Electrical stimulation of motor cortex for pain control: a combined PET-scan and electrophysiological study

Pain 83 (1999) 259±273 www.elsevier.nl/locate/pain Electrical stimulation of motor cortex for pain control: a combined PETscan and electrophysiological study q L. GarcõÂa-Larrea a,b,*, R. Peyron b,c, P. Mertens d, M.C. Gregoire b, F. Lavenne b, D. Le Bars b, P. Convers c, F. MauguieÁre a, M. Sindou d, B. Laurent c b a Functional Neurology Unit, UPRES-EA 1880, Claude Bernard University, and Institut FeÂdeÂratif de Neurosciences of Lyon (IFNL), Lyon, France Human Neurophysiology Laboratory, CERMEP (Positron Emission Tomography Center), and Institut FeÂdeÂratif de Neurosciences of Lyon (IFNL), Lyon, France c Pain Center and Neurology Department, St Etienne University Hospital, St Etienne, France d Neurosurgery Department, Neurological Hospital, and Institut FeÂdeÂratif de Neurosciences of Lyon (IFNL), Lyon, France Received 22 October 1998; received in revised form 6 April 1999; accepted 24 May 1999 Abstract Although electrical stimulation of the precentral gyrus (MCS) is emerging as a promising technique for pain control, its mechanisms of action remain obscure, and its application largely empirical. Using positron emission tomography (PET) we studied regional changes in cerebral ¯ood ¯ow (rCBF) in 10 patients undergoing motor cortex stimulation for pain control, seven of whom also underwent somatosensory evoked potentials and nociceptive spinal re¯ex recordings. The most signi®cant MCS-related increase in rCBF concerned the ventrallateral thalamus, probably re¯ecting cortico-thalamic connections from motor areas. CBF increases were also observed in medial thalamus, anterior cingulate/orbitofrontal cortex, anterior insula and upper brainstem; conversely, no signi®cant CBF changes appeared in motor areas beneath the stimulating electrode. Somatosensory evoked potentials from SI remained stable during MCS, and no rCBF changes were observed in somatosensory cortex during the procedure. Our results suggest that descending axons, rather than apical dendrites, are primarily activated by MCS, and highlight the thalamus as the key structure mediating functional MCS effects. A model of MCS action is proposed, whereby activation of thalamic nuclei directly connected with motor and premotor cortices would entail a cascade of synaptic events in painrelated structures receiving afferents from these nuclei, including the medial thalamus, anterior cingulate and upper brainstem. MCS could in¯uence the affective-emotional component of chronic pain by way of cingulate/orbitofrontal activation, and lead to descending inhibition of pain impulses by activation of the brainstem, also suggested by attenuation of spinal ¯exion re¯exes. In contrast, the hypothesis of somatosensory cortex activation by MCS could not be con®rmed by our results. q 1999 International Association for the Study of Pain. Published by Elsevier Science B.V. Keywords: Pain control; Electrical stimulation; PET-scan; Electrophysiological study; Nociceptive re¯exes 1. Introduction Experimental studies in animals have repeatedly demonstrated the strong inhibitory in¯uences that electrical stimulation of the nervous system can exert on pain transmission (e.g. Melzack and Wall, 1965; Handwerker et al., 1975; Lindblom et al., 1977; Gerhart et al. 1983; Carstens and Campbell, 1988), thus prompting the use of neurostimulaq Part of these results were presented to the World Congress of Stereotactic and Functional Neurosurgery (Lyon, 1997), and published with the Proceedings of the Congress. * Corresponding author. Tel.: 133-4-72-68-86-00; fax: 133-4-72-6886-10. E-mail address: [email protected] (L. GarcõÂa-Larrea) tion strategies for the relief of chronic pain in humans. The neural targets of stimulation procedures have been mostly the sensory pathways mediating transmission of nonnoxious information (e.g. large afferent peripheral ®bers, spinal dorsal columns and thalamic sensory nuclei) and to a lesser extent brainstem structures exerting descending antinociceptive in¯uences (reviews in Gybels et al., 1995; Jessurun et al., 1996; Holsheimer, 1997). Although stimulation of central motor ®bers was also shown to inhibit afferent transmission in the dorsal horn (Lindblom and Ottosson, 1957; Andersen et al., 1962) and to produce analgesic effects in man (Adams et al., 1974) the use of motor cortex stimulation for pain control was documented only during the 1990s by Tsubokawa et al. (1991, 1993a). 0304-3959/99/$20.00 q 1999 International Association for the Study of Pain. Published by Elsevier Science B.V. PII: S 0304-395 9(99)00114-1 260 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 These authors described a procedure for chronic motor cortex stimulation (MCS) which, in preliminary studies, provided satisfactory control of central post-stroke pain with a better risk/bene®t ratio than stimulation of deeper structures such as the thalamus. Since then, the use of MCS for analgesic purposes is being increasingly used for both poststroke pain (Herregodts et al., 1995; Katayama et al., 1998) and other conditions including trigeminal neuropathic pain (Meyerson et al., 1993) and central pain after lateral medullary infarct (Wallenberg's syndrome) (Katayama et al., 1994). In spite of the encouraging results of motor cortex stimulation, the reasons accounting for its analgesic ef®cacy have not yet been elucidated, and its clinical application remains, therefore, largely empirical. Tsubokawa et al. (1993a) have suggested that, in cases of thalamic pain, MCS is superior to thalamic stimulation due to its more rostral level of application, which ensures activation of preserved functional zones acting upon deafferented structures. In these authors view, precentral gyrus stimulation could entail analgesia through secondary activation of non-nociceptive neurons in the sensory cortex, via backward excitation of axons connecting somatosensory and motor areas. Thus, precentral gyrus stimulation would activate, through cortico-cortical ®bers, non-nociceptive somatosensory neurons which in turn would inhibit