Aberrant salience? Brain hyperactivation in response to pain onset and offset in fibromyalgia

Abstract

Objective.

While much brain research on fibromyalgia (FM) focuses on the study of hyper-responsiveness to painful stimuli, some studies suggest that the increased pain-related brain activity often reported in FM studies may be in part explained by stronger responses to salient aspects of the stimulation rather than, or in addition to, its painfulness. We thus hypothesized that FM patients would demonstrate elevated brain responses to both pain onset and offset, two salient sensory events of opposing valences.

Methods.

38 FM patients (mean age ± SD = 46.1 ± 13.4; 33 females) and 15 healthy controls (mean age ± SD = 45.5 ± 12.4; 10 females) received a moderately painful pressure stimulus to the leg during blood-oxygen-level-dependent (BOLD) functional MRI scanning. Stimulus onset and offset transients were modeled in a general linear model as stick functions.

Results.

During pain onset, FM patients demonstrated higher BOLD signal response compared to healthy controls, in dorsolateral, ventrolateral prefrontal and orbitofrontal cortices (DLPFC, VLPFC, OFC), frontal pole and precentral gyrus (PrCG). During pain offset, patients demonstrated higher and more widespread BOLD signal response compared to controls, including in frontal regions significantly hyperactivated in response to onset. In the patients, some of these responses were positively correlated to pain unpleasantness ratings (VLPFC, r = .35, p = .03; onset), pain catastrophizing scores (DLPFC, r = .33, p = .04; offset) or negatively correlated with stimulus intensity (OFC, r = −.35, p = .03; PrCG, r = −.39, p = .02; offset).

Conclusions.

Our results suggest that the increased sensitivity exhibited by FM patients in response to painful stimuli may reflect a more generalized hypersensitivity to salient sensory events, and that brain hyperactivation may be a mechanism potentially involved in the generalized hypervigilance to salient stimuli in FM.

1. Introduction

Fibromyalgia (FM) is a poorly-understood condition characterized by a constellation of symptoms including chronic widespread musculoskeletal pain and tenderness, extreme fatigue, disturbances in mood, cognition, sleep, and memory (1-3). While the pathogenesis of FM is not well understood, current consensus is that this condition is principally a disorder of central origin, arising from sensitized afferent nociceptive circuits and/or disrupted descending pain modulatory signaling, which in turn leads to widespread amplification of pain (1, 4-6) (although some studies have provided evidence of peripheral changes in a subgroup of FM patients) (7, 8). This persistent state of heightened CNS reactivity or central pain amplification often manifests clinically with increased sensitivity to painful stimuli (hyperalgesia) and the tendency to perceive non-painful stimuli as painful (allodynia). Psychophysical studies utilizing quantitative sensory testing (QST) have shown that FM patients, at equivalent stimulus intensity levels as healthy controls (CTL), report greater perceived pain to a variety of sensory stimuli, including mechanical (deep blunt pressure), thermal (heat and cold) and electrical stimuli (9, 10). Moreover, compared to CTL, FM patients’ show markedly reduced pain thresholds, potentiated temporal summation, and attenuated descending pain modulatory responses (11, 12). Neuroimaging studies complement and extend these findings by providing a glimpse into the putative neural substrates underlying the pathophysiology of FM. For example, FM patients relative to CTL display altered structural, neurochemical, neuroinflammatory and brain network connectivity patterns, as well as show augmented brain responses to painful and non-painful somatosensory stimuli in sensory-discriminative (e.g., primary and secondary somatosensory cortex), affective-motivational (e.g., cingulate and insular-opercular regions), and cognitive-attentional (e.g., dorsolateral prefrontal cortex) pain processing areas, as well as regions involved in processing of punishing and rewarding events (ventral tegmental area) (13-20).

