Patients with fibromyalgia display less functional connectivity in the brain’s pain inhibitory network

A total of 42 female subjects (28 FM patients and 14 healthy controls) were analyzed in the present study. Patients were recruited as part of a pharmacological multicenter study (EudraCT # 2004-004249-16). All three sites – one in England, Sweden and Germany, respectively - also performed the experiment in healthy controls, consecutively recruited throughout the study. A total of 92 female FM patients aged 25 to 55 years, mean age 44 (SD = 8.2) were enrolled. Of the 92 patients 9 were excluded from fMRI analyses due to image artifacts or en passant findings of intracranial anomalies, leaving 83 patients for matching with healthy controls. A total of 14 healthy female controls were available, aged 24 to 48 years, mean age 34 (SD = 8.6). Due to the relatively large number of patients, each healthy control could be age-matched with two FM patients (n = 28, range 24–48 years, mean 38, SD = 6.8). In cases where there were more than two patients with the same age, the two patients with the most similar duration of FM symptoms were chosen. The study was performed in accordance with the Helsinki Declaration and was approved by the local Ethical Committee at each of the three participating sites. All patients and healthy controls gave written informed consent. Parts of the present dataset were used in 2 previous publications; see Jensen et al. 2010 [30] where all 92 available FM patients were analyzed, and Jensen et al. 2009 [22], where 11 FM patients and 14 controls from the present study were included.

All subjects were first screened and then scheduled for one behavioral session and one fMRI session on two adjacent days. On day one, subjects were familiarized with the equipment that was used to evoke experimental pain, calibrated for subjective pain ratings and asked to fill out self-report questionnaires. FMRI scanning was performed on day two.

All subjects were right-handed. The study protocol required that all medications that could influence pain perception, including psychopharmacological medications, were washed out. Patients had to withdraw from all central nervous system acting therapies, including antidepressants, anticonvulsants, mood stabilizers, opioids, and narcotic patches, and to discontinue treatment with transcutaneous electrical nerve stimulation, biofeedback, tender and trigger point injections, acupuncture, and anesthetics. Required periods off medication were dependent on the actual pharmacological characteristics to ensure complete washout. All analgesics were prohibited during the study, except for paracetamol, dipyrone and nonsteroidal anti-inflammatory agents (NSAIDs), which were used as rescue medications. The rescue medications had to be prescribed at the lowest available dose and for the shortest period of time necessary to manage the patient’s acute pain. Use of any analgesic or narcotic drug had to be discontinued 48 hours prior to the assessments of pain sensitivity. Zolpidem was allowed for treatment of insomnia. Exclusion criteria included the following: severe psychiatric illness (including severe melancholic depressive episode); serious suicide risk; history or behavior that would prohibit study compliance; history of substance, drug, or alcohol abuse; heavy cigarette smoking (> 25 cigarettes/day); presentation of an intracranial anomaly; significant cardiovascular, pulmonary, gastrointestinal, hepatic, or renal disease; history of autoimmune disease; current systemic infection; active cancer (except basal cell carcinoma) or current cancer therapy; unstable endocrine disease; severe sleep apnea; or pregnancy or breastfeeding. FM patients were mainly recruited from primary care and were diagnosed according to the 1990 American College of Rheumatology criteria [2]. All patients had a self-reported average pain intensity of at least 40 mm on a 100 mm visual analogue scale (VAS) over the previous week. For healthy controls, the exclusion criteria included presence and/or history of any clinical pain problem.

Calibration of pain and stimulations during fMRI were performed using a laptop-controlled tool that delivered pressure to the right thumbnail. This specially designed tool, which consists of a plastic piston that applies pressure via a 1 cm² hard rubber probe, has been proven effective in several previous publications [9, 31, 32]. Calibration was performed one day before fMRI scanning. Subjects were instructed to rate the intensity of the pain evoked by each stimulus by putting a mark on a 0–100 mm horizontal VAS ranging from “no pain” to “worst imaginable pain”. Each subject’s pain threshold and tolerance were determined in an ascending series of pressures, using 50 kPa increments (tolerance was determined as the first pressure where the patient rated > 60 mm VAS). These values were then used to create a randomized series of five different pressure intensities within the range of each subject’s threshold and tolerance. For example, if the pain threshold was represented by 200 kPa, and the maximum pain rating by 600 kPa, the randomized series would consist of pressures of 200 kPa, 300 kPa, 400 kPa, 500 kPa and 600 kPa. In total, 15 stimuli were delivered in a randomized order at 30 seconds intervals. The duration of each pressure stimulus was 2.5 seconds. Pain ratings from the randomized series were used to calculate each subject’s calibrated pressure at 50 mm VAS. This calibrated pressure was used during fMRI scanning on the following day.

