Brain imaging of neuropathic pain

https://doi.org/10.1016/j.neuroimage.2007.03.054Get rights and content

Abstract

Many studies have focused on defining the network of brain structures involved in normal physiological pain. The different dimensions of pain perception (i.e., sensory discriminative, affective/emotional, cognitive/evaluative) have been shown to depend on different areas of the brain. In contrast, much less is known about the neural basis of pathological chronic pain. In particular, it is unclear whether such pain results from changes to the physiological “pain matrix”. We review here studies on changes in brain activity associated with neuropathic pain syndromes—a specific category of chronic pain associated with peripheral or central neurological lesions. Patients may report combinations of spontaneous pain and allodynia/hyperalgesia—abnormal pain evoked by stimuli that normally induce no/little sensation of pain. Modern neuroimaging methods (positron emission tomography (PET) and functional MRI (fMRI)) have been used to determine whether different neuropathic pain symptoms involve similar brain structures and whether these structures are related to the physiological “pain matrix”. PET studies have suggested that spontaneous neuropathic pain is associated principally with changes in thalamic activity and the medial pain system, which is preferentially involved in the emotional dimension of pain. Both PET and fMRI have been used to investigate the basis of allodynia. The results obtained have been very variable, probably reflecting the heterogeneity of patients in terms of etiology, lesion topography, symptoms and stimulation procedures. Overall, these studies indicated that acute physiological pain and neuropathic pain have distinct although overlapping brain activation pattern, but that there is no unique “pain matrix” or “allodynia network”.

Introduction

Modern non-invasive brain-imaging techniques helped to change our perception of the processing of pain in the brain. Many studies based on positron emission tomography (PET) or functional magnetic resonance imaging (fMRI) have investigated changes in brain activity in response to various experimental stimuli inducing pain. This has led to the characterization of a network of brain areas forming a “pain matrix” involved in the different dimensions of pain perception. Activation in response to pain has consistently been reported in six areas: the primary and secondary somatosensory cortices (S1, S2), the insular cortex (IC), the anterior cingulate cortex (ACC), the thalamus and the prefrontal cortex (PFC) (Treede et al., 1999, Peyron et al., 2000, Davis, 2000, Apkarian et al., 2005, Tracey, 2005). Activation of the lateral thalamus, S1 and S2 (and probably the posterior insula) seems to be preferentially associated with the sensory-discriminative aspect of pain. ACC and the anterior insular cortex, which are components of the limbic system, seem to be more involved in the affective-emotional aspect of pain. The activation of the prefrontal areas may be related to more cognitive aspects of pain. Other brain areas, including the brainstem, the basal ganglia and the cerebellum have also been shown to be activated, but their role in pain processing remains unclear. Many studies have also investigated the modulation of brain correlates of pain perception by psychological factors such as attention, distraction, expectation, anticipation, emotions and hypnotic suggestions (Porro, 2003, Villemure and Bushnell, 2002, Rainville, 2002, Kupers et al., 2005, Apkarian et al., 2005).

Changes in brain activity associated with clinical pain have been much less thoroughly investigated (Apkarian et al., 2005, Kupers and Kehlet, 2006, Schweinhardt et al., 2006a, Schweinhardt et al., 2006b). In particular, it is unknown whether the “pain matrix” associated with acute experimental pain is involved in pathological chronic pain syndromes. It is unlikely that all chronic pain syndromes depend on the same brain mechanisms, as many such syndromes exist and they are highly diverse.