In addition to heightened pain perception and augmented pain-related brain responses to tactile stimuli, FM patients also show evidence of generalized hypersensitivity to visual, auditory, and olfactory stimuli (10, 21-24). Given that FM patients appear to be hypersensitive to different types of sensory stimuli, we hypothesized that this increased sensitivity in response to noxious stimuli may partly reflect a more generalized hypersensitivity to salient sensory events. Thus, we implemented an analysis approach to evaluate the distinct brain responses to evoked pain onset and offset, two salient sensory events with opposing hedonic value. We reasoned that heightened responses to both onset and offset would support the view of a generalized hypersensitivity to salient stimuli in FM. Moreover, given that FM patients tend to report higher levels of negative affect, and particularly pain catastrophizing, compared to CTL, we hypothesized that exaggerated brain responses to pain onset/offset in FM would be positively associated with pain catastrophizing.

2. Patients and Methods

2.1. Subject characteristics

A total of 53 FM patients (mean age ± SD = 46.3 ± 11.4) and 17 CTL (mean age ± SD = 44.1 ± 14.8) initially entered the study. Each subject provided written informed consent prior to commencement of the study, and all study procedures were approved by the local Institutional Review Board. Inclusion criteria included an age of at least 18 yrs. and diagnosis of FM by a rheumatologist (for ≥1 yr.) according to Wolfe et al. (2010) American College of Rheumatology classification criteria [49]. Exclusion criteria included a history of psychiatric, neurological or autoimmune disorders, cardiac events and/or head injury, claustrophobia or MRI contraindication, current recreational drug use, including opioids, and pregnancy or plans to become pregnant. Patients were instructed to continue their medication regimens throughout the course of the study, which included antidepressants, gabapentin, nonsteroidal anti-inflammatory drugs, and acetaminophen. Healthy control subjects were frequency matched for age and gender to the patient group. It should be mentioned that while the number of CTL subjects in the present investigation is typical for group comparisons in fMRI studies, we have elected to include a significantly larger number of FM patients. This unbalanced design was adopted to maximize statistical power and dynamic range for regression analyses evaluating the association between brain activations and behavioral variables within the patient group (see below), thereby enhancing our ability to understand the clinical significance of the functional changes observed across groups, as in previous studies (19, 25).

2.2. Experimental design and procedures

In our previous report (25), we examined brain responses to the prolonged painful cuff stimulation period, as well as the 15-s post-stimulus offset period to model painful after-sensations in the same sample of FM patients and CTL. In contrast, herein we investigated brain response to cuff stimulus onset and offset, both rapid transitory events, in order to probe the degree to which FM patients are hypersensitive to salient sensory stimuli represented by pain onset (cuff inflation) and offset (cuff deflation).

Subjects participated in a behavioral visit performed at Brigham Women’s Hospital and an MRI visit held at Athinoula A. Martinos Center for Biomedical Imaging. At the behavioral visit subjects were asked to rate the severity and extent of their pain using a numerical rating scale (NRS) followed by administration of the Brief Pain Inventory (26), Neuropathic Pain Questionnaire (27), Widespread Pain Inventory, Symptom Severity Index (2), Pain Catastrophizing Scale (PCS) (28), Beck Depression Inventory (29) and a verbal anxiety NRS. Given the significant association between PCS scores and perceptual differences in painful after-sensations in FM patients previously reported by our group (25), we focused on PCS to further evaluate the role between catastrophizing and brain processing of salient aspects of the painful stimulation (i.e., pain onset/offset). Upon completion of self-report measures, subjects underwent QST, which included a cuff pain threshold assessment. Each subject’s cuff pain threshold was individually determined using cuff pain algometry with a Rapid Cuff Inflation System (E20; Hokanson Inc., Bellevue WA), which was adapted to inflate (i.e., reach the target pressures) and deflate (i.e., return to baseline) in approximately 2 seconds, in order to minimize the risk of startling the participants. For cuff pain threshold assessment, a 13 x 85 cm wide vascular pressure cuff was placed around the subject’s left calf and secured with a Velcro strip. The cuff was connected to the E20 device and inflated to a pressure (mmHg) individually calibrated to elicit a target pain intensity rating of approximately 40 out of 100 on a NRS with 0 representing ‘no pain’ and 100 representing ‘worst pain imaginable’. The pressure at which the subject rated a pain intensity of 40 out of 100 on the NRS was then used during the MRI cuff pain paradigm.

