Research Signpost 37/661 (2), Fort P.O., Trivandrum-695 023, Kerala, India

Nociceptive and Neuropathic Pain : mechanisms and treatments, 2009:177-204 ISBN:978-81-308-0368-5 Editors: Bai-Chuang Shyu and Chih-Cheng Chien

Brain imaging of muscle pain in humans

David M. Niddam1,2 and Jen-Chuen Hsieh1,2,3,4

1Brain Research Center, National Yang-Ming University, Taipei, Taiwan 2Laboratory of Integrated Brain Research, Department of Medical Research and Education, Taipei Veterans General Hospital, Taipei, Taiwan 3Faculty of Medicine, School of Medicine, National Yang-Ming University, Taipei, Taiwan 4Institute of Brain Science, School of Medicine, National Yang-Ming University, Taipei, Taiwan

Abstract Brain imaging has advanced our understanding of how acute and chronic pain is processed in humans. It is now well accepted that the pain experience is generated by distributed networks consisting of multiple cortical and subcortical areas. Where considerable data has been generated about the supraspinal organization of cutaneous pain little is known as to how nociceptive information from musculoskeletal tissue is processed in the brain. This is in spite of the fact that pain from musculoskeletal tissue is much more frequently encountered in clinical

Niddam et al. 2

practice. Differences are known to exist between acute pain from cutaneous and muscular tissue systems in both psychophysical responses as well as in physiological characteristics. In this chapter, characteristics of acute and chronic muscle pain will be presented together with a brief introduction to the methods of induction and psychophysical assessment of experimental muscle pain. Results from the brain imaging literature concerned with phasic and tonic muscle pain will be reviewed.

Correspondence/Reprint request: Dr. David M. Niddam, Brain Research Center, National Yang-Ming University, No 115, Sec. 2, Linong St., Taipei 112, Taiwan. E-mail: niddam@vghtpe.gov.tw

Introduction

Since the beginning of last decade, numerous studies have explored the brain mechanisms of both acute and chronic pain. Advances in non-invasive brain imaging combined with novel methods to induce pain have provided the substrate for this rapidly growing field of research. It is now well accepted that the pain experience is the final product of activity in distributed networks consisting of multiple cortical and subcortical areas. Brain regions contributing to processing of the different aspects (sensory/discriminative, affective/ motivational, cognitive/ attentive, motor) of the pain experience as well as to allodynia and hyperalgesia have been mapped in humans. However, due to the complex nature of the pain experience, a single cerebral representation of pain does not exist. Instead, pain depends on the context in which it is experienced and is generated through variable expression of the different aspects of pain in conjunction with modulatory influences.

As a conscious subjective experience pain can be influenced by a multitude of factors such as cognition, emotion, and mood. Such processes engage efficient endogenous pain modulating systems of the brain that can be of both inhibitory and facilitatory nature. For example, anticipation of a painful stimulus of uncertain magnitude may result in increased pain sensitivity (hyperalgesia) while anticipation of pain relief may result in reduced pain sensitivity (placebo analgesia). Under normal circumstances, the expression of pain is a delicate balance between these opposing influences and can be considered in advantage for the survival of the individual. However, when pain becomes chronic, maladaptive changes in the brain take place that may shift this balance in a fashion that is no longer advantageous to the individual. Accumulating evidence from brain imaging

Brain imaging of muscle pain 3

studies suggest that chronic pain may lead to brain atrophy, induces changes in brain chemistry and alters brain function. However, chronic pain may also result from or be maintained by these central changes. A better understanding of the endogenous mechanisms of pain modulation and the central effects of pain chronicity may provide novel and more effective therapeutic targets as well as novel diagnostic markers leading to more satisfactory treatment of pathological pain.

Where considerable data has been generated about the supraspinal organization of cutaneous pain little is known as to how nociceptive information from musculoskeletal tissue is processed in the brain. This is in spite of the fact that pain from musculoskeletal tissue is much more frequently encountered in clinical practice. Further on, treatment of pain from deep tissue posses a bigger challenge than superficial cutaneous pain. The existing treatments are often insufficient [1] and the impact by which the patients are affected is much more severe.

This chapter outlines the current knowledge about brain mechanisms of muscle pain in humans and how it differs from cutaneous pain processing. It is a review based on the brain imaging literature concerned with acute and chronic muscular pain from diverse imaging platforms such as electroencephalography (EEG), functional magnetic resonance imaging (fMRI), and positron emission tomography (PET). With the existing non-invasive imaging techniques, it is possible to study complementary aspects of brain function encompassing biochemical changes, hemodynamic changes, and electro- and magneto-physiological changes. Each method has its advantages and disadvantages in terms of temporal and spatial resolution and requires specific study designs for optimal utilization. The choice of pain induction method, imaging technique and study design is tightly interconnected. A short overview is provided of the physiology of muscular nociception and clinical muscle pain. The methods of induction and psychophysical assessment of experimental human muscle pain are then considered. These sections form the essential prerequisites for understanding brain imaging studies on human muscle pain. Characteristics of acute and chronic muscle pain Acute muscle pain

Acute muscle pain is characterized by activation of muscle nociceptors in the periphery under normal conditions and results in a sensation of pain that it closely related to the duration of the noxious stimulus. Differences are known to exist between acute pain from cutaneous and muscular tissue systems in both psychophysical responses as well as in physiological characteristics. This is for example clear when

Niddam et al. 4

comparing sensory manifestations. Pain from muscles is perceived as diffuse, nagging and cramp-like and is often also referred to other somatic structures distant from the site of origin. In contrast, cutaneous pain is perceived as sharp and burning and is well localized. Also, cutaneous pain is rarely referred to other sites [2]. The two tissue systems may also differ in pain sensitivity to the same stimuli. For example, intradermal administration of the algogenic substance capsaicin results in stronger pain intensity rating of local pain than for intramuscular administration of the same concentration [2]. Also, the two tissue systems differ in their response to analgesic substances. Acute muscle pain is inhibited more strongly by opioids than cutaneous pain [3]. These differences can partially be explained by different peripheral and spinal mechanisms such as fiber and sensory receptor distribution and spinal projection sites. For a thorough review of the mechanisms of acute muscle pain we refer to the work by Graven-Nielsen [4]. Chronic muscle pain

The transition from acute to chronic pain involves functional and morphological changes in the peripheral and central nervous system and is accompanied by release of sensitizing agents. Chronic muscle pain may develop from tissue damage provoked by e.g. trauma, ischemic contractions or inflammation. It is characterized by spontaneous ongoing pain, lowered pain threshold turning non-nociceptive input into pain (allodynia), and enhanced pain sensitivity (hyperalgesia). However, chronic muscle pain conditions are often also associated with evolving motor dysfunction such as muscle stiffness, muscle weakness, and restricted range of motion [5].

Chronic muscle pain conditions frequently met in the clinic include myofascial pain syndrome, tension type headache, low back pain, and temporomandibular disorder. These conditions are accompanied by structural changes in the muscle often leading to the formation of focal allodynic/ hyperalgesic contractures termed myofascial trigger points (MTPs)(Figure 1). MTPs are thought to arise from trauma, overload or strain resulting in local ischemia and increased metabolism [5]. However, the exact pathophysiologic mechanism remains controversial. Since acute and chronic stress often results in increased muscle tone, stress may not only serve as an initiating cause but also as a pain exacerbating factor [6]. Pain from MTPs is often also referred to distant regions albeit in characteristic patterns. Since MTPs can remain undetected pain referral can be a cause of confusion in diagnostication [5].

