Arthritic pain is processed in brain areas concerned with emotions and fear




Functional neuroimaging studies have shown that experimentally induced acute pain is processed within at least 2 parallel networks of brain structures collectively known as the pain matrix. The relevance of this finding to clinical pain is not known, because no direct comparisons of experimental and clinical pain have been performed in the same group of patients. The aim of this study was to compare directly the brain areas involved in processing arthritic pain and experimental pain in a group of patients with osteoarthritis (OA).


Twelve patients with knee OA underwent positron emission tomography of the brain, using 18F-fluorodeoxyglucose (FDG). Scanning was performed during 3 different pain states: arthritic knee pain, experimental knee pain, and pain-free. Significant differences in the neuronal uptake of FDG between different pain states were investigated using statistical parametric mapping software.


Both pain conditions activated the pain matrix, but arthritic pain was associated with increased activity in the cingulate cortex, the thalamus, and the amygdala; these areas are involved in the processing of fear, emotions, and in aversive conditioning.


Our results suggest that studies of experimental pain provide a relevant but quantitatively incomplete picture of brain activity during arthritic pain. The search for new analgesics for arthritis that act on the brain should focus on drugs that modify this circuitry.

It is now well established that pain is processed within a network of brain structures collectively known as the pain matrix (1–4). Within this matrix, 2 parallel systems have been identified: the medial pain system and the lateral pain system, each of which subserves different functions. The medial pain system comprises structures including the medial thalamus and the perigenual, midcingulate, and insular cortices (5) and is involved in processing the affective (emotional) aspects of pain. The lateral pain system is involved in processing the sensory-discriminative (intensity, location, and duration) aspects of pain and comprises the lateral thalamus and its projections to the primary and secondary somatosensory cortices in addition to thalamoinsular projections from the ventromedial posterior nucleus of the thalamus (6). This division of function between the medial and lateral pain systems was first suggested by Bowsher in 1957 (1) and has now been demonstrated directly and noninvasively in humans, using functional brain imaging techniques (2, 3, 7, 8).

We recently showed that when healthy volunteers received painful experimental heat stimuli, selectively attending to the affective aspects of the pain resulted in increased regional cerebral blood flow in areas of the medial pain system (the perigenual anterior cingulate cortex, prefrontal cortex, orbitofrontal cortex, and amygdala). However, selectively attending to the location of the pain was associated with increased regional cerebral blood flow in the contralateral primary somatosensory cortex and the inferior parietal cortex of the lateral pain system.

It remains unclear whether these brain responses to experimental pain are relevant to the processing of ongoing clinical pain. Several groups of investigators have compared brain responses to acute experimental pain in healthy volunteers and patients with chronic pain (9–12) and have shown some differences, predominantly within areas of the medial pain system. For example, in patients with atypical facial pain, a type of chronic regional pain syndrome, anterior cingulate responses to painful stimulation were exaggerated compared with these responses in healthy control subjects (11). In contrast, patients with rheumatoid arthritis showed diminished responses to pain compared with those demonstrated by age- and sex-matched healthy control subjects, particularly in the anterior cingulate and prefrontal cortices (12). These differences between patients and healthy control subjects in the response of the medial pain system to experimental pain are likely to be associated with greater emotional salience of pain in patients with clinical pain and possible differential recruitment of the limbic and medial pain circuitry in different clinical pain conditions. However, no studies to date have compared brain responses to experimental and clinical pain in the same group of patients, and therefore, the relevance of these studies to ongoing clinical pain is unclear.

We compared the brain activity of 12 patients with osteoarthritis (OA) between 3 different pain states: arthritic knee pain, experimental knee pain (experimental pain applied on an occasion when patients were not experiencing arthritic pain), and pain-free. Arthritic pain, by its nature, is usually acute and recurrent, in contrast to many chronic pain conditions (e.g., neuropathic pain) in which pain is more often continuous. Arthritic pain therefore provides an ideal model for comparing experimental and common clinical pain. We used positron emission tomography (PET) to measure 18F-fluorodeoxyglucose (FDG) uptake in the brain as a correlate of neuronal activity (13, 14). Our hypothesis was that both arthritic pain and acute experimental pain would be processed within the same areas of the pain matrix, but that the arthritic pain state would be associated with increased activation within the medial pain system.


