATR, and CST, (3) stronger functional connectivity between the sensorimotor RSN, and S1 and M1 and (4) stronger functional connectivity between the salience RSN, and the precuneus and PCC. Our study in a large sample of subjects builds on our initial discovery in a small study sample that some individuals had reduced task performance during pain while others had enhanced performance [76]. Our findings of less gray matter and functional

2068

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

Fig. 4. Greater sensorimotor and salience resting-state network (RSN) functional connectivity in the P group than the A group. (A) Group ICA revealed a sensorimotor RSN that consisted of S1, M1, SMA, and premotor cortex. (B) The P group had greater functional connectivity between the sensorimotor RSN, and a large cluster encompassing S1 and M1 compared to the A group. (C) Group ICA revealed a salience RSN with major nodes in the dACC/aMCC, pre-SMA, bilateral Ins, bilateral TPJ, and PFC. (D) The P group had greater functional connectivity between the salience RSN, and the precuneus and PCC compared to the A group. The bar graphs indicate group averages ± SE. All results were significant at P < .025 corrected for multiple comparisons using threshold-free cluster enhancement (TFCE) and Bonferroni correction. aMCC, anterior midcingulate cortex; dACC, dorsal anterior cingulate cortex; Ins, insula; ICA, independent component analysis; M1, primary motor cortex; PCC, posterior cingulate cortex; PFC, prefrontal cortex; PM, Premotor cortex; TPJ, temporoparietal junction; S1, primary somatosensory cortex; SMA, supplementary motor area.

connectivity in pain and salience regions in the A group (compared to the P group) could explain the flexibility in the A group in our previous study to attenuate pain-related responses in S1, S2 and aIns during task performance [76]. Also of note is that our newly modified numerical interference task ([34], modified from [32]) provides a more powerful method to assess individuals’ ability to switch attention between pain or task performance, allowing us

to identify individuals as either A- or P-type with substantial between-group and between-condition RTs compared to the modest RT effects we previously reported using a Stroop task [76]. Therefore, our current approach provides a sensitive method to determine if an individual fits the A or P profile and could thus be used in the future to direct therapeutic options. Although our results indicated that most of our subjects (ie, 87% of subjects) clearly

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

fit either behavioural group, it is conceivable that there are individuals that are not impacted in their cognitive performances by pain, which then potentially could be classified into a third group. However, because we were specifically interested in characterizing individuals according to their typical A- and P-type strategies and their characteristics in brain function and structure, we excluded these subjects with small RT differences (ie, <10 ms). Although the range of RTs suggest there may be a continuum of A/P behaviour (Fig. 1), nonetheless pain did seem to either interfere or facilitate task performance and this was linked to brain function and structure. Pain engages brain areas that both represent and modulate pain, including S1, S2, Ins, ACC/MCC, PFC, basal ganglia, and thalamus [1,26] and distraction away from pain can attenuate activity in many of these areas [3,15,63–65]. However, pain modulation studies rarely consider individual differences in pain coping strategies, despite the general sense that such differences must exist. The discovery of A- and P-type behavioural strategies prompted us to explore a series of analyses as to their underlying etiology. Differences in pain sensitivity have been linked to sex [45], personality [25], psychological factors (eg, pain catastrophizing and neuroticism) [4,22,43,54,55,82] and differences in cortical thickness to thermal sensitivity [35]. Conversely, here we found that the A and P groups were indistinguishable based on group demographics, personality traits including pain catastrophizing and neuroticism, as well as thermal detection and HP thresholds, and TS of heat pain. It is impossible to know whether subjects prioritized task vs pain via a conscious or unconscious process. However, the lack of any identifiable personality factor in contributing to the A vs P types suggests that our subjects were not aware of their behaviour and did not voluntarily choose one coping response over another to deal with the cognitive demanding task during pain. It is also possible that other psychological variables (ie, personality, motivation, and/or coping) could determine an individual’s behaviour. Lower CP thresholds (albeit not significant after multiple comparisons correction) in the P group could represent vulnerability to develop cold allodynia after injury (as it is a hallmark of sympathetically maintained pain). Given that demographic and personality factors did not significantly contribute to the A- or P-type behaviour, we next considered that brain gray matter may underlie differences in pain coping strategies when dealing with challenging cognitive tasks. Indeed, A subjects had less cortical gray matter in the aIns, aMCC and PFC compared to P subjects; areas implicated in pain, emotion, cognition, and salience functions [26]. These findings may also reflect different ways in which A and P individuals engage brain regions that balance bottom-up and top-down processes to switch between pain and attentional needs. Therefore, gray matter is a key structural feature that differentiates A and P subject groups, but the precise underlying mechanisms need further study. Furthermore, we considered whether A- and P-type subjects are affected by how the brain is wired. Interestingly, we found distinct white matter differences between A and P groups—specifically, the A group exhibited greater FA in the IC, EC and ATR. The IC is a major tract for perceptual, motor and higher cognitive functions [74] with interconnections between the MD thalamus and PFC, and fibers that join the lentiform nuclei and the caudate. The EC consists of fibers connecting the basal ganglia and a large part of the cerebral cortex, and the ATR projects from the medial and dorsal thalamic nuclei to the PFC and the frontal pole. Together, these tracts contain ascending nociceptive fibers and are part of the medial pain system, believed to contribute to the cognitive–affective dimension of pain [66]. This may indicate that A-type individuals have greater white matter in tracts that process cognitive– emotional information and thus express more effective motor