Brain imaging of muscle pain 5

Figure 1. Hyperalgesia in myofascial pain syndrome patients is demonstrated by (A) lower pressure pain threshold and (B) a leftward shift in stimulus-response function for intramuscular electrostimulation (IMES). Stimuli are applied to a myofascial trigger point in the upper trapezius muscle in patients (PTS, squares) and in an equivalent site in healthy controls (CTR, triangles). From Niddam et al. [7]. Peripheral and central mechanisms

Muscle nociceptors are unincapsulated “free” nerve endings formed by thin myelinated (group III / Aδ) and unmyelinated (group IV/ C) afferent fibers. A large proportion of these fibers are polymodal responding to chemical, mechanical and thermal stimulation [8]. Overall, the majority of sensory fibers from muscle nerves consist of unmyelinated afferent fibers [9]. Pulses from muscle nociceptors are transmitted past the sensory nerve bodies in the dorsal root ganglion into the spinal cord where they are integrated and modulated by interneurons before being transmitted to higher brain centers or onto reflex pathways. Some differences exist at spinal cord level between nociceptive input from superficial and deep tissue. Where projections from cutaneous nociceptors mainly terminate in lamina I and II in the spinal cord [10] projections from muscle nociceptors mainly terminate in lamina I and V [11-14]. Also, nociceptive input from superficial and deep tissue is known to influence spinal reflex pathways differently [15-17]. Activation of unmyelinated nociceptive afferents from deep tissue such as muscles and joints evoke longer lasting changes in excitability (prolonged facilitation) of the flexion reflex than cutaneous unmyelinated nociceptive afferents [17]. Since a strong sensory convergence from joint, muscle and cutaneous afferents exist in this pathway [15] deep tissue input can potentially influence other input long after they were applied.

Muscle nociceptors can be sensitized by muscle lesions resulting in hyperexcitability of the receptors and spontaneous resting activity

Niddam et al. 6

(peripheral sensitization). Prolonged input from muscle nociceptors leads to hyperexcitability of spinal neurons (central sensitization). In clinical muscle pain, peripheral and central sensitization acts together to induce spontaneous pain, allodynia, hyperalgesia, and referred pain. With further pain chronification complex changes occur that may include extra-segmental expansion of receptive fields, de novo synthesis of membrane channel proteins, sprouting of spinal terminals of afferent fibers, formation of new synaptic contacts, and altered balance in descending influences. Once developed central sensitization becomes independent of input from the lesioned muscle. Induction of experimental muscle pain

The choice of experimental pain induction method depends on several factors including the tissue system to be stimulated, the sensory and pain modalities to be activated, and the temporal characteristics of the pain. Several methods exist to experimentally induce pain. For muscle pain, these methods can be either endogenous (ischemia or exercise) or exogenous (electrical, mechanical or chemical) [4,18]. The methods can also be divided according to the time course of pain sensation they evoke, namely phasic and tonic methods. For phasic methods, each stimulus is relatively short and the whole series of stimuli is perceived as a pulsed sensation. Phasic muscle pain can be evoked by pressure to the muscle [7] and electrical stimulation applied within the muscle [19,20] or directly to the muscle afferents [21]. In contrast, for tonic methods, the stimulus is perceived as a continuous sensation over a longer period of time. Tonic muscle pain can be evoked by pressure to the muscle, by ischemia and exercise, via fast repetitive electrical stimuli (~20Hz) as well as by injection of chemical substances. These substances include hypertonic saline (5%), capsaicin (the active ingredient in chili peppers), bradykinin, glutamate, and nerve growth factor [2,22-25]. Intramuscular administration of nerve growth factor does not induce pain per se but sensitizes the tissue to mechanical stimuli. To date, only electrical stimulation, chemical substances and exercise have been applied in brain imaging studies of muscle pain.

Electrical stimulation is a non-specific method that activates the nerve fiber directly. At low stimulus intensities electrical stimulation activates the large nerve fibers and at increasing intensities nerve fibers of decreasing diameter are activated. Consequently, at the high intensities needed to evoke pain additional responses from a wide range of fibers, other than the smaller nociceptive fibers, will also be evoked. In addition, muscle twitches and possible limb movement accompany the stimulation.

Brain imaging of muscle pain 7

On one hand, electrical stimuli allow for precise control of many stimulus parameters. On the other hand, it can be argued that since electrical stimuli do not constitute an adequate stimulus for the tissue, other methods such as injection of algogenic substances are more closely related to pathological pain.

Only two chemical substances have been used in conjunction with brain imaging of muscle pain. These are hypertonic saline and capsaicin. When applied within the muscle, both hypertonic saline and capsaicin induce long lasting cramp-like diffuse pain which closely mimic chronic pain observed in patients. Apart from pain in the primary region, sensory manifestations such as referred pain, superficial mechanical hyperesthesia and autonomic reactions also occur [2,22,26]. Where capsaicin is known to activate the small unmyelinated nerve fibers via vanilloid receptors (TRPV1) the exact receptor transduction mechanism remains unknown for hypertonic saline. However both substances excite group III and IV afferent fibers and within the later group both low- and high-threshold mechanosensitive units are excited [4]. From a brain imaging point of view phasic and tonic stimuli requires different recording paradigms and to some degree also different method of analysis. Only phasic stimuli can yield information about short lasting evoked brain processing via its time-locked nature. Psychophysical assessment

In human pain research, psychophysics is considered an important tool for assessment of the multidimensional subjective pain experience. Within brain imaging, it provides an additional powerful tool for functional characterization of specific brain regions through correlational approaches. Psychophysical methods commonly applied in this field of research include visual analogue scales (VAS), discrete numerical rating scales, verbal descriptor scales, and pain questionnaires. Scales can be used to quantify sensory attributes of a stimulus or a state. However, some scales are limited by not having all the properties of the number system. A commonly used method to measure the subjective response from individuals is the VAS. It is a continuous scale presented as a line anchored by descriptors in each end. Typically, the anchors are “no pain” in one end and “the worst imaginable pain” in the other end. VAS has the advantage of providing a continuum of values. From a subjective point of view, pain is not perceived to change in discrete jumps but rather changes over a continuum of values. Another advantage of VAS is its highly desirable ratio-scale properties [27]. A ratio scale has a natural origin which represents no amount of the property and it has all three basic properties of the number system making

Niddam et al. 8

numerical manipulations easier. Reliable ratio scales have also been generated from pain descriptors [28]. Other scales and questionnaires might have achieved less powerful properties such as identity, ordinal or interval scales. One of the most commonly used questionnaires, the McGill Pain Questionnaire [29], specifies the subjective pain experience using sensory, affective and evaluative word descriptors. Three major measures are used: the present pain intensity based on a 1-5 intensity scale; the pain rating index, based on two values that can be assigned to each word descriptor; the number of words chosen. This type of questionnaire has ordinal and interval properties implying that descriptors can be ranked and that differences between scale values can be represented as differences between amounts of pain measured.

Of the multiple pain dimensions and aspects, pain intensity and pain unpleasantness have predominantly been assessed. Pain intensity is characterized by words such as mild, moderate and strong and is considered part of the sensory-discriminative dimension of pain, which also encompasses other characteristics such as quality, duration and location. Pain unpleasantness is considered part of the affective dimension of pain and is characterized by words such as annoying, distressing and uncomfortable. The two pain dimensions are closely related to each other and the scores on a numeric rating scale are typically highly correlated. However, unpleasantness is not unique to the pain experience. In fact, it has been shown that separate manipulation of the two pain dimensions is possible [30].

Classical psychophysical methods can be used to measure pain thresholds, pain tolerance and stimulus-response functions. Threshold values can be determined by the “Method of Constant Stimuli”, the “Method of Limits”, or the “Method of Adjustment”. The most frequently used technique for determining pain thresholds is the “Method of Limits” in which the stimulus intensity is either manipulated in ascending or descending series. In each series the stimulus intensity is increased or decreased in small amounts until the threshold sensation is reached. In increasing series, the threshold has been reached when the pain sensation occur for the first time in that series. In decreasing series, the threshold is reached when pain is no longer present. The threshold value is then calculated as the average of the transition values across all the series. In general, it is important to avoid expectancy/ predictability and habituation since this may influence pain scores. To avoid this, ascending and descending series should be randomized. The staircase method is a variation of the “Method of limits” that has been widely used. If, for example, an ascending series is first presented then when the threshold is reached the value is noted and the stimulus series is decreased until the

Brain imaging of muscle pain 9

stimulus is below threshold. This is done until a sufficient number of response transitions have been obtained. Transition scores are then averaged to yield the pain threshold. With this method the size of the steps should be chosen with care.