hyperactive nociceptive units within SI (Tsubokawa et al., 1993a). Although this hypothesis remains speculative and has not received objective con®rmation in humans, it is supported by the ®nding of histochemical changes within the primary sensory cortex of rats subject to chronic motor stimulation (Tsubokawa et al., 1993b). On the other hand, experimental data also point to the thalamus as a possible target of MCS, since this procedure (unlike SI stimulation) was able to attenuate thalamic hyperactivity after spinothalamic transection in cats (Hirayama et al., 1990; Tsubokawa et al., 1991). Whatever the precise mechanisms underlying MCS clinical effects, these are likely to be mediated by regional changes in synaptic activity, which should, in turn, be re¯ected by corresponding changes in cerebral blood ¯ow (CBF) (Sokoloff et al., 1991). Assessment of in vivo CBF changes under different therapeutic or experimental conditions can be performed with positron emission tomography (PET), using 15O-labelled water as the tracing compound. In a previous study we used PET to assess the regional changes in CBF of two patients undergoing MCS for central intractable pain (Peyron et al., 1995). CBF increased from 6 to 16% during MCS in the thalamus ipsilateral to MCS, as well as in orbito-frontal and cingulate gyri and in the upper brainstem. These changes appeared to be related to MCS since they were reversible after discontinuation of the procedure; however, their relationship with clinical effects remained dubious, since only one of the two patients experienced a good analgesic effect from MCS, while both of them exhibited similar CBF changes (Peyron et al., 1995). This suggests that at least some of the observed modi®cations might not be directly related to the analgesic ef®cacy of the procedure. In order to gain insight into the mechanisms of pain relief by MCS, to extend our previous observations and to test them for statistical signi®cance, we studied CBF regional changes induced by this procedure in ten consecutive patients with intractable pain, for whom we have now a minimal follow-up of 36 months. To complete PET data and provide a comprehensive view of the effects of MCS, we also recorded polysynaptic spinal (nociceptive) re¯exes and somatosensory evoked responses, which allowed to assess the in¯uence of the procedure on, respectively, spinal and cortical excitability to external inputs, and gather evidence for a possible implication of descending inhibitory mechanisms in the analgesic action of MCS. 2. Patients and methods 2.1. Patients Ten consecutive patients (5 women) implanted with a MCS device were investigated. Their age ranged between 30 and 55 years, and they all suffered from long-lasting, unilateral neuropathic pain that had resisted pharmacological therapy for more than 2 years. Pain was secondary to vascular lesions of the central nervous system in seven cases (six thalamo-cortical, one mesencephalic), and to brachial plexus avulsion in the three remaining patients. Patients' clinical data are summarized in Table 1. 2.2. Surgical procedure After general anaesthesia and craniotomy, somatosensory evoked potentials (SEPs) were recorded by epidural electrodes (see below). Then, the stimulating electrode (`Resume' Medtronic w 4-pole electrode, 10 mm interelectrode distance) was placed epidurally, anterior and parallel to the central sulcus, over the convexity of the pre-central gyrus (Fig. 1). In four patients with pain in both upper and lower limbs a second electrode was placed subdurally over the medial aspect of the pre-central cortex. The electrode wires were then tunneled under the skin down to the lateral neck and connected to the system antenna (a passive receiver activated by external radiofrequency transmitter), in subclavicular position. 2.3. Stimulation parameters adjustment and clinical assessment of pain relief Parameters of stimulation were not standardized, neither in the immediate postimplant period nor later in the patients follow-up. A number of combinations of frequency, intensity and electrode polarities were tested during the ®rst weeks after surgery, in order to select the parameters providing the best analgesic effect. These parameters were kept constant and used continuously during 2±6 months, after 261 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 Table 1 Patients' clinical data a Sex/age of patient Etiology of pain Distribution of pain Duration of pain (years) Associated conditions Unsuccessful therapies F/71 (Chas...) Stroke (ischemic) (parietal) Stroke (haemorr) (capsulo-thalamic) Plexus avulsion Right hemibody 2.5 epilepsy, AE NSAID, Tr, Bzd, 90 Satis®ed Right upper and lower limbs Left phantom limb (C5±C7) Right tipper limb and trunk (C5±T3) Left hemibody 3.5 Dystonia, AE, Op NSAID, Tr, Bzd. 100 Satis®ed 80 Satis®ed 50 Rather satis®ed F/39 (Mor...) M/44 (Bou...) M/46 (Ant...) M/25 (Mar...) Stroke (haemorr) (capsulo-thalamic) Stroke (ischemic) (capsulo-thalamic) Stroke (haemorr) (capsulo-thalamic) Stroke (ischemic) (parietal) Stroke (ischemic) (mesencephalic) Plexus avulsion M/53 (Aou...) Plexus avulsion M/54 (Ruh...) F/50 (Via...) F/48 (Sch...) F/54 (God...) % Pain relief under MCS b Global estimation of ef®cacy 2 Epilepsy, AE Tr; NSAID DREZ-tomy NSAID, Tr; Bzd, 3 Neuropathy, AE NSAID, Ti; Bzd, 40 Rather unsatis®ed Left hemibody 3 Depression, AE NSAID, Tr, Bzd, 30 Unsatis®ed c Left hemibody 6 NSAID, Tt; Bzd 20 Unsatis®ed Left lower limb 4 Porencephalic cyst Breast cancer NSAID, Tr 20 Unsatis®ed NSAID, Tr, Bzd, , 20 AE, Drez-tomy NSAID, Ti; Bzd ?? Nlp, Drez-tomy, ALC Unsatis®ed Left phantom limb (C5 ±C7) Left C6±T1 .10 10 16 Depression, amputation Neuropathy Unsatis®ed d a Abbreviations: Ti; tricyclic antidepressants; Bzd, benzodiazepines; Nlp, neuroleptics; AE, antiepileptics; NSAID, non-steroidal antlin¯ammatory drugs, ALC, anterolateral cordotomy; DREZ-tomy, microsurgical lesion at the dorsal root entry zone. b Data obtained the same week of PET investigation. c Signi®cant reduction of evoked pain, but no effect on spontaneous pain. d Severe cognitive impairment, patient unable to provide VAS