The ability to avoid pain is crucial for survival because it minimizes harm and injury.17 Although individuals typically have a strong urge to avoid or escape pain,32 pain avoidance behavior is promptly adjusted after recent pain experiences43 and can even be disrupted when individuals repeatedly experience inescapable pain.33,46 Malfunctioning of pain avoidance is important because it is associated with diminished quality of life and higher treatment costs.6,26 However, studies investigating changes in the ability to avoid pain and their underlying neural correlates in humans are still sparse.
The preclinical literature provides convincing evidence that the periaqueductal gray (PAG) plays a central role in pain avoidance.1,12,13 Specifically the dorsal or dorsolateral aspect of the rodent PAG has been associated with risk assessment before defense behavior12 and the consequential readiness to actively avoid physical threats and to show defense behaviors.12,24,25,30,31,53 Typical patterns of defense behaviors are established by the PAG through sensorimotor circuits,28 possibly regulating a cortical effector network associated with motor preparation and execution. Behavioral responses are accompanied by autonomic changes, such as increases in blood pressure, heart rate, respiration, and muscle tone.30,39 This suggests that the rodent PAG is essential for the preparation of the body to avoid or escape pain. Currently, it is unknown whether the PAG and its cortical effector network subserve a similar role in human pain avoidance.
In this study, we examined the neural networks underlying the influence of successful and unsuccessful pain avoidance on subsequent pain avoidance behavior, using functional magnetic resonance imaging (fMRI). Because (subconscious) behavioral decisions not only depend on the most recent experience but are also influenced by the individual's learning history,35 participants with a broad experience in unsuccessful pain avoidance attempts previous to study participation were included. Specifically, participants with episodic migraine were recruited as part of the sample, ie, individuals who experience a frequent reoccurrence of unavoidable headaches as a key symptom of their condition (International Classification of Headache Disorders ICHD-II). Although individuals with migraine generally report feelings of helplessness about their migraine attacks, many still highlight a positive aspect of their pain condition and might even develop a psychological resilience to its negative consequences.36 Therefore, a wide range of helplessness levels was expected across the participants with migraine. This would allow to assess the influence of individual differences in helplessness on pain avoidance behavior.
We hypothesized that (1) participants (ie, individuals with and without migraine) show greater pain avoidance behavior subsequent to successful than unsuccessful avoidance attempts and (2) the readiness to avoid pain is driven by the PAG. Assuming that the experience of difficulties in coping with clinical pain can cause a generalized tendency to decrease pain avoidance behavior when having been unsuccessful on previous attempts, we further hypothesized that (3) reductions in pain avoidance behavior and in neural activation subsequent to unsuccessful avoidance attempts are greater with higher levels of self-reported helplessness in participants with migraine.
2. Materials and methods
2.1. Sample size calculation
This was the first formal investigation on the interaction between unsuccessful pain avoidance and subsequent alterations in pain avoidance behavior, and therefore, the effect size was unknown. Consequently, we based the sample size calculation for a behavioral effect on a desired medium effect size. In the context of analyses of variance (ANOVAs), effect size can be expressed as Cohen f2, and f2 = 0.06 (f = 0.25) is considered a medium effect size.9
A repeated measure ANOVA with 2 within-subject factors (“difficulty level” and “outcome on preceding trial”) with three-by-two levels and 1 between-subject factor (group) with 2 levels was conducted. To detect a medium interaction effect between the within and between factors, with a 5% probability of committing a type 1 error (alpha = 0.95) and a 20% probability of committing a type 2 error (beta = 0.8), a minimum of 28 participants were needed to be tested (GPower 3.1). To allow for potential data loss due to issues, such as technical difficulties with MRI or pain equipment, or potential incompliances of the participants, it was decided to recruit an additional 6 participants (20% extra), totaling in a sample of 34 participants.
Thirty-four participants were recruited for the study (6 males and 28 females; mean age 27 ± 5 years). Half of the initially recruited sample were participants with episodic migraine; however, 2 of them were excluded because of incompliance with the experimental instructions. This resulted in a final study sample of 32 (6 males and 26 females; mean age 27 ± 5 years, 15 participants with and 17 without episodic migraines).
The exclusion criteria included excessive smoking (more than 10 cigarettes a day); regular use of recreational drugs (more than once in 3 months); alcohol consumption of >10 UK units per week; the presence or history of chronic pain conditions other than episodic migraine; major medical, neurological, or psychiatric conditions; and MRI contradictions. Additional exclusion criteria for participants with migraine included less than 1 or more than 15 migraine attacks per month to ensure that they fulfilled the criteria for episodic migraine (ICHD-II) while excluding participants with an infrequent reoccurrence of migraine headaches. Participants could not have a headache (or any pain) at the time of testing to avoid the possibility that ongoing pain would affect their ability to focus on the tasks within the experiment or confound the fMRI signal. If pain was reported to be present on the day of testing, they were rescheduled. All included participants with migraine reported taking acute pain medication on demand when migraine attacks commence. One participant reported the daily use of preventative migraine medication (Table 1), including the day of testing. The remaining participants (N = 31) did not take any pain medication on the day of testing. All included participants with migraine met the criteria of episodic migraine23; 4 of them had migraines with aura and 11 had migraine without aura. All participants with migraine perceived their migraine attacks as unavoidable, aversive events but as distinctive attacks with pain-free periods in between.
Table 1 -
Characteristics of the study sample.
||Without migraine (n = 17)
|With migraine (n = 15)
| Duration (y)
| Typical duration of an attack (hours)
| Experience of auras (n participants)
|Migraine medication, acute (n participants)
|Migraine medication, prevention (n participants)
| Depressive symptoms (BDI)
| Hopelessness and helplessness (HHS)
BDI, Beck Depression Inventory; HHS, Hopelessness and Helplessness Scale of the Avoidance-Endurance Questionnaire; NSAIDs, nonsteroidal anti-inflammatory drugs.
Note that acute migraine medication is listed; however, none of the participants experienced any migraine attack during the study and therefore, did not use any acute medication. Only 1 participant was taking preventative medication. Groups were compared using 2-sided 2-sample t tests (age and questionnaire scores) and χ2 analysis (sex and handedness).
The study was approved by the local ethics committee, and informed consent was obtained from all participants according to the revised Declaration of Helsinki (2013).