Several types of questionnaires are now commonly applied in the field of brain imaging of pain to evaluate personality traits, coping strategies, predispositions to enhanced pain sensitivity, and co-morbid disorders. Here we mention one specific psychological construct that is increasingly being used, namely pain catastrophizing. Pain catastrophizing is a coping strategy employed by individuals experiencing acute or chronic pain. It has been associated with increased pain sensitivity and can be evaluated by the Pain Catastrophizing Scale (PCS) [31-34]. It denotes a tendency of rumination, pain magnification, and feelings of helplessness in coping with pain. Originally, catastrophizing was a concept included in the Coping Strategies Questionnaire [34]. However, since chronic pain syndromes often are associated with mood disorders it is important to remove the effect of confounding factors such as depression and anxiety status. This was done by Sullivan and colleagues and resulted in the PCS [31,33]. Finally, a large variety of questionnaires exist that specifically evaluate the level of depression or anxiety. These should especially be applied in studies where patient populations are being studied. Brain hemodynamics Hemodynamic imaging techniques are based on the principle that regional cerebral blood flow (rCBF) is redistributed to regions with neuronal activation. Regional increase in rCBF is closely related to an increase in the concentration of oxygenated blood. By injection of radioactive oxygen-15 labeled water (H2

15O), PET can measure changes in rCBF directly. Functional MRI is typically based on the blood oxygenation level dependent (BOLD) technique. Both H2

15O-PET and BOLD fMRI have extensively been used to explore the cerebral processing of cutaneous pain. Phasic stimuli are by far the most commonly applied type of stimuli used to induce acute pain in imaging studies. The brain regions most consistently responding include thalamus, primary and secondary somatosensory areas (SI and SII), inferior parietal lobule (IPL) posterior, middle and anterior portions of insular and cingulate cortices, and medial and lateral prefrontal cortices (see Figure 2)[35,36]. Other less frequently observed brain regions include cerebellum, basal ganglia and supplementary motor area (SMA) [35,36]. Patterns of functional activity are often interpreted as contributing to the generation of a specific aspect of the pain experience. Activation of the lateral portion of thalamus, SI, SII, IPL,

Niddam et al. 10

and posterior insula are mainly associated with the sensory-discriminative component of pain which include encoding of stimulus properties (intensity, localization, duration, and spatial and temporal discrimination). The medial portion of thalamus, anterior insula and anterior cingulate cortex (ACC) are thought to subserve affective-motivational processing and prefrontal regions are thought to be involved in top-down modulatory processing including planning and cognitive mechanisms. Finally, cerebellum, basal ganglia, SMA, and cingulate motor area (CMA) possibly subserve motor-related behavior e.g. planning and execution of motor-defensive strategies. It is noteworthy to mention, activation in pain modulatory regions such as the periaqueductal grey in the brain stem only rarely is observed in “nociceptive specific” contrasts. Instead, activation in such areas often emerges in contrasts between experimental conditions manipulating e.g. cognitive factors such as attention and anticipation. Since specific brain regions may participate in the processing of several different aspects of the pain experience, the division into pain components may be regarded as too simplistic. Nonetheless, this approach does provide a useful framework to guide functional interpretations. Tonic muscle pain

Only a limited number of imaging studies have investigated pain originating from the muscle. Of these, the majority has been concerned with tonic pain induced by intramuscular injection of hypertonic saline. Other studies have used intramuscular electrostimulation (IMES) to induce tonic or phasic pain. The first brain imaging study investigating pain from muscles used rCBF PET to compare tonic muscle pain, induced by high-frequency (20Hz) IMES in the brachioradialis muscle of the forearm, with cutaneous pain, induced by phasic laser stimuli delivered to the skin above the same muscle [37]. Muscle pain (noxious – innocuous) resulted in increased rCBF in SII/ posterior insula, IPL, posterior mid-cingulate (ACC/ BA 24), and cerebellum. These areas completely overlapped with regions engaged by cutaneous pain (noxious – innocuous). Trends towards significance indicated increased rCBF in the cutaneous pain condition in thalamus, premotor cortex, SII/ posterior insula, and prefrontal cortex. Thus, the authors concluded that a similar set of brain regions mediate muscle and cutaneous pain.

More recently, Korotkov and Colleagues induced tonic pain in the triceps brachii muscle by infusion of hypertonic saline and found pain related increased rCBF in mid insula, putamen and superior temporal gyrus (STG) and decreased activity in IPL and middle frontal gyrus (BA9/10) [38]. Non-specific activity was observed in a wide range of regions. In another

Brain imaging of muscle pain 11

study from the same group employing similar methodology, tonic pain was induced in the erector spinae muscle and activity associated with different phases (pre pain, early pain, late pain, post pain) of the tonic pain profile was investigated [39]. In comparison with the other conditions, increased activity was consistently observed for the early acute pain phase in mid-posterior insula and STG. Increased pain related activity was also observed in putamen and inferior frontal gyrus when the early phase was compared to the later phases. Decreased IPL activity was only observed when the early phase was compared to the pre-pain baseline. Also, decreased anterior mid-cingulate (ACC/ BA 24 and 32) activity was observed in the late pain phase compared to initial pain or baseline. Although pain from hypertonic saline is salient and unpleasant, these results surprisingly did not involve persistent increases in the limbic regions typically found in pain processing. Nonetheless, the results emphasize the dynamic and complex nature of the pain experience involving engagement of specific circuitries at distinct phases. This is in agreement with the notion that ongoing pain is associated with changes in coping strategies, vigilance and attentiveness. Although Kupers and Colleagues [40] employed similar methodology as in the studies by Korotkov and Thunberg (hypertonic saline and rCBF PET) they found different results for tonic orofacial pain induced in the masseter muscle. In their muscle pain condition changes in activity was observed in a wide range of regions some of which differed from those responding to non-painful cutaneous mechanical stimuli. Increased pain specific activity was found in cerebellum, anterior insula, posterior mid-cingulate (caudal CMA/ BA 24), and perigenual cingulate (ACC/ BA 32). Decreased responses were found in amygdala and subgenual cingulate (ACC/ BA 25). Non-specific loci shared between pain and mechanical stimuli included posterior insula, IPL, dorsolateral prefrontal cortex, anterior mid-cingulate (ACC/ BA 32), and the brain stem. A non-specific decrease in activity was found in retrosplenial posterior cingulate which is possibly related to default-state activity (also known as resting-state activity). It is difficult to explain the differences between the above mentioned studies using hypertonic saline induced muscle pain. However, some methodological differences exist. First, variations in scanning onset relative to hypertonic saline injection onset may play a role. Where scanning was initiated 20 seconds after injection of hypertonic saline in Kupers et al., scanning was initiated 2 minutes or 4 minutes post-injection in the other two studies [38,39]. Second, in Kupers et al. scanning conditions were presented in a semi-randomized order while in the other two studies a fixed order was used. Third, pain was induced in different muscles. The later, however, is an unlikely explanation of the substantial

Niddam et al. 12

differences. Overall, the findings from Kupers et al. are more in line with known regions contributing to acute pain processing. Common to all three studies is the absence of sensory-discriminative activity in thalamus, SI and SII during tonic muscle pain. Although at first glance this may be speculated to be due to the tonic character of the stimulus as well as being specific to muscle pain another more likely explanation may be found in the PET methodology (see below). Interestingly, when mechanical allodynia (mechanical stimuli plus muscle pain) was compared to muscle pain (without mechanical stimuli) enhanced activity was found in multiple regions involved in sensory discrimination. These regions included thalamus, SI, SII, inferior parietal lobule, and superior parietal lobule [40]. In this respect, discrepancies in Svensson et al. [37] may be accounted for by the non-specific stimulation method used.