2069

reactions, a hypothesis supported by the finding that the A group had higher FA in the CST adjacent to SMA/BA6 compared to the P group. The P group exhibited greater gray matter than the A group in regions implicated in pain and cognitive–affective processes in addition to lower white matter integrity in tracts associated with cognitive–affective and motor processes. This constellation of gray matter–white matter findings in the P group has been linked to cognitive and emotional deficits, as well as altered pain sensitivity states observed in chronic pain populations [36,41]. Resting-state fMRI analysis revealed that A and P groups had different patterns of task-free functional connectivity. Task-free states involve spontaneous neuronal fluctuations that occur throughout the brain and show temporally coherent patterns between anatomically and functionally related brain regions [48,80,89]. It is commonly assumed that resting-state brain activity, among other functions, supports interoception and prepares responses for upcoming events [69]. Consistent with earlier research [10], our group ICA revealed that the sensorimotor RSN consisted of a coherent network of the bilateral S1, M1, SMA and premotor cortex. Using dual regression analysis, we found that P subjects had greater functional connectivity between the sensorimotor RSN, and a large cluster that extended over S1, M1, paracentral lobule, SMA and premotor cortex compared to A subjects. These regions are activated during pain experiences [1]. Therefore, strong functional connectivity between the sensorimotor RSN and sensorimotor areas may provide a neural basis for an increased pain-oriented behaviour in P subjects. Our second RSN of interest was the salience RSN that, in line with previous research [75], consisted of major nodes in the dACC/aMCC, precuneus, preSMA, bilateral aIns and PFC. Interestingly, P subjects had greater functional connectivity between the salience RSN, and the precuneus and PCC compared to A subjects. These regions are implicated in guiding attentional focus towards salient stimuli and in processing self-relevant emotional and nonemotional information [16,90]. They are also part of the default-mode network, a RSN that is suppressed during cognitive task performance [68]. Accordingly, our results indicate that painful stimuli may be perceived as more salient in the P group and may therefore explain their enhanced engagement in pain that diminished their task performance. Additionally, a greater connectivity between the salience RSN and key brain regions of the default-mode network in the P group suggests that it may be more difficult for P-type individuals to switch between different state of minds that balance task-negative and task-positive processes. It has been a quandary that some studies report that cognitive tasks and distraction attenuate pain [14,33,61,88], yet chronic pain can interfere with cognitive performance [28,29,31,32, 72,91]. Our findings suggest that differences in the availability of anatomical and functional resources may contribute to an individuals’ attentional switching/multitasking ability. Our characterization of A and P individuals suggests that previous studies could have been hampered by subject/patient heterogeneities. Thus, our data highlight the importance of understanding individual neural and behavioural factors that may impact the effectiveness of specific chronic pain therapies, such as interventions that manipulate coping skills to improve pain management. The findings broadly link behaviour to brain structure and network functionality and contribute to a better understanding of brain–behaviour relationships.

Conflict of interest statement The authors report no conflict of interest.