Neuropathic pain, caused by a lesion or dysfunction of the peripheral or central nervous system, is generally chronic and disabling, and is one of the most difficult pains to treat because it responds poorly to conventional analgesics (Dworkin et al., 2003, Finnerup et al., 2005). Neuropathic pain is associated with a large variety of peripheral and central nervous system lesions, corresponding to highly heterogeneous clinical conditions. Patients with neuropathic pain display certain specific clinical characteristics, making it possible to distinguish this type of pain from the chronic nociceptive/inflammatory pain associated with somatic lesions (Bouhassira et al., 2005, Bennett et al., 2007). However, neuropathic pain is a multidimensional clinical entity (Bouhassira et al., 2004), probably mediated by a number of different pathophysiological mechanisms (Woolf and Mannion, 1999, Baron, 2006). The heterogeneity of neuropathic pain syndromes is apparent from the clinical examination of patients who may present with various painful symptoms including spontaneous pain, which may be continuous or paroxysmal, and evoked pains (see Fig. 1). Evoked pain can be even more distressing than spontaneous pain. Evoked pain is described as allodynia when triggered by normally innocuous (mechanical or thermal) stimuli and hyperalgesia when it corresponds to an exaggerated response to a stimulus that would normally cause pain.

We review here studies specifically investigating changes in brain activity associated with neuropathic pain by hemodynamic imaging methods (PET and fMRI). We considered only studies (not case reports) concerning definite neuropathic pain, which could clearly be attributed to a peripheral or central nervous system lesion. Studies relating to chronic pain conditions of uncertain origin, such as fibromyalgia, complex regional pain syndrome (CRPS) type I (former reflex sympathetic dystrophy), irritable bowel syndrome and burning mouth syndrome, were not included in this review, although some of the mechanisms underlying these conditions may be common to neuropathic pain. Several recent neuroimaging studies were conducted in patients with CRPS (e.g., Maihöfner et al., 2005, Maihofner et al., 2006a, Maihofner et al., 2006b, Pleger et al., 2006). However, these studies have concerned almost exclusively the much more common CRPS type I, which, in contrast to CRPS type 2 (formally called causalgia), is not associated with a nerve lesion and is generally not regarded as a typical neuropathic pain syndrome (Janig and Baron, 2003, Janig and Baron, 2006).

A few studies have used PET to investigate changes in basal regional cerebral blood flow (rCBF) related to spontaneous continuous neuropathic pain. Some have compared variations in rCBF within specific regions of interest (ROI) between the normal and affected sides, or between affected subjects and healthy volunteers, whereas others have investigated changes in brain due to analgesic procedures.

Laterre et al. (1988) suggested that central neuropathic pain (post-ischemic “thalamic syndrome”) in one patient was associated with a decrease in basal thalamic activity on the affected side. This unexpected result was confirmed in subsequent studies. Di Piero et al. (1991) studied five patients presenting chronic intractable unilateral chronic pain in a leg or arm due to cancer, before and after percutaneous cervical cordotomy. Comparisons of basal cerebral blood flow within three ROIs (the thalamus, the primary sensory cortex and the prefrontal cortex) between the patients (before surgery) and a group of five age-matched healthy volunteers showed a significant decrease in rCBF in the hemithalamus (dorsal anterior quadrant) on the side contralateral to the pain. This decrease in thalamic activity was reversed by cervical cordotomy providing significant pain relief. In contrast, no significant change in rCBF was observed in the prefrontal or somatosensory cortices. Unfortunately, the description of pain (intensity, duration, quality etc.) before and after surgery was poor in this study. In particular, it is unclear whether the patients experienced any pain during the PET scan. It is therefore difficult to correlate the functional imaging data to specific pain characteristics. In addition, although the authors considered the pain to result primarily from the invasion of the brachial or lumbar plexus by the tumor, the mechanisms of pain operating in these patients were probably not purely neuropathic.