Images were collected using three different 1.5 Tesla scanners: in London a General Electric HDx scanner was used, in Stockholm a General Electric Twinspeed Signa Horizon, and in Cologne a PHILIPS scanner was used. Multiple T2*-weighted single-shot gradient echo EPI sequences were used to acquire blood oxygen level dependent (BOLD) contrast images. The following parameters were used: repetition time: 3000 ms (35 slices acquired), echo time: 40 ms, flip angle: 90 degrees, field of view: 24 × 24 cm, 64 × 64 matrix, 4 mm slice thickness with 0.4 mm gap and sequential image acquisition order. In the scanner, cushions and headphones were used to reduce head movement and dampen scanner noise. Placing a blank screen in front of the patient’s field of view from inside the scanner minimized visual distraction during scans.

Two types of stimulations were used during the functional scans: individually calibrated painful pressure, representing each subject’s 50 mm VAS, and a non painful pressure perceived as light touch, representing 0 mm VAS. All stimulations were randomly jittered over the scanning time, preventing subjects from anticipating the onset time and event type. The time interval between consecutive events was randomized with a mean stimulus onset asynchronicity (SOA) of 15 seconds (range 10–20 seconds). The total duration of the scans was approximately 35 minutes. Before scanning, subjects were instructed to focus on the pressures on the thumb and not to use any distraction or coping techniques.

In addition to the functional scans, high-resolution T1-weighted structural images were acquired in coronal orientation for anatomical reference purposes and screening for cerebral anomalies. Parameters were: Spoiled Gradient Recalled 3D sequence, repetition time: 24 ms, echo time: 6 ms, flip angle 35 degrees with a voxel size of 0.9 × 1.5 × 0.9 mm³.

The scanning procedure was standardized between sites by the use of written manuscripts for the oral instructions as well as practical training for all investigators involved in the study. In order to ensure calibrated experimental procedures and scanner settings, several visits were done at the three sites by a central coordinator.

The day before fMRI-scanning, all patients used a 0–100 mm VAS to rate their average clinical pain during the previous week and their current pain. Additionally, patients completed the Fibromyalgia Impact Questionnaire (FIQ); a 20-question questionnaire that assesses the overall symptom severity in patients with FM [33]. A high score on the FIQ corresponds to high self-reported FM severity.

The differences between FM patients and healthy controls regarding age and pressure pain sensitivity were assessed using two independent samples t-tests (two-tailed). A polynomial regression was used to determine each individual’s representation of VAS 50 mm, built on the 15 ratings from the randomized series of thumb pressures. The SPSS statistics software, version 18, was used for statistical analyses of behavioral data.

