The role of the central nervous system in the generation and maintenance of chronic pain in rheumatoid arthritis, osteoarthritis and fibromyalgia

Pain is a key component of most rheumatologic diseases. In fibromyalgia, the importance of central nervous system pain mechanisms (for example, loss of descending analgesic activity and central sensitization) is well documented. A few studies have also noted alterations in central pain processing in osteoarthritis, and some data, including the observation of widespread pain sensitivity, suggest that central pain-processing defects may alter the pain response in rheumatoid arthritis patients. When central pain is identified, different classes of analgesics (for example, serotonin-norepinephrine reuptake inhibitors, α2δ ligands) may be more effective than drugs that treat peripheral or nociceptive pain (for example, nonsteroidal anti-inflammatory drugs and opioids).

Although pain is commonly the patients' utmost priority and the reason most patients seek rheumatologic consultation, the medical community has historically had a poor understanding of the etiology, mechanisms and treatment of pain. Rheumatologists often consider pain a peripheral entity, but there is great discordance between pain severity and purported peripheral causes of pain, such as infl ammation and structural joint damage (for example, cartilage degradation, erosions).
In recognition of the importance of pain in the rheumatic diseases, the American College of Rheumatology Pain Management Task Force established an initia tive to increase awareness and call for organized research and education [1]. Th is initiative emphasizes the need for high-quality, quantitative research to understand the mechanisms underlying individual diff erences in pain among patients with rheumatic disease. Currently, most advances in the study of pain mechanisms have been in non-infl ammatory diseases, such as fi bromyalgia [2]. Th ese studies have highlighted the role of central painprocessing mechanisms, such as loss of descending analgesic activity and central pain augmentation or sensitization. Some pain researchers also believe that these mechanisms may have a signifi cant impact on pain severity among patients with osteoarthritis (OA) and rheuma toid arthritis (RA), diseases that have historically been associated with peripheral pain due to joint damage and infl ammation.
In the present review we give a brief overview of the basic biology of acute and chronic pain, including the role of central pain-processing defects. We discuss the role of these mechanisms in diseases commonly seen in rheumatology practices (for example, fi bromyalgia, OA and RA) and consider potential treatments that may correct defi cits in central pain processing.
Specialized receptors sense these stimuli and transport the signals to the central nervous system (CNS) via nerve fi bers that extend into the dorsal horn of the spinal cord. Th e specialized receptors include low-threshold receptors that respond to non-noxious levels of stimuli, and high-threshold receptors that sense noxious stimuli (nociceptors). Both nerve fi bers reside in soft tissue throughout the body, including the muscle, skin and internal organs.
Two types of nociceptors, the Aδ aff erent and the C aff erent, are responsible for the sensation and diff erentiation of mechanical, chemical and heat stimuli. Th e Aδ nerve fi ber has two classes, Type I and Type II, which respond to mechanical and heat stimuli. Type I fi bers have higher heat thresholds than Type II fi bers, while Type II fi bers have higher mechanical thresholds than Type I fi bers [3]. Consequently, the Type I Aδ aff erents usually transmit noxious mechanical stimuli while the Type II Aδ aff erents often transmit noxious heat stimuli. Th e C nerve fi bers detect mechanical and heat stimuli, as well as chemical stimuli. Compared with pain mediated by Aδ fi bers, pain mediated by unmyelinated C fi bers tends to be poorly localized [4].

Chronic pain
Chronic pain is associated with many rheumatologic conditions, varying from non-infl ammatory syndromes, such as fi bromyalgia, to systemic infl ammatory diseases, such as RA. Depending on the condition, as well as individual factors, diff ering pain mechanisms are involved. Mechanisms of chronic pain can be divided into peripheral mechanisms and central mechanisms.
Peripheral pain mechanisms stem from abnormalities in the peripheral nerves, leading to local areas of enhanced pain sensitivity. Th e most commonly cited peripheral pain mechanism besides direct nociceptive input is peripheral sensitization, which probably plays important roles in chronic pain mediated by OA and RA. Th is topic is covered in depth by Schaible and colleagues in an earlier manuscript in this Biology of Pain review series [5].