A series of recent studies employed a novel analysis approach to investigate BOLD fMRI responses to tonic pain stimuli [41-44]. Henderson and Colleagues compared responses to tonic pain induced by similar modalities, namely subcutaneous and intramuscular infusion of hypertonic saline [41]. They found activity in the same cortical areas including somatosensory, insular and cingulate cortices albeit with some regional differences in signal intensities and cluster locations. Muscle pain but not cutaneous pain evoked a signal intensity decrease in the perigenual cingulate region and signal intensity increases in the primary motor cortex, caudal CMA and anterior insular cortex. In SI, more widespread activity was observed in response to muscle pain. Furthermore, signal intensity increases were observed in anterior and posterior portions of mid-cingulate and insular cortex for both cutaneous and muscle pain. These results were largely confirmed in another study [42] but with additional increases in SII and cerebellum and decrease in lateral prefrontal cortex during both tonic muscle and cutaneous pain induced in the forearm as well as in the leg.

Somatotopic organization of the body is a necessity for accurate localization of peripheral somatosensory stimuli. Cortical somatotopy has previously been shown to exist for cutaneous pain in SI and posterior insula [45]. However, it has not been clear from the previous literature if somatotopic organization also exists for muscle pain. In a recent study [42], tonic muscle pain induced in the forearm and the leg resulted in similar somatotopy as for tonic cutaneous pain in SI and posterior insula. Additionally, in anterior insula a somatotopic representation not only existed for both muscle and cutaneous pain but the muscle pain representation was also located more rostral to that of cutaneous pain. Thus, anterior insula was suggested to subserve encoding of pain (unpleasantness) location as well as pain quality (muscle pain: cramping and dull; cutaneous pain: sharp, pricking and hot).

Brain imaging of muscle pain 13

It is well-known that muscle pain is perceived as less well-localized than cutaneous pain and is often referred to distant regions of the body. Since SI is known to encode stimulus location and intensity, the more widespread SI activity in response to muscle pain may be speculated to underlie the perceived spread of pain compared to the more focal cutaneous pain [41]. This issue was further elaborated on in the study by Macefield et al. [44] in which a subgroup of subjects exhibiting pain referral also exhibited a greater spread of SI activation. In fact, a positive correlation between peak intensity and pain area was observed in a specific SI cluster that only appeared in the referred pain group. Interestingly, this specific area corresponded to the somatotopic representation in SI of the area on the body to which pain was referred. Similar changes were also observed in anterior insula and cerebellum and an inverse relationship was observed in the perigenual cingulate. As mentioned above, somatotopic organization was previously also observed in anterior insula for muscle pain [42] and the region specifically engaged during muscle pain may also be related to pain referral.

In summary, tonic muscle pain engages a set of brain regions known to be involved in the processing of pain from other tissue. Several of these regions are also involved in the processing of non-painful stimuli. Finer spatial analysis revealed subtle differences between pain of muscular and cutaneous origin albeit within the same brain regions. Also, tonic muscle pain may more effectively engage motor related brain regions including MI and caudal CMA. The mid-posterior insula is consistently activated during tonic muscle pain. This is in agreement with posterior insula being among the most frequently observed regions in brain imaging of pain. Although it is a locus that is not exclusively involved in pain processing, lesions encompassing the posterior insula may result in altered pain perception [46] and direct stimulation elicit painful sensations [47]. The specific role of the posterior insula remains to be established. Functionally, the posterior insula rely on information related to pain intensity and stimulus location and may be involved in integrating aspects of pain and other somatosensory sensations of different tissue origin. Several other issues are worth mentioning. The activation of SI and SII in the above mentioned fMRI studies and the absence of these two regions in all the PET studies using hypertonic saline injection emphasizes differences between the two imaging modalities including a better spatial and temporal resolution of fMRI compared to PET. The relatively large smoothing kernel used in many PET studies makes it difficult to differentiate between insula, SII and STG. Another issue is the decreased activity in ACC observed in many of the studies. This was not located homogeneously across studies. Kupers et al. [40] found decreased activity in subgenual cingulate; Thunberg et al. [39]

Niddam et al. 14

in anterior mid-cingulate and Henderson et al. [41] in perigenual cingulate. It is unclear if these results represent the same mechanism. Finally, it is worth mentioning that none of the above mentioned studies found activity in thalamus during tonic pain.

Figure 2. BOLD fMRI responses to non-painful IMES (upper), painful IMES (middle) and painful – non-painful IMES (lower). The Z-coordinate denotes the axial section in superior-inferior position relative to the commissural line position (given in millimeters; positive = superior). Based on Niddam et al. [48]. Phasic muscle pain Due to the reasonable good temporal resolution of fMRI, it is possible to record the brain response to transient or phasic stimuli on a trial-by-trial basis. In these so-called event-related designs, randomly intermixed types of stimuli can be presented thereby avoiding a specific cognitive set associated with a specific stimulus e.g. a fixed attentional state. In a series of event-related fMRI studies, phasic muscle pain was induced by intramuscular electrostimulation (IMES) [7,48,49]. Phasic muscle pain resulted in a bilaterally distributed network covering the most commonly observed brain regions responding to pain (Figure 2). These regions included inferior, middle and medial frontal cortices, SI, SII, IPL, STG, anterior, middle and posterior portions of cingulate and insula, basal ganglia, thalamus, and cerebellum [7,48]. Furthermore, non-painful IMES clearly activated a subset of brain structures engaged by painful IMES [48]. Increases in stimulus intensity have previously been shown to result in more bilateral engagement of brain regions [50]. In two recent studies, the central effects of hyperalgesia [7] and therapeutic intervention [49] was investigated in patients with myofascial

Brain imaging of muscle pain 15

pain syndrome (MPS) using painful IMES in conjunction with event-related fMRI. MPS is characterized by local and regional pain emanating from hyperalgesic contractures in the muscle termed myofascial trigger points (MTPs). Pain evoked from an MTP was compared to pain from an equivalent site in controls induced by the same stimulus current magnitude as well as at the magnitude inducing the same subjective pain intensity (see figure X). Effects associated with hyperalgesia was then revealed by comparing patients and controls at matched current magnitudes while adaptive mechanisms was revealed by comparing patients and controls at matched subjective pain intensity. Hyperalgesia in MPS patients was associated with enhanced activity in regions involved in sensory-discriminative (SI, SII, mid-insula, and IPL) and affective (anterior insula) processing. A highly significant decrease in activity was observed in dorsal hippocampus. At matched subjective pain intensity, enhanced activity in the same sensory-discriminative regions remained but no anterior insular activity was found. The later is possibly due to similar levels of pain unpleasantness resulting from the matched pain intensity. Enhanced sensory-discriminative processing is a robust finding in hyperalgesia and allodynia of various etiologies and is not specific to muscular hyperalgesia. Although speculative, adaptive changes in these regions may indicate longer lasting central changes. However, peripheral and spinal influences cannot be ruled out from this study alone. The meaning of suppressed hippocampal activity in patients with MPS remains unknown. A similar result was found for allodynia in patients with mononeuropathy [51] and in experimentally induced allodynia by capsaicin [40]. Hippocampus is a limbic structure known to process nociceptive-related information. Electrical stimulation of hippocampus can evoke both pro- and anti-nociception [52,53]. This dual role in analgesia and hyperalgesia is in agreement with a key role in modulating nociceptive responses under acute and chronic stress [6,54]. During acute stress hippocampus inhibits brain regions involved in generating the stress response [55,56]. However, stress-related hormones may influence neurotransmission within hippocampus as well as promote stress-related atrophy [57] leading to impaired function. Moreover, ongoing pain may also serve as an emotional and physical stressor further promoting chronification [58]. In this context, patients with myofascial pain exhibit exaggerated stress responses in the hypothalamus-pituitary-adrenocortical system [59]. MPS is notoriously difficult to treat effectively on a long-term basis. Several therapeutical interventions provide short-term pain release. A relatively novel approach uses low-intensity (above motor threshold) low-frequency electrostimulation within the MTP. This method result in immediate reduction in pain and long-term effects after periodic