The study protocol was approved by the Salford and Trafford Research Ethics Committee, and permission to administer radioactive substances was obtained from the Administration of Radioactive Substances Advisory Committee of the Department of Health, UK. All patients gave informed written consent.


Twelve patients with knee OA (6 women and 6 men; mean age 59 years [range 52–67 years]) participated in this study. All patients fulfilled the American College of Rheumatology criteria for the classification of OA (15). Six patients had predominant left knee arthritis, and the other 6 had predominant right knee arthritis. All patients were right-handed and had no psychiatric condition and no other medical condition. None of the patients had received opiates or antidepressants for at least 1 year prior to the study. The arthritic pain was treated mainly by nonpharmacologic measures and with simple analgesics such as acetaminophen. All analgesics were discontinued at least 12 hours prior to PET scanning and were restarted after the scanning session. All patients were instructed to continue their routine activities prior to undergoing the arthritic pain condition and were asked not to engage in unaccustomed physical activities prior to undergoing the pain-free and experimental pain conditions. The order of the 3 conditions was randomized between patients.

PET protocol.

We used the radioisotope FDG, which, as a glucose analog, provides an image of the rate of regional glucose metabolism (13, 14). The metabolic product of FDG, FDG-6-phosphate, remains trapped within neuronal tissue for a longer duration (half-life 6.5 hours) compared with radioactive water. This unique metabolic behavior makes FDG an excellent marker for mapping regional function in the brain in response to various stimuli.

All patients underwent PET scanning under 3 conditions, in randomized order, at least 24 hours apart: arthritic pain, pain-free, and experimental pain. The arthritic pain condition assessed ongoing knee pain experienced during patients' accustomed physical activities (patients reported this pain as being moderately strong). The pain-free and experimental pain conditions were assessed during a period when patients were free from arthritic pain. In the absence of arthritic pain, acute experimental heat pain stimuli were delivered to the skin over the arthritic knee, using a Marstock thermal threshold stimulator (Thermotest type 1; Somedic, Stockholm, Sweden) (16). This device delivers reproducible intermittent ramps of increasing heat via a water-cooled probe (2.5 cm × 5 cm).

In a training session prior to PET scanning, patients were given a standard explanation of the study procedure. For each patient, the temperature that was reproducibly experienced as moderately painful was carefully established using a 0–100-point scale (mean temperature 46.3°C). This temperature was used during scanning to induce acute experimental pain. The standard instruction given to patients was to rate the perceived intensity and unpleasantness of pain using a 0–100-point scale. All patients completed the Hospital Anxiety Depression Scale (17), the Center for Epidemiologic Studies Depression Scale (18), the Spielberger State Anxiety Inventory (19), and the Pain Catastrophizing Scale (20) during the training session and before each scanning session. These questionnaires were used to screen for current symptoms of mood and behavioral disorders, because altered emotional states might influence the processing of pain in the brain (Table 1).

Table 1. Demographic and clinical characteristics of the patients*
CharacteristicBaselineSession 1Session 2Session 3
  • *

    Except where indicated otherwise, values are the mean ± SD. HADS = Hospital Anxiety Depression Scale; PCS = Pain Catastrophizing Scale; CES-D = Center for Epidemiologic Studies Depression Scale; STAI = Spielberger State Anxiety Inventory.

Age, mean (range) years59.16 (52–67)
No. men/no. women6/6
Arthritis, no. left knee/no. right knee6/6
HADS depression score3.36 ± 3.833.25 ± 3.793.66 ± 4.613.16 ± 3.29
HADS anxiety score4.86 ± 3.694.83 ± 3.615.25 ± 3.724.50 ± 4.03
PCS12.78 ± 13.714.09 ± 15.6511.63 ± 12.8012.63 ± 13.77
CES-D score12.66 ± 2.1613.09 ± 2.4212.66 ± 2.1812.09 ± 2.11
STAI score47.66 ± 3.6647.5 ± 2.1148.00 ± 4.2847.54 ± 4.45

All patients were asked to fast for 4 hours prior to scanning, to ensure a relatively stable basal blood glucose level. To ensure that the patients were experiencing the condition of interest, pain ratings were obtained prior to the injection of FDG and every 10 minutes during the FDG uptake period (activation condition), which lasted for 30 minutes following the injection of FDG. The patients lay down on the scanning table, outside the scanner, in a quiet, dimly lit room. Rating scales were suspended in the patient's field of view to allow verbal reporting of perceived pain intensity and unpleasantness. The thermode was in place on the skin 5 cm above the arthritic knee during all conditions, but heat stimuli were delivered only during the experimental pain condition. Following the activation condition, each patient was positioned in the scanner with his or her head secured firmly, using a molded headrest and a head-restraining Velcro band to reduce motion artifact.