2.3. MRI acquisition, preprocessing and statistical analyses

fMRI data were acquired using a 3T Siemens Tim Trio scanner equipped with a 32-channel head coil (Siemens Healthcare GMbH, Erlangen, Germany). A high-resolution structural scan was collected using a multiecho magnetization-prepared rapid gradient-echo pulse sequence (TR/TE = 2.53 s/1.64 ms and TE2/TE3/TE4 = 3.5/5.36/7.22 ms, flip angle = 7°, voxel size = 1 x 1 x 1 mm). A T2*-weighted echo-planar image pulse sequence was also used to obtain high-resolution functional images during the cuff pain paradigm (TR/TE = 2 s/30 ms, 37 slices, voxel size = 3.1 x 3.1 x 3.6 mm). A total of four blood-oxygenation level-dependent (BOLD) runs were acquired (25). For each run, two block cuff pressure pain stimuli were delivered at the pressure previously determined during the threshold assessment procedure. The cuff pressure stimuli were delivered with a variable duration (75-105s; average ± SD: 90 ± 10s) to limit predictability. Following the end of each run, subjects were asked to rate the average pain intensity and unpleasantness of the stimulus using the NRS.

fMRI data preprocessing and analyses was performed with fMRI Expert Analysis Tool (FEAT; version 6), part of FMRIB Software Library (FSL; www.fmrib.ox.ac.uk/fsl). Our imaging pipeline included slice timing (slicetimer) followed by motion correction (MCFLIRT)(30), skull stripping using FSL’s brain extraction tool (BET)(31), realignment of mean fMRI volume with FLIRT(30, 32), grand-mean intensity normalization by a single multiplicative factor, high-pass temporal filtering with Gaussian-weighted least squares straight line fitting (with sigma = 136-164 s depending on the run, estimated using cutoffcalc), and spatial smoothing with a full width at half maximum of 5 mm. Time-series statistical analysis was conducted using FMRIB's Improved Linear Model (FILM) with local autocorrelations correction(33). Freesurfer’s bbregister tool was used to reconstruct the cortical surface and improve coregistration of functional with structural images (34).

A within-subject analysis using a general linear model (GLM) was performed by modeling the stimulus onset and offset transients as stick functions (35-37) each lasting a single TR in duration, corresponding to the approximate time of the cuff inflation and deflation (i.e., 2 s). In addition, the sustained tonic response between the onset and the offset periods was modeled as a boxcar function and designated as a regressor of no interest in the design matrix, lasting 75-105 s in duration. The model also included the 15 s post stimulus period after stimulus offset, which we previously used to evaluate brain activity associated with painful after-sensations (25). All regressors were convolved with a canonical double-gamma hemodynamic response function (HRF). To minimize the effect of motion in our estimates of brain responses to pain onset and offsets, the six head motion parameters (6 translations and 6 rotations) as well as a regressor of no interest for each volume determined to be an outlier in terms of motion (computed using fsl_motions_outliers) were entered into the design matrix. Time points within each run were flagged as outliers if they were deemed to have been significantly affected by motion based on the root means square frame displacement (38), as done in our previous study (25). The relatively conservative approach of scrubbing motion outliers was used given our specific focus on stimulus onset and offset, transition phases that might be particularly vulnerable to stimulus-correlated motion. However, group comparisons revealed no significant differences in head motion (25). The resulting first-level parameter estimates and variance maps were registered to Montreal Neurological Institute (MNI) 152 standard space using FMRIB Nonlinear Image Registration Tool (FNIRT)(39). Group maps were generated for the cuff pain onset and offset periods using a series of whole-brain voxelwise GLMs with FMRIB Local Analysis of Mixed Effects (FLAME) 1+2 (40), and automatic outlier detection enabled. The use of FLAME1+2 is well suited for unbalanced designs such as this one, because of its ability to model different variances using Metropolis-Hastings Markov Chain Monte Carlo sampling (41, 42) (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FEAT/UserGuide). The statistical maps were cluster-corrected for multiple comparisons using FSLs default cluster forming voxelwise threshold of Z > 2.3, and a cluster-corrected significance threshold of p < 0.05.