2.3. Experimental procedure
Individual data collection was performed in a single session. On arrival, informed consent was obtained, and the study aims, pain rating scale, and tasks were explained. Once positioned on the scanner bed, pain sensitivity was tested (details below) to familiarize participants with the sensation of the electrical stimuli and calibrate individualized stimulation intensities. Participants were reminded of the task instructions and practiced 1 trial of the pain avoidance task (details given below) using visual feedback instead of electrical shocks. Before testing commenced, participants received 1 more painful and 1 more nonpainful stimulus to confirm the calibrated stimulation intensities. During functional image acquisition, participants performed 2 tasks—first the “pain avoidance task” and second a different behavioral task that is not further reported here. The behavioral tasks were followed by a high-resolution anatomical scan of the brain and finally another functional scan throughout which participants were engaged in a “motor–visual control task.” After participants had been removed from the scanner, they completed the Beck Depression Inventory-II2 and the avoidance-endurance questionnaire, of which only the hopelessness and helplessness subscale (HHS) was of interest here.20 The HHS asks specifically for helplessness perceived when dealing with clinical or reoccurring pain. Thus, participants with migraine were asked to relate the questions to their migraine, whereas participants without migraine were instructed to think of common pains they may experience every once in a while, such as occasional headaches, tooth ache, or period pain (in women).
2.4. Electrical stimulation
As part of the pain avoidance task, participants received transcutaneous electrical stimuli using pairs of 1-cm2 MRI-compatible surface electrodes (Vermed). Depending on the participant's task performance, stimuli were either painful or nonpainful. (1) Painful stimuli were applied to degreased skin over the retromalleolar path of the right sural nerve using a Grass S48 stimulator (AstroNova, Inc., West Warwick, RI). Each stimulation consisted of 4 repetitions of a 45-ms train (with 15 repetitions of 1-ms long pulses per train). The time in between the 4 repetitions was 500 ms. (2) Nonpainful stimuli were applied to degreased skin over the right anterior tibialis muscle using a Constant Voltage Isolated Linear Stimulator (STMISOLA; Biopac Systems, Inc, Goleta, CA). Each stimulation consisted of 3 repetitions of three 1-ms long pulses. The time in between the 3 repetitions was 750 ms.
Stimulation intensities for painful and nonpainful stimuli were individually determined before the scan. The painful stimuli were intended to be strongly aversive and painful, rated as 75 on a scale ranging from 0 (“no sensation”) to 100 (“extremely painful/unpleasant”) with 10 being the pain threshold (“just painful/unpleasant”). Stimulation intensities for the nonpainful stimuli had to be perceivable (>0) but nonpainful (<10). Stimulation intensities were determined using a staircase method. The intensity rated around 75 three consecutive times was used for the painful stimulation during the pain avoidance task. For the nonpainful stimuli, the stimulation intensity that the participant rated consistently around 5 was used.
2.5. Behavioral tasks
2.5.1. Pain avoidance task
We used an adaptation of the incentive delay task, which has previously been used to investigate motivated behavior in humans including by our laboratory.18,27 In the modified version (ie, “pain avoidance task”), the participants' goal was to avoid a painful stimulus and to receive a nonpainful one instead. At the beginning of each trial, participants would see 1 of 3 words (safe, easy, or difficult) displayed for 2 to 12 seconds (average 6 seconds). Then, a target cue would appear in the center of the screen (Fig. 1). Participants were instructed to press a button on a button box as fast as possible when the target cue appeared. They were told that they had to press the response button quickly in the easy condition, and even more quickly in the difficult condition, to avoid the painful stimulus; in “safe” trials participants knew they would receive the nonpainful stimulus as long as they pressed the button eventually. Should participants not have responded (which did not occur), the trial would have automatically continued and initiated the nonpainful stimulation after a maximum of 2 seconds. Despite being safe in “safe” trials regardless of reaction times, we asked participants to respond to the target cue immediately to continue the task without delays. After their response, participants received a painful or a nonpainful electrical stimulus, depending on their reaction time. Participants received the nonpainful stimulus (ie, they avoided the painful stimulus successfully) if they responded to the target cue within 310 ms for the easy and within 230 ms for the difficult condition. Conversely, they received the painful stimulus if they failed to press the button within the allotted time window of each condition. To ensure that the safe, easy, and difficult conditions were clearly distinguishable, we aimed for a success rate of 100%, 67%, and 33%, respectively. The order of the 3 conditions was pseudorandom but identical for all participants. After the stimulation, a fixation cross appeared on the screen for 6 to 12 seconds (average 9 seconds), before the next trial commenced. Figure 1 shows a schematic illustration of 1 trial. In total there were 36 trials, with 12 repetitions per difficulty level (“safe,” “easy,” and “difficult”). The duration of the task was approximately 11 minutes.
2.5.2. Motor–visual control task
The control task served the purpose to compare the hemodynamic responses between participants with and without migraine. Importantly, in this task behavior was not motivated by explicit threats or incentives and, thus, allowed us to assess potential differences in hemodynamic responses between the 2 groups that were unrelated to alterations in the neural correlates of pain avoidance. This seemed important because migraine is a neurological disorder,5 including alterations in central processes such as increased blood flow,21 altered brain excitability, intracranial arterial dilation, and sensitization of the trigeminovascular pathway.40,52 Such alterations might entail changes in neurovascular coupling or the hemodynamic response, albeit evidence to support this is still sparse.14
During the control task, participants first saw a fixation cross projected onto a screen for 1 second. The fixation cross was followed by the presentation of a circular checkerboard that flickered at a frequency of 3 Hz for a duration of 1.7 seconds. The participants were asked to press a button on a button box using their right index finger once per trial, namely, when they saw the presentation of the flickering checkerboard commencing. The 2 left-handed participants both reported daily use of a computer mouse with the right hand so that the button presses in this task with their right (nondominant) hand was similarly familiar to them as it was to the right-handed participants. The presentation of the checkerboard was followed by an intertrial interval of 6 to 20 seconds. During the intertrial interval, a fixation cross was displayed on the screen again. There were 30 trials in total.
2.6. Magnetic resonance imaging data acquisition
Brain images were acquired using a 3 T Siemens Magnetom TRIO scanner (Siemens, Erlangen, Germany) with a standard 32-channel head coil. Functional MRI data were acquired using a blood oxygenation level–dependent protocol with a T2*-weighted multiband accelerated gradient echo planar imaging (EPI) sequence (TR = 854 ms, TE = 30 ms, flip angle 52°, resolution 2 × 2 × 2 mm, field of view 208 mm, matrix size 104 × 104, and acceleration factor 6). Axial slices were oriented 30° from the line between the anterior and posterior commissure, covering the entire brain. After discarding the first 3 volumes to allow for steady-state magnetization, 721 volumes were acquired for the pain avoidance task and 436 volumes for the control task. Anatomical images were acquired using a T1-weighted 3D magnetization prepared rapid acquisition by gradient echo sequence (TR = 2300 ms, TE = 2.98 ms, flip angle 9°, field of view 256 mm, and resolution 1 × 1 × 1 mm). Throughout the session, participants wore earplugs and their heads were stabilized.