estimates, declared himself unsatis®ed but refused to discontinuate MCS. which a new clinical assessment was performed and parameters changed if needed. The temporal pattern of MCS stimulation varied among patients, from multiple short stimulation periods (10±15 min) every 2 h, to almost continuous stimulation during daytime in one case. However, MCS was discontinued in all patients during bedtime. PET scans were obtained during a period of stable parameters, and after 18 h of MCS arrest. Clinical ef®cacy of MCS was assessed in each case by two investigators, on the basis of the patients reports obtained during MCS activation. Patients were asked to estimate the analgesic effects of the procedure in three ways: Firstly, they were asked to report a `percentage of pain decrease' during MCS ranging between 0 (no relief at all) to 100 (total relief). Secondly, they marked on a 100 mm Visual Analog Scale (VAS) the average pain level during periods with and without (12±24 h) MCS. Lastly, they expressed verbally their feeling in simple colloquial terms, namely satis®ed, rather satis®ed, rather unsatis®ed or unsatis®ed. Percentage estimations of pain and pain relief obtained by each investigator were averaged and are presented in Table 1. No disagreement between investigators was observed in the qualitative (verbal descriptors) estimation of pain relief, also reported in Table 1. Since subjective estimations could change over time, the values reported here correspond to those obtained during the week preceding the PET-scan session. 2.4. PET investigations The patient was positioned in the scanner so that the slices were parallel to the bicommissural line (AC-PC), as estimated by external landmarks (infraorbital-meatus and inion-nasion lines (Merlet et al., 1999). Patient movements were constrained by means of a personalized polyurethane foam helmet. As no arterial catheter was used, the reconstructed images were not converted to rCBF; however, on the tested range, blood ¯ow has been shown to be linearly related to the observed activity (Herscovitch et al., 1983). Therefore, responses reported here are changes in linear radioactive distribution but will be referred to as changes in regional blood ¯ow. During the PET-scan session patients had their eyes covered and their ears plugged to maximize visual and auditory deprivation. PET assessment of rCBF was performed 2±12 months following surgery, using 15O-labelled water as the tracing compound and a time-of-¯ight PET camera (TTV03, LETI, France). This latter generated seven 9 mm-thick slices separated by a 3 mm gap. Images were reconstructed with a Hanning ®lter providing a spatial resolution of 7 mm at the center of the ®eld of view (Trebossen and Mazoyer, 1991). Attenuation correction was performed using measured coef®cients derived from a 20-min transmission acquisition. Four CBF scans were recorded in each patient after injec- 262 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 Fig. 1. Postoperative coronal cranial radiograph showing the 4-pole epidural electrode over the convexity of the right motor cortex. tion of a 45 mCi dose of 15O-labelled water in the antecubital vein on the non-painful side. The ®eld of view was adjusted to study brain regions from the upper brainstem (z coordinates in Talairach and Tournoux (1988) atlas: 220 mm) to the sensorimotor cortex (z coordinates: 60 mm). PET investigations were performed before, during and after a 20-min motor cortex stimulation session. The sequence included four conditions organized as follows: ² ®rst control (condition A) : CBF assessed in basal state, 15 min before MCS (stimulator turned off for 18 h); ² motor cortex stimulation (MCS, conditions B and C): Two consecutive scans performed respectively after 5 and 20 min of continuous MCS; ² Second control (condition D): a fourth scan obtained 30 min after MCS discontinuation. Pain ratings were assessed with a VAS scale immediately after each scan. The standardised question was: ``please rate in this scale the pain intensity during the preceding minutes''. 2.5. PET data analysis Data analysis was performed using the Statistical Para- metric Map (SPM95) software developed at the Functional Imaging Laboratory of London. Seven out of ten patients were being stimulated over the right hemisphere; the images of the three patients with left sided MCS were therefore ¯ipped in order to homogenize the side of stimulation to the right before pooling interindividual data. Patient movement between scans was corrected by a realignment procedure. Then all data were spatially normalized (Friston et al., 1991a) according to the stereotaxic space described by Talairach and Tournoux (1988) so as to homogenize brain morphology and atrophy and to allow interindividual pooling. Images were smoothed with a gaussian ®lter (FHWM 12 mm) to account for anatomo-functional variability and a proportional scaling was performed to normalize images and remove the effect of differences in global activity (Friston et al., 1990). Control and MCS conditions were compared by SPM95, through ANOVA followed by t-test (Worsley et al., 1992; Friston et al., 1991b). The analysis was based on the estimation of the covariates introduced in the general linear model for each and every pixel (Friston et al., 1991b). The resulting set of voxel values (t-map) was then transformed to the unit normal distribution (Z-map). The foci were then characterized in terms of peak height and number of pixels above a given threshold (Worsley et al., 1992). Pairwise comparisons included `between control', `between MCS', and `control vs. MCS' comparisons. The statistical signi®cance of CBF pairwise comparisons was estimated using a two-step procedure based on that of Decety et al. (1994). Thus, two thresholds for signi®cance of CBF changes were set as a function of Z-scores, respectively at Z $ 3:5 and Z $ 4. The ®rst (lower) threshold corresponded to a relatively `liberal' analysis, and therefore subject to possible false positives with overestimation of effects, while the second, much more stringent and including correction for multiple comparisons, implied the possibility of false negatives occluding physiologically signi®cant effects. The two analyses were conducted for each comparison, so as to get a meaningful picture of the location and extent of CBF effects linked to MCS activation. On individual inspection of PET-scan results it appeared that patient #7, with a cortico-subcortical infarct due to middle-cerebral artery thrombosis presented also a very extensive porencephalic cyst in the parietal lobe (Table 1) entailing a decrease of CBF comprising almost the whole right hemisphere gray matter. This patient could not be entered into group analysis without entailing a bias in the resulting pool, and therefore the remaining nine patients were analysed using statistical parametric mapping (SPM). 