Given that group differences (FM > CTL) for onset and offset yielded significant effects in overlapping regions, we generated an intersection mask of both contrast maps and parcellated it using anatomical labels derived from the Harvard Oxford Atlas in FSL. Using an arbitrary threshold of 30, this parcellation method resulted in five subregions, which included the dorsolateral prefrontal cortex (DLPFC), ventrolateral prefrontal cortex (VLPFC), orbitofrontal cortex (OFC), precentral gyrus (PrCG) and frontal pole. These subregions were used as masks to extract and display percent signal change and time courses, for illustrative purposes. To calculate percent signal change, the contrasts for parameter estimates from the onset and offset phases for each subject, along with the peak-peak height of the regressor and the mean of the functional data time series were extracted from each voxel within our masks for each subject and averaged within groups. To visualize differences in percent signal change between groups for each mask, bar graphs were created to display means (±SEM) for cuff pain onset and offset. In addition, in patients, we performed two-tailed Pearson’s bivariate correlational analyses between percent signal change values obtained from each of the aforementioned five masks (i.e., DLPFC, VLPFC, OFC, PrCG, and frontal pole) for onset and offset and PCS scores. Lastly, given that salience can be determined both by the intensity of a stimulation as well as its painfulness, we performed an additional series of within-group exploratory correlational analyses (two-tailed Pearson’s) between percent signal change values extracted from each of the five masked subregions during both onset and offset, cuff pressure levels (mmHg) and pain ratings. Since these were not planned comparisons but instead post-hoc adjunctive correlational analyses, correction for multiple comparisons was not performed.

3. Results

A total of 43 FM patients and fifteen CTL participated in the fMRI visit. Five subjects were excluded from analyses due to either technical difficulties encountered during scanning, incorrect scan parameters used, incomplete scan sessions, and in one case, due to subject falling asleep during scanning. Therefore, the final sample for all subsequent analyses consisted of 38 FM patients (mean age ± SD = 46.1 ± 13.4; 33 females) and fifteen CTL (mean age ± SD = 45.5 ± 12.4; 10 females). Demographic and behavioral data from the final sample of FM patients and CTL are shown in Table 1. As expected and previously reported (25), the pressure required to elicit comparable pain ratings was significantly lower in FM compared to CTL.

Table 1.

Demographic, clinical and psychosocial characteristics

	Controls (n = 15)		FM Patients (n = 38)
Variable	Mean	SD	Mean	SD	p
Age (years)	45.53	12.40	46.13	13.44	0.882
PCS	5.93	5.97	23.16	13.08	<0.001
CPA thresholds	190.67	85.81	101.18	57.20	<0.001
NRS ratings	42.54	3.82	44.84	7.70	0.309

Abbreviations: PCS = pain catastrophizing scale; CPA = cuff pressure algometry; NRS = numerical rating scale.

3.1. fMRI response to pain onset and offset

During cuff pain onset (Fig. 1), both groups showed widespread increases in BOLD signal, including primary somatosensory/motor (S1/M1; leg area), secondary somatosensory (S2), anterior and posterior insular, posterior parietal, pregenual anterior, middle and posterior cingulate, lateral and medial prefrontal cortices, as well as cerebellum, basal ganglia, thalamus, and brainstem. Both groups also showed deactivation in S1/M1 (outside of the leg area) and higher-order visual cortices (e.g., lateral occipital cortex).

Figure 1.