2.7. Statistical analysis
Participants with migraine were included for their pre-existing experience with unavoidable pain and for the anticipated individual differences across this group in feeling helpless when dealing with their pain. Potential group differences were tested for by including “group” as a between-subject factor in all analyses. To further assess whether observed group differences were associated with the extent to which participants feel helpless when dealing with their pain, individual helplessness scores were included in addition to “group” to behavioral and imaging analyses as appropriate (see below).
2.7.1. Statistical analysis of the behavioral data
Only trials with reaction times greater than 150 ms and below 1000 ms were included to exclude reaction times that were unlikely to reflect motivated, physiologically plausible behavior.57 This resulted in the exclusion of 10 trials across all participants (ie, 10 of 1152 trials in total).
Reaction times were corrected for nonnormality by applying f(x) = 1/x (kurtosis and skewness <2 after correction). This measure will be referred to as “response speed” throughout this article, with a high response speed indicating a faster reaction to the target cue and thereby increasing the chance to avoid the pain. To test the effects of difficulty level, preceding outcome and group on the response speed, a repeated measurement ANOVA design was used, using mixed model procedures with the between-subject factor “group” (2 levels: participants with migraine and participants without migraine) and the within-subject factors “difficulty level” (3 levels: safe, easy, and difficult) and “outcome on preceding trial” (2 levels: painful vs nonpainful). The ANOVA analysis was followed by post hoc ANOVA designs (ie, “group” by “difficulty level,” separately for “outcome on the preceding trial,” and “difficulty level” by “outcome on the preceding trial,” separately for the 2 groups) or pairwise comparisons, and calculation of Cohen d as a measure of effect size9 when appropriate.
To test for linear relations between different clinical characteristics of the participants with migraine and pain avoidance behavior, as well as between helplessness scores (HHS, log transformed) and (1) pain avoidance behavior for each group separately (participants with and without migraine) and (2) mean activation of brain regions identified within the fMRI analysis, Pearson product-moment correlation coefficients (Pearson r) were calculated between the respective variables.
The significance level was set to 5% for all analyses and results were Bonferroni-corrected to account for multiple testing. Outliers were defined as exceeding the group mean value by more than 2 SDs and were excluded from the analyses. All statistical analyses were performed using IBM SPSS Statistics 25 (SPSS Inc, Chicago).
2.7.2. Statistical analysis of functional magnetic resonance imaging data
188.8.131.52. Blood oxygenation level–dependent functional magnetic resonance imaging analysis of the “pain avoidance task”
All image processing and statistical analysis were performed using the software package FSL 5.0.8 (FMRIB Software Library; http://www.fmrib.ox.ac.uk/fsl).48 For 1 participant without migraine, only 500 volumes (instead of 721) were available for the analysis of the pain avoidance task because the scan stopped early because of technical failure.
184.108.40.206.1. Subject-level analysis
The following preprocessing steps were applied to each functional data set: motion and distortion correction using MCFLIRT;22 denoising using the independent component analysis (ICA) Multivariate Exploratory Linear Optimized Decomposition into Independent Components;4 and an automated algorithm to remove ICA components identified as motion-related noise (ICA-based Automatic Removal Of Motion Artifacts42), spatial smoothing (Gaussian kernel, full width at half-maximum: 5 mm), and temporal high-pass filtering (Gaussian-weighted least squares straight line fitting with sigma = 90 seconds). Susceptibility-related distortions were corrected using FSL field map correction routines. T1-weighted structural images were segmented into white matter (WM), gray matter, and cerebrospinal fluid (CSF). White matter and CSF maps were subsequently transformed to the individual's EPI space. Mean time series of WM and CSF were extracted from the individual EPIs and added to the first-level analyses for each participant as regressors of no interest.
A general linear model (GLM) was applied to each functional data set, modeling the time of the display of the difficultly level in 6 conditions (safe subsequent to successful pain avoidance, easy subsequent to successful pain avoidance, difficult subsequent to successful pain avoidance, safe subsequent to unsuccessful pain avoidance, easy subsequent to unsuccessful pain avoidance, and difficult subsequent to unsuccessful pain avoidance; “basic GLM”) to test the brain activation related to participants' preparatory state while waiting for the target cue, separately for each condition. This “preparatory phase” corresponds to the 2 to 12 seconds (6 seconds on average) when participants were presented with the cue indicating the difficulty level (“safe,” “easy,” or “difficult”) but before the onset of the target cue. The time points of the target cue, button press, electrical stimulation, and WM and CSF time series were included in the model as nuisance variables. A second GLM was applied to each functional data set, modeling 2 conditions (preceding outcome nonpainful and preceding outcome painful) with each trial weighted according to the level of difficulty (safe = 1, easy = 2, and difficult = 3; “linear GLM”) to test increasing brain activation with increasing task difficulty while waiting for the target cue. The resulting network of brain regions associated with a linear increase in task difficulty across the 2 conditions of the outcome on the previous trial will be referred to as the “preparatory matrix.” The same nuisance variables as described for the basic GLM were included. In both GLMs, all regressors other than the CSF and WM time series were convolved with a gamma hemodynamic response function (HRF) (phase = 0 seconds, SD = 3 seconds, and mean lag = 6 seconds) and the first temporal derivatives were included. Voxel-wise parameter estimates (PEs) were derived using the appropriate contrasts. Individuals' functional images were first registered to their own structural scan and subsequently to the International Consortium for Brain Mapping 152 nonlinear sixth generation symmetric template in MNI standard space using linear (FLIRT22) and nonlinear transformations (FNIRT, warp resolution = 6 mm).
220.127.116.11.2. Group-level analysis
Second-level analyses were performed using a mixed-effects model, implemented in FLAME.3 Statistical inference was based on a voxel-based threshold of z = 2.3, cluster corrected for spatial extent across the whole brain at P < 0.05.
To identify brain areas in which activation correlated with self-reported helplessness, a regressor was added to the second-level analysis of the contrasts “difficult subsequent to unsuccessful pain avoidance” and “difficult subsequent to unsuccessful vs successful pain avoidance” (“basic GLM + helplessness regressor”). This regressor coded the individual level of helplessness that participants reported using the HHS subscale of the avoidance-endurance questionnaire. The aim of the analysis was to assess which brain areas were (positively or negatively) associated in their activation with the individual level of helplessness dependent on the success on the previous trial. It was assumed that helplessness would be associated with relatively reduced activation of brain areas relevant to prepare pain avoidance behavior, especially when previous avoidance attempts had been unsuccessful. The scores on the HHS were normally distributed after log transformation. Log-transformed HHS scores were then demeaned across the sample before entered as a regressor.