2.6. Electrophysiological investigations 2.6.1. Somatosensory evoked potentials (SEPs) The purpose of SEP recordings was, on the one hand, to assess possible changes in cortical somatosensory responses L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 during MCS, which would support one of the hypotheses put forward by previous authors (see Section 1). On the other hand, epidural SEP recordings were also obtained intraoperatively to determine the position of the central sulcus before implantation of the stimulating electrodes (Wood et al., 1988). For pre- and postoperative scalp-recorded SEPs we stimulated the median and posterior tibial nerves at 2 Hz, with intensity slightly above the motor threshold for abductor pollicis brevis (median) or abductor hallucis (posterior tibial). Cortical responses were recorded with Ag-AgCl or tin electrodes attached at least to the P3, P4, Cz, F3 and F4 electrode positions of the International 10-20 system (Jasper, 1958), using an earlobe reference and bandpass ®lters of 1±1500 Hz (23 dB, 12 dB/octave). Peripheral and spinal responses were recorded using Erb's point and posterior cervical electrodes in case of upper limb stimulation, and popliteal and lumbar electrodes for lower limb SEPs, with identical ®lter settings as above. During the postoperative session, SEPs were obtained whenever possible with the MCS device in `on' and `off' positions to assess any systematic variation in cortical somatosensory excitability. Intraoperative location of the central sulcus with SEPs was done using the phase-reversal of the SEP wave N20/ P20 in epidural recordings, following the procedure described by Wood et al. (1988). The primary somatosensory response to median nerve stimulation was recorded with platinum electrodes placed at various putative preand post-central locations over the dura. Successive recordings permitted determining the sites of maximal cortical primary response, as well as the line of optimal phase-reversal which delineated the position of the central sulcus. 2.6.2. Nociceptive re¯exes Flexion nociceptive re¯exes (RIII re¯exes, Willer, 1977, 1984) were obtained in the seven patients with central poststroke pain (in the three other patients complete peripheral deafferentation prevented the recording of any re¯ex response). The electromyographic (EMG) response was recorded with surface electrodes placed over the biceps femoris (lower limb) or the ¯exor carpi ulnaris (upper limb) after electrical stimulation of, respectively, the ipsilateral sural or ulnar nerves. The eliciting stimulus was a train of ®ve shocks of 1 ms each, delivered at 300 Hz (see Willer, 1977, 1984; and GarcõÂa-Larrea et al., 1989). After estimating the threshold of the nociceptive re¯ex by the method of limits, the stimulation intensity was kept at about 1.5 times this threshold. Then, several consecutive responses of steady intensity were obtained following a procedure similar to that described for PET investigation, comprising (a) a control condition before MCS; (b) a test condition during MCS, and (c) a second control situation following MCS discontinuation. A minimum of six consecutive re¯exes were averaged in each of the three experimental conditions. After full-wave recti®cation of the EMG response, the area under the curve was measured and 263 compared between the control and MCS conditions by a ttest. Whenever a signi®cant depression of nociceptive responses was observed during MCS the whole 3-step sequence was repeated to ensure reproducibility of the results. 3. Results 3.1. Analgesic effects of MCS (Table 1) Four patients (40) were experiencing satisfactory or very satisfactory clinical effects at the time of investigation. VAS levels of self-estimated analgesia were, respectively, of 100, 90, 80 and 50. Three of these patients had post-stroke central pain, and the remaining one had pain after plexus avulsion. One further patient with post-stroke pain estimated pain relief during MCS as consistent and of 40, but declared that the results did not meet his expectancies and was, therefore, classed as rather unsatis®ed. Four other patients reported 10±30 VAS estimations of pain decrease, which were clearly insuf®cient for clinical satisfaction. Interestingly, one of these patients reported a good effect of MCS on provoked pain only (hyperalgesia and allodynia), but not on the spontaneous component of her pain, which was predominant. As to the remaining patient, with severe cognitive impairment, he was unable to provide consistent VAS estimates of any aspect of his pain, declared himself unsatis®ed but yet refused to discontinue the procedure. For a complete description of clinical characteristics and results see Table 1. Pain ratings were also obtained during the PET scan sessions. These were of 4.8 ^ 2.6 during the ®rst control condition, 4.3 ^ 2.9 and 3.96 ^ 3, respectively, in the 1st and 2nd stimulation scans, and 3.69 ^ 2.8 in the second control condition. In spite of a trend to pain decrease from the ®rst to the last condition, differences were not signi®cant (ANOVA, F 3; 8† ˆ 3:71, NS). 