Niddam et al. 16

intervention include release of sarcomere shortening and the associated ischemia resulting in lasting pain relief. Since the antinociceptive effect of non-painful electroacupuncture and transcutaneous electrical stimulation at least in part is mediated by the endogenous opioid system it was hypothesized that the short-term effects of low-intensity low-frequency IMES within an MTP also engages the opioid system as represented by activation of the periaqueductal grey (PAG) matter in the brain stem [49]. The PAG is a well-known key region in the central pain modulating circuitry. As defined by the change in pain threshold relative to intervention patients were divided into responders and non-responders to intervention. Interestingly, responders engaged the PAG and affect regulating regions more effectively than non-responders during painful IMES to the MTP. Also, the PAG activity was positively correlated with increases in pain threshold across the whole patient group. Thus, the individual response to intervention may depend on the capacity to engage the central pain modulatory circuitry. Although several factors were taken into account to avoid expectation of pain relief, this study was not placebo controlled. In a recent study, placebo analgesia was shown to be mediated by the opioid system [60]. Conclusions must therefore be made with caution. Receptor-binding PET imaging

Receptor-binding PET imaging has been used to map the in vivo distribution of distinct neurotransmitter systems in both the healthy and dysfunctional human brain. With dynamic imaging paradigms, it is possible to examine the function of specific neurotransmitter systems in response to experimental pain challenges in conjunction with cognitive and emotional modulation. Recent studies examined the role of the opioid [60-63] and dopamine [64] systems during sustained experimental muscle pain. In these studies, hypertonic saline was continuously infused in the masseter muscle to maintain the subjective pain intensity within a specific range. A placebo condition consisting of infusion of isotonic saline was included for comparison. Placebo and algesic substances were administered in a blind or double-blind, randomized and counterbalanced fashion. µ-Opioid receptor imaging

In the first study to examine the involvement of the µ-opioid receptor system in response to tonic muscle pain several cortical and subcortical regions were found to be activated [61].These regions encompassed dorsal ACC (posterior mid-insula/ caudal CMA), lateral prefrontal cortex, anterior insula, thalamus, hypothalamus, and amygdala. Subjective sensory scores (subscale of the McGill Pain Questionnaire) were negatively correlated

Brain imaging of muscle pain 17

with activation in nucleus accumbens, thalamus and amygdala and subsignificantly with activation in PAG. µ-Opioid activation was also negatively correlated with affective scores in dorsal ACC, thalamus and nucleus accumbens. These results indicate that the individual capability to engage the opioid system may account for variability in sensory and affective components of the pain experience. Individuals that more successfully activated the opioid system also reported reduced sensory and affective ratings of the pain experience. Furthermore, amygdala and PAG are part of a circuitry known to mediate antinociception. Interestingly, both gender [62] and genotype [63] have recently been shown to influence the opioid neurotransmission to acute pain. Compared to females in their early follicular phase of the menstrual cycle, male subjects could more effectively activate the µ-opioid system in the above mentioned regions [62]. In another study, the COMT polymorphism (met/ met, met/ val, val/ val) was shown to influence the individual experience of pain through differential activation of µ-opioid neurotransmission [63]. Individuals with the val/ val genotype engaged the µ-opioid system more effectively and also reported lower pain ratings than individuals with the met/met genotype. Individuals with the met/ val genotype fell in-between the other two genotypes. In yet another study, it was further shown that cognitive factors such as expectation of pain relief (placebo analgesia) is also capable of modulating the pain experience through the µ-opioid receptor system [60]. µ-Opioid related brain regions involved in placebo analgesia included pregenual and subgenual ACC, dorsolateral prefrontal cortex, insular cortex, and nucleus accumbens. Dopamine receptor imaging

The dopaminergic neurotransmitter system is best known for its role in the mediation of reward reinforcement. However, substantial evidence also points to a role in the mediation of stress and tonic pain [65]. It has in fact been suggested that rewarding and aversive stimuli may rely on similar neural substrate. A study by Scott et al. [64] demonstrated the involvement of dopaminergic neurotransmission in response to tonic muscle pain. Pain specific activation was observed in the dorsal caudate nucleus and putamen and was associated with enhanced subjective sensory and affective ratings of the pain experience. Non-specific activation was observed in nucleus accumbens and was associated with enhanced negative affect. These findings support the involvement of the dopaminergic receptor system in responses to aversive painful stimuli and the contribution to individual expression of the pain experience.

Niddam et al. 18

Somatosensory evoked potentials Somatosensory evoked potentials (SEP) can be elicited by many stimulus modalities depending on the tissue system and afferent fibers to be investigated. However, only a limited number of methods can be applied phasically to induce well-defined SEPs from the muscle. Large myelinated afferent fibers conveying proprioceptive information (group I) can be activated phasically by passive movement or by instantaneous changes in weight loads [66]. Thin myelinated and unmyelinated fibers can only be activated phasically by intramuscular electrostimulation (IMES) applied within the muscle by a pair of needle electrodes (typically uninsulated at the tip) or by direct stimulation of muscle afferents. Although electrical stimulation is a non-specific method with respect to afferent fiber activation [67] and evokes muscle twitches it has the advantage that it can be applied to both the skin and muscle thereby facilitating direct comparisons between the two tissue systems. Another advantage is that it can evoke activity from the median nerve innervated thenar muscle which can then be compared to the well characterized SEP activity from mixed median nerve stimulation at the wrist. Early SEPs and muscle afferent projections

According to the conventional view of somatosensory afferent projections to primary somatosensory cortex, large myelinated cutaneous afferents (Aβ/ group II) mainly projects to Brodmann Area (BA) 3b and 1 and large myelinated afferents from deep tissue (group I/ Aα) mainly projects to BA 3a and BA 2 [68]. In SEP recordings, non-painful cutaneous electrostimulation and non-painful electrical median nerve stimulation at the wrist result in the well-known early N20 component. It is a negative component that can be found in contralateral centro-parietal electrodes around 20 ms following stimulus onset. A “short-latency” paradigm consisting of short interstimulus intervals and hundreds of repetitive stimuli is needed in order to obtain a clear response with sufficient signal-to-noise ratio. The N20 component is the earliest cortical response in SEPs and reflects large afferent projections to the hand area of primary somatosensory cortex in the posterior bank of the central sulcus (BA 3b) [68,69]. Controversies still exist over whether projections of muscle afferents to cortex are reflected in the N20 component. While both group I and II muscle afferents have been suggested to contribute to the generation of the response [70-73] others have suggested that the early N20 component is mainly generated by cutaneous large afferent input (Aβ/ group II) [74,75]. In support of the later, proprioceptive (group I) input engaged by passive movement [68,76] or by weight change [66,77] does not result in an N20

Brain imaging of muscle pain 19

component in accord with their main projection to BA 3a in primary somatosensory cortex. More recently, a series of studies applied “short-latency” paradigms in conjunction with IMES and failed to observe the N20 component [19,78-81]. Although this issue remains unresolved, we interpret the body of evidence in favor of the later explanation. That is, the presence of an N20 response after muscle stimulation most likely indicates contamination by activated cutaneous Aβ-afferents.

Figure 3. Time sequence of grand mean topographies. Activation from non-painful (upper) and painful (lower) intramuscular electrostimulation given at 0.2 Hz. Stimulus intensity-effects are clear at several loci. The latency is given below each map. The color scales applies to the topographies above and are given in units of microvolts. Processing of somatosensory input from the muscle The earlier studies on cortical processing of muscle afferent evoked activity were only concerned with the early latency SEPs (0-50 ms post stimulus onset), used a limited number of scalp electrodes, and did not investigate responses from nociceptive afferents. Due to the slower conduction velocity of nociceptive afferents cortical activation from these occur at a later latency than from non-nociceptive afferents. To record these late responses longer interstimulus intervals are required resulting in longer recording sessions and fewer evoked responses. Although the late responses (100-500 ms post stimulus onset) can be detected with few trials (20-30) this is not possible for earlier responses evoked by non-specific electrostimulation. The first studies comparing late-latency SEPs from painful muscle and skin stimulation used “long-latency” paradigms consisting of long interstimulus intervals (30-40 seconds) and few stimulus trials (20-40) albeit with relative low number (21-32) of recording electrodes [82-84]. In these studies, similar topographic patterns were found in skin and muscle conditions at both non-painful and painful stimulus

Niddam et al. 20

intensities indicating similar generating sources in cortex. Initial source analysis concluded that the same regions were involved in processing of skin and muscle pain [83]. Increases in peak amplitudes were found for several components reflecting the increase in stimulus intensity between painful and non-painful stimuli. However, no unique topography or source was found to be activated in the painful condition compared to the non-painful condition also indicating shared generators (Figure 3). The later is not surprising since it is well-known that several brain regions process both non-nociceptive and nociceptive input.