Central pain mechanisms operate at the level of the CNS, leading to enhanced widespread pain sensitivity. Individuals with augmented central pain processing will display diff use hyperalgesia (increased pain in response to normally painful stimuli) and allodynia (pain in response to normally nonpainful stimuli).
Abnormalities in central pain processing are divided into abnormalities in the descending facilitatory and inhibitory pain pathways, and central sensitization ( Figure 1). Th e descending pain pathways descend from the brainstem, hypothalamus and cortical structures, and modulate sensory input from primary aff erent fi bers and projection neurons in the dorsal horn of the spinal cord [6]. Th e best characterized descending analgesic pathways are the serotonergic-noradrenergic pathway and the opioidergic pathway. Th ese pathways lead to the release of serotonin, norepinephrine and endogenous opioids, which inhibit the release of excitatory neurotransmitters such as glutamate. Th ese pathways are activated in response to noxious stimuli, leading to a widespread decrease in pain sensitivity after exposure to an acutely painful stimulus. In chronic pain syndromes, descending analgesic activity is often impaired or absent -hence the term loss of descending analgesia.
In the present review, loss of descending analgesia is used synonymously with the term loss of diff use noxious inhibitory controls. Experimentally, diff use noxious inhibi tory control is commonly assessed by exposing subjects to two types of stimuli: the conditioning stimulus, and the test stimulus. Th e conditioning stimulus is an acute noxious stimulus that activates descending analgesic pathways, leading to a diff use decrease in pain sensitivity throughout the body [7]. In healthy controls, a wide range of noxious stimuli -including ice-cold water, contact heat and tourniquet ischemia -are all eff ective conditioning stimuli, produc ing increased pain thres holds throughout the body [7]. Th e test stimulus is a painful stimulus that is applied at baseline and during/after exposure to the conditioning stimulus. Th e magnitude of the descending analgesic response is the diff erence between the pain rating of the test stimulus before exposure to the conditioning stimu lus and the pain rating of the test stimulus after exposure to the conditioning stimulus [7].
When evaluating these studies, it is important to under stand that, although commonly used to assess descen ding analgesia, these studies do not specifi cally localize the areas of pain modulation to the descending spinal tracts. Changes in pain threshold after noxious pain stimulation may also partly refl ect changes in attention (for example, distraction) or other processes that infl uence pain perception. To directly assess the descen ding spinal pathways, electrophysiologic assessments of the spinal nociceptive fl exion refl ex must be performed.
While descending analgesic pathways are typically toni cally active and inhibit the upward transmission of pain signals, other descending pain-processing mechanisms involve enhanced activity down the descending facilitatory pain pathways that lead to generalized increases in sensory sensitivity [8]. Th e role of these facilitatory pathways, however, has not been well established in human studies.
In addition to descending inhibitory and facilitatory pathways, central sensitization also leads to enhanced CNS neuron excitability and increased transmission of pain signals. In the literature, the term central sensi ti zation may be used in two ways: to describe general abnormalities in central pain processing (which, in the present review, we will refer to as central augmentation); and to describe a specifi c defect in central pain process ing associated with activation of N-methyld-aspartate (NMDA) receptor channels (which we will refer to as central sensitization).
Central sensitization occurs largely as a result of enhanced release of glutamate and substance P at the level of the spinal cord. Glutamate is the major excitatory neurotransmitter in the nervous system, and it acts on three receptor subsets: the α-amino-3-hydroxy-5-methyl-4-isoxazeloproprionic acid receptor, the NMDA receptor and the G-protein-coupled metabotropic family of receptors. While the α-amino-3-hydroxy-5-methyl-4isoxazelo proprionic acid receptor is responsible for the baseline response to noxious stimuli, the NMDA receptor enhances and extends the pain response [9]. NMDA receptor activation results in calcium infl ux, stimulating calcium/calmodulin-dependent kinases and extracellular signal-regulated kinases. Th ese changes modulate CNS plasticity, resulting in the hyperalgesia and allodynia that characterize central sensitization [9].