Statistical maps showing within group brain responses to cuff pain onset for healthy controls (CTL) and fibromyalgia (FM) patients displayed in the top panel, with group difference map (FM > CTL) shown in bottom panel. Patients showed increased brain activity in response to pain onset in the frontal cortical areas compared to controls including the dorsolateral prefrontal (DLPFC), ventrolateral prefrontal (VLPFC), orbitofrontal (OFC) cortices, precentral gyrus (PrCG) and frontal pole (FP).

Results from the whole-brain voxelwise group comparison analyses demonstrated that in response to cuff pain onset FM patients, relative to CTL, showed significantly greater activation in the frontal cortex, including the VLPFC (i.e., the inferior frontal gyrus), the DLPFC (i.e., middle and superior frontal gyri), the OFC, the PrCG, and frontal pole (see Table 2 for cluster information).

Table 2.

Group Differences in Brain Responses to Pain Onset and Offset

cluster size	cluster	side	label	zstat	peak
(# voxels)	p-value
					x	y	z
Pain onset
FM > CTL
1397	0.000404	R	middle frontal gyrus	4.41	42	14	36
		R	superior frontal gyrus	4.32	24	28	50
		R	middle frontal gyrus	4.17	40	26	46
		R	inferior frontal gyrus	3.64	60	18	18
		R	frontal pole	3.57	48	48	14
		R	precentral gyrus	3.58	48	0	28
		R	orbital frontal gyrus	3.29	28	32	−8
CTL > FM	n.s.
Pain offset
FM > CTL
17067	1.79E-23	R	precuneus	4.97	8	−76	48
		R	superior frontal gyrus	4.87	24	10	48
		L	precentral gyrus	4.84	−2	−16	58
		R	frontal pole	4.72	50	44	16
		R	orbital frontal gyrus	3.71	32	40	−8
8567	1.63E-14	R	middle temporal gyrus	4.72	56	−48	−10
		L	fusiform gyrus	4.7	−20	−90	−20
		L	lingual/parahippocampal gyrus	3.25	−30	−46	−6
1711	0.000212	L	thalamus	4.45	−14	−6	14
		R	posterior cingulate gyrus	4.22	4	−50	10
CTL > FM	n.s.

During pain offset (Fig. 2), both groups demonstrated activations in S2, anterior and posterior insulae, and anterior middle cingulate cortex, basal ganglia, thalamus, and brainstem, as well as deactivations in S1/M1 (outside of the leg representation) as well as occipital, medial prefrontal and dorsolateral prefrontal cortices. As for the onset contrast, in response to stimulus offset FM patients demonstrated elevated BOLD signal in the frontal cortex including, DLPFC, VLPFC, OFC, PrCG, and frontal pole, compared to CTL. FM patients relative to CTL also showed greater BOLD signal increases in regions not statistically significant in the onset contrast, including the dorsomedial prefrontal cortex, supplementary motor area, paracentral lobule, posterior cingulate gyrus, precuneus, posterior parietal cortex, fusiform and lingual gyri, middle temporal gyrus, bilateral thalamus, caudate nuclei and cerebellum (see Table 2 for cluster information).

Figure 2.

Statistical maps illustrating within group brain responses to cuff pain offset for healthy controls (CTL) and fibromyalgia (FM) patients (top panel), and the group difference map displaying the FM > CTL contrast (bottom panel). Patients showed increased brain activity in response to pain offset in the lateral prefrontal cortical areas including the dorsolateral prefrontal (DLPFC) and ventrolateral prefrontal (VLPFC) cortices in addition to orbitofrontal cortex (OFC), precentral gyrus (PrCG) and frontal pole (FP) compared to controls. Abbreviations: cereb. = cerebellum; IPL = inferior parietal lobule; LOC = lateral occipital cortex; MPFC = medial prefrontal cortex; M1 = primary motor cortex; OP = occipital pole; PCC = posterior cingulate cortex; precun. = precuneus; S1 = primary somatosensory cortex; SPL = superior parietal lobule; thal. = thalamus.