More recent studies have investigated SEPs to IMES in the full time-interval from early to late latency by applying shorter interstimulus intervals and more trials [19,78-81]. Apart from differences between skin and muscle SEPs in the early phase, as discussed above, differences were also observed in middle latency components (50-100 ms), and in late components. High-density EEG source analysis revealed a temporal cascade of cortical activity (Figure 4) involving the same generators for painful and non-painful IMES. The cascade was initiated by activity in vicinity of the contralateral central sulcus (either BA 3b or 4) about 60-90 ms after IMES. At both stimulus intensities this was followed by activity in the dorsal premotor cortex, bilateral operculo-insular cortices (including secondary somatosensory cortices), caudal cingulate motor area, and concluded with activation in vicinity of posterior cingulate/ parietal cortex [81]. Interestingly, the prominent contralateral fronto-central negativity around 60-90 ms (Figure 3), generated by the initial source, was not as clearly defined in responses to surface stimulation of the skin [81] or to intracutaneous stimulation (see Figure 2 in Diers et al. [79]). In contrast, a similar prominent waveform can be found in proprioceptive SEPs from both passive movement and weight change [66,77,85,86]. This component has been localized more frontal than the N20 source in BA 3b, most likely corresponding to primary motor cortex [85,86]. It is noteworthy to mention that proprioceptive information is passed on from area 3a to both BA 3b and BA 4 (motor cortex) [87,88]. Source analysis further revealed that the dorso-lateral premotor area and cingulate motor area (CMA) were specifically activated during IMES compared to skin stimulation (Figure 4) [81]. Since CMA has been shown to be activated in other cutaneous pain paradigms [89], its activation might represent part of the motor dimension of pain such as e.g. motor defensive strategies involving preparation of orientating the body towards the painful stimulus. In this respect, the CMA activity might represent the motor response to escapable pain and thus be specific to phasic pain. Such motor processing likely takes place in parallel to processing of other sensory-pain dimensions [81]. Interestingly, motor defensive strategies rely crucially on redirecting attention towards potential

Brain imaging of muscle pain 21

harmful external stimuli. In accord with this, it was recently found that the late positive vertex potential corresponding to CMA activation could be modulated by redirecting attention towards the stimuli [90].

Figure 4. Temporal cascade of dipole sources in response to non-painful cutaneous stimulation (Surf), non-painful intramuscular electrostimulation (NP IMES), and painful intramuscular electrostimulation (P IMES). The white line indicates mean peak values across subjects and the endpoints of the colored bars indicate start and end of source activity. SMI, primary sensorimotor cortex; PM, premotor cortex, cSII, contralateral secondary somatosensory cortex; iSII, ipsilateral secondary somatosensory cortex; CMA/ ACC, cingulate motor area (for muscle stimulation/ anterior cingulate cortex (for skin stimulation); PCC/ PPC, posterior cingulate cortex/ parietal cortex. Based on Niddam et al. [81]. Latency differences between SEPs from skin and muscle stimuli

In studies where SEP responses to electrical skin and muscle stimulation were compared longer latencies were found for the major mid-

Niddam et al. 22

and late-latency peaks for muscle input. One explanation of this finding could be involvement of different neural pathways as indicated above. Cortical input from muscle afferents as observed with SEPs may reflect more motor-related processing involving a larger number of relay sites than for cutaneous input. In contrast, Svensson et al. [82] found considerable shorter latencies for the major peaks after painful IMES compared to purely nociceptive laser stimuli. This, on the other hand, could be explained by the contribution of large-diameter fiber types to the late components evoked by painful IMES. Modulation of SEPs in patients Intramuscular electrostimulation was recently applied to study changes in SEPs in several different patient populations including patients with fibromyalgia [80], chronic low back pain (CLBP) [79], and chronic tension-type headache (CTTH) [78]. While both CLBP and CTTH may have muscular origin fibromyalgia is considered a more general disorder of central sensory processing. In CLBP patients relative to healthy controls, a higher response was found for the negativity around 80 ms (N80) while a diminished response was found for the late positive potential around 260 ms (P260; see Figure 3) [79]. However, these findings were not only limited to the affected muscle but were in fact even more pronounced for a more distant arm muscle. Further analysis revealed a positive relationship between a perceptual sensitization measure and the N80 amplitude and a negative relationship between habitual pain intensity and the P260 amplitude. As mentioned above, the contralateral N80-component is most likely generated either in the posterior or anterior bank of the central sulcus corresponding to primary somatosensory (area 3b) and motor cortex, respectively [81]. Since pain intensities were matched between patients and controls, increased N80 amplitude may indicate an adaptive mechanism subserving sensitization in CLBP patients as part of the chronicity process. In this respect, primary somatosensory cortex is known to encode stimulus intensity related information. The late P260-component is most likely generated in CMA [81]. Diminished activity was interpreted as a gating effect where chronic ongoing pain inhibits phasic pain [79]. Experimental tonic muscle pain has also been found to exert a similar effect on the P260 component to non-painful phasic muscle stimuli [91]. However, due to the salient nature of ongoing pain, it cannot be ruled out that an altered cognitive set contributed to the observed interaction. Ongoing pain could serve as a distracter from the less salient phasic stimuli. In agreement with this, the late positive SEP component has been shown to be reduced by distraction [92].

Brain imaging of muscle pain 23

The same non-specific increase in N80 activity as observed for CLBP patients was also found in fibromyalgia patients [80]. However, significantly lower levels of stimulus intensity were required to induce subjective pain ratings of similar magnitude. Taken together, this indicates a highly sensitized state in fibromyalgia underpinned by augmented cortical sensorimotor processing. In a somewhat different approach, Buchgreitz et al. [78] used IMES of the trapezius muscle in patients with CTTH before, during and after experimentally induced homotopic tonic muscle pain and compared the results with healthy controls. Single discrete stimuli were given as well as trains of stimuli consisting of 5 discrete pulses. The later was implemented in order to induce temporal summation. In healthy controls, the P200 dipole strength (which corresponds to the P260 component due to shorter conduction distance from the muscle) was reduced after the first and fifth train stimuli during and after tonic muscle pain. This was not the case in patients. However, abnormal central processing was not paralleled by any differences in temporal summation at the perceptual level. As for CLBP patients, the result may indicate a deficient pain-inhibiting effect of phasic painful stimuli in patients with CTTH. Spontaneous brain oscillations Ongoing spontaneous oscillations can be recorded by EEG from many brain regions and provide an alternative approach to study brain functions. Since brain oscillations arise independently of external stimuli and are mainly non-phase locked to these they are suitable for modulation by experimental tonic pain stimuli. At rest, brain oscillations arising from sensory cortices or motor cortex represent an idling state in which the region does not receive or process any input. The amplitude of these oscillations increases during inactivation (inhibition) of the specific modality and decreases during activation (excitation).

In an early study using intramuscular administration of hypertonic saline to induce tonic muscle pain only non-specific effects were found possibly attributable to contamination by muscle tension in head musculature [93]. Later studies found alterations in oscillatory activity specific to tonic muscle pain albeit with seemingly contradictory results. Increased power has been observed in the lower alpha band (8-11 Hz) over the contralateral parietal region during hypertonic saline induced muscle pain compared to a baseline [94]. This effect remained when compared directly to continuous vibration. In contrast, other studies using hypertonic saline and capsaicin to induce muscle pain resulted in decreased lower alpha power over parieto-occipital regions compared to a baseline [95-97].