Data acquisition.

Imaging was performed using a GE Advance PET scanner (General Electric Medical Systems, Milwaukee, WI). An initial 30-second scout scanning was performed to ensure correct positioning of the patients within the scanner, before the commencement of the main scanning (40 minutes after injection of the FDG). Scanning consisted of a 20-minute 3-dimensional (3-D) emission scan preceded by a 3-minute 2-D transmission scan (to correct for tissue attenuation) and a 2-minute 2-D emission scan (to correct for emission contamination of the transmission scan). Thirty-five slices at 4.25-mm intervals were obtained to cover the whole brain. From the acquired data, images of FDG distribution were produced in 2 15-minute blocks during the 3-D emission scan. Images were reconstructed by fully 3-D filtered back projection, with reprojection into 128 × 128 × 35 image matrices (voxel size 1.95 × 1.95 × 4.25 mm3), using measured attenuation correction.

Statistical analysis.

The goal of this study was to compare pain-related responses in the brain during different pain conditions. Statistical parametric mapping software implemented in MatLab (MathWorks, Natick, MA; online at was used to compare alterations in regional cerebral glucose metabolism (rCMRGlu) between conditions. Individual data were preprocessed, followed by statistical analysis of the group data (see the following web site: To correct for head movement between scans, all images were realigned to the first one to create a mean PET image. Each realigned set of images was then normalized to enable data from an individual patient to be transformed onto a standardized stereotaxic anatomic space using an FDG template (21, 22). This standard space conforms to the space described by Talairach and Tournoux in their atlas of the human brain (23).

To increase the signal-to-noise ratio and to accommodate variability in functional anatomy, each image was smoothed with a Gaussian filter with a width of 16 mm full width at half maximum. A correction was made for global changes in rCMRGlu between scans. Differences between one condition and another were assessed in the appropriate contrast, using t statistics. This analysis was performed for each pixel, and the resulting set of t values constituted a statistical parametric map (t). A further analysis was performed with subjective unpleasantness ratings as covariates, and differences between arthritic pain and experimental pain were assessed. Only results that were significant (P < 0.05) after correction for multiple comparisons across the brain were reported.

A second-step analysis was performed in 2 subgroups (6 patients with right knee arthritis and 6 patients with left knee arthritis), to identify any laterality effects. Direct statistical comparisons were made between the 2 subgroups to demonstrate brain activity in the arthritic pain condition versus the pain-free condition and the experimental pain condition versus the pain-free condition.


Questionnaires for current symptoms of depression, anxiety, or catastrophizing pain (see Table 1) were administered before each scanning session, and the results did not vary significantly from those obtained at baseline (training session), suggesting no significant change in mood states during different conditions.

Behavioral data.

All patients rated their perceived pain intensity and unpleasantness on separate 0–100-point scales at 10-minute intervals, from 10 minutes before the FDG injection to the end of the scan. In the arthritic pain condition, the mean ± SD intensity rating was 62.9 ± 11.4, and the mean ± SD unpleasantness rating was 57.7 ± 13.8. In the experimental pain condition, the mean ± SD intensity rating was 62.4 ± 15.3, and the mean ± SD unpleasantness rating was 49.2 ± 12.4. In the pain-free condition, the mean ± SD intensity rating was 4.0 ± 10, and the mean ± SD unpleasantness rating was 3.6 ± 6.4. These differences in the mean intensity and unpleasantness ratings between the arthritic and experimental pain conditions were not statistically significant (P = 0.9 and P = 0.08 by paired t-test, respectively). Both the mean intensity and the mean unpleasantness ratings during the arthritic and experimental pain conditions were significantly higher than those during the pain-free condition (P = 0.00 [to 2 decimal places] by paired t-test for both comparisons).