While the FM>CTL contrast maps for both onset and offset demonstrated similar regions of statistical significance in DLPFC, VLPFC, OFC, PrCG and frontal pole (Fig. 3A-B), different activity patterns drove such group differences in the two phases. In response to the stimulus onset, both groups generally responded with activations in these regions, which were larger in patients (Fig. 3C). During stimulus offset, on the other hand, the FM patients demonstrated activations in regions where CTL demonstrated deactivations (Fig. 3D). Examination of the time courses extracted from these regions (Fig. 3E) revealed marked differences in neural response signatures between groups. The most striking effect among the regions evaluated was the observed increases in BOLD signal at the termination of cuff pain in the DLPFC, VLPFC, PrCG and frontal pole in FM compared to CTL.

Figure 3.

Statistical maps for group contrasts (FM > CTL) during pain onset and offset showing shared regions comprising the lateral prefrontal cortex (IPFC) along with derived region of interest (ROI) masks used for analysis. (A) Statistical maps showing group differences in brain responses for FM > CTL contrasts during pain onset (green) and pain offset (red), with overlapping regions common to both groups shown in yellow. (B) ROIs masks included the dorsolateral prefrontal cortex (DLPFC) shown in light blue, the ventrolateral prefrontal cortex (VLPFC) shown in orange, the precentral gyrus (PrCG) shown in pink, the frontal pole (FP) shown in purple, and the orbitofrontal cortex (OFC) shown in dark blue. (C) Bar graphs representing mean percent signal changes extracted for anatomically parcellated ROIs created from the intersection mask for pain onset, including the DLPFC, VLPFC, PrCG, and FP. (D) Bar graphs representing mean percent signal changes extracted for anatomically parcellated ROIs (DLPFC, VLPFC, PrCG, and FP) created from the intersection mask for pain offset. (E) Time courses for extracted blood-oxygenation level-dependent signal responses from anatomically parcellated subregions derived from the IPFC intersection mask for pain onset and offset.

3.2. Correlational analyses

Results from the correlational analysis revealed a significant positive association between PCS scores and DLPFC signal changes during pain offset (r = .33, p = .04) in FM patients (Fig. 4A). There was a trend towards a significant positive association between PCS scores and frontal pole signal changes during pain offset as well (r = .30, p = .07). No other correlations with PCS reached statistical significance (p > .07).

Figure 4.

Scatter plots depicting correlations between brain activations extracted from anatomically parcellated region of interest (ROIs) created from the intersection mask for pain onset and offset and pain catastrophizing scale (PCS) scores, cuff pressure (mmHg) and pain unpleasantness ratings in fibromyalgia patients only. (A) Scatter plot illustrating dorsolateral prefrontal cortex (DLPFC) percent signal change in response to pain offset was positively correlated with pain catastrophizing scores in FM patients. Scatter plots showing orbitofrontal cortex (OFC) (B) and precentral gyrus (PrCG) (C) percent signal change in response to pain offset negatively correlated with cuff pressure (mmHg). (D) A scatter plot of ventrolateral prefrontal cortex (VLPFC) percent signal change in response to pain onset positively correlated with ratings of pain unpleasantness.

Correlational analyses revealed no significant correlations between brain response to cuff onset and cuff pressure in patients or CTL. During cuff offset, however, we observed significant negative correlations between cuff pressure and OFC (r = −.35, p = .03) and PrCG (r = −.39, p = .02) activation in FM patients only (Fig. 4B-C). No significant relationship between pain intensity ratings and brain response to cuff onset or offset in either group emerged, although there was a trend toward significance for an association between pain intensity and PrCG activation during cuff onset (r = .31, p = .06). There was a significant positive correlation between pain unpleasantness ratings (r = .35, p = .03) and VLPFC activity in patients during cuff onset (Fig. 4D), whereas in CTL a positive correlation between pain unpleasantness and VLPFC activation (r = .6, p = .02) for cuff offset was found.