Experimentally, central sensitization is characterized by diff use pain sensitivity and increased pain severity during and after repeated stimuli. Individuals with central sensitization have low thermal and mechanical thresholds in a diff use pattern, refl ecting enlargement of the spinal cord neuron receptive fi elds [4]. Repeated stimulation results in painful after-sensations that persist after a stimulus is withdrawn; and results in enhanced temporal summation of pain such that the pain rating for the last stimulus is higher than the pain rating for the fi rst stimulus, even though the stimuli are exactly the same. NMDA receptor antagonists, such as dextromethorphan and ketamine, inhibit temporal summation [10][11][12].
Studies sugge st that maintenance of central augmentation requires persistent noxious peripheral input, even in syndromes such as fi bromyalgia, which is characterized by the absence of well-defi ned, localized, pain-causing lesions [13,14]. A recent study of 68 fi bromyalgia patients with myofascial pain syndromes and 56 fi bromyalgia patients with regional joint pain showed that peripheral trigger point injections and hydroelectrophoresis ameliorate fi bromyalgia pain and increase pain thresholds at sites distant from the therapeutic interventions [15], providing further evidence that painful peripheral stimuli contribute to the perpetuation of central augmentation.

Fibromyalgia
Fibromyalgia is the prototypical non-infl ammatory chronic pain syndrome. Th e disease is characterized by chronic widespread pain and associated symptoms, including sleep problems, fatigue, cognitive dysfunction and depression. Quantitative sensory testing methods have consistently identifi ed abnormalities in pain percep tion among fi bromyalgia patients (Table 1). Most notably, patients with fi bromyalgia have diff usely lower pressure pain thresholds than healthy controls [16]. Th is diff use hyperalgesic state of central augmentation of pain processing has been repeatedly identifi ed using functional neuroimaging techniques [17,18] and may partly be due to specifi c defects such as loss of descending analgesic activity and central sensitization.
Evidence for the role of defects in descending analgesic activity in fi bromyalgia comes from studies of conditioned pain modulation [19][20][21]. In a study of 26 healthy controls and 25 fi bromyalgia patients, heat stimulation of the foot increased pain thresholds to electric stimulation of the forearm among healthy controls but not among fi bromyalgia patients [19]. Similarly, tourniquet ischemic pain increased pressure pain threshold in 10 healthy controls but not in 10 fi bromyalgia patients [20], and a noxious cold stimulus reduced heat pain ratings among 20 healthy controls but not among 45 fi bromyalgia patients [21].
Th ese defects in inhibitory pain responses may be due to blunted activity of the descending serotonergic-noradrenergic system. Fibromyalgia patients have reduced serum levels of serotonin and its precursor, l-tryptophan, as well as reduced levels of the principal serotonin metabolite, 5-hydroxyindoleacetic acid, in their cerebral spinal fl uid [22]. Levels of 3-methoxy-4-hydroxyphenethylene, the principal metabolite of norepinephrine, are also lower in the cerebral spinal fl uid of fi bromyalgia patients compared with healthy controls [22]. In contrast, biochemical and imaging fi ndings suggest that fi bromyalgia patients actually have increased activity of endogenous opioidergic systems, which is consistent with anecdotal experience that opioids are ineff ective analgesics in patients with fi bromyalgia and related conditions [23,24].
Th e evidence for central sensitization in fi bromyalgia predominantly consists of studies comparing the magnitude of temporal summation in fi bromyalgia patients with healthy controls. Although both fi bromyalgia patients and healthy controls experience temporal summa tion, the magnitude of temporal summation may be slightly greater in fi bromyalgia patients [25]. Th e magnitude of temporal summation is decreased by treatment with either fentanyl injections or ketamine, an NMDA antagonist [10,12].