Alterations in rCMRGlu.

The reported results represent the pooled data for all 12 patients. However, 6 patients had predominant left knee arthritis, and the other 6 had predominant right knee arthritis. In the experimental pain condition, when patients were free from arthritic pain, the experimental pain stimulus was applied to the skin of the arthritic knee. Therefore, no inferences can be made from the pooled results for either the experimental pain or arthritic pain condition regarding the laterality of any effects seen. However, this issue was addressed in further analyses comparing the 2 subgroups of patients (see Patients and Methods). Those analyses revealed no consistently lateralized activations within any areas of the pain matrix.

The pooled data for all 12 patients were subjected to 4 statistical comparisons to investigate significant differences between the arthritic pain and experimental pain conditions, as described below and shown in Tables 2 and 3 and Figures 1 and 2. The threshold for activations was uncorrected P < 0.01 (extent threshold = 25 voxels). Only significant activations (corrected P < 0.05) within areas of the pain matrix (based on a priori hypotheses) are reported.

Table 2. Brain areas that were significantly more active during arthritic pain compared with the pain-free condition*
Brain regionMNI coordinatesBAZ scoreKEP
  • *

    For all brain areas reported, corrected P < 0.05. MNI = Montreal Neurological Institute; BA = Brodmann's area; KE = cluster size.

Right posterior cingulate cortex10−3636314.216260.01
Left posterior cingulate cortex−10−3636314.216260.01
Right anterior midcingulate cortex10242024/324.012560.02
Left anterior midcingulate cortex−10242024/324.012560.02
Left perigenual anterior cingulate cortex−8401232/243.922430.03
Right orbitofrontal cortex2740−17114.252350.01
Left orbitofrontal cortex−2240−17114.252350.01
Right prefrontal cortex184020104.426480.01
Left prefrontal cortex−164220104.426480.01
Right primary motor cortex8−2764, 63.943950.03
Left primary motor cortex−8−2764, 63.943950.03
Right primary somatosensory cortex55−22153, 1, 23.943250.01
Left primary somatosensory cortex−51−22153, 1, 23.943250.01
Left amygdala−20−8−144.362260.01
Right insula/secondary somatosensory cortex45264.453250.02
Left insula/secondary somatosensory cortex−352364.323250.02
Right inferior parietal cortex54−3830405.026500.01
Left inferior parietal cortex−60−4528405.026500.01
Left thalamus−13−14−34.362350.01
Right supplementary motor area4−25063.262350.03
Table 3. Brain areas that were significantly more active during experimental pain compared with the pain-free condition*
Brain regionMNI coordinatesBAZ scoreKEP
  • *

    For all reported brain areas, corrected P < 0.05. MNI = Montreal Neurological Institute; BA = Brodmann's area; KE = cluster size.

Right perigenual cingulate cortex10448325.838500.01
Left perigenual cingulate cortex−830−8323.523930.03
Right orbitofrontal cortex455−16113.282280.04
Left orbitofrontal cortex−150−16113.333930.01
Right prefrontal cortex454012105.238500.01
Left primary motor cortex−600204, 65.328500.01
Right primary motor cortex34−26624, 65.233830.01
Right primary somatosensory cortex25−28603, 1, 24.261,0540.01
Left secondary somatosensory cortex−40−32563, 1, 25.231,0540.03
Right putamen251013.833430.01
Left putamen−221013.423630.01
Right insula/secondary somatosensory cortex46465.851,2500.01
Left insula/secondary somatosensory cortex−402083.651,1060.01
Right inferior parietal cortex50−2820405.838500.01
Left inferior parietal cortex−60−3020405.838500.01
Left thalamus−16−2523.842430.03
Left supplementary motor area−2415464.251540.00
Figure 1.

Brain areas of increased activation in the arthritic pain (AP) condition and the experimental pain (EP) condition, both relative to the pain-free (PF) condition. The images show statistical parametric maps (SPMs) of each data comparison, with color coding reflecting the Z scores. The SPMs are displayed on the sagittal (top row), coronal (middle row), and axial (bottom row) sections of a standard Montreal Neurological Institute (MNI) template. The MNI coordinates (in millimeters) are indicated for the sagittal sections on the x-axis, the coronal sections on the y-axis, and the axial sections on the z-axis. The key areas for each comparison are labeled. For descriptive purposes only, the threshold for SPM images was uncorrected P < 0.01. PCC = posterior cingulate cortex; aMCC = anterior midcingulate cortex; pACC = perigenual anterior cingulate cortex; SII = secondary somatosensory cortex.