3.2. Somatosensory responses (SEPs) 3.2.1. Intraoperative SEPs Absence of central SEP responses in patients with plexus avulsion obviously precluded intraoperative monitoring in them. Conversely, cortical somatosensory responses could be obtained intraoperatively with epidural electrodes in all seven cases with central lesions, this in spite of various degrees of deafferentation that obliterated scalp responses in three of them. Epidural SEP responses to stimulation of the median nerve in the affected side were used in these seven patients to localize the central sulcus. Electrophysiological localization of the motor cortex in the brain convexity was considered successful in all of them, as none experienced paresthesiae or other sensory phenomena in upper limbs during MCS activation. Localization of the interhemispheric portion of the sulcus presented more dif®culties due to the lack of a clear phase-reversal of the SEP 264 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 of MCS induced systematically a high-frequency sinusoidal artifact that contaminated scalp responses, and had to be eliminated with digital low-pass ®lters (zero-phase shift) applied off-line. Parietal somatosensory responses up to 50 ms post-stimulus did not exhibit any signi®cant change in amplitude, latency or topography in relation to motor cortex stimulation (Fig. 2). Fig. 2. Somatosensory evoked potentials (SEPs) obtained in patient no. 1. Superimposed traces (bottom) correspond to parietal recordings ipsilateral to MCS, respectively, with the stimulating device `on' and `off'. The `N20' response (arrow) is the primary parietal response generated in SI area 3b. Maps at the top depict the scalp distribution of the N20 response with the MCS in `on' and `off' positions. No signi®cant difference in SEP traces or maps was observed in relation with MCS activity. This patient was obtaining a very satisfactory (90%) pain relieving effect from MCS. response to lower limb stimulation (Wood et al., 1988). Of the four patients with lower limb pain in whom a second (interhemispheric) stimulating plate was placed, one experienced MCS-related paresthesiae in the lower limb when one particular electrode combination (out of six tested possibilities) was applied, thus suggesting that the stimulator impinged on the posterior bank of the central sulcus. 3.2.2. Scalp-recorded SEPs during MCS (Fig. 2) Although the seven patients with central lesions had sizable epidural SEPs during intraoperative monitoring, only four of them retained scalp-recorded SEPs of enough amplitude to permit assessment of MCS effects. Activation 3.2.3. Changes in cerebral blood ¯ow during MCS The size, anatomical location and magnitude of CBF changes during MCS are summarised in Table 2. Comparison of the two stimulation conditions did not yield any signi®cant modi®cation of regional CBF. Conversely, comparison of the control vs. the stimulation scans demonstrated signi®cant and anatomically restricted CBF changes during MCS, illustrated in Fig. 3 and Fig. 4. The low-threshold analysis (Z-score $3.5) of our twostep procedure yielded four regions of signi®cant CBF increase, corresponding respectively to (a) the thalamus; (b) the anterior cingulate/orbitofrontal area; (c) a region comprising the insula and descending towards the medial temporal lobe, and (d) the subthalamic and upper brainstem region. Both thalamic, subthalamic and brainstem effects were clearly ipsilateral to the site of MCS.Cingulate activation reached the midline but clearly predominated in the hemisphere contralateral to MCS. Finally, the insular and medial temporal lobe effects were exclusively contralateral to the side of MCS (see Fig. 3, bottom). The second (high-threshold) step of the analysis (Z-score $4) restricted spatially the above results, and limited the anatomical region of signi®cant CBF increase to a small area of the lateral thalamus ipsilateral to MCS. Projection of this region onto the standardized anatomical atlas of Talairach and Tournoux (1988) was centered on the nucleus ventralis lateralis (VL) (coordinates x ˆ 014, y ˆ 28 and z ˆ 04) with extensions toward the anterior thalamus (region of the Ventral Anterior (VA) nucleus) and the subthalamic region. Surprisingly, the ventro-postero-lateral thalamus (somatosensory thalamus) was clearly outside the region of increased blood ¯ow, this in both the high- and the low-threshold analyses (Fig. 3, top). Table 2 Coordinates of regions where maximal rCBF effects were observed during motor cortex stimulation a Coordinates of maxima (x, y, z; Talairach atlas) Anatomical region Type of rCBF effect Z-score 14, 28, 4 16, 26, 28 10, 212, 8 228, 2, 16, 230, 26, 12 24, 32, 8 6, 216, 28 30, 268, 8, 240, 66, 8 Ventrolateral thalamus (i) Subthalamic area (i) Medial thalamus (i) Anterior insula (c) Anterior cingulate (BA 32) (c?) Upper brainstem (i) Bi-occipital (i,c) (BA18119) Increase Increase Increase Increase Increase Increase Decrease 4.22 4.07 3.67 3.59 3.55 3.35 4.32 a Note that several of these maxima were grouped in single regions of statistical signi®cance, as shown in Figs. 2 and 3. This was the case of medial and lateral thalamus, subthalamic area and upper brainstem (see Fig. 3, bottom). i, ipsilateral to MCS; c, contralateral to MCS. L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 265 Fig. 3. Coronal, axial and sagittal MRI brain sections showing regions with signi®cant blood ¯ow increase during MCS. MRIs were spatially normalized (Friston et al., 1991a) according to the stereotaxic space described by Talairach and Tournoux (1988). The region showing maximally signi®cant ¯ow increases corresponded to ventral-lateral thalamus, extending toward the subthalamic area (high-threshold analysis, top of the ®gure). In addition to the above area, the lower-threshold analysis (z . 3:5, bottom) highlighted a restricted network of structures ipsilateral and contralateral to MCS, including the medial thalamus, anterior cingulate, insula/MTL and brainstem regions. The temporal dynamics of rCBF changes in the different regions showing MCS-related rCBF enhancement are shown in Fig. 5. In all of them was observed an abrupt CBF increase during the ®rst scan under MCS (5 0 after onset), which remained stable during the 2nd scan (20 0 after MCS onset). These effects were reversible during the second control condition (30 0 after stimulation offset) in all sites except in anterior cingulate, where rCBF had not yet reverted to pre-stimulation values 30 min after MCS discontinuation. This aspect of long-lasting persistence of CBF 266 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 Fig. 4. Consecutive axial MRI sections normalized according to the Talairach space, showing regions with signi®cant (z . 