In addition to heightened sensitivity to pain, fi bromyalgia patients are also more sensitive to a variety of other sensory stimuli [26,27]. Th is polysensory augmentation may partly be due to enhanced neural activity that has been consistently observed in brain regions such as the insula, a region known to code for the intensity of all sensory information [17]. Previous studies suggest that the anterior insula is involved in the aff ective/emotional modulation of pain processing, while the posterior insula is involved in the sensory/discriminative processing of pain [28]. Compared with controls, fi bromyalgia patients have higher levels of glutamate in the posterior insula, and changes in glutamate levels in the posterior insula are correlated with changes in pain and tenderness after acupuncture [29,30]. Th ese studies suggest that at least a component of pain in fi bromyalgia is a result of sensory amplifi cation, rather than just aff ective processing.
Genetic studies also support an association between the serotonergic-noradrenergic system and fi bromyalgia. In candidate gene studies, polymorphisms in the meta bolism and transport of monoamines (for example, catechol amine-o-methyltransferase, serotonin 5-hydroxy trypta mine type 2a receptor, serotonin transporter) have been associated with the diagnosis or severity of fi bromyalgia [31][32][33][34][35]. Most of these studies were small, however, and confl icting data exist -with some studies reporting no association between these genes and fi bromyalgia [31,[36][37][38]. Future studies, incorporating a larger number of fi bromyalgia patients and/or utilizing meta-analysis techniques, are needed.
In addition to genetic studies, a recent surge of interest has surrounded the use of functional magnetic resonance imaging (fMRI) to study pain in a more quantitative, objec tive manner. Th is area of research, however, is still relatively new. As such, we present the following results as preliminary evidence for the role of the CNS in pain modulation, rather than as well-established facts.
In one of the early studies of fMRI in fi bromyalgia, Gracely and colleagues reported that fi bromyalgia patients, compared with controls, exhibit enhanced activation in the contralateral primary somatosensory cortex (SI), inferior parietal lobe, insula, anterior cingulate cortex, posterior cingulate cortex, ipsilateral secondary somatosensory cortex (SII) cortex, bilateral superior temporal gyrus and cerebellum when exposed to experimental pain of the same magni tude (for example, same pressure) [17]. When exposed to experimental pain stimuli rated of similar intensity (moderate), however, fi bromyalgia patients exhibited activation in the same neural struc tures (contralateral SI, SII, contralateral superior temporal gyrus, inferior parietal lobe, contralateral putamen, ipsilateral cerebel lum and contralateral insula) as controls. Th ese observations provided the fi rst fMRI-based evidence for central augmentation of pain sensitivity in fi bromyalgia.
Cook and colleagues noted similar fi ndings in a study examining responses to heat stimuli [39]. In addition, their study reported post hoc analyses showing no neural activation in the periaqueductal gray region of fi bromyalgia patients exposed to painful heat stimuli but signifi cant activity in the periaqueductal gray region of healthy controls exposed to painful heat stimuli. Because previous studies have suggested that the periaqueductal gray region is involved in descending pain modulation, these fi ndings were interpreted as possible evidence for loss of descending analgesia among fi bromyalgia patients. A more recent article by Jensen and colleagues showed similar decreases in neuronal activation in the anterior cingulate cortex, a region also involved in pain modulation [40].
Th e fMRI techniques examining resting-state functional connectivity have also identifi ed the default mode network as a potential modulator of spontaneous clinical pain in fi bromyalgia patients. Th e default mode network consists of neural regions (medial frontal gyri, hippocampus, lateral temporal cortex, posterior cingulate cortex, precuneus, inferior parietal lobe) that are active at rest and may be involved in self-referential thought. In a study of 18 fi bromyalgia patients and 18 age-matched and sex-matched controls, Napadow and colleagues noted that connectivity between the default mode network and the insula was positively correlated with clinical pain severity [41].