Iadarola et al. (1995) compared right and left thalamic activity in five patients with unilateral chronic pain and in 13 healthy subjects. Four patients presented post-traumatic pain and one patient presented postherpetic neuralgia. Mean rCBF was significantly lower (12 ± 1.6%) in the thalamus on the side contralateral to the pain than on the ipsilateral side in the patients, whereas healthy volunteers displayed only 1.5% right/left asymmetry in thalamic activity. The most robust decreases in activity were observed in the anterior dorsal quadrant and the posterior ventral quadrant. However, the neuropathic nature of the pain experienced by the patients included in this study is questionable. Neurological examination was normal in the four patients with post-traumatic pain, and no nerve lesions were identified. These patients probably presented a type I complex regional pain syndrome (CRPS), involving both neuropathic and non-neuropathic mechanisms. Furthermore, no clear information was provided concerning the intensity of pain experienced by these patients during brain imaging.

Hsieh et al. (1995) carried out a more comprehensive study in eight patients with chronic painful mononeuropathy. PET scans were not restricted to the thalamus area and were acquired before and after regional nerve block with lidocaine providing significant pain relief. The intensity and quality of pain were monitored during scanning. A comparison of scans showed a bilateral decrease in rCBF in the anterior insula, the posterior parietal cortex, the inferior and lateral prefrontal cortex (Brodmann area (BA) 10/47), the cerebellar vermis and in the right posterior sector of anterior cingulate cortex (BA 24) after treatment indicating that the initial levels of rCBF were abnormally high. In contrast, treatment increased rCBF in the posterior thalamus contralateral to the painful side. No significant changes were observed in the primary and secondary somatosensory cortices (SI and SII).

Despite the uncertainty regarding the neuropathic origin of pain in some of the patients studied, these data suggest that changes in basal brain activity in patients with chronic neuropathic pain involve only some parts of the “pain matrix” associated with acute pain. The lack of change in rCBF in the somatosensory cortices SI and SII (Di Piero et al., 1991, Hsieh et al., 1995), together with the increase in activity in the insula and anterior cingulate cortex (Hsieh et al., 1995), suggests that chronic continuous pain is principally related to changes in areas associated with the affective/emotional dimension of pain. The most striking observation common to all three of these studies and confirmed in several case reports (Pagni and Canavero, 1995, De Salles and Bittar, 1994, Kupers et al., 2000) was the decrease in rCBF in the thalamus contralateral to the pain, probably reflecting a decrease in thalamus neuronal activity in these patients. This result, which contrasts sharply with the results of acute pain studies, is not easy to reconcile with electrophysiological data. Microelectrode recordings from the thalamus of chronic patients undergoing stereotaxic surgery and in animal models have generally demonstrated abnormal hyperactivity of the thalamic neurons (Hua et al., 2000). Abnormally high levels of metabolic activity, as assessed by 2-deoxyglucose autoradiography, have also been demonstrated in animal models of focal neuropathy (Mao et al., 1993). There is no satisfactory explanation for the observed relative decrease in thalamic rCBF in neuropathic pain patients. It has been suggested that this decrease corresponds to a compensatory mechanism for inhibiting excessive nociceptive inputs or to the uncoupling between blood flow and neuronal activity in patients experiencing chronic pain (Iadarola et al., 1995). The reversal of this decrease in thalamic activity following various types of analgesic procedure suggests that this decrease results from functional impairment rather than degenerative processes. The changes in thalamic activity induced by electrical motor cortex stimulation or thalamic stimulation for analgesic purposes in patients with refractory neuropathic pain (Peyron et al., 1995, Peyron et al., 2007, Garcia-Larrea et al., 1997, Duncan et al., 1998) are consistent with this conclusion.

Further studies, including larger number of patients with definite neuropathic pain and a thorough description of pain characteristics are needed to confirm the pattern of brain activity associated with spontaneous neuropathic pain. Further investigations, both in humans and animals, are also required into the mechanisms underlying the decrease in thalamic rCBF, which is the most reliable finding in neuroimaging studies. These mechanisms may involve the opioidergic systems. A series of studies in which PET was used to analyze changes in opioid receptor distribution in patients with peripheral or central neuropathic pain (Jones et al., 1999, Jones et al., 2004, Willoch et al., 2004, Maarrawi et al., 2007) showed significant decrease in opioid receptor binding, not only in the thalamus contralateral to the pain, but also in the insula, anterior cingulate cortex, secondary somatosensory cortex and prefrontal cortex.