Parameter estimates were extracted for areas that were associated with HHS scores for trials after successful pain avoidance and for those after unsuccessful pain avoidance using FEATquery as implemented in FSL. Extracted PEs for each of these regions were used to depict the correlation between them and change in response speed due to an unsuccessful pain avoidance attempt on the previous trial. Specifically, the mean differences in brain activation of the HHS-associated brain regions between trials after a successful vs unsuccessful pain avoidance attempt in the difficult condition were correlated with the differences in response speed after a successful vs unsuccessful pain avoidance attempt in the difficult condition.
18.104.22.168. Motor–visual control task: estimation of the hemodynamic response function
Data from 1 female participant with migraine were included in the analysis of the visual responses but excluded from the analysis of the motor responses because she failed to give the required behavioral motor responses.
Using FSL 5.0.8 (FMRIB Software Library; http://www.fmrib.ox.ac.uk/fsl),48 the following preprocessing steps were applied to each functional data set: spatial smoothing (Gaussian kernel, full width at half-maximum: 5 mm), motion correction, and temporal high-pass filtering (Gaussian-weighted least squares straight line fitting with sigma = 90 seconds). After this, a GLM was applied to each functional data set, modeling 1 event (visual–motor) to test the pattern of brain activation related to participants' visual and motor responses while seeing the flickering checkerboard and simultaneously pressing the response button. Hereafter, we refer to this analysis as the “initial GLM.”
For the actual evaluation of the HRF, we performed a finite impulse response estimation11,41 as implemented in the software NeuroLens 2 (scripted in Python). We applied this model-free analysis to estimate the shape of the HRF in the primary visual cortex (V1) in response to the onset of the visual stimulus and in the hand area of the contralateral primary motor cortex (M1), as well as in the ipsilateral putamen in response to the button press. The coordinates of the regions of interest (RoI) were based on the respective peak activations resulting from the “initial GLM.” Voxels exceeding a z-value of 2.3 were included in the RoIs and multiplied with anatomical masks of the left primary visual cortex, the contralateral primary motor cortex, and the ipsilateral putamen as defined by the “Juelich Histological Atlas” implemented in FSL (using a probabilistic threshold of 10% for each anatomical mask). RoIs for V1 and M1 were further manually restricted (based on anatomy) to avoid large RoIs exceeding the area of V1 and the hand area of M1, respectively.
To estimate the shape of the HRF for each RoI, an additional GLM was fitted for every individual data set using 41 delta functions as regressors with one sample estimate per repetition time (TR = 854 ms, resulting in a time window of 41 × 0.854 seconds = 35 seconds) and a third-order polynomial function as a regressor of no interest to control for slow drifts in the signal over time. The resulting HRF estimates were then further analyzed using a repeated measure ANOVA with “trial number” as a within-subject factor and “group” (participants with and without migraine) as a between-subject factor, performed separately for the 3 RoIs. This analysis served the aim to compare the HRF shapes between participants with and without migraine and was performed using IBM SPSS Statistics 25 (SPSS Inc, Chicago, IL).
3.1. Participants with migraine had a wide spread of helplessness levels across the group and, overall, showed many similarities to participants without migraine
In participants with migraine, helplessness scores on the HHS ranged from 0 to 48 (with 54 being the highest possible score on the HHS) and had a mean score of 21 (SD = 13), thus showing the anticipated wide spread of perceived helplessness when dealing with their migraine pain. Helplessness scores correlated with the length of a typical migraine attack (r = 0.59, P = 0.021), meaning that patients who experience longer attacks feel more helpless.
Comparing participants with and without migraine showed that they were similar in age and sex and had comparable depression scores (P's > 0.09, Table 1).
We further tested whether the HRF was comparable between participants with and without migraine. The HRF was quantified in 2 cortical and 1 subcortical brain region (ie, the primary visual cortex, the hand area of the contralateral primary motor cortex, and the ipsilateral putamen) in response to a nonincentive, motor–visual control task (Fig. 2A). We found no differences for the HRF between participants with and without migraine in any of the tested brain regions (Fig. 2B). More specifically, a repeated measure ANOVA estimating the influence of time and group on the HRF revealed no significant effect of group and no significant interaction between time and group in any of the regions tested. Therefore, it can be concluded that participants with and without migraine studied here had a comparable HRF, indicating similarity of the neurovascular coupling.
3.2. Overall, success rates in the pain avoidance task and the electrical stimulation intensities were comparable for participants with and without migraine
Applied stimulation intensities did not differ between participants with and without migraine for neither painful (participants without migraine 6.5 ± 2.8 mA (mean ± SD) and participants with migraine 7.3 ± 3.5 mA; P = 0.46) nor nonpainful stimuli (participants without migraine 1.1 ± 1.0 mA and participants with migraine 1.1 ± 1.0 mA; P = 0.94), indicating that both groups received comparable physical input throughout the pain avoidance task.
Furthermore, success rates for all 3 conditions (safe, easy, and difficult; independent of preceding pain avoidance success) were comparable between groups. Per default, the success rate in the safe condition was 100% for all participants, and no painful shocks were received. In the easy condition, participants with migraine received on average 4.00 ± 0.68 and participants without migraine 3.65 ± 0.51 painful shocks (P = 0.677, Cohen d = 0.15), equivalent to an overall success rate of 68.25%, which was close to the 67% success rate aimed for. In the difficult condition, participants with migraine received on average 8.07 ± 0.81 and participants without migraine 8.17 ± 0.69 painful shocks (P = 0.919, Cohen d = −0.04), equivalent to an overall success rate of 32.25%, which was close to the 33% success rate aimed for. Therefore, the goal of distinct success rates across conditions was achieved.
3.3. Response speed is altered after unsuccessful pain avoidance
Response speed was influenced by the difficulty level as well as by the outcome on the preceding trial. It should be noted that response speed is the inverse to reaction time (see Statistical analysis of the behavioral data section), meaning that higher response speed can be interpreted as an indication of greater alertness or motivation to avoid the painful stimulus. As depicted in Figure 3A (left panel), participants reacted faster with increasing difficulty for both types of preceding outcome (main effect of “difficulty level” F2,78 = 43.61, P < 0.001). However, the response speed was altered when participants failed to avoid the painful stimulus on the preceding trial (main effect of “outcome on preceding trial” F1,614 = 6.79, P = 0.009). Furthermore, analysis revealed an interaction between “difficulty level” and “outcome on preceding trial” (F2,692 = 3.48, P = 0.031). Post hoc pairwise comparisons between trials after unsuccessful vs successful pain avoidance showed a significant decrease for the difficult condition (mean difference −35.86 ms, P < 0.001, Cohen d = 0.49, Fig. 3A), but not for the safe or easy condition.