Niddam et al. 24

This finding remained when compared to aversive auditory stimuli [95]. One explanation of this discrepancy may be that an initial phase with decrease in power is followed by a phase with augmented power [94]. In Le Pera et al. [94], EEG recordings were performed about 5-6 minutes after onset of infusion while in Chang et al. [98] it was performed when a certain subjective rating was obtained, possibly almost immediately, and lasted 2 minutes. However, later injections of hypertonic saline in a series of injections resulted in increased lower alpha power [96]. Thus, tonic muscle pain seemingly initially enhances cortical excitability (decreased spontaneous oscillations) but this turns into decreased excitability (increased spontaneous oscillations) or inhibition at a later stage. In accord with this, phasic pain briefly suppresses oscillations from other sensory modalities such as e.g. occipital visual alpha oscillations [99]. The later augmentation may reflect habituation to the painful stimuli. In the series of papers by Chang et al. changes in lower alpha power were more or less paralleled by changes in upper alpha power (11-14 Hz). In contrast, high beta power (25-35 hz) over the vertex of the head was found to increase throughout the pain phase [96,98]. Apparently, the substantial augmentation of high beta power is specific to muscle pain relative to subcutaneous pain while effects in alpha bands are tissue non-specific [97]. The neurophysiologic significance of these findings remains to be clarified in terms of involved brain regions. Where occipital low alpha oscillations originate from visual cortex, high alpha oscillations mainly originate from primary somatosensory cortex and are typically modulated by contralateral somatosensory input. However, the later appear as small focal changes above the rolandic area and can easily be superimposed by the more dominating occipital low alpha oscillations if eyes are kept closed during EEG recordings. Also, movement during continuous scoring of VAS may suppress low beta activity (15-25 Hz) mainly originating from primary motor cortex. Finally, oscillatory brain activity occurs in different bands that may vary between individuals. More refined results may be obtained by tailoring frequency bands to each individual. Conclusions

Acute muscle pain has been studied by diverse imaging techniques revealing different aspects of brain function. Many commonalities between brain processing of cutaneous and muscle pain emerged from these studies. It is reasonable to argue that in the acute state, pain of cutaneous and muscular origin is largely processed in the same brain regions. However, within these regions finer spatial and temporal analysis has revealed subtle differences that may reflect differences at the

Brain imaging of muscle pain 25

perceptual level. More substantial differences are likely to occur from future work on chronic muscle pain since different types of chronic pain may result in different maladaptive plastic changes. In this respect, chronic muscle pain is often associated with motor dysfunction and stress which are known to affect specific brain circuits. Suppressed hippocampal function in patients with myofascial pain syndrome may represent one such example. More detailed longitudinal studies are needed to monitor the progress of the chronicity process. At present, it remains unknown whether central changes are reversible or not and to what degree they contribute to the generation of pain. These questions are important to answer since they may provide new neural targets for interventions. These targets do not necessarily have to be directly involved in pain processing per se since many types of chronic muscle pain have a substantial motor component. Acknowledgments This work was supported by National Science Council grants, NSC 97-2314-B-010-050-MY2 and NSC 96-2752-B-010-008-PAE. References 1. Curatolo M, Bogduk N. Clin J Pain 2001; 17: 25-32. 2. Witting N, Svensson P, Gottrup H, Arendt-Nielsen L, Jensen TS. Pain 2000;

84: 407-412. 3. Curatolo M, Petersen-Felix S, Gerber A, Arendt-Nielsen L. Br J Anaesth 2000;

85: 529-532. 4. Graven-Nielsen T. Scand J Rheumatol Suppl 2006; 1-43. 5. Travell JG, Simons DG. Myofascial pain and dysfunction. The trigger point

manual Williams & Wilkins: Baltimore, first ed., 1983. 6. Gameiro GH, da Silva Andrade A, Nouer DF, Ferraz de Arruda Veiga MC.

Clin Oral Investig. 2006; 7. Niddam DM, Chan RC, Lee SH, Yeh TC, Hsieh JC. Neuroimage 2008; 39:

1299-1306. Epub 2007 Oct 1211. 8. Kumazawa T, Mizumura K. J Physiol 1977; 273: 179-194. 9. Stacey MJ. J Anat 1969; 105: 231-254. 10. Swett JE, Woolf CJ. J Comp Neurol 1985; 231: 66-77. 11. Hoheisel U, Lehmann-Willenbrock E, Mense S. Neuroscience 1989; 28: 495-

507. 12. Cervero F, Iggo A, Ogawa H. Pain 1976; 2: 5-24. 13. Mense S, Prabhakar NR. Neurosci Lett 1986; 66: 169-174. 14. Mense S, Craig AD, Jr. Neuroscience 1988; 26: 1023-1035. 15. Schomburg ED, Steffens H, Kniffki KD. Neurosci Res 1999; 33: 195-206. 16. Wall PD, Woolf CJ. J Physiol 1984; 356: 443-458. 17. Woolf CJ, Wall PD. J Neurosci 1986; 6: 1433-1442. 18. Staahl C, Drewes AM. Basic Clin Pharmacol Toxicol 2004; 95: 97-111.

Niddam et al. 26

19. Niddam DM, Graven-Nielsen T, Arendt-Nielsen L, Chen AC. Brain Topogr 2001; 13: 283-292.

20. Svensson P, Beydoun A, Morrow TJ, Casey KL. Exp Brain Res 1997; 114: 390-392.

21. Torebjork HE, Ochoa JL, Schady W. Pain 1984; 18: 145-156. 22. Graven-Nielsen T, McArdle A, Phoenix J, Arendt-Nielsen L, Jensen TS,

Jackson MJ, et al. Pain 1997; 69: 137-143. 23. Babenko VV, Graven-Nielsen T, Svensson P, Drewes AM, Jensen TS, Arendt-

Nielsen L. Eur J Pain 1999; 3: 93-102. 24. Svensson P, Cairns BE, Wang K, Arendt-Nielsen L. Pain 2003; 104: 241-247. 25. Svensson P, Cairns BE, Wang K, Hu JW, Graven-Nielsen T, Arendt-Nielsen L,

et al. Pain 2003; 101: 221-227. 26. Svensson P, Graven-Nielsen T, Arendt-Nielsen L. Pain 1998; 74: 93-100. 27. Price DD, McGrath PA, Rafii A, Buckingham B. Pain 1983; 17: 45-56. 28. Gracely RH, McGrath F, Dubner R. Pain 1978; 5: 5-18. 29. Melzack R. Pain 1975; 1: 277-299. 30. Malone MD, Kurtz RM, Strube MJ. Am J Clin Hypn 1989; 31: 221-230. 31. Sullivan MJ, D'Eon JL. J Abnorm Psychol 1990; 99: 260-263. 32. Sullivan MJ, Thorn B, Haythornthwaite JA, Keefe F, Martin M, Bradley LA,

et al. Clin J Pain 2001; 17: 52-64. 33. Sullivan MJ, Rodgers WM, Kirsch I. Pain 2001; 91: 147-154. 34. Rosenstiel AK, Keefe FJ. Pain 1983; 17: 33-44. 35. Peyron R, Laurent B, Garcia-Larrea L. Neurophysiol Clin 2000; 30: 263-288. 36. Apkarian AV, Bushnell MC, Treede RD, Zubieta JK. Eur J Pain 2005; 9: 463-

484. Epub 2005 Jan 2021. 37. Svensson P, Minoshima S, Beydoun A, Morrow TJ, Casey KL. J

Neurophysiol 1997; 78: 450-460. 38. Korotkov A, Ljubisavljevic M, Thunberg J, Kataeva G, Roudas M, Pakhomov

S, et al. Neurosci Lett 2002; 335: 119-123. 39. Thunberg J, Lyskov E, Korotkov A, Ljubisavljevic M, Pakhomov S, Katayeva

G, et al. Eur J Pain 2005; 9: 185-194. 40. Kupers RC, Svensson P, Jensen TS. Pain 2004; 108: 284-293. 41. Henderson LA, Bandler R, Gandevia SC, Macefield VG. Pain 2006; 120: 286-

296. Epub 2006 Jan 2019. 42. Henderson LA, Gandevia SC, Macefield VG. Pain 2007; 128: 20-30. Epub

2006 Oct 2002. 43. Henderson LA, Gandevia SC, Macefield VG. Neuroimage 2008; 39: 1867-

1876. Epub 2007 Nov 1812. 44. Macefield VG, Gandevia SC, Henderson LA. Cereb Cortex 2007; 17: 2050-

2059. Epub 2006 Nov 2013. 45. Brooks JC, Zambreanu L, Godinez A, Craig AD, Tracey I. Neuroimage 2005;

27: 201-209. 46. Greenspan JD, Lee RR, Lenz FA. Pain 1999; 81: 273-282. 47. Ostrowsky K, Magnin M, Ryvlin P, Isnard J, Guenot M, Mauguiere F. Cereb

Cortex 2002; 12: 376-385. 48. Niddam DM, Yeh TC, Wu YT, Lee PL, Ho LT, Arendt-Nielsen L, et al.