2070

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071

Acknowledgments This study was funded by a grant from the Canadian Institutes of Health Research (MOP 10626). Dr Erpelding was supported by the Fonds National de la Recherche Luxembourg (PDR-09-023). The authors thank Dr Massieh Moayedi and Aaron Kucyi for helpful insights with data analysis, Eugen Hlasny and Keith Ta for expert technical assistance and Drs Adrian Crawley, Tim Salomons and Mary Pat McAndrews for statistical consultation. References [1] Apkarian AV, Bushnell MC, Treede RD, Zubieta JK. Human brain mechanisms of pain perception and regulation in health and disease. Eur J Pain 2005;9:463–84. [2] Ashburner J, Friston KJ. Voxel-based morphometry—the methods. Neuroimage 2000;11:805–21. [3] Bantick SJ, Wise RG, Ploghaus A, Clare S, Smith SM, Tracey I. Imaging how attention modulates pain in humans using functional MRI. Brain 2002;125:310–9. [4] Baum C, Huber C, Schneider R, Lautenbacher S. Prediction of experimental pain sensitivity by attention to pain-related stimuli in healthy individuals. Percept Mot Skills 2011;112:926–46. [5] Beckmann CF, Smith SM. Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans Med Imaging 2004;23:137–52. [6] Beckmann M, Johansen-Berg H, Rushworth MFS. Connectivity-based parcellation of human cingulate cortex and its relation to functional specialization. J Neurosci 2009;29:1175–90. [7] Behrens TEJ, Johansen-Berg H, Woolrich MW, Smith SM, Wheeler-Kingshott CAM, Boulby PA, Barker GJ, Sillery EL, Sheehan K, Ciccarelli O, Thompson AJ, Brady JM, Matthews PM. Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging. Nat Neurosci 2003;6:750–7. [8] Behrens TEJ, Woolrich MW, Walton ME, Rushworth MFS. Learning the value of information in an uncertain world. Nat Neurosci 2007;10:1214–21. [9] Bingel U, Rose M, Gläscher J, Büchel C. FMRI reveals how pain modulates visual object processing in the ventral visual stream. Neuron 2007;55:157–67. [10] Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med 1995;34:537–41. [11] Blankstein U, Chen J, Diamant NE, Davis KD. Altered brain structure in irritable bowel syndrome: potential contributions of pre-existing and disease-driven factors. Gastroenterology 2010;138:1783–9. [12] Blankstein U, Chen JYW, Mincic AM, McGrath PA, Davis KD. The complex minds of teenagers: neuroanatomy of personality differs between sexes. Neuropsychologia 2009;47:599–603. [13] Buhle J, Wager TD. Performance-dependent inhibition of pain by an executive working memory task. PAINÒ 2010;149:19–26. [14] Bushnell MC, Duncan GH, Dubner R, Jones RL, Maixner W. Attentional influences on noxious and innocuous cutaneous heat detection in humans and monkeys. J Neurosci 1985;5:1103–10. [15] Bushnell MC, Duncan GH, Hofbauer RK, Ha B, Chen JI, Carrier B. Pain perception: is there a role for primary somatosensory cortex? Proc Natl Acad Sci U S A 1999;96:7705–9. [16] Cavanna AE, Trimble MR. The precuneus: a review of its functional anatomy and behavioural correlates. Brain 2006;129:564–83. [17] Coen SJ, Aziz Q, Yágüez L, Brammer M, Williams SCR, Gregory LJ. Effects of attention on visceral stimulus intensity encoding in the male human brain. Gastroenterology 2008;135:2065–74. 2074.e1. [18] Cordes D, Haughton VM, Arfanakis K, Carew JD, Turski PA, Moritz CH, Quigley MA, Meyerand ME. Frequencies contributing to functional connectivity in the cerebral cortex in ‘‘resting-state’’ data. AJNR Am J Neuroradiol 2001;22:1326–33. [19] Costa PT, McCrae R. Professional manual: Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI). Odessa, Florida, Psychological Assessment Resources; 1992. [20] Crombez G, Baeyens F, Eelen P. Sensory and temporal information about impending pain: the influence of predictability on pain. Behav Res Ther 1994;32:611–22. [21] Crombez G, Eccleston C, Baeyens F, Eelen P. Attentional disruption is enhanced by the threat of pain. Behav Res Ther 1998;36:195–204. [22] Crombez G, Eccleston C, Van den Broeck A, Goubert L, Van Houdenhove B. Hypervigilance to pain in fibromyalgia: the mediating role of pain intensity and catastrophic thinking about pain. Clin J Pain 2004;20:98–102. [23] Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage 1999;9:179–94. [24] Damoiseaux JS, Rombouts SARB, Barkhof F, Scheltens P, Stam CJ, Smith SM, Beckmann CF. Consistent resting-state networks across healthy subjects. Proc Natl Acad Sci U S A 2006;103:13848–53. [25] Davis KD. Neuroimaging of pain: what does it tell us? Curr Opin Support Palliat Care 2011;5:116–21.