4. Discussion

Our findings demonstrated that patients with FM, compared to CTL, show extensive brain hyperactivity in response to both cuff pain onset and offset. While an increased response to pain onset was not surprising, particularly given the extensive literature demonstrating overall stronger brain responses to pain stimuli in FM, the large group differences observed at pain offset were striking. Such differences were noted in frontal regions that were also differentiated between FM and CTL for pain onset (i.e., DLPFC, VLPFC, OFC, PrCG and frontal pole), as well as additional parietal, temporal and occipital areas. Not only were group differences at offset more widespread than at onset, they reflected different activation/deactivation patterns: during offset patients demonstrated significant activations in many regions whereas CTL exhibited deactivations. Moreover, the magnitude of such hyper-responsiveness in FM patients was rather remarkable, particularly considering that intensity of stimuli presented to this group was 37.5% less than that for CTL (mean ± SD: 100 ± 43 mmHg in FM, 160 ± 74 mmHg in CTL)(25), due to individualized calibration of the stimulus to achieve a ~40/100 pain intensity rating.

Because FM patients demonstrated increased brain responses to both onset (i.e., the event signaling the beginning of pain) and offset (i.e., its termination), our results are compatible with the notion that FM patients might generally be more sensitive to salient events (although this interpretation awaits proper corroboration with behavioral data). Several prior studies, reported a generalized increased sensitivity to non-noxious, non-somatic sensations, including to auditory, gustatory and olfactory stimuli (21-24). For instance, when presented with intense auditory stimuli of varying intensities, FM patients’ demonstrated shorter auditory evoked potential (N1 and P2) latencies compared to CTL (24). Other studies using self-report questionnaires have found that FM patients tend to report greater sensitivity to everyday sounds and smells than their non-FM counterparts (22, 23). Altogether, the results from these studies, as well as our own demonstration of brain hyper-responsiveness to both cuff pain onset and offset, suggest that increased pain-related brain activity often reported in FM studies might perhaps reflect a more generalized hypersensitivity to salient aspects of the pain stimuli, rather than (or in addition to) their painful quality per se. In addition, the positive association between pain unpleasantness ratings and VLPFC activation during cuff onset in FM patients may hint at some sort of dysregulated modulatory processing with regard to perceived controllability over pain in FM patients, given a few studies have found greater VLPFC-related activity during self-controlled painful stimulation (43) and less perceived pain during uncontrollable compared to controllable pain conditions in healthy subjects (44). Clinically, these results would support the use of cognitive and behavioral interventions focused on training salience detection/stimulus driven orienting of attention, such as mindfulness meditation (59).

A notable finding gleaned from the current study is that FM patients showed frontal hyper-responsivity to both cuff pain onset and offset compared to CTL, particularly in the lateral prefrontal cortex (IPFC). The IPFC is comprised of multiple brain regions that together subserve the integration of cognitive inhibitory control functions and regulatory processes associated with threat detection (45, 46). There is growing evidence that augmented activity in IPFC, and the DLPFC in particular, may be associated with increased sensitivity to painful and non-painful somatic stimuli, perhaps indicative of impaired somatosensory gating in these patients (10, 47). While our data may be compatible with the hypothesis of dysregulation of salience detection, it should be noted that we did not observe group differences in canonical salience network nodes such as the dorsal ACC or anterior insula, rather both groups showed similar levels of activations in these regions. Regardless, our findings warrant further investigation to determine the mechanism underlying the PFC hyperactivation in FM and whether or not this is driven by an overactivity in salience-related circuits and/or an inability to sufficiently regulate/dampen these responses. Alternate explanations for greater prefrontal activation in response to stimulus onset/offset may correspond to a slower hemodynamic response recovery at offset (i.e., aberrant neurovascular coupling) and/or disruption in top-down control mechanisms including greater catastrophizing or altered stimulus appraisal.