Osteoarthritis
OA is a common degenerative joint disease, characterized by damage to cartilage and bone, which aff ects approximately 27 million people in the United States [42]. Individuals with OA often suff er from chronic pain, ultimately leading to signifi cant disability and healthcare costs. Despite the signifi cant impact of pain in OA patients, little is known about the causes of the pain associated with OA.
On a population level, pain intensity (via patient selfreport) correlates poorly with peripheral joint damage assessed by the Kellgren-Lawrence radiologic classification criteria [43]. Within individuals, however, pain severity is strongly associated with radiographic damage [44]. Taken together, these studies suggest that other mechanisms of pain that are not knee specifi c (for example, enhanced pain sensitivity due to alterations in central pain processing) may play a role in the variability in pain severity across individuals.
Studies utilizing quantitative sensory testing indicate that OA patients are more sensitive to experimental pain stimuli than healthy controls (Table 1). Most studies have focused on pain sensitivity at sites close to aff ected joints, showing that OA patients have lower mechanical and thermal pain thresholds (for example, higher pain sensi tivi ty) than healthy controls [45][46][47][48][49]. Intriguingly, O'Driscoll and Jayson also reported low-pressure pain thresholds at the forehead, a clinically nonpainful site, unaff ected by OA [50]. Similarly, among 15 patients with OA of the hip, Kosek and Ordeberg noted increased sensitivity to pressure, ischemia and innocuous warm stimuli at the aff ected hip and at the contralateral hip, indicating a diff use process extending beyond just the aff ected joint. Th ese studies suggest that OA pain, histori cally considered a peripheral entity, may also be modulated via wide spread mecha nisms controlled by the CNS.
Assessments of the wide spread nature of pain sensitivity in OA have provided further support for the role of central pain mechanisms in OA. Bajaj and colleagues infused hypertonic saline into the tibialis anterior muscles of 14 OA patients and of 14 agematched and sex-matched controls. OA patients reported increased pain intensity and larger pain areas, extending to the toes, whereas healthy controls reported lower pain intensity with the distribution of pain ending near the ankle. Th e authors attributed these fi ndings to changes in central pain mechanisms [51]. In a larger study of 62 female knee OA patients and 22 age-matched healthy controls, Imamura and colleagues highlighted the widespread distribution of pain sensitivity, showing subcutaneous hyperalgesia to pressure stimuli at seven dermatome levels, myotomal hyperalgesia at nine lower extremity muscle groups, and sclerotomal hyperalgesia at eight sites across the lower back and legs. Th e authors speculated that both peripheral and central mechanisms contribute to the chronic pain state, with peripheral mechanisms being more important in the early stages, and central mechanisms dominating in the later stages [52].
Additional evidence for defects in central pain processing comes from studies assessing specifi c painprocessing mechanisms, such as loss of descending analgesic activity. In a study of 48 knee OA patients and 24 age-matched and sex-matched controls, OA patients exhibited greater loss of descending analgesic activity than healthy controls [49] -a fi nding similar to the previous study by Kosek and Ordeberg of 15 hip OA patients [47]. Th e study by Kosek and Ordeberg was particularly interesting because it showed that loss of descending analgesic activity is contingent upon the chronic pain state and that loss of descending analgesic activity can be reversed [47]. After the initial evaluation, 13 out of 15 hip OA patients underwent surgery, resulting in signifi cant clinical pain relief. When the patients were reassessed 6 to 14 months after surgery (when pain-free), they exhibited signifi cant increases in pain thresholds com pared with pre surgery. Postsurgery pain thresholds were similar to pain thresholds among healthy controls. Furthermore, modulation of pain through descending analgesic pathways was restored. Th ese results suggest that dysfunctional central pain mechanisms are associated with the chronic pain state, and removal of the inciting pain stimulus may lead to normalization of central pain processing [47].
In addition to loss of descending analgesic activity, central sensitization may also alter pain processing among OA patients. In a study examining the eff ects of repeated pressure stimulation on pain sensitivity, temporal summation at the knee and tibialis anterior muscle was signifi cantly greater among patients with knee OA compared with controls [49].