Allodynia, pain elicited by normally non-painful stimuli, is not specific to neuropathic pain syndromes. However, it is frequently associated with spontaneous (continuous and/or paroxysmal) pain in patients with a peripheral or central neurological lesion. Allodynia can be elicited by mechanical or thermal low intensity stimuli. The two most frequent subtypes are dynamic mechanical allodynia, induced by moving tactile stimuli such as brushing, and allodynia induced by cold. Several studies have investigated changes in brain activity associated with allodynia, to determine whether this aberrant pain results from an abnormal activation of the physiological “pain matrix”.

Cesaro et al. (1991) were probably the first to attempt to analyze changes in brain activity due to abnormal evoked pain. They used single photon emission computerized tomography (SPECT) to compare changes in thalamus activity in four patients with central post-stroke pain (two with and two without mechanical allodynia) and in one patient with algodystrophy. In each patient, SPECT scans were acquired before and after the application of mechanical stimuli to the painful area. These scans provided evidence of hyperactivity of the thalamus contralateral to the pain in the two patients with mechanical allodynia, but not in the patients without allodynia. These results were interpreted as reflecting the pathological disinhibition of nociceptive processes in the thalamus during allodynia. More specifically, the authors suggested that allodynia might result from a decrease in the inhibition normally exerted by the reticular nucleus of the thalamus on the medial thalamic nuclei.

Seven other studies have used modern neuroimaging techniques to investigate changes in brain activity associated with allodynia in groups of patients with definite neurological lesions. Four studies focused on patients with peripheral nerve injury: Petrovic et al. (1999) and Witting et al. (2006) used PET, whereas Schweinhardt et al., 2006a, Schweinhardt et al., 2006b and Becerra et al. (2006) used fMRI. Two studies were performed in patients with central pain: one used PET (Peyron et al., 1998) and the other fMRI (Ducreux et al., 2006). Finally, one study based on fMRI included a mixed group of patients with peripheral and central lesions (Peyron et al., 2004). These seven studies included patients with dynamic mechanical allodynia (pain induced by brushing the painful area). This subtype of allodynia is the most frequent in patients with peripheral or central lesions, but underlying mechanisms are poorly understood (Baron, 2006). In particular, it is unclear whether the brush-evoked allodynia associated with different types of lesions results from similar mechanisms in all cases. However, the fact that this type of allodynia is elicited by light tactile stimuli and mediated by peripheral large myelinated fibers indicates that it depends on central alterations in somatosensory processes (Koltzenburg et al., 1992). Other forms of allodynia (cold allodynia) were analyzed in three studies (Ducreux et al., 2006, Peyron et al., 1998, Peyron et al., 2004).

The changes observed in the six areas forming the so-called “pain matrix” are summarized in Table 1 and in Fig. 2.

The first two studies, performed in parallel, by Peyron et al. (1998) in nine patients with unilateral central pain due to lateral medullary infarct (Wallenberg’s syndrome), and by Petrovic et al. (1999) in five patients with traumatic peripheral nerve injury, concluded that dynamic allodynia was predominantly associated with changes in the lateral system relating to sensory discriminative aspects of pain (lateral thalamus, S1, S2), rather than the medial system, which relates to the affective-emotional dimension of pain (medial thalamus, ACC and insula). Peyron’s conclusion was based on the observation of a reduction of ACC activity during allodynia in patients with Wallenberg syndrome (Peyron et al., 1998). In contrast, Petrovic et al. (1999) observed no significant changes in the ACC or insula in patients with a peripheral lesion, but they found a correlation between rCBF in the ACC and anterior IC and pain intensity. However, as this correlation took into account both continuous pain intensity and allodynia intensity, it is difficult to draw firm conclusions concerning the specific involvement of these elements in allodynia.