Another key finding was the significant positive association between pain catastrophizing and BOLD signal response in the DLPFC during cuff pain offset in FM patients. It is well understood that catastrophizing is an important contributing factor to the experience and expression of pain and its chronification (48, 49). Indeed, patients that tend to catastrophize about their pain report overall greater pain severity, rating higher in pain intensity and unpleasantness, than those that do not (23). Moreover, the degree of catastrophizing has been shown to be predictive of whether or not an acute pain event actually develops into a chronic pain state (50). However, much less is known about the neurobiological mechanism that drives this phenomenon. Our recent studies have found that patients’ engaging in catastrophizing thoughts about their clinical pain activates medial prefrontal and posterior cingulate cortices (51), however, different circuits may support how catastrophizing influences perception of evoked pain in these patients. One theory posited is that catastrophizers are unable to disengage attention away from their pain, and direct more attentional resources toward non-painful or salient stimuli encountered in their environment. This attentional bias, over time, may sensitize the system overwhelming it to the point to which it can no longer compensate via descending inhibitory pathways. This, in turn, could lead to a host of pathological downstream effects, including but not limited to pain amplification (hyperalgesia), the development of allodynic responses to previously innocuous signals, and/or a generalized hypersensitivity manifest across multiple sensory modalities. Our previous study found that greater PCS scores were associated with greater connectivity between somatosensory (i.e. S1) and salience (i.e. anterior insula) processing regions during sustained evoked pain, which compared to a resting state elevated S1-insula connectivity (52, 53). A confluence of evidence also points to the DLPFC as a possible candidate region responsible for driving this interaction given its involvement in cognitive inhibitory control functions and attentional processes related to pain perception, as well as the detection and mediation of adaptive behavioral responses to aversive threats (54). Our data provide support for this theory by demonstrating a significant association between DLPFC activity and catastrophizing during cuff pain offset in FM patients, suggesting that the tendency to catastrophize may be linked to an inability to appropriately disengage attention away from the salient sensory event, in this case, the termination of the cuff pain stimulus – despite the fact that the stimulus has ended and is therefore no longer noxious or threatening. Other neuroimaging studies have reported findings that are consistent with our results. For example, Gracely and colleagues found that FM patients high in catastrophizing showed greater DLPFC activity during pain perception, a finding that persisted even when depressive symptoms were statistically controlled (55). Ellingson and colleagues reported a significant positive correlation between catastrophizing, pain ratings, and DLPFC signal responses in FM patients, but not performance, during a distracting cognitive attention task (i.e., Stroop task)(56). Specifically, FM patients’ ability to modulate their pain was impaired and varied depending upon the degree of catastrophizing reported, and the magnitude of this relationship was linked to DLPFC signal responding; those patients higher in catastrophizing showed greater DLPFC activity. One interpretation put forth by Ellingson et al. was that catastrophizing likely interferes with the pain modulatory system via descending pathways arising from the DLPFC, by weakening engagement of attentional resources to inhibit incoming nociceptive signals. Our result, in context of their findings, provides further support for this theory although more research is needed.

Several limitations should be considered when interpreting the present findings. Firstly, given our study utilized a cross-sectional design, drawing predictive conclusions with regard to the relationship between brain response to cuff pain offset and the degree of pain catastrophizing is not possible. Future studies employing a longitudinal design to investigate the causal nature of these relationships are needed. Secondly, medication usage was not controlled. Some patients were undergoing antidepressant therapy or taking analgesics (gabapentin, NSAIDs and/or acetaminophen). As such, it is unclear to what extent medication might have affected our results. Moreover, it remains to be determined whether the patterns of hyperactivation observed in FM patients can also be observed in other groups, including pain-free participants with similarly elevated levels of catastrophizing. Unfortunately, we were unable to evaluate the effects of PCS on the BOLD signal in our CTL participants given that the dynamic range in their PCS scores was too narrow. Lastly, due to the secondary nature of this study, we did not collect behavioral data directly measuring salience, and our interpretation about differences in salience detection is only speculative at this point, and will need to be confirmed in future investigations.