Studies utilizing fMRI during quantitative sensory testing have also shown enhanced activity in the periaqueductal gray matter of OA patients compared with healthy controls [48]. Th is fi nding was interpreted as an increase in activity of the descending facilitatory pathways, a mechanism that would have the same net eff ect as a decrease in descending analgesic activity. Notably, this fi nding is the opposite of that found by Cook and colleagues in fi bromyalgia patients [39]. Cook and colleagues reported lower levels of activity in the periaqueductal gray matter of fi bromyalgia patients compared with pain-free controls, which the authors interpreted as an impairment in descending analgesic pathways. Other studies using fMRI have suggested that OA-related knee pain is modulated by the medial pain system, a network of brain structures associated with the aff ective dimension of pain processing [53].

Rheumatoid arthritis
In contrast to fi bromyalgia and OA, RA is characterized by systemic infl ammation. Although infl ammation contri butes to pain in RA, it may not be the only factor. For some patients, pain does not improve despite treatment with anti-infl ammatory disease-modifying anti-rheumatic drugs. In a cross-sectional analysis of 12,090 RA patients recruited from rheumatology practices, pain levels were almost constant over RA duration, even though most participants were treated with a disease-modifying anti-rheumatic drug, an anti-TNF agent or both [54]. A large longitudinal study, consisting of 882 RA patients, reported that pain initially decreased during the fi rst 3 years after diagnosis but subsequently increased over time. Th e authors speculated that the initial decrease in pain was due to control of infl ammation while the later rise in pain was attributed to other pain pathways [55].
Although few studies have specifi cally examined the role of central pain-processing mechanisms in RA, studies utilizing dolorimetry to assess pain thresholds suggest that these other pathways may include defi cits in central pain processing. Defi cits in central pain processing are characterized by enhanced pain sensitivity in a widespread distribution, and studies have consistently shown that RA patients have lower pressure pain thresholds (higher pain sensitivity) than healthy controls at joint and nonjoint sites [56][57][58].
Only one study has directly examined the role of descending analgesic activity in RA patients [59]. Th e study compared the magnitude of descending analgesic activity in 11 patients with RA of short duration to 11 healthy controls and in 10 patients with RA of long duration to 10 healthy controls. Th e magnitude of descending analgesic activity in both groups of RA patients was less than the magnitude of descending analgesic activity in healthy controls. Th ese diff erences were not statistically signifi cant [59], but given the small samples sizes it was diffi cult to determine whether there really was no diff erence between the two groups or whether the study was underpowered to detect an eff ect.
A few small studies have provided support for the role of central sensitization in pain augmentation among RA patients. Wendler and colleagues demonstrated using electroencephalography that, compared with age-matched and sex-matched controls, RA patients had enhanced cortical responses to repeated noxious stimu lation, suggesting changes in CNS modulation of pain [60]. Morris and colleagues showed that capsaicin induces a larger area of hyper algesia among RA patients compared with healthy controls [61]. Th is area of enhanced hyper algesia may correspond to the enlargement of spinal cord neuron receptive fi elds, characteristic of central sensitization.
In addition to central augmentation of pain through central sensitization and/or loss of descending analgesia, functional neuroimaging studies suggest that structures in the medial pain system may modulate pain processing in RA. Using positron emission tomography, Jones and Derbyshire observed that regional cerebral blood fl ow in the dorsolateral prefrontal cortex, anterior cingulate cortex and cingulofrontal transition cortex was lower in RA patients compared with healthy controls exposed to heat pain [62]. More recently, Schwienhardt and colleagues showed that fMRI signal intensity in the medial prefrontal cortex was signifi cantly associated with depression severity among 20 RA patients with provoked joint pain [63]. Th ese diff erences in cortical activity may refl ect enhanced cortical opioid peptide release in patients with RA [64].