Figure 2.

Brain areas of increased activation in the arthritic pain versus the experimental pain condition and in the experimental pain condition versus the arthritic pain condition. The key areas for each comparison are labeled. SGC = subgenual cingulate cortex (see Figure 1 for other definitions).

Arthritic pain versus pain-free condition.

During the arthritic pain condition (compared with the pain-free condition), rCMRGlu was enhanced in all areas of the pain matrix. Most areas were activated bilaterally, including the posterior cingulate cortex (Brodmann's area [BA] 31), anterior midcingulate cortex (BA 24/32), prefrontal cortex (BA 10), orbitofrontal cortex (BA 11), inferior parietal cortex, as well as the primary motor cortex, insula/secondary somatosensory cortex, and primary somatosensory cortex (Table 2 and Figure 1, left). Unilateral activations were seen in the left thalamus (extending to the midline), left perigenual cingulate cortex (BA 32/24), left amygdala, and right supplementary motor area (SMA). Activations in these areas were significant at P < 0.05 (corrected for multiple comparisons).

Experimental pain versus pain-free condition.

The experience of experimental heat pain, compared with the pain-free condition, was associated with increased brain activation in most areas of the pain matrix, including the bilateral perigenual cingulate cortex, insula/secondary somatosensory cortex, primary somatosensory cortex, primary motor cortex, orbitofrontal cortex, inferior parietal cortex, putamen and left thalamus, SMA, and right prefrontal cortex (P < 0.01) (Table 3 and Figure 1, right). Activations in the right-sided regions were generally more extensive than those in the left hemisphere.

Arthritic pain versus experimental pain.

When compared with experimental knee pain, arthritic pain in the same knee was associated with increased activation of several areas of the pain matrix and associated limbic and motor areas. Bilateral activations were seen in the perigenual cingulate (BA 32/24), anterior midcingulate (BA 24b), posterior midcingulate (BA 23b), subgenual cingulate (BA 25) and posterior cingulate (BA 31), the amygdala, and the orbitofrontal cortex (Table 4 and Figure 2, left). Left-sided activations were seen in the posterior insula/secondary somatosensory cortex, prefrontal cortex, inferior parietal cortex, and the thalamus, whereas the SMA, primary motor cortex, caudate nucleus, and putamen were activated only in the right hemisphere. All of these areas were significant at P < 0.01. The most statistically significant activations were in the anterior cingulate cortex, where the cluster extended from the anterior midcingulate to the posterior midcingulate.

Table 4. Brain areas that were significantly more active during arthritic pain compared with experimental pain*
Brain regionMNI coordinatesBAZ scoreKEP
  • *

    For all reported brain areas, corrected P < 0.01. MNI = Montreal Neurological Institute; BA = Brodmann's area; KE = cluster size.

Right posterior cingulate cortex6−3640315.255430.01
Left posterior cingulate cortex−10−3636315.025430.01
Right anterior midcingulate perigenual anterior cingulate cortex1401232/245.341,0540.01
Left anterior midcingulate perigenual anterior cingulate cortex−240532/245.451,0540.01
Right posterior midcingulate cortex2−238243.652450.01
Left posterior midcingulate cortex−4−238244.953430.03
Right subgenual cingulate cortex810−16255.022450.01
Left subgenual cingulate cortex−910−16255.022450.01
Right orbitofrontal cortex628−27115.256450.01
Left orbitofrontal cortex−816−25115.026450.01
Left prefrontal cortex−45−452105.458450.01
Right primary motor area32−14474, 65.255430.02
Right amygdala26−2−215.256450.01
Left amygdala−20−6−185.256450.01
Left insula/secondary somatosensory cortex−40−23124.026450.03
Left inferior parietal cortex−50−3240404.955430.01
Right putamen27−11−63.652450.03
Left thalamus−20−1313.652450.04
Right caudate nucleus20−12213.852450.04
Right supplementary motor area3205063.452450.04

The correlation analysis with the unpleasantness ratings during this comparison also demonstrated increased activity in the bilateral subgenual (BA 25) and perigenual cingulate (BA 32) cortices, bilateral posterior cingulate cortices (BA 31), amygdalae, and left-sided insula.