A lack of activation on the ACC during brush-induced allodynia was reported in two more recent studies. Witting et al. (2006) studied nine patients with traumatic nerve injury. A significant increase in rCBF was observed bilaterally in S2, in the ipsilateral anterior insula, the contralateral orbitofrontal cortex (BA 11) and cerebellum, but not in the ACC. Surprisingly, no activation was observed in S1 or the thalamus in this study. In six patients with syringomyelia, Ducreux et al. (2006) reported a significant increase in activity bilaterally in S1, S2 and the prefrontal cortex (PFC) and in the thalamus contralateral to the stimulation side, but not in the ACC. In addition, consistent with the results obtained by Petrovic et al. (1999), no significant activation was observed in the insula in patients with allodynia associated with syringomyelia.

The predominant activation of the lateral pain system during dynamic mechanical allodynia reported in these four studies contrasts with the reduced activation of these areas associated with spontaneous continuous pain in patients with mononeuropathy (Hsieh et al., 1995). Although the results are not unequivocal, these studies tend to indicate that dynamic mechanical allodynia does not involve the whole physiological “pain matrix” and that its mechanisms are different from those of spontaneous continuous pain. However, these results were obtained with small populations of patients, and should therefore be interpreted with caution, because a lack of statistical significance does not mean that these areas were not activated. The existence of this bias is supported by a complementary analysis carried out by Petrovic et al. (1999), showing a relationship between (global) pain intensity and rCBF in the ACC and IC. Alternatively, such covariation may indicate that changes in these areas are involved in the evaluation of pain intensity (i.e., a sensory discriminative aspect) rather than the determination of the type of pain (spontaneous continuous pain or allodynia). The apparent lack of activation in part of the medial pain system may also reflect variability in the intensity of spontaneous ongoing pain. Strong activation of the ACC and IC, due to spontaneous pain during the “rest” period in scanning paradigms, may have prevented further activation during allodynic stimulation.

Consistent with this hypothesis, two other studies in patients with peripheral nerve injury (Schweinhardt et al., 2006a, Schweinhardt et al., 2006b, Becerra et al., 2006) reported activation in all structures of the “pain matrix” during brush-evoked allodynia (Table 1, Fig. 2). However, these results do not necessarily imply that allodynia is simply the result of an amplification or “leftward shift” of the stimulus–response function in physiological pain systems. Although the spatial activation pattern seems similar, a difference in the temporal pattern – in the dynamics of the activation of these structures – cannot be excluded. Also, these authors emphasized the differences between the patterns of brain activity associated with allodynia and acute physiological pain. Schweinhardt et al., 2006a, Schweinhardt et al., 2006b analyzed more specifically the correlation between changes in the activity of different sectors of the insular cortex and ratings of ongoing pain intensity and allodynia intensity. They observed three different patterns of insula activation. In the first, activation was observed in the posterior insula; according to studies of experimental pain, this activation is related to the intensity of the stimulus rather than to the perception of pain (Apkarian et al., 2005, Peyron et al., 2000). The abnormal activation of this area by innocuous stimuli in patients with allodynia may reflect plastic changes in somatosensory systems. The pattern of activation in the anterior portion of the insula was more complex. Consistent with studies of experimental pain, the magnitude of activation in the caudal sector of the anterior insula covaried with the intensity of allodynia, independently of continuous pain intensity. However, in allodynic conditions, the peak of activation occurred in a more rostral portion of the anterior insula. Thus, allodynia and the coding of pain intensity were dissociated within the anterior insula, suggesting a functional subdivision of this area and changes in cortical processing due to clinical pain. In contrast, there was no correlation between the intensity of allodynia and ACC activation.