Th e relationships between infl ammation, psychosocial factors and peripheral and central pain processing are intricately entwined. In a recent study of 59 female RA patients, we showed that C-reactive protein levels were inversely associated with pain thresholds at joint sites but not nonjoint sites, consistent with peripheral sensiti zation [65]. Sleep disturbance, on the other hand, was associated with pain thresholds at both joint and nonjoint sites, indicating a central mechanism linking pain sensitivity and sleep problems. Recent studies in healthy women [66] and in patients with temperomandibular joint disorder [67] support this hypothesis, showing that short sleep duration and forced awakenings are associated with loss of descending analgesic activity.

Mechanism-based treatment
Th e rheumatologist's approach to pain management has historically focused on treatment of the underlying disease process. With recent advances in the study of pain mechanisms, it has become clear that pain is multifactorial in origin, and successful treatment may require a combination of medications with diff erent mechanisms of action. Although most rheumatologists are familiar with the use of nonsteroidal anti-infl ammatory drugs for pain, few are experienced with newer classes of medications, such as antidepressants and anticonvulsants, that target central pain-processing mechanisms. Current treatments for central pain have mainly been used in the fi bromyalgia population, although a few studies have examined these agents in OA patients and RA patients. In the remainder of the present review we give an overview of the medications that are likely to play an increasing role in pain management among patients with rheumatic disease.

Tricyclic antidepressants
Tricyclic antidepressants (TCAs) work by inhibiting serotonin and norepinephrine reuptake. Th e most commonly used TCA is amitriptyline. Other TCAs include dothiepin and imipramine.
Ten randomized, double-blinded, placebo-controlled trials have examined the effi cacy of amitriptyline in fi bromyalgia [68]. A meta-analysis of these studies revealed poor to moderate evidence for the effi cacy of amitriptyline 25 mg daily over 6 to 8 weeks but no evidence for the effi cacy of amitriptyline at higher doses or longer treatment durations. Outcome measures included patient and physician global assessment of disease, the visual analog pain scale and the tender point count [68]. Although these studies were classifi ed as of high methodo logical quality by Jadad's score, other quality issues (for example, sample size, duration of follow-up and retention rates) were not considered and may limit the strength of these results.
Studies of TCAs in OA and RA have been limited. To our knowledge, no studies have specifi cally assessed the role of TCAs in the treatment of pain in OA -although one study examined the effi cacy of imipramine in the treatment of pain in a mixed population of 66 OA, RA and ankylosing spondylitis patients, showing signifi cant pain relief in patients treated with imipramine compared with placebo [69]. In RA, four out of six studies reported signifi cant improvements in pain among RA patients taking TCAs compared with RA patients on placebo [70][71][72][73]. Th e largest study, including 184 RA patients, showed a decrease in pain among patients treated with dothiepin, but the change in pain scores was not statistically diff erent from the change in pain scores among patients treated with placebo [74]. Studies examining the eff ects of TCAs on depression and pain showed that improve ments in pain were independent of improvements in depression [70,73].
In clinical practice, the use of TCAs is often problematic because TCAs are associated with substantial adverse eff ects, and compliance with these medications is low. In addition to inhibiting serotonin and nor epi nephrine reuptake, TCAs also block cholinergic, histaminic and α-adrenergic receptors. As a result, many patients taking TCAs experience side eff ects such as sedation, dizziness, blurred vision, constipation and dry mouth. Dry mouth is particularly problematic in the RA population because many patients also have secondary Sjogren's syndrome, an infl ammatory disorder characterized by decreased salivary gland function.

Serotonin norepinephrine reuptake inhibitors
Serotonin norepinephrine reuptake inhibitors (SNRIs) have similar noradrenergic/serotonergic reuptake ratios compared with TCAs. While TCAs have many eff ects other than inhibiting serotonin and norepinephrine reuptake, however, SNRIs are selective. A selective SNRI, such as duloxetine or milnacipran, could thus show greater overall benefi t by enhancing the serotonergic and noradrenergic eff ects that lead to drug effi cacy, while minimizing the dose-limiting eff ects of toxicity.