Experimental pain versus arthritic pain.

Experimental pain, compared with arthritic pain, was associated with bilateral increases in rCMRGlu in the posterior/inferior parietal cortices, primary motor cortex, and SMA and with right-sided increases in the insula/secondary somatosensory cortex (extending from the anterior insula to the posterior insula), primary somatosensory cortex, subgenual cingulate (BA 25), and orbitofrontal cortex. All of these areas were significant at P < 0.01 (Table 5 and Figure 2, right).

Table 5. Brain areas that were significantly more active during experimental pain compared with arthritic pain*
Brain regionMNI coordinatesBAZ scoreKEP
  • *

    For all reported brain areas, corrected P < 0.01. MNI = Montreal Neurological Institute; BA = Brodmann's area; KE = cluster size.

Right subgenual cingulate520−11253.353430.03
Right orbitofrontal cortex664−13114.285460.01
Right primary motor cortex50−7554, 63.655650.03
Left primary motor cortex−48−13554, 63.454840.03
Right primary somatosensory cortex50−23553, 1, 23.823430.02
Right insula/secondary somatosensory cortex465−64.858650.01
Right inferior parietal cortex50−4055403.655460.02
Left inferior parietal cortex−46−4540403.484850.02
Right supplementary motor area4145063.452430.04
Left supplementary motor area−505065.144850.00


The results of this study show that both the medial and lateral pain systems are activated during arthritic and experimental pain, but that the medial system is more active during arthritic pain. To our knowledge, this study is the first to document the cerebral substrates of the perception of arthritic pain. It is also the first study to directly compare experimental pain with ongoing clinical pain in the same group of patients at the same level of perceived pain intensity. The finding that both experimental pain and arthritic pain activate the medial and lateral pain systems suggests that there is no unique brain network for processing arthritic pain.

The present study confirms that we are justified in using experimental pain in humans as a tool for investigating some of the generalized mechanisms of pain perception. However, it also demonstrates the possibility of using FDG-PET to measure subtle changes in brain activity in patients experiencing different types of clinical pain, as has been achieved for mood disorders (24). Because arthritic pain is rarely continuous and generally is characterized by recurrent acute pain, it is particularly suited to this experimental approach.

The present study showed that arthritic pain, when compared with experimental pain, was associated with increased activity within the medial pain system of the brain, including most of the cingulate, insula/secondary somatosensory cortex, and thalamus. It has been previously shown that activity within the medial pain system and related structures is increased during selective attention to the unpleasantness of an experimental pain stimulus (8, 25). Therefore, the present findings are consistent with the idea that, for these patients, arthritic pain has more emotional salience than does experimental pain. This is also consistent with the trend toward higher scores for the unpleasantness of arthritic pain compared with the unpleasantness of experimental pain observed in this study, although the difference was not significant. Although both mood and pain are processed within the orbitofrontal, perigenual, and midcingulate cortices (25, 26), patients' mood states did not differ between scans, and therefore, the changes in brain activation observed are likely to be specific to changes in the pain state.

During arthritic pain, activation of the cingulate cortex (including the perigenual, midcingulate, and posterior cingulate cortices) was much more extensive. The greater perigenual and midcingulate activity is consistent with increased affective processing (5, 7, 25). The posterior cingulate cortex is traditionally thought to be involved in visuospatial processing (27), although there is increasing evidence for a role in processing affect and pain memory (28, 29). This area has been shown to have increased blood flow during ongoing neuropathic pain (compared with a pain-free condition) following regional nerve block in patients with neuropathic pain (30), supporting a possible role in the affective processing of pain. The adjacent retrosplenial cortex (BAs 29 and 30) has been shown to be activated in response to unpleasant stimuli such as fearful and sad words and images (31). Activity in this region might therefore be related to the emotional salience during arthritic pain. Increased activations in the cingulate cortex during arthritic pain were accompanied by greater activity in the amygdala, orbitofrontal cortex, and putamen. These areas are associated with aversive conditioning, reward, and fear (32–36), suggesting that processing the fear of further injury and disability has a possible role in arthritic pain. This circuitry is less commonly revealed during experimental pain (2, 3, 5, 37) and therefore may be important in the development of pain chronicity.