3:5) CBF increases during motor cortex stimulation. In addition to the structures depicted in Fig. 3, note the absence of any signi®cant CBF change in the right motor or somatosensory cortices directly underlying the stimulator (see details in text). changes in anterior cingulate/orbitofrontal structures was also suggested by SPM comparisons between the pre- and post-stimulation control conditions. Two spots of increased rCBF during the post-MCS control condition, as compared with the pre-MCS baseline, appeared in the right (coordinates 22, 40, 20) and left (28, 40, 0) anterior cingulate/ orbitofrontal boundaries (z ˆ 3:97 and 3.22, respectively), suggesting a remnant effect on CBF within these structures still 30 min after MCS offset. No signi®cant rCBF change related to MCS was observed within the primary sensory or motor cortices. In particular, no signi®cant or subsigni®cant change was noted within the cortical motor area lying directly beneath the stimulating electrodes (precentral region, BA 4). Conversely, a signi®cant decrease of rCBF during cortical stimulation was noted in both occipital regions, affecting visual areas 18 and 19 but sparing the calcarine sulcus and primary visual areas. This effect (z ˆ 4:5) was totally reversible with MCS discontinuation. On the basis of the above results and previous data (Peyron et al., 1995), regional normalized activity was measured in each subject by means of ellipsoidal regions of interest (ROIs) placed over the lateral thalamus and the anterior perigenual cingulate gyrus (BA 32). The same procedure could not be applied to other regions, such as anterior insula and upper brainstem, due to their irregular shape in the three space axes. The results issued from ROI measurements in patients with good or very good clinical effect of MCS (pain relief $80) and those with poor or very poor ef®cacy of the procedure (pain relief ,30) are shown in Fig. 6. CBF increase within the lateral thalamus during MCS was not different in the groups of patients with good or poor analgesic ef®cacy. Blood ¯ow increase in the anterior cingulate gyrus during MCS was signi®cantly higher (2tailed t-test) in patients with good/very good analgesic ef®cacy than in the others. Also, the temporal dynamics of cingulate changes in the two groups of patients were markedly different. 3.2.4. Nociceptive re¯exes during motor cortex stimulation (Fig. 7) Signi®cant modulation of spinal nociceptive re¯exes was observed during MCS in three out of the seven patients in whom it was studied, while the re¯ex remained unchanged in the other four. Modi®cation of nociceptive re¯exes corresponded in every case to attenuation of the response during cortical stimulation (see example in Fig. 7); in no instance was an enhancement of nociceptive re¯exes observed during MCS. Two of the three patients with MCS-related re¯ex attenuation were experiencing a good/very good clinical pain relief from the procedure (##1 and 2, see Table 1), while the remaining patient (#6) reported a selective decrease in provoked pain during MCS (.60), but unsatisfactory effect of the procedure on spontaneous pain (30, see L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 267 Fig. 5. Temporal dynamics of rCBF changes in the regions with MCS-related rCBF increase. x axis, experimental conditions; y axis, normalised radioactivity within regions studied. In all of them was observed an abrupt CBF increase during the ®rst MCS scan (5 min after onset), remaining rather stable during the 2nd scan (20 min after MCS onset). These effects were reversible during the second control condition (30 0 after stimulation offset) except in anterior cingulate, where rCBF had not yet reverted to pre-stimulation values 30 min after MCS discontinuation. Table 1). None of the four patients whose nociceptive re¯exes remained unmodi®ed by MCS was satis®ed with the clinical effect of neurostimulation. 4. Discussion 4.1. Introductory remarks Motor cortex stimulation increased CBF with different intensity and temporal dynamics in a restricted set of cortical and subcortical regions. On statistical grounds the most signi®cant changes were those observed in the ventrolateral thalamus ipsilateral to stimulation (Fig. 3, top), but other loci of signi®cant rCBF increase included the rostral cingulate gyrus (BA 32), medial thalamus, upper brainstem and contralateral insula (Fig. 3, bottom). Conversely, no signi®cant change in blood ¯ow was evidenced at the site of cortical stimulation, i.e. in motor cortex. Since changes in blood ¯ow are considered to re¯ect regional changes in synaptic function (Sokoloff et al., 1991) the lack of rCBF increase at the site of cortical stimulation suggests that MCS did not modify the cortical synaptic activ- ity beneath the stimulator, and therefore that the structures ®rst activated by the procedure were not apical dendrites but rather subcortical ®bers. This conclusion is consistent with our previous PET-scan ®ndings in two patients undergoing MCS (Peyron et al., 1995), and also in line with the conclusions of Tsubokawa et al. (1993a) that cortical activation beneath the surface electrode is probably not responsible for the effects of MCS, the excitation of white matter axons being a more likely mechanism. This pattern of activation, orthogonal to the cortical layers, is also in accordance with the inter-electrode distance of our stimulating system, as it has been shown that surface stimulation of the motor cortex using either monopolar currents or interpolar distances greater than 10 mm activate preferentially descending axons (Amassian et al., 1987; Katayama et al. 1988). 