SNRIs modulate the descending serotonin-norepineph rine pathways involved in central pain-inhibiting mechanisms and are eff ective in the treatment of conditions characterized by defects in central pain processing (for example, fi bromyalgia). In a group of 40 healthy individuals with low descending analgesic activity at baseline, treatment with duloxetine 60 mg daily resulted in an increase in descending analgesic activity from 0.15 to 19.35 within 1 week [75].
Two SNRIs, duloxetine and milnacipran, are approved by the Food and Drug Administration for the treatment of fi bromyalgia. In three large, randomized placebocontrolled trials of fi bromyalgia patients, duloxetine was associated with signifi cant improvements in clinical pain [76][77][78]. Similar results have been reported in studies examining the eff ects of milnacipran on fi bromyalgia pain [79][80][81]. Th e pain-relieving eff ects of these agents have been observed in depressed patients and in nondepressed patients [79].
Recent studies have expanded the potential use of SNRIs to other chronic painful conditions, including OA. In a 13-week, randomized, double-blind, placebocontrolled trial of 231 patients with knee OA, duloxetine 60 to 120 mg daily signifi cantly reduced mean 24-hour pain scores [82]. Duloxetine was also associated with signifi cant improvements in the Western Ontario and McMasters physical function scores. To date, no studies have examined the eff ect of SNRIs on pain in RA.

The α 2 δ ligands
Th e α 2 δ ligands, pregabalin and gabapentin, are anticonvulsants used to treat chronic pain conditions such as postherpetic neuralgia and diabetic neuropathy. Pregabalin and gabapentin bind to the α 2 δ subunit of calcium channels, inhibiting the release of neuro transmitters, including glutamate, noradrenaline, serotonin, and substance P. Th ese compounds could thus work in individuals with central sensitization, as well as decreased descending analgesic response due to low serotonergicnoradrenergic activity.
Among fi bromyalgia patients, pregabalin has con sistently been associated with improvements in pain severity [83,84]. A Cochrane systematic review including 1,376 fi bromyalgia patients treated with pregabalin 300 to 450 mg daily reported a relative benefi t between 1.5 (95% confi dence interval 1.2 to 1.9) and 1.7 (95% confidence interval 1.4 to 2.1) for a 50% decrease in pain [85]. Th e authors concluded that although some patients will experience moderate pain relief from pregabalin, few will experience a large eff ect [85]. No studies have examined the eff ect of pregabalin on pain in OA or RA patients, although a recent animal study suggested that pregabalin decreased pain sensitivity in a rat model of OA [86].

Conclusions
Central pain mechanisms play important roles in widespread pain syndromes, including fi bromyalgia. Th e role of these mechanisms in rheumatologic diseases such as OA and RA is not well understood. A few small studies, utilizing quantitative sensory testing and fMRI, have documented loss of descending analgesic activity and alterations in CNS activity among OA patients, and a couple of small studies suggest a role for central sensitiza tion in RA (Table 1). Th e data regarding loss of descending analgesic activity in RA, however, remain inconclusive.
Larger studies involving extensive pain phenotyping and comprehensive information about disease charac teristics are necessary to better understand the impact of central pain mechanisms in OA and RA. Studies are also neces sary to determine whether these patients, or a subgroup of these patients, may benefi t from treatment with drugs such as SNRIs and α 2 δ ligands that target central pain mechanisms. If central pain mechanisms do play a signifi cant role in pain processing among OA and RA patients, these medi ca tions may be attractive adjunctive treatments to manage pain in patients with rheuma tologic disease.

Competing interests
YCL receives grant support from Forest Laboratories and holds stocks in Merck and Company, Inc., Novartis, and Elan Corporation. DJC has acted as a consultant for Pfi zer, Lilly, Forest Laboratories, Cypress Biosciences, Pierre Fabre, UCB, Johnson and Johnson, Nuvo, Merck and Company, Inc. and Wyeth. DJC has also received grant support from Pfi zer, Cypress Bioscience, and Forest. NJN declares that he has no competing interests.