In their study in five patients with facial neuropathic pain due to traumatic nerve injury or post herpetic neuralgia, Becerra et al. (2006) highlighted the role of the prefrontal cortex (BA 44, 45, 46) in brush-evoked allodynia. The PFC was also preferentially activated in patients with syringomyelia studied by Ducreux et al. (2006). Changes in this area have been reported in almost all studies of brush-induced allodynia (see Table 1, Fig. 2), although different sectors of the PFC have been implicated. Interestingly, according to a recent meta-analysis (Apkarian et al., 2005), the prefrontal cortex is the area most frequently reported to be activated in neuroimaging studies of different types of chronic pain (81% of studies). In particular, recent studies have highlighted the role of the prefrontal cortex in chronic low back pain (e.g., Baliki et al., 2006). Thus, like other types of chronic pain, neuropathic pain may principally involve areas associated with the cognitive/evaluative dimension of pain, which is generally linked to frontal lobe activity (Casey, 1999, Peyron et al., 1999). However, the frontal cortex may also be more directly involved in pain perception, particularly through the modulation of diencephalon or brainstem structures involved in pain modulation (Lorenz et al., 2002, Lorenz et al., 2003). The modulatory function of the prefrontal cortex in pain perception is further illustrated by the role of this structure in the placebo effect (Wager et al., 2004). Thus, the aberrant painful sensation evoked by tactile stimuli may result from alterations in pain modulatory systems. Interestingly, a series of studies using the capsaicin model of pain in healthy volunteers came to similar conclusions (Iadarola et al., 1998, Baron et al., 1999, Witting et al., 2001, Lorenz et al., 2002, Lorenz et al., 2003). Although this experimental model cannot be considered as truly representative of neuropathic pain, it has repeatedly been shown that topical applications or intradermal injections of capsaicin transiently reproduce certain pathological painful symptoms, such as heat or brush-evoked allodynia, in healthy volunteers.

The study conducted by Peyron et al. (2004) included the largest reported cohort of patients with dynamic allodynia (n = 27). However, these patients were highly heterogeneous, with various types of peripheral and central neurological lesions, and the authors used in most of their patients an ambiguous mode of stimulation combining tactile and thermal stimuli (i.e., cold rubbing), making comparison with other studies difficult. This study suggested that allodynia is characterized by “bilateralization” of the pattern of brain activation, with the recruitment of ispilateral areas, including the primary somatosensory cortex and anterior insula, in particular. The functional significance and pathophysiological role of the shift in brain activity are poorly understood, although this shift may reflect some reorganization and neuroplasticity in the somatosensory processes. Interestingly, ispilateral activation was also reported in most of the other studies (Table 1, Fig. 2). Such activation was also demonstrated in a spectacular case report of central pain in a patient who had undergone hemispherectomy (Olausson et al., 2001).

All studies devoted to brush allodynia have reported additional increases or decreases in the activation of structures not usually included in the physiological “pain matrix” (Table 2). The areas most frequently affected include: the motor/premotor cortex and supplementary motor area (SMA) (BA 4, 6, 8), the inferior/posterior parietal cortex (BA 39/40), the basal ganglia (caudate, globus pallidus/putamen), the cerebellum and various nuclei located in the brainstem, such as the periaqueductal gray matter (PAG).

Traditionally, most of these structures were not considered to be associated with pain processing. Activations of the premotor, SMA and cerebellum, which have also been reported in several studies of acute experimental pain (Peyron et al., 2000, Apkarian et al., 2005), may be related to the preparation, selection (or inhibition) of motor responses to painful stimuli. The inferior parietal cortex is generally associated with attentional processes. However, the almost systematic activation of these areas during brush-evoked allodynia suggests that they may play a more direct role in pain perception. Similarly, the basal ganglia may be involved in pain modulation in addition to their role in motor control (Chudler and Dong, 1995). The major role of brainstem nuclei (notably the PAG) in descending controls of the spinal transmission of nociceptive signals is well documented in both animals and humans (Millan, 2002).