[26] Davis KD, Moayedi M. Central mechanisms of pain revealed through functional and structural MRI. J Neuroimmune Pharmacol 2013;8:518–34. [27] Destrieux C, Fischl B, Dale A, Halgren E. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. Neuroimage 2010;53:1–15. [28] Dick BD, Eccleston C, Crombez G. Attentional functioning in fibromyalgia, rheumatoid arthritis and musculoskeletal pain patients. Arthritis Care Res 2002;47:639–45. [29] Dick BD, Rashiq S. Disruption of attention and working memory traces in individuals with chronic pain. Anesth Analg 2007;104:1223–9. [30] Dowman R. Distraction produces an increase in pain-evoked anterior cingulate activity. Psychophysiology 2004;41:613–24. [31] Eccleston C. Chronic pain and attention: a cognitive approach. Br J Clin Psychol 1994;33:535–47. [32] Eccleston C. Chronic pain and distraction: an experimental investigation into the role of sustained and shifting attention in the processing of chronic persistent pain. Behav Res Ther 1995;33:391–405. [33] Eccleston C. The attentional control of pain: methodological and theoretical concerns. PAINÒ 1995;63:3–10. [34] Erpelding N, Davis KD. A robust method to assess cognitive interference using a modified version of the numerical interference task. In: Paper presented at 41st Annual Meeting of the Society for Neuroscience. Washington, DC; 2011. [35] Erpelding N, Moayedi M, Davis KD. Cortical thickness correlates of pain and temperature sensitivity. PAINÒ 2012;153:1602–9. [36] Farmer MA, Chanda ML, Parks EL, Baliki MN, Apkarian AV, Schaeffer AJ. Brain functional and anatomical changes in chronic prostatitis/chronic pelvic pain syndrome. J Urol 2011;186:117–24. [37] Filippini N, MacIntosh BJ, Hough MG, Goodwin GM, Frisoni GB, Smith SM, Matthews PM, Beckmann CF, Mackay CE. Distinct patterns of brain activity in young carriers of the APOE-e4 allele. Proc Natl Acad Sci U S A 2009;106:7209–14. [38] Fischl B, Dale AM. Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proc Natl Acad Sci U S A 2000;97:11050–5. [39] Fischl B, Sereno MI, Dale AM. Cortical surface-based analysis. II: Inflation, flattening, and a surface-based coordinate system. Neuroimage 1999;9:195–207. [40] Frankenstein UN, Richter W, McIntyre MC, Rémy F. Distraction modulates anterior cingulate gyrus activations during the cold pressor test. Neuroimage 2001;14:827–36. [41] Geha PY, Baliki MN, Harden RN, Bauer WR, Parrish TB, Apkarian AV. The brain in chronic CRPS pain: abnormal gray–white matter interactions in emotional and autonomic regions. Neuron 2008;60:570–81. [42] Good CD, Johnsrude IS, Ashburner J, Henson RN, Friston KJ, Frackowiak RS. A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage 2001;14:21–36. [43] Goubert L, Crombez G, Van Damme S. The role of neuroticism, pain catastrophizing and pain-related fear in vigilance to pain: a structural equations approach. PAINÒ 2004;107:234–41. [44] Greicius MD, Srivastava G, Reiss AL, Menon V. Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: evidence from functional MRI. Proc Natl Acad Sci U S A 2004;101:4637–42. [45] Hashmi JA, Davis KD. Women experience greater heat pain adaptation and habituation than men. PAINÒ 2009;145:350–7. [46] Hoffman HG, Richards TL, Coda B, Bills AR, Blough D, Richards AL, Sharar SR. Modulation of thermal pain-related brain activity with virtual reality: evidence from fMRI. Neuroreport 2004;15:1245–8. [47] Hoffman HG, Sharar SR, Coda B, Everett JJ, Ciol M, Richards T, Patterson DR. Manipulating presence influences the magnitude of virtual reality analgesia. PAINÒ 2004;111:162–8. [48] Honey CJ, Sporns O, Cammoun L, Gigandet X, Thiran JP, Meuli R, Hagmann P. Predicting human resting-state functional connectivity from structural connectivity. Proc Natl Acad Sci U S A 2009;106:2035–40. [49] Houlihan ME, McGrath PJ, Connolly JF, Stroink G, Allen Finley G, Dick B, Phi TT. Assessing the effect of pain on demands for attentional resources using ERPs. Int J Psychophysiol 2004;51:181–7. [50] Jenkinson M, Bannister P, Brady M, Smith S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 2002;17:825–41. [51] Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5:143–56. [52] Kóbor I, Gál V, Vidnyánszky Z. Attentional modulation of perceived pain intensity in capsaicin-induced secondary hyperalgesia. Exp Brain Res 2009;195:467–72. [53] Kuhajda MC, Thorn BE, Klinger MR, Rubin NJ. The effect of headache pain on attention (encoding) and memory (recognition). PAINÒ 2002;97:213–21. [54] Kunz M, Chatelle C, Lautenbacher S, Rainville P. The relation between catastrophizing and facial responsiveness to pain. PAINÒ 2008;140:127–34. [55] Lautenbacher S, Huber C, Kunz M, Parthum A, Weber PG, Griessinger N, Sittl R. Hypervigilance as predictor of postoperative acute pain: its predictive potency compared with experimental pain sensitivity, cortisol reactivity, and affective state. Clin J Pain 2009;25:92–100. [56] Lautenbacher S, Prager M, Rollman GB. Pain additivity, diffuse noxious inhibitory controls, and attention: a functional measurement analysis. Somatosens Mot Res 2007;24:189–201.