The connectivity fMRI analyses were performed in line with the methods previously described by our group [26, 27, 34] and other groups [35, 36]. In brief, functional data was preprocessed to decrease image artifacts, between-slice timing differences, and differences in odd/even slice intensity. The data was then spatially smoothed, using a Gaussian kernel of 6 mm full-width at half-maximum and temporally filtered (0.009 Hz < f < 0.08 Hz). Several spurious or nonspecific sources of variance were removed by including the following variables in the overall statistical model: 1) six movement parameters computed by rigid body translation and rotation during preprocessing, 2) mean whole brain signal, 3) mean brain signal within the lateral ventricles, and 4) the mean signal within a deep white matter ROI. Inclusion of the first temporal derivatives of these regressors within the linear model accounted for the time-shifted versions of spurious variance. The functional connectivity analysis produced coefficients for each seed-to-voxel correlation, and Fisher’s r-to-z transformation was used to convert these correlation maps into z maps. Group effects were tested with random-effects analyses using Matlab 7 (Mathworks) and the Statistical Parametric Mapping 5 software (SPM5) (http://​www.​fil.​ion.​ucl.​ac.​uk/​spm/​). Two separate one-sample t-tests were performed for the rACC and thalamus, assessing the connectivity between the seeds and the rest of the brain in FM patients (n = 28) and healthy controls (n = 14), respectively. Differences in rACC and thalamus connectivity between FM patients and healthy controls were assessed by two separate two-sample t-tests. The two seeds used in this study were 1) the rACC, centered at x = 8, y = 46, z = 4 (Montreal Neurological Institute [MNI] coordinates) with a 2 mm radius, and 2) the thalamus, centered at x = −14, y = −34, z = 12 MNI with a 2 mm radius. The coordinates were based on the results from a previous publication where we compared the pain-evoked fMRI signal in FM patients and healthy controls [22].

According to our hypothesis and previous studies of the DPMS [37], the PAG/brain stem [38], amygdala [39], nucleus accumbens [40], hippocampus [41], and prefrontal cortex [42] were defined as regions of interest. For predefined anatomical regions of interest (ROI’s), the statistical threshold was set at voxel-wise p < .005, uncorrected, with a minimum cluster size of 20 contiguous voxels. For all other brain regions, a threshold of voxel-wise p < .05, corrected for multiple comparisons, was used.

Each healthy control subject (n = 14) was age-matched with two FM patients (n = 28). A two-sample t-test validated that there was no significant difference in age between the two groups (t = −1.59, p = 0.126). In accordance with our previous findings, the sensitivity to pressure was significantly higher in FM patients, represented by lower amounts of pressure required to evoke pain at 50 mm VAS (t = 4.14, p < .001) (see Figure 1).

In a previous study, we found that patients with FM failed to recruit the rACC during evoked pressure pain [22]. Using the same region as seed for the present connectivity fMRI analysis, we found that healthy controls displayed higher rACC connectivity than FM patients to the hippocampus, a cluster that extended into the PAG area of the brainstem. It is well known that descending inhibitory pathways in the brain play a crucial role in pain modulation. The PAG receives direct projections from regions within the limbic forebrain such as ACC and the amygdala [37, 39, 44, 45] and can modulate pain perception through brainstem structures, such as the Rostral Ventromedial Medulla (RVM), that directly communicate with nociceptive neurons in the dorsal horn of the spinal cord [37, 45]. In an exploratory analysis of our data, using a lower threshold for continuous voxels, we found that the RVM had higher rACC connectivity in healthy controls, compared to FM. Given the crucial role of the rACC-PAG-RVM network in pain inhibition [38], our findings provide neuroimaging evidence for dysfunction of descending pain inhibition in FM. It is important to note that some brain structures, such as the ACC, may be divided in sub-regions that serve different aspects of pain processing [46], and therefore a region can sometimes be associated with increases or decreases in response to pain modulation [47], depending on the exact anatomical location and experimental paradigm. Recently published animal studies found augmented synaptic transmission in the ACC in response to long-term exposure to peripheral pain [48] and there is evidence suggesting that the rACC is required for the reward associated with pain relief [49]. Activation of the rostral ACC in humans has been demonstrated during inhibition of pain [50, 51], whereas activation of the more posterior parts of the ACC has been associated with increased pain affect [52‐54] in humans.

The role of the hippocampus in pain modulation is less explored but we speculate that the connectivity between the rACC and the hippocampus is likely related to the aversive drive and motivational dimension of pain [41, 55]. For example, increased activation of the hippocampus has been reported in response to anxiety-induced exacerbation of pain [41] and nocebo-induced hyperalgesia [56]. It is possible that the relatively lower connectivity between the rACC and the hippocampus in FM patients represents an attenuated defensive response to pain. Similar to reports of lower opioid binding potentials in FM [14], it is possible that the constantly ongoing nociceptive activity in patients with FM causes a ceiling effect for homeostatic pain control, expressed as relatively lower connectivity between pain defensive regions like the rACC and the hippocampus, in response to experimental pain.