A few studies have analyzed changes in brain activity associated with cold allodynia (see Table 1). Ducreux et al. (2006) reported striking differences between the changes in brain activity associated with cold and brush allodynia in patients with syringomyelia. Unlike dynamic allodynia, cold allodynia activated a network of structures similar to that activated by normal cold pain in healthy volunteers. Thus, the different subtypes of allodynia may be associated with distinct patterns of brain activity, reflecting different pathophysiological mechanisms. Consistent with this hypothesis, neurological sensory examination and quantitative sensory testing showed that patterns of thermal and mechanical sensory deficits differed between areas of cold and brush-evoked allodynia in these patients (Ducreux et al., 2006). In contrast, Peyron et al., 1998, Peyron et al., 2004 found a different pattern of activation in patients with Wallenberg’s syndrome and concluded that different types of allodynia induced similar changes in brain activity. The discrepancies between the results obtained in these studies may be due primarily to the ambiguous mode of stimulation (i.e., mixed brushing and cold stimuli) used by Peyron et al., 1998, Peyron et al., 2004. Becerra et al. (2006) applied cold stimuli to the affected or the unaffected side in patients with trigeminal neuropathy. However, the responses to these stimuli could not be considered allodynic because, at least in some patients, the same stimulus induced a painful sensation on the unaffected control side. In any case, too few data are available for any firm conclusions to be drawn regarding the mechanisms of cold allodynia and their putative similarities or differences from those of other types of allodynia.

In conclusion, neuroimaging data relating to neuropathic pain tend to confirm that acute physiological pain and clinical pain are associated with different, but overlapping patterns of brain activation (Apkarian et al., 2005, Schweinhardt et al., 2006a, Schweinhardt et al., 2006b). The results summarized here also suggest that there is no unique network associated with neuropathic pain and that the different components of neuropathic pain syndromes (spontaneous and evoked pains) probably involve different mechanisms. This is best illustrated by the striking difference in thalamic activity between spontaneous continuous pain and allodynia. Differences in the pattern of activation induced by mechanical and cold allodynia also suggest that there is no unique “allodynia network”. From a clinical perspective, these data suggest that the different neuropathic symptoms may respond differently to treatment. Consistent with this, it has been shown that several pharmacological agents do not act uniformly on neuropathic pain, but have preferential effects on various symptoms (Attal et al., 2000, Attal et al., 2002, Attal et al., 2004). These studies have revealed new potential pathophysiological mechanisms. Several studies have highlighted the activation of the prefrontal cortex, which role in pain processing would have to be studied thoroughly in experimental studies both in animals and humans. The pathophysiological significance of the bilateralization of cortical activation should also be addressed in experimental studies. Thus, neuropathic pain might involve more subtle alterations in central somatosensory processes than those described in animal studies (Woolf and Mannion, 1999, Baron, 2006).

However, these conclusions remain tentative, because too few neuroimaging studies have been carried out on neuropathic pain. The number of patients studied is also currently too small, and considerable differences in results have been observed between studies. This variability may reflect the high level of heterogeneity of patients in terms of the topography and etiology of the lesion, pain location, intensity, duration and quality, and associated symptoms. The inclusion criteria for these studies were not standardized and very few studies provided quantitative (or even qualitative) information about the sensory deficits associated with pain. Thus, variations in the activation of somatosensory systems in the brain may have been due to differences in the magnitude of sensory deafferentation, independent of pain. The large proportion of patients on analgesic treatments may also have biased the results. Other confounding factors are related to the methods used in these studies. For example, the variable proportion of patients presenting spontaneous continuous pain of variable intensity may have masked the effects of allodynia. In addition, in all but one of the studies (Ducreux et al., 2006), the contralateral side was used as a “control”, even though bilateral alterations have been described following unilateral neurological lesions.