In the present study, arthritic pain (compared with experimental pain) was also associated with increased activation of the prefrontal cortex and the inferior posterior parietal cortex. Coactivation of these areas is consistent with the finding of dense interconnections between them and their projections to common cortical and subcortical areas (38). Evidence suggests a role of these regions in the supervision of attention (39). Therefore, their activation during the experience of arthritic pain may reflect the recruitment of appropriate coping strategies in these patients (40). This is also consistent with reduced responses to painful experimental stimuli within BA 10 of the dorsolateral prefrontal cortex in patients with chronic regional pain and depression, who tend to have poor coping strategies (11).

The greater activation of the primary motor cortex that was observed in the present study during arthritic pain compared with experimental pain, even in the absence of actual movement, may be attributable to an increased motivational response or the inhibition of a desire to move the painful knee during arthritic pain. Increased activation of the primary motor cortex was also seen in our previous study, when healthy volunteers selectively attended to the unpleasantness of acute experimental pain stimuli in the absence of actual movement (25). The close association between motivational and affective responses, particularly within the midcingulate cortex (5), is likely to explain the more extensive activations within this cortical area during arthritic pain compared with experimental pain.

The present study demonstrates the importance of the medial pain system during the experience of arthritic pain and suggests that it is a likely target for both pharmacologic and nonpharmacologic interventions. The endogenous opioid system provides one possible candidate for modulation (41, 42). High concentrations of opiate receptors have been identified in the medial pain system compared with the concentrations in the lateral pain system (43). Changes in the occupation of opioid receptors, reflecting their occupation by endogenous opioid peptides such as enkephalins, have also been demonstrated during chronic arthritic pain and neuropathic pain (44, 45). The identification of brain-specific inhibitors of enkephalin breakdown (46) raises the possibility of enhancing such endogenous responses without the gastrointestinal side effects associated with most of the currently available analgesics. Considering the recent concerns about the long-term safety of cyclooxygenase inhibitors (47), we hope that our current findings will stimulate partnerships between academia and the pharmaceutical industry to develop a new class of analgesics for arthritic pain that specifically target the medial pain system.

The main limitation of the present study is the small number of subjects. Previous functional imaging studies in patients and healthy volunteers have been performed in as few as 6 subjects and have reproducibly demonstrated the cortical areas involved in nociceptive processing. Larger studies need to be performed, particularly to look at further details of nociceptive processing with variables such as depression, anxiety, and coping strategies, which can affect chronic pain conditions.

One of the possible criticisms of the study is that we are comparing 2 different types of pain stimulations, such as heat pain and arthritic pain. Most of the evidence for cortical nociceptive processing comes from studies that have used experimentally induced heat pain (3, 7, 11, 12, 25, 48). Those studies reproducibly activated the structures of the pain matrix and demonstrated a division of function within this pain matrix. This suggests that heat pain is sensitive enough to activate both the medial and lateral pain systems. Thus far, however, there has not been a direct comparison between clinical pain and artificially induced pain in the same group of individuals. Our studies aimed to examine possible differences in cortical processing between artificial pain and clinical pain, and there is no other way to compare them directly.

In summary, the present study demonstrates that although both experimental pain and arthritic pain predominantly activate the same areas of the brain (confirming the relevance of experimental pain studies), arthritic pain is also associated with increased activity in areas of the brain that play a role in affect, aversive conditioning, and motivational responses. This suggests that researchers should be moving toward more naturalistic studies in patients, in order to fully understand the perception of different types of clinical pain.


Dr. Kulkarni had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Kulkarni, Bentley, Elliott, Julyan, Watson, Jones.

Acquisition of data. Kulkarni, Julyan, Boger, Watson, Boyle.

Analysis and interpretation of data. Kulkarni, Elliott, Julyan, El-Deredy, Jones.

Manuscript preparation. Kulkarni, Bentley, Elliott, Julyan, El-Deredy, Jones.

Statistical analysis. Kulkarni, Elliott.


We are grateful to all of the radiographers and technicians at the Manchester PET Centre, North Western Medical Physics, Christie Hospitals NHS Trust, Manchester.