In addition to the functional connectivity between the rACC, hippocampus and PAG, control subjects showed significantly higher connectivity between the rACC and bilateral amygdalae, a limbic structure with direct projections to the PAG [37]. Previous studies found that the amygdala is highly involved in opioid-dependent analgesia and has been described as a crucial subcortical node for endogenous pain inhibition [57]. In line with the proposed role of the hippocampus, the activation of amygdala in relation to pain is proposed to represent a defensive mechanism that contributes to the recruitment of descending inhibition [50]. Thus, the reduced connectivity between the rACC and amygdala in FM patients may indicate a weakened defensive response to pain stimuli, compared to healthy controls.

In one of our previous studies [30], including 92 FM patients, we investigated the impact of depression and anxiety on cerebral pain processing in FM, not including any healthy controls. We found no difference in pain ratings or cerebral processing in patients with high depression and anxiety, suggesting that experimentally evoked pain in FM patients is not confounded by comorbid affective symptoms. The results from the present study are derived from the same patient cohort but concerns comparisons between FM patients and healthy controls, using connectivity analyses, which is different from any of our previous studies. However, our previous results may suggest that the present study is not confounded by patients’ affective symptoms.

There is increasing evidence for altered thalamic function in pain patients with chronic pain such as neuropathic pain [58, 59] and FM [9, 23, 24]. Previous literature suggests that lowered thalamic function in FM patients represents a ceiling effect of descending pain inhibition [9] maintained by the persistent excitatory input of pain signals. Support for this mechanism was found in a study where normalization of reduced thalamic activity was seen in response to analgesic treatment (nerve blockade) in patients with peripheral neuropathic pain [58]. When comparing the functional connectivity between FM patients and controls, using the thalamus as a seed, we found significantly higher connectivity between the thalamus and the lateral OBFC in healthy controls. This finding supports the general idea that chronic pain alters the thalamocortical connections, causing a disruption of thalamic feedback and the view of chronic pain as thalamocortical dysrhythmia [28].

The OBFC has been proposed to be involved in sensory integration, reward processing, decision-making and expectation [60]. A meta-analysis of previously published imaging data revealed that the medial part of the OBFC is related to reward whereas the lateral areas are more related to evaluation of noxious events and motivation to act upon these [61]. In a recent pain study, the OBFC was associated with reduction in pain unpleasantness after meditation training [62], indicating OBFC involvement in the evaluation of incoming pain signals. Moreover the lateral OBFC is central for placebo analgesia (compared to opioid analgesia) [42] and for anticipation of pain relief [63], furthering the notion that the OBFC plays an essential role in evaluation and inhibition of pain. However, there are also studies suggesting that the OBFC inhibits pain through neural projections that do not require conscious evaluation [64], arguing against the cognitive role of OBFC in pain inhibition. This alternate theory suggests that the OBFC is part of a functionally connected network that inhibits pain through homeostatic regulation of nociceptive input and endogenous inhibition [65]. Nevertheless, we believe that the reduced functional connectivity between the thalamus and OBFC in FM patients reflects lower function of pain regulation, possibly acting through a combination of cognitive evaluative processes and homeostatic control. A previous neuroimaging study by Guedj et al. [66], found that FM patients had lowered resting state perfusion of the prefrontal cortex, supporting our findings of impaired connectivity between crucial nodes of the pain inhibitory network of FM patients.

Due to the uneven number of FM patients and healthy controls available for this study, we chose to match each healthy control (n = 14) with two FM patients (n = 28). We estimate that the benefit of increasing our power by matching two patients per control, instead of one, is larger than the risk of getting unfair variance estimations for the two groups.

In the present study, we used subjectively calibrated pressure stimuli during fMRI scans, leading to matched levels of experienced pain in FM patients and controls. As a result, the pressure needed to evoke pain was significantly different for FM patients and controls and could possibly affect some aspects of our results. It is possible, but unlikely, that certain parts of the central nervous system can detect information about absolute stimulus magnitudes and thereby confound the comparison between FM patients and controls.