Ò

N. Erpelding, K.D. Davis / PAIN 154 (2013) 2060–2071 [57] Lautenbacher S, Roscher S, Strian F. Tonic pain evoked by pulsating heat: temporal summation mechanisms and perceptual qualities. Somatosens Mot Res 1995;12:59–70. [58] Lowe MJ, Mock BJ, Sorenson JA. Functional connectivity in single and multislice echoplanar imaging using resting-state fluctuations. Neuroimage 1998;7:119–32. [59] Martino de F, Gentile F, Esposito F, Balsi M, Di Salle F, Goebel R, Formisano E. Classification of fMRI independent components using IC-fingerprints and support vector machine classifiers. Neuroimage 2007;34:177–94. [60] Melzack R. From the gate to the neuromatrix [review]. PAINÒ 1999:S121–6. [61] Miron D, Duncan GH, Bushnell MC. Effects of attention on the intensity and unpleasantness of thermal pain. PAINÒ 1989;39:345–52. [62] Moayedi M, Weissman-Fogel I, Crawley AP, Goldberg MB, Freeman BV, Tenenbaum HC, Davis KD. Contribution of chronic pain and neuroticism to abnormal forebrain gray matter in patients with temporomandibular disorder. Neuroimage 2011;55:277–86. [63] Nakamura Y, Paur R, Zimmermann R, Bromm B. Attentional modulation of human pain processing in the secondary somatosensory cortex: a magnetoencephalographic study. Neurosci Lett 2002;328:29–32. [64] Petrovic P, Petersson KM, Ghatan PH, Stone-Elander S, Ingvar M. Pain-related cerebral activation is altered by a distracting cognitive task. PAINÒ 2000;85:19–30. [65] Peyron R, García-Larrea L, Grégoire MC, Costes N, Convers P, Lavenne F, Mauguière F, Michel D, Laurent B. Haemodynamic brain responses to acute pain in humans: sensory and attentional networks. Brain 1999;122:1765–80. [66] Price DD. Psychological and neural mechanisms of the affective dimension of pain. Science 2000;288:1769–72. [67] Pud D, Sapir S. The effects of noxious heat, auditory stimulation, a cognitive task, and time on task on pain perception and performance accuracy in healthy volunteers: a new experimental model. PAINÒ 2006;120:155–60. [68] Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GL. A default mode of brain function. Proc Natl Acad Sci U S A 2001;98:676–82. [69] Raichle ME, Snyder AZ. A default mode of brain function: a brief history of an evolving idea. Neuroimage 2007;37:1083–90. [70] Raudenbush B, Koon J, Cessna T, McCombs K. Effects of playing video games on pain response during a cold pressor task. Percept Mot Skills 2009;108:439–48. [71] Rémy F, Frankenstein UN, Mincic A, Tomanek B, Stroman PW. Pain modulates cerebral activity during cognitive performance. Neuroimage 2003;19:655–64. [72] Roelofs J, Peters ML, Zeegers MPA, Vlaeyen JWS. The modified Stroop paradigm as a measure of selective attention towards pain-related stimuli among chronic pain patients: a meta-analysis. Eur J Pain 2002;6:273–81. [73] Schlereth T, Baumgärtner U, Magerl W, Stoeter P, Treede RD. Left-hemisphere dominance in early nociceptive processing in the human parasylvian cortex. Neuroimage 2003;20:441–54. [74] Schmahmann JD, Pandya DN. Disconnection syndromes of basal ganglia, thalamus, and cerebrocerebellar systems. Cortex 2008;44:1037–66. [75] Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, Reiss AL, Greicius MD. Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 2007;27:2349–56. [76] Seminowicz DA, Mikulis DJ, Davis KD. Cognitive modulation of pain-related brain responses depends on behavioral strategy. PAINÒ 2004;112:48–58.