Neuroimage 2002; 17: 1437-1450.

Brain imaging of muscle pain 27

49. Niddam DM, Chan RC, Lee SH, Yeh TC, Hsieh JC. Clin J Pain 2007; 23: 440-448.

50. Coghill RC, Sang CN, Maisog JM, Iadarola MJ. J Neurophysiol 1999; 82: 1934-1943.

51. Petrovic P, Ingvar M, Stone-Elander S, Petersson KM, Hansson P. Pain 1999; 83: 459-470.

52. Halgren E, Walter RD, Cherlow DG, Crandall PH. Brain 1978; 101: 83-117. 53. Prado WA, Roberts MH. Brain Res 1985; 340: 219-228. 54. Bodnar RJ, Kelly DD, Brutus M, Glusman M. Neurosci Biobehav Rev 1980; 4:

87-100. 55. Feldman S, Conforti N. Neuroendocrinology 1980; 30: 52-55. 56. Fendler K, Karmos G, Telegdy G. Acta Physiol Acad Sci Hung 1961; 20: 293-

297. 57. Sousa N, Lukoyanov NV, Madeira MD, Almeida OF, Paula-Barbosa MM.

Neuroscience 2000; 97: 253-266. 58. Duric V, McCarson KE. J Pain 2006; 7: 544-555. 59. Yoshihara T, Shigeta K, Hasegawa H, Ishitani N, Masumoto Y, Yamasaki Y. J

Orofac Pain 2005; 19: 202-208. 60. Zubieta JK, Bueller JA, Jackson LR, Scott DJ, Xu Y, Koeppe RA, et al. J

Neurosci 2005; 25: 7754-7762. 61. Zubieta JK, Smith YR, Bueller JA, Xu Y, Kilbourn MR, Jewett DM, et al.

Science 2001; 293: 311-315. 62. Zubieta JK, Smith YR, Bueller JA, Xu Y, Kilbourn MR, Jewett DM, et al. J

Neurosci 2002; 22: 5100-5107. 63. Zubieta JK, Heitzeg MM, Smith YR, Bueller JA, Xu K, Xu Y, et al. Science

2003; 299: 1240-1243. 64. Scott DJ, Heitzeg MM, Koeppe RA, Stohler CS, Zubieta JK. J Neurosci 2006;

26: 10789-10795. 65. Altier N, Stewart J. Life Sci 1999; 65: 2269-2287. 66. Arnfred S, Chen AC, Eder D, Glenthoj B, Hemmingsen R. Neurosci Lett 2000;

288: 111-114. 67. Laursen RJ, Graven-Nielsen T, Jensen TS, Arendt-Nielsen L. Muscle Nerve

1999; 22: 1564-1570. 68. Mima T, Ikeda A, Terada K, Yazawa S, Mikuni N, Kunieda T, et al.

Electroencephalogr Clin Neurophysiol 1997; 104: 103-107. 69. Buchner H, Fuchs M, Wischmann HA, Dossel O, Ludwig I, Knepper A, et al.

Brain Topogr 1994; 6: 299-310. 70. Restuccia D, Valeriani M, Insola A, Lo Monaco M, Grassi E, Barba C, et al.

Muscle Nerve 2002; 26: 44-54. 71. Gandevia SC, Burke D, McKeon B. Brain 1984; 107: 1-13. 72. Gandevia SC, Burke D. Brain 1988; 111: 389-403. 73. Gandevia SC, Burke D. Electroencephalogr Clin Neurophysiol 1990; 77: 353-

361. 74. Kunesch E, Knecht S, Schnitzler A, Tyercha C, Schmitz F, Freund HJ. J Clin

Neurophysiol 1995; 12: 476-487. 75. Halonen JP, Jones S, Shawkat F. Electroencephalogr Clin Neurophysiol 1988;

71: 331-335.

Niddam et al. 28

76. Desmedt JE, Ozaki I. Electroencephalogr Clin Neurophysiol 1991; 80: 513-521.

77. Arnfred SM. Clin Neurophysiol 2005; 116: 849-860. Epub 2004 Dec 2025. 78. Buchgreitz L, Egsgaard LL, Jensen R, Arendt-Nielsen L, Bendtsen L. Brain

2008; 79. Diers M, Koeppe C, Diesch E, Stolle AM, Holzl R, Schiltenwolf M, et al. J

Clin Neurophysiol 2007; 24: 76-83. 80. Diers M, Koeppe C, Yilmaz P, Thieme K, Markela-Lerenc J, Schiltenwolf M,

et al. J Clin Neurophysiol 2008; 25: 153-160. 81. Niddam DM, Chen LF, Wu YT, Hsieh JC. Neuroimage 2005; 25: 942-951. 82. Svensson P, Beydoun A, Morrow TJ, Casey KL. Electroencephalogr Clin

Neurophysiol 1997; 104: 343-350. 83. Shimojo M, Svensson P, Arendt-Nielsen L, Chen AC. Brain Topogr 2000; 13:

43-58. 84. Chen AC, Shimojo M, Svensson P, Arendt-Nielsen L. Brain Topogr 2000; 13:

59-72. 85. Seiss E, Praamstra P, Hesse CW, Rickards H. Exp Brain Res 2003; 148: 308-

319. Epub 2002 Nov 2030. 86. Seiss E, Hesse CW, Drane S, Oostenveld R, Wing AM, Praamstra P.

Neuroimage 2002; 17: 461-468. 87. Huerta MF, Pons TP. Brain Res 1990; 537: 367-371. 88. Heath CJ, Hore J, Phillips CG. J Physiol 1976; 257: 199-227. 89. Kwan CL, Crawley AP, Mikulis DJ, Davis KD. Pain 2000; 85: 359-374. 90. Valeriani M, Tonali P, De Armas L, Mariani S, Vigevano F, Le Pera D.

Neurosci Res 2008; 60: 170-175. Epub 2007 Oct 2030. 91. Niddam DM, Hsieh JC. J. Med. Biol. Eng. 2008; 28: 135-138. 92. Yamasaki H, Kakigi R, Watanabe S, Hoshiyama M. Brain Res Cogn Brain

Res 2000; 9: 165-175. 93. Veerasarn P, Stohler CS. Pain 1992; 49: 349-360. 94. Le Pera D, Svensson P, Valeriani M, Watanabe I, Arendt-Nielsen L, Chen AC.

Clin Neurophysiol 2000; 111: 2130-2137. 95. Chang PF, Arendt-Nielsen L, Chen AC. Exp Brain Res 2002; 147: 387-393.

Epub 2002 Oct 2017. 96. Chang PF, Arendt-Nielsen L, Graven-Nielsen T, Chen AC. Brain Res Bull

2003; 59: 533-543. 97. Chang PF, Arendt-Nielsen L, Graven-Nielsen T, Svensson P, Chen AC. Int J

Psychophysiol 2004; 51: 117-126. 98. Chang PF, Arendt-Nielsen L, Graven-Nielsen T, Svensson P, Chen AC. Exp

Brain Res 2001; 141: 195-203. 99. Ploner M, Gross J, Timmermann L, Pollok B, Schnitzler A. Cereb Cortex

2006; 16: 537-540. Epub 2005 Jul 2020.