Response speed was altered in participants with migraine compared with those without migraine but only after unsuccessful pain avoidance on the preceding trial (significant interaction between the outcome of the preceding trial and group; F1,614 = 4.43, P = 0.036). When pain avoidance attempts were successful on the preceding trial both groups showed comparable response speed (nonsignificant main effect of group; F1,215 = 1.34, P = 0.249), independent of the difficulty level (nonsignificant interaction between group and difficulty level; F2,253 = 1.84, P = 0.162, Fig. 3A, middle panel). After unsuccessful pain avoidance on the preceding trial, however, response speed was slower in participants with migraine compared with participants without migraine (main effect of “group” F1,88 = 9.45, P = 0.003). Post hoc pairwise comparisons revealed a significant difference between groups for the difficult condition after unsuccessful pain avoidance on the preceding trials only (mean difference = 32.65 ms, P = 0.007, Cohen d = 0.59), but not for easy and safe (Fig. 3A, right panel). However, the interaction effect for group and difficulty level was not significant (F2,132 = 1.33, P = 0. 269). Furthermore, plotting the individual data points for all participants separately for all conditions (Fig. 3A, middle and right panel) depicts that participants with and without migraine showed large overlaps in their response speed—even for the difficult condition after unsuccessful pain avoidance attempts. This implies that many participants with migraine reacted to changes in demand (indicated by the difficulty level) and in previous experiences (modulated by previous pain avoidance success) in a comparable fashion to participants without migraine. The change in response speed after unsuccessful pain avoidance attempts on the previous trial was correlated to self-reported helplessness in participants with migraine (r = 0.67, P = 0.007, Fig. 4A), where a greater reduction in response speed after unsuccessful vs successful pain avoidance attempts on the preceding trial was associated with higher helplessness scores. For participants without migraine, by contrast, there was no association between the change in response speed after unsuccessful pain avoidance on the previous trial and HHS scores (r = −0.47, P = 0.057, Fig. 4A) (the scatterplot demonstrates that the trend for the negative association is related to an HHS score of 0 of many participants without migraine).
3.4. Diminished activation of the posterior parietal cortex during pain threats underpins behavioral disruptions, especially in patients with migraine reporting high helplessness
The brain activity was analyzed for the preparatory phase, ie, during the period before the actual avoidance behavior was performed. During this phase, participants were presented with the cue indicating the difficulty level to avoid the upcoming pain stimulus (safe, easy, or difficult) and knew that they would have to react very promptly with a button press as soon as the presentation of the difficulty cue switched to the presentation of the target cue. Although they needed to prepare for the required motor behavior, no actual movement was performed during this phase yet. Therefore, brain activity observed during the preparatory phase is not only correlated with the observed behavior but can be interpreted as a predictor of the subsequent motor response.
Activation during the preparatory phase increased with increasing task difficulty and resulted in a vast “preparatory matrix” activation. This “preparatory matrix” included regions related to alertness, motor preparation or execution, and cognitive control, such as bilateral dorsolateral prefrontal cortex, anterior insula, posterior parietal cortex (PPC), anterior cingulate cortex (ACC), premotor cortex, supplementary motor area, left primary motor cortex, basal ganglia, cerebellum, PAG, and superior colliculus (SC) (Fig. 3B). Comparing brain activation after unsuccessful vs successful pain avoidance on the preceding trial showed significantly reduced activation of most areas of the preparatory matrix, including bilateral PPC, SMA, ACC, premotor cortex, secondary somatosensory cortex, and insula (Fig. 3C). This suggests that alertness and motor preparation were decreased during the preparatory phase subsequent to unsuccessful pain avoidance. This is consistent with the behavioral result of a reduced increase in response speed with increasing difficulty after unsuccessful pain avoidance. Furthermore, after unsuccessful pain avoidance on the previous trial, participants without migraine show a greater increase in right parietal cortex activation with increasing task difficulty compared with participants with migraine (Fig. 3D). This is consistent with the behavioral finding of significantly reduced response speed in participants with migraine compared with those without after unsuccessful pain avoidance on the previous trial for the difficult condition, but not for easy and safe.
Being specifically interested in the question whether helplessness is associated with reduced brain activation and whether it might explain the decreased response speed in participants with migraine after unsuccessful pain avoidance, over and above a group difference, participants' HHS scores were added as a regressor to the contrast “previous avoidance attempt unsuccessful,” for the difficult condition only (basic GLM). Using whole-brain analysis, a significant negative correlation between the activation magnitudes after unsuccessful pain avoidance and the helplessness scores was found for participants with migraine in a network of cortical areas associated with motor behavior, attention, and memory, ,ie, bilateral primary motor cortex, left PPC, bilateral secondary somatosensory cortex, and left hippocampus and parahippocampus (Fig. 4B, left panel), with higher HHS scores being associated with less activation of these areas after unsuccessful pain avoidance on the previous trial. In participants without migraine, a significant negative correlation between the activation magnitudes after unsuccessful pain avoidance and the helplessness scores was found for PPC, occipital lobe, and right premotor cortex (Fig. 4B, middle panel). Comparison of participants with and without migraine showed that this negative association was stronger for participants with migraine for right PPC or posterior cingulate cortex, bilateral primary motor cortex, bilateral secondary somatosensory cortex, and right posterior insula (Fig. 4B, right panel). This suggests that not only having clinical pain but also feeling helpless about it is associated with decreased brain activation relevant to prepare for pain avoidance.
Because of the positive correlation between HHS scores and a reduction in response speed due to unsuccessful avoidance attempts on the previous trial in the difficult condition that was found for participants with migraine only, it was next assessed whether the change in activation within the HHS-associated brain regions of participants with migraine (bilateral primary motor cortex, left PPC, bilateral secondary somatosensory cortex, and left hippocampus or parahippocampus) after successful vs unsuccessful pain avoidance was also correlated with changes in pain avoidance behavior. Changes in brain activation of the HHS-associated brain regions were indexed by the difference score of the extracted PEs for each of the 2 contrasts, ie, after successful minus unsuccessful pain avoidance in the difficult condition; accordingly, changes in avoidance behavior were indexed by the difference in response speed after successful minus response speed after unsuccessful pain avoidance in the difficult condition. This analysis yielded a significant positive correlation (r = 0.53, P = 0.042). Assessing each cluster of this network (Fig. 4B, left panel) separately showed this association to be primarily driven by the PPC (r = 0.668, P = 0.007, Fig. 4C), with the correlations of change in response speed after successful vs unsuccessful pain avoidance and chance in activation of the other areas not being statistically significant (P's > 0.49).