4.2. Thalamic blood ¯ow changes induced by MCS If, as discussed above, thalamic CBF changes were due to direct activation of descending axons from the motor cortex, then the lateral thalamus appears not only as the region with greatest CBF increase, but also the one that was, in chronological terms, ®rst activated by MCS. When projected onto 268 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 Fig. 6. ROI analysis of anterior perigenual cingulate (BA32) and lateral thalamic areas in patients with very good (.80) or very insuf®cient (,20) pain relief. x axis, conditions; y axis, normalised radioactivity within regions studied. While lateral thalamic CBF appears to increase in all patients (albeit to a greater extent in those with good clinical effect), anterior cingulate CBF shows very different trends in patients with good and bad clinical effect. the normalized anatomical atlas (Talairach and Tournoux, 1988) the locus of maximal thalamic rCBF increase was centered on the ventral lateral nucleus (VL) with extension towards the ventral anterior (VA) nucleus and the subthalamic region. Even allowing for the limited spatial resolution of PET-scan relative to MRI, we believe that this restricted localization of thalamic CBF changes is reliable, not only because of very high statistical signi®cance in SPM analysis, but also since it makes very strong sense from the anatomical and physiological points of view. Indeed, the Ventral Lateral and Ventral Anterior nuclei are the only thalamic nuclei directly connected with the motor and premotor cortices. Speci®cally, studies in the monkey have shown that area 4 projects to, and receives axons from VL: while motor cortex projections reach all the subdivisions of the VL nucleus (VLo, VLc and VPLo), ascending ®bers from VL to area 4 come preferentially from the socalled cell-sparse zone, consisting of VLc and VPLo (Carpenter, 1995). Since the lateral thalamic regions with signi®cant blood ¯ow changes also receive connections from the premotor cortex (BA6), immediately anterior to the motor area, excitation of axons from this adjacent cortical region by MCS may be also possible. Interestingly, thalamic regions with such highly signi®cant rCBF increase are not involved in pain integration, but rather in motor control. Thus, subcortical afferents to VL and VA arise from motor-related structures, namely the globus pallidus via the thalamic fasciculus (Kim et al., 1976), the substantia nigra (Ilinsky et al., 1985) and the deep cerebellar nuclei (see Percheron et al., 1996 for a recent review). VL cells receiving cerebellar afferents are somatotopically arranged, as are their projections to motor area 4. The ventral lateral thalamus is, therefore, part of the motor thalamus (Percheron et al., 1996) and functional changes induced by MCS in this region might explain improvement in motor function in patients submitted to this procedure. Indeed, a transient and not quanti®ed relief of spasticity during MCS was ®rst described by Tsubokawa et al. (1993a), and was also observed in some of our patients with stroke-related spasticity, although this point was not speci®cally assessed by EMG. Very recently, Nguyen et al. (1998) have reported long-term improvement of severe action tremor after chronic MCS originally intended for pain relief, and concluded that chronic MCS was an effective treatment for controlling such kind of motor disorder. In view of our PET results, with maximal localization of thalamic blood ¯ow increase within the motor thalamus, it may be hypothesized that MCS may in¯uence thalamo-cerebellar and thalamo-striatal connections, which may in turn be responsible for changes in muscle tone, rigidity and tremor. 4.3. Possible mechanisms of MCS-induced pain relief The above results suggest that the principal modi®cations of thalamic blood ¯ow re¯ected the anatomical connections between the motor cortex and the thalamus, their relation with pain relief not being probably a direct one. A better insight into the possible mechanisms of MCS-induced pain relief was provided by the lower-threshold SPM analysis, which delineated a network of structures where rCBF was also increased by MCS, although with lower signi®cance than in ventrolateral thalamus, and which are known to be involved in pain processing and control (Fig. 3, bottom, and Fig. 4). This set of areas comprised (a) thalamic regions medial to those discussed above, and covering the whole width of the thalamus, (b) a segment of the anterior cingulate region (BA 32) almost at the orbitofrontal boundary, (c) the upper brainstem and (d) the insula and inferomedial temporal lobe contralateral to MCS. The participation of medial thalamus in pain mechanisms is well known from both animal and human studies (e.g. Boivie, 1979; Rinaldi et al., 1991; Jeanmonod et al., 1993) and has been substantiated recently by in vivo functional imaging studies in man. In particular, chronic states of pain have been associated to decreased thalamic blood ¯ow (Di Piero et al., 1991; Cesaro et al., 1991; Hsieh et al., 1995; Iadarola et al., 1995; Pagni and Canavero, 1995), while increase of thalamic ¯ow has been described in response to a number of pain-relieving procedures as disparate as L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 269 Fig. 7. Changes in spinal ¯exion polysynaptic re¯exes during MCS. Patient no. 2. Histograms at the ®gure top are integrated surface values (in nanovolts £ s) of each re¯ex response, depicted over a time axis of 20 min. Grey and black histograms correspond respectively to `on' and `off' MCS periods. The actual EMG re¯ex responses are illustrated under the histograms. Note the decrease of response magnitude when MCS is `on' (white frames), and the progressive increase of the re¯ex response after MCS discontinuation (grey ®lled frames). The patient was obtaining a very satisfactory ( . 