2071

[77] Smith SM. Fast robust automated brain extraction. Hum Brain Mapp 2002;17:143–55. [78] Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mackay CE, Watkins KE, Ciccarelli O, Cader MZ, Matthews PM, Behrens TEJ. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage 2006;31:1487–505. [79] Smith SM, Johansen-Berg H, Jenkinson M, Rueckert D, Nichols TE, Miller KL, Robson M, Jones D, Klein J, Bartsch A, Behrens TEJ. Acquisition and voxelwise analysis of multi-subject diffusion data with tract-based spatial statistics. Nat Protoc 2007;2:499–503. [80] Smith SM, Miller KL, Moeller S, Xu J, Auerbach EJ, Woolrich MW, Beckmann CF, Jenkinson M, Andersson J, Glasser MF, Van Essen DC, Feinberg DA, Yacoub ES, Ugurbil K. Temporally-independent functional modes of spontaneous brain activity. Proc Natl Acad Sci U S A 2012;109:3131–6. [81] Smith SM, Nichols TE. Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage 2009;44:83–98. [82] Sullivan MJL, Adams H, Sullivan ME. Communicative dimensions of pain catastrophizing: social cueing effects on pain behaviour and coping. PAINÒ 2004;107:220–6. [83] Sullivan MJL, Bishop S, Pivik J. The pain catastrophizing scale: development and validation. Psychol Assess 1995;7:524–32. [84] Taylor KS, Anastakis DJ, Davis KD. Cutting your nerve changes your brain. Brain 2009;132:3122–33. [85] Terkelsen AJ, Andersen OK, Mølgaard H, Hansen J, Jensen TS. Mental stress inhibits pain perception and heart rate variability but not a nociceptive withdrawal reflex. Acta Physiol Scand 2004;180:405–14. [86] Valet M, Sprenger T, Boecker H, Willoch F, Rummeny E, Conrad B, Erhard P, Tolle TR. Distraction modulates connectivity of the cingulo-frontal cortex and the midbrain during pain—an fMRI analysis. PAINÒ 2004;109:399–408. [87] Veldhuijzen DS, Kenemans JL, De Bruin CM, Olivier B, Volkerts ER. Pain and attention: attentional disruption or distraction? J Pain 2006;7:11–20. [88] Villemure C, Bushnell MC. Cognitive modulation of pain: how do attention and emotion influence pain processing? PAINÒ 2002;95:195–9. [89] Vincent JL, Patel GH, Fox MD, Snyder AZ, Baker JT, Van Essen DC, Zempel JM, Snyder LH, Corbetta M, Raichle ME. Intrinsic functional architecture in the anaesthetized monkey brain. Nature 2007;447:83–6. [90] Vogt BA, Vogt L, Laureys S. Cytology and functionally correlated circuits of human posterior cingulate areas. Neuroimage 2006;29:452–66. [91] Weissman-Fogel I, Moayedi M, Tenenbaum HC, Goldberg MB, Freeman BV, Davis KD. Abnormal cortical activity in patients with temporomandibular disorder evoked by cognitive and emotional tasks. PAINÒ 2011;152: 384–96. [92] Windes JD. Reaction time for numerical coding and naming of numerals. J Exp Psychol 1968;78:318–22. [93] Yamasaki H, Kakigi R, Watanabe S, Hoshiyama M. Effects of distraction on pain-related somatosensory evoked magnetic fields and potentials following painful electrical stimulation. Brain Res Cogn Brain Res 2000;9:165–75. [94] Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging 2001;20:45–57.