In this study, pain avoidance behavior was increased with increasing task difficulty and compromised when previous attempts had been unsuccessful. Although observed for both groups, participants with migraine were even more affected by preceding unsuccessful avoidance attempts as they showed greater reductions of their response speed in the difficult condition in trials after unsuccessful pain avoidance compared with trials after successful pain avoidance. Interestingly, patients' helplessness scores were associated with a greater reduction in response speed after unsuccessful attempts, suggesting that coping style is an important determinant of pain avoidance behavior.
Increasing task difficulty was reflected in increased activation of a large brain network during the preparatory phase, ie, when participants prepared to react to the upcoming target cue, including areas associated with attention, motor behavior, cognition, and defense preparation (“preparatory matrix”). In parallel to the reduction in response speed after unsuccessful compared with successful pain avoidance, preparatory matrix activation was also reduced. After unsuccessful pain avoidance, participants with migraine showed a reduced increase in right PPC activation with increasing task difficulty compared with participants without migraine, possibly causing the difference in response speed between the groups observed for the difficult condition. Of several brain regions that were negatively associated with helplessness scores in participants with migraine, only left PPC activation predicted subsequent response speed, thereby highlighting its link to pain avoidance behavior. We conclude—in disagreement with the original hypothesis—that PPC rather than PAG plays a key role in human pain avoidance, with diminished PPC activation during pain threats underpinning behavioral disruptions, especially in migraine patients with elevated helplessness levels.
Preclinical studies describe a major role for the PAG in pain avoidance.12,24,25,30,31,53 Although this study identified a cluster spanning PAG and SC as part of the preparatory matrix and, thus, suggesting PAG-SC involvement in the preparation to avoid pain also in humans, no specific link of its activation to reduced pain avoidance behavior after unsuccessful pain avoidance attempts nor an association with helplessness scores was observed. This suggests that PAG-SC might be unaffected by immediate performance feedback as well as by long-term experiences with unavoidable pain. Periaqueductal gray stimulation in rodents leads to autonomic adaptations and changes in muscle tone,30,39 the latter likely regulated through the cerebellum,28 probably to facilitate defense behavior. In this study, the cerebellum was also part of the preparatory matrix but—similar to PAG-SC—unaffected by the outcome of previous avoidance attempts as well as by clinical pain. These findings suggest that the PAG and its effectors consistently reacted to pain threats, implying unaltered PAG-dependent defense mechanisms even after unsuccessful attempts. The apparent discrepancy in findings between the current and preclinical studies is likely related to different experimental designs: In a majority of preclinical studies, the PAG was either stimulated directly to elicit defense behavior1,13,53 or its neural activity was measured in response to noxious input.24,25,28,30,31Thus, these studies rather assessed the role of the PAG in pain defense behaviors than in the avoidance of imminent pain, in contrast to this study. One study in mice did provide evidence for PAG activation already during threat,12 which is more comparable with the current design. However, neural activity in PAG was recorded while the animals executed risk assessment behavior, whereas the participants here had to ensure not to miss the visual target cue during the preparatory phase to avoid pain, meaning that they had to be highly attentive and prepare for, but not execute, any motor behaviors. This suggests that different threat processing networks are involved depending on the exact contextual and behavioral requirements. Neural changes that did depend on unsuccessful pain avoidance on the previous trial were restricted to cortical areas in the current study and included the PPC, right insula, premotor cortex, and SMA. Although the premotor cortex and SMA are typically associated with motor planning and preparing the motor system for exact movements,15 the insula and PPC have been linked to regulatory and attention-related functions.34,38,47 The right insula, specifically, has been suggested to link brain areas encoding task difficulty and attention16,49; it has further been demonstrated to play a key role in evaluating task performance.16 Accordingly, the reduction in right insula activation after unsuccessful vs successful pain avoidance might encode a perceived decrease in task performance, ultimately causing a decreased signal in PPC as one of the brain's attention areas.34,47 Functional connectivity between the right insula and PPC has been shown specifically for visual attention tasks7,16,37 during which participants have to react to visual cues in a similar way as in this study's paradigm. The PPC has further been associated with generating nocifensive behaviors.29 Taken together, alertness in the current study might have been diminished after unsuccessful pain avoidance attempts due to the perception of a poor task performance, leading to delays in the initiation of nocifensive behavior. This effect was greater in participants with migraine compared with those without because they showed significantly reduced increases in PPC activation with increasing task difficulty after unsuccessful pain avoidance. This might have consequently led to the reduced response speed observed specifically in the difficult condition.
Interestingly, changes in left PPC activation after unsuccessful pain avoidance were associated with higher helplessness scores in participants with migraine, suggesting that helpless behavior within a pain context may be based on individual differences in the underlying neurobiology when facing threats. It has been described that, despite experiencing similar levels of pain, not everyone develops the same level of helplessness; some patients with pain are even resilient.36,50 In line with the current findings, studies in healthy individuals and patients with pain showed that helplessness is associated with functional and anatomical brain measures.8,44,45,56 Extending these earlier studies, the present results link the relevant brain function that scales with individual helplessness scores to coping relevant behavior (ie, pain avoidance) because a reduction in left PPC activation after unsuccessful vs successful pain avoidance was also associated with a reduction in response speed.
The observation that participants with migraine were even more affected by previous unsuccessful avoidance attempts suggests that, on a cognitive level, they might have appraised the benefit-cost ratio as lower compared with participants without migraine, making it less favorable to exhibit strong efforts of staying attentive to avoid the next painful stimulus.19 In addition to being associated with a stronger reduction in response speed after unsuccessful pain avoidance attempts, higher helplessness scores were related to a longer duration of migraine attacks. This reflects that patients with high helplessness scores experience a longer exposure to unavoidable pain with each migraine attack. This previous experience of uncontrollable pain can influence behavior in response to current threats.51 It is thus plausible that patients with high helplessness appraise the benefit-costs ratio differently after unsuccessful preceding attempts.35 This inaccurate appraisal of the benefit-cost ratio can explain the present findings and can also resolve an apparent contradiction of the present results with previous literature that has pointed to increased tendencies of pain avoidance behavior in patients with chronic pain.10,54,55 Patients with pain who avoid activities they had previously cherished out of fear to experience more pain might do so due to an overestimation of the involved cost (ie, increased pain). In this study, the inaccurate appraisal of the benefit-cost ratio might be more related to a misjudgment of their own ability to avoid pain, leading to an underestimation of the potential benefit (pain avoidance) when the effort of doing so had previously remained unrewarded.