Future studies should include larger number of patients, with standardized and well-characterized neurological lesions. Pain characteristics should be defined precisely and sensory deficits thoroughly evaluated by quantitative sensory testing. For estimation of the effects related to the lesion itself, a comparative control group should be studied, including patients with a similar lesion (and sensory deficit) but no pain. As recently suggested (Kupers and Kehlet, 2006), variability could also be reduced by carrying out longitudinal studies in specific clinical models (e.g., postsurgical pain), making possible the prospective comparisons of individual patients during the transition from acute to chronic neuropathic pain. Finally, it would also be useful to include the systematic assessment of a number of psychological factors, such as anxiety, depression and catastrophizing, which are known to modulate the brain activation associated with pain strongly, in future studies.

Section snippets

Acknowledgment

Xavier Moisset was supported by the INSERM MD–PhD program.

References (65)

  • W. Janig et al.

    Complex regional pain syndrome: mystery explained?

    Lancet Neurol.

    (2003)
  • W. Janig et al.

    Is CRPS I a neuropathic pain syndrome?

    Pain

    (2006)
  • A.K. Jones et al.

    Cerebral decreases in opioid receptor binding in patients with central neuropathic pain measured by [11C]diprenorphine binding and PET

    Eur. J. Pain

    (2004)
  • M. Koltzenburg et al.

    Dynamic and static components of mechanical hyperalgesia in human hairy skin

    Pain

    (1992)
  • R. Kupers et al.

    Brain imaging of clinical pain states: a critical review and strategies for future studies

    Lancet Neurol.

    (2006)
  • R.C. Kupers et al.

    Positron emission tomography study of a chronic pain patient successfully treated with somatosensory thalamic stimulation

    Pain

    (2000)
  • J. Lorenz et al.

    A unique representation of heat allodynia in the human brain

    Neuron

    (2002)
  • J. Maarrawi et al.

    Differential brain opioid receptor availability in central and peripheral neuropathic pain

    Pain

    (2007)
  • C. Maihöfner et al.

    Brain processing during mechanical hyperalgesia in complex regional pain syndrome: a functional MRI study

    Pain

    (2005)
  • M.J. Millan

    Descending control of pain

    Prog. Neurobiol.

    (2002)
  • H. Olausson et al.

    Central pain in a hemispherectomized patient

    Eur. J. Pain

    (2001)
  • P. Petrovic et al.

    A PET activation study of dynamic mechanical allodynia in patients with mononeuropathy

    Pain

    (1999)
  • R. Peyron et al.

    Electrical stimulation of precentral cortical area in the treatment of central pain: electrophysiological and PET study

    Pain

    (1995)
  • R. Peyron et al.

    Functional imaging of brain responses to pain. A review and meta-analysis

    Neurophysiol. Clin.

    (2000)
  • R. Peyron et al.

    Motor cortex stimulation in neuropathic pain. Correlations between analgesic effect and hemodynamic changes in the brain. A PET study

    NeuroImage

    (2007)
  • B. Pleger et al.

    Patterns of cortical reorganization parallel impaired tactile discrimination and pain intensity in complex regional pain syndrome

    NeuroImage

    (2006)
  • P. Rainville

    Brain mechanisms of pain affect and pain modulation

    Curr. Opin. Neurobiol.

    (2002)
  • P. Schweinhardt et al.

    An fMRI study of cerebral processing of brush-evoked allodynia in neuropathic pain patients

    NeuroImage

    (2006)
  • I. Tracey

    Nociceptive processing in the human brain

    Curr. Opin. Neurobiol.

    (2005)
  • R.D. Treede et al.

    The cortical representation of pain

    Pain

    (1999)
  • C. Villemure et al.

    Cognitive modulation of pain: how do attention and emotion influence pain processing?

    Pain

    (2002)
  • F. Willoch et al.

    Central poststroke pain and reduced opioid receptor binding within pain processing circuitries: a [11C]diprenorphine PET study

    Pain

    (2004)
  • Cited by (302)

    View all citing articles on Scopus
    View full text