The required sample size for this study was calculated for the anticipated behavioral effect. It is acknowledged that for a brain imaging study a sample of 32 participants (17 participants without and 15 with migraine) was relatively small. Nevertheless, strong effects within the fMRI analysis were detected, including interactions (eg, between group and difficulty level after unsuccessful pain avoidance on the previous trial and between difficulty level and previous outcome across the whole group), which often remain unobserved if statistical power is lacking. Furthermore, the findings related to self-reported helplessness are based on the subsample of participants with migraine and, owing to its relatively small number, are to be interpreted with caution. Nevertheless, the findings provide interesting insights into maladaptive coping with clinical pain and how this may influence attentional processing during pain threats that may be deemed difficult or even impossible to avoid dependent on individuals' previous experiences.
Individuals adjust pain avoidance behavior promptly, depending on their previous success in doing so. Participants with frequent unavoidable migraines were even more affected in their ability to quickly respond to pain threats when previously unsuccessful compared with participants without migraine. This behavioral group difference was underpinned by reduced PPC activation in participants with migraine, suggesting diminished alertness after unsuccessful pain avoidance. Helplessness in participants with migraine was related to both a greater reduction in PPC activation and in pain avoidance behavior after unsuccessful pain avoidance. This implies that not only the mere experience of clinical pain compromises pain avoidance but also an interaction of it with the individual's coping capacity.
Conflict of interest statement
The authors have no conflicts of interest to declare.
Supplemental video content
A video abstract associated with this article can be found at https://links.lww.com/PAIN/B485.
The authors thank Drs Catherine Bushnell and Richard Gracely for their valuable feedback and intellectual input to an earlier version of the manuscript. This work was supported by a Merit Scholarship Program for Foreign Students (Ministere de l’Education et de l’Enseignement superieur, MELS, Quebec), a Quebec Bio-Imaging Network (QBIN) scholarship for foreign students, and a The Louise and Alan Edwards Foundation's Edwards PhD Studentship in Pain Research to W. Gandhi and a Canadian Institutes of Health Research (CIHR) Operating Grant to P. Schweinhardt.
. Bandler R, Carrive P. Integrated defence reaction elicited by excitatory amino acid microinjection in the midbrain periaqueductal grey region of the unrestrained cat. Brain Res 1988;439:95–106.
. Beck AT, Steer RA, Brown GK. Manual for the Beck Depression Inventory–II. San Antonio: Psychological Corporation, 1996.
. Beckmann CF, Jenkinson M, Smith SM. General multilevel linear modeling for group analysis in FMRI. Neuroimage 2003;20:1052–63.
. Beckmann CF, Smith SM. Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans Med Imaging 2004;23:137–52.
. Burstein R, Noseda R, Borsook D. Migraine: multiple processes, complex pathophysiology. J Neurosci 2015;35:6619–29.
. Camacho EM, Verstappen SM, Chipping J, Symmons DP. Learned helplessness predicts functional disability, pain and fatigue in patients with recent-onset inflammatory polyarthritis. Rheumatology 2013;52:1233–8.
. Casey BJ, Thomas KM, Welsh TF, Badgaiyan RD, Eccard CH, Jennings JR, Crone EA. Dissociation of response conflict, attentional selection, and expectancy with functional magnetic resonance imaging. Proc Natl Acad Sci USA 2000;97:8728–33.
. Christidi F, Karavasilis E, Michels L, Riederer F, Velonakis G, Anagnostou E, Ferentinos P, Kollias S, Efstathopoulos E, Kelekis N, Kararizou E. Dimensions of pain catastrophising and specific structural and functional alterations in patients with chronic pain: evidence in medication-overuse headache. World J Biol Psychiatry 2020;21:726–38.
. Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed. Hillsdale: Lawrence Earlbaum Associates, 1988.
. Crombez GP, Eccleston CP, Van Damme SP, Vlaeyen JWSP, Karoly PP. Fear-avoidance model of chronic pain: the next generation. Clin J Pain 2012;28:475–83.
. Dale AM, Buckner RL. Selective averaging of rapidly presented individual trials using fMRI. Hum Brain Mapp 1997;5:329–40.
. Deng H, Xiao X, Wang Z. Periaqueductal gray neuronal activities underlie different aspects of defensive behaviors. J Neurosci 2016;36:7580–8.
. Depaulis A, Keay KA, Bandler R. Longitudinal neuronal organization of defensive reactions in the midbrain periaqueductal gray region of the rat. Exp Brain Res 1992;90:307–18.
. Descamps B, Vandemaele P, Reyngoudt H, Deblaere K, Leybaert L, Paemeleire K, Achten E. Absence of haemodynamic refractory effects in patients with migraine without aura: an interictal fMRI study. Cephalalgia 2011;31:1220–31.
. Donoghue JP, Sanes JN. Motor areas of the cerebral cortex. J Clin Neurophysiol 1994;11:382–96.
. Eckert MA, Menon V, Walczak A, Ahlstrom J, Denslow S, Horwitz A, Dubno JR. At the heart of the ventral attention system: the right anterior insula. Hum Brain Mapp 2009;30:2530–41.
. Esch T, Stefano GB. The neurobiology of pleasure, reward processes, addiction and their health implications. Neuro Endocrinol Lett 2004;25:235–51.
. Gandhi W, Becker S, Schweinhardt P. Pain increases motivational drive to obtain reward, but does not affect associated hedonic responses: a behavioural study in healthy volunteers. Eur J Pain 2013;17:1093–103.
. Gandhi W, Morrison I, Schweinhardt P. How accurate appraisal of behavioral costs and benefits guides adaptive pain coping. Front Psychiatry 2017;8:103.
. Hasenbring MI, Hallner D, Rusu AC. Fear-avoidance- and endurance-related responses to pain: development and validation of the Avoidance-Endurance Questionnaire (AEQ). Eur J Pain 2009;13:620–8.
. Hodkinson DJ, Veggeberg R, Wilcox SL, Scrivani S, Burstein R, Becerra L, Borsook D. Primary somatosensory cortices contain altered patterns of regional cerebral blood flow in the interictal phase of migraine. PLoS One 2015;10:e0137971.
. 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.
. Katsarava Z, Buse DC, Manack AN, Lipton RB. Defining the differences between episodic migraine and chronic migraine. Curr Pain Headache Rep 2012;16:86–92.
. Keay KA, Bandler R. Parallel circuits mediating distinct emotional coping reactions to different types of stress. Neurosci Biobehav Rev 2001;25:669–78.
. Keay KA, Bandler R. Distinct central representations of inescapable and escapable pain: observations and speculation. Exp Physiol 2002;87:275–9.
. Keefe FJ, Rumble ME, Scipio CD, Giordano LA, Perri LM. Psychological aspects of persistent pain: current state of the science. J Pain 2004;5:195–211.
. Knutson B, Westdorp A, Kaiser E, Hommer D. FMRI visualization of brain activity during a monetary incentive delay task. NeuroImage 2000;12:20–7.
. Koutsikou S, Watson TC, Crook JJ, Leith JL, Lawrenson CL, Apps R, Lumb BM. The periaqueductal gray orchestrates sensory and motor circuits at multiple levels of the neuraxis. J Neurosci 2015;35:14132–47.
. Lloyd D, Morrison I, Roberts N. Role for human posterior parietal cortex in visual processing of aversive objects in peripersonal space. J Neurophysiol 2006;95:205–14.
. Lovick T, Bandler R. The organizationof the midbrain periaqueductal grey and the integration of pain behaviors. In: Hunt S, Koltzenburg M, editors. The Neurobiology of Pain. New York: Oxford University Press Inc., 2005. p. 267–88.
. Lumb BM. Inescapable and escapable pain is represented in distinct hypothalamic-midbrain circuits: specific roles for Adelta- and C-nociceptors. Exp Physiol 2002;87:281–6.
. Lynn MT, Demanet J, Krebs RM, Van Dessel P, Brass M. Voluntary inhibition of pain avoidance behavior: an fMRI study. Brain Struct Funct 2016;221:1309–20.
. Maier SF, Seligman MEP. Learned helplessness—theory and evidence. J Exp Psychology-General 1976;105:3–46.
. Malhotra P, Coulthard EJ, Husain M. Role of right posterior parietal cortex in maintaining attention to spatial locations over time. Brain 2009;132:645–60.
. Marsella SC, Gratch J. EMA: a process model of appraisal dynamics. Cogn Syst Res 2009;10:70–90.
. Martelletti P, Schwedt TJ, Lanteri-Minet M, Quintana R, Carboni V, Diener HC, Ruiz de la Torre E, Craven A, Rasmussen AV, Evans S, Laflamme AK, Fink R, Walsh D, Dumas P, Vo P. My Migraine Voice survey: a global study of disease burden among individuals with migraine for whom preventive treatments have failed. J Headache pain 2018;19:115.
. Menon V, Adleman NE, White CD, Glover GH, Reiss AL. Error-related brain activation during a Go/NoGo response inhibition task. Hum Brain Mapp 2001;12:131–43.
. Menon V, Uddin LQ. Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct 2010;214:655–67.
. Mokhtar M, Singh P. Neuroanatomy, Periaqueductal Gray. StatPearls. Treasure Island: StatPearls Publishing Copyright © 2020, StatPearls Publishing LLC., 2020.
. Noseda R, Burstein R. Migraine pathophysiology: anatomy of the trigeminovascular pathway and associated neurological symptoms, cortical spreading depression, sensitization, and modulation of pain. PAIN 2013;2013:25.
. Orban P, Doyon J, Petrides M, Mennes M, Hoge R, Bellec P. The richness of task-evoked hemodynamic responses defines a pseudohierarchy of functionally meaningful brain networks. Cereb Cortex 2015;25:2658–69.
. Pruim RHR, Mennes M, van Rooij D, Llera A, Buitelaar JK, Beckmann CF. ICA-AROMA: a robust ICA-based strategy for removing motion artifacts from fMRI data. Neuroimage 2015;112:267–77.
. Roy M, Shohamy D, Daw N, Jepma M, Wimmer GE, Wager TD. Representation of aversive prediction errors in the human periaqueductal gray. Nat Neurosci 2014;17:1607–12.
. Salomons TV, Moayedi M, Weissman-Fogel I, Goldberg MB, Freeman BV, Tenenbaum HC, Davis KD. Perceived helplessness is associated with individual differences in the central motor output system. Eur J Neurosci 2012;35:1481–7.
. Salomons TV, Nusslock R, Detloff A, Johnstone T, Davidson RJ. Neural emotion regulation circuitry underlying anxiolytic effects of perceived control over pain. J Cogn Neurosci 2015;27:222–33.
. Seligman ME, Maier SF. Failure to escape traumatic SHOCK. J Exp Psychol 1967;74:1–9.
. Sestieri C, Shulman GL, Corbetta M. Attention to memory and the environment: functional specialization and dynamic competition in human posterior parietal cortex. J Neurosci. 2010;30:8445–56.
. Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TE, Johansen-Berg H, Bannister PR, De Luca M, Drobnjak I, Flitney DE. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 2004;23:S208–19.
. Sridharan D, Levitin DJ, Menon V. A critical role for the right fronto-insular cortex in switching between central-executive and default-mode networks. Proc Natl Acad Sci U S A 2008;105:12569–74.
. Stonnington CM, Kothari DJ, Davis MC. Understanding and promoting resiliency in patients with chronic headache. Curr Neurol Neurosci Rep 2016;16:6.
. Tabor A, Van Ryckeghem DML, Hasenbring MI. Pain unstuck: the role of action and motivation. Clin J Pain 2020;36:143–9.
. Vecchia D, Pietrobon D. Migraine: a disorder of brain excitatory-inhibitory balance? Trends Neurosci 2012;35:507–20.
. Vianna DM, Graeff FG, Brandao ML, Landeira-Fernandez J. Defensive freezing evoked by electrical stimulation of the periaqueductal gray: comparison between dorsolateral and ventrolateral regions. Neuroreport 2001;12:4109–12.
. Vlaeyen JW. Learning to predict and control harmful events: chronic pain and conditioning. PAIN 2015;156(suppl 1):S86–93.
. Vlaeyen JW, Linton SJ. Fear-avoidance model of chronic musculoskeletal pain: 12 years on. PAIN 2012;153:1144–7.
. Wiech K, Kalisch R, Weiskopf N, Pleger B, Stephan KE, Dolan RJ. Anterolateral prefrontal cortex mediates the analgesic effect of expected and perceived control over pain. J Neurosci 2006;26:11501–9.
. Woods DL, Wyma JM, Yund EW, Herron TJ, Reed B. Factors influencing the latency of simple reaction time. Front Hum Neurosci 2015;9:131.