90%) pain relieving effect from MCS at the moment of the recording session. anterolateral cordotomy (Di Piero et al., 1991), regional nerve block (Hsieh et al., 1995), relief of spinal cyst (Pagni and Canavero, 1995) and motor cortex stimulation (Peyron et al., 1995). On the other hand, the implication of anterior cingulate and insular cortices in the processing of pain is supported by lesion studies in animals and humans (Foltz and White, 1962; Magnusson and Vaccarino, 1996) as well as by recent functional activation studies using PET (Jones et al., 1991; Talbot et al., 1991; Casey et al., 1994; Hsieh et al., 1996), fMRI (Davis et al., 1995; Dostrovsky et al., 1995) and evoked potentials (Treede et al., 1995; Valeriani et al., 1996; GarcõÂa-Larrea, 1998). Anterior cingulate is connected to medial and anterior thalamus, and thus MCSinduced blood ¯ow increases in these two structures might be functionally related. Direct thalamic input to anterior cingulate comes mainly from the anterior nuclear group and the midline nuclei (Baleydier and MauguieÁre, 1980; Carpenter, 1995; Devinsky et al., 1995), and projections to anterior cingulate and medial orbitofrontal cortices from the VA nucleus have also been described (Carmel, 1970; Carpenter, 1995; Devinsky et al., 1995). Most of these thalamic regions are likely to be included within the area of increased blood ¯ow shown in bottom of Fig. 3 and in Fig. 4, and their projections to anterior cingulate may therefore explain secondary changes of synaptic activity in this latter area. In turn, the anterior cingulate cortex, and notably its perigenual portion, has very pronounced projections to the upper brainstem, notably to the periaqueductal gray of the mesencephalon (Hardy and Leichnetz, 1981; Devinsky et al., 1995), which might explain in part the brainstem rCBF increase detected in our study. However, although it is tempting to describe a medial thalamus-anterior cingulate-brainstem network as a result of MCS, images of blood ¯ow changes in the brainstem did not cover the localization of the periaqueductal gray, where anterior cingulate mainly projects. This, along with the continuity observed between the thalamic and subthalamic (brainstem) activation (Fig. 4), would rather favor the hypothesis of a direct effect on brainstem from thalamic structures. More dif®cult to explain on purely connective grounds was CBF increase in the insular and inferomedial temporal regions (MTL), including the amygdala. Although such regions are also known to be involved in pain mechanisms and processing (Craig et al., 1996; Cesaro and Ollat, 1997; Iadarola et al., 1998), and receive projections from other areas activated by MCS, their rCBF activation concerned exclusively the hemisphere contralateral to the stimulated motor cortex (Fig. 3). A putative explanation of this may be that six patients out of ten had cortico-subcortical lesions involving the hemisphere where MCS was applied. Although one of them was dropped from PET analyses due to extensive decrease of blood ¯ow covering the whole damaged hemisphere (see Section 2), residual hemodynamic basal changes in structures surrounding the lesion may have prevented in the others the statistical demonstration of rCBF changes in the insular/MTL region of the 270 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 affected hemisphere. This assumption is of course speculative and based on electrophysiologic and PET data on normal subjects suggesting bilateral implication of insular and MTL cortices in pain processing (Coghill et al., 1994; Valeriani et al., 1996; Iadarola et al., 1998). The lower statistical signi®cance of rCBF increase in the pain related network, as compared with that observed in the motor thalamus, suggests a greater heterogeneity of the haemodynamic response in these regions than in the ventrolateral thalamus. In turn, this is reminding of the important heterogeneity of clinical outcome concerning pain relief (Table 1) and prompted us to perform a subject-by-subject analysis within two regions of increased rCBF that were accessible to ROI placement. Such analysis, although of limited statistical power due to the small number of patients and the shortcomings of ROI measurement, allowed an individualised approach to CBF changes in subjects with and without clinical relief. The results, which must be considered as exploratory and deserving con®rmation, suggest a possible relationship between blood ¯ow increase in perigenual cingulate gyrus and pain relief by MCS. As shown in Fig. 6, the pro®le of cingulate CBF changes in patients with good or very good (.80) clinical effect was different from that observed in patients totally unsatis®ed, the ¯ow increase during stimulation being signi®cantly higher in the former. Conversely, rCBF changes in lateral thalamus were similar in patients with or without good clinical effect. 4.4. Functional hypotheses on MCS effects This set of results lend support to some functional hypotheses on the pain-relieving effects of motor cortex stimulation. First of all, we did not ®nd evidence favoring the notion of a MCS-related activation of sensory cortex, as previously suggested (Tsubokawa et al., 1993a,b), since all signi®cant rCBF changes detected in our patients occurred far from somatosensory areas. Failure to detect rCBF changes within SI may have been due to technical factors linked to PET technology. Possible rCBF changes may have been diluted because of transient and spatially dispersed neural activation. Also, MCS-related inhibition of local circuits within SI may result in both increase of synaptic activity (through release of inhibitory neurotransmitters) and decrease of such activity (one synapse downstream the inhibited neurons). These coexisting effects without SI might not be resolved spatially with current PET-scan resolution. In view of the complexity of intrinsic circuits within SI this result must be taken with caution, and deserves to be replicated using new PET cameras with enhanced sensitivity to small rCBF changes and increased spatial resolution. We should stress, however, the lack of any signi®cant change in somatosensory evoked potentials (SEPs) during MCS in any of the recorded patients with central lesions. This observation is in accordance with the PET results and also argues against any acute modi®cation of SI excitability during the MCS procedure (Fig. 2). We propose that thalamic activation is a necessary step allowing the pain-relieving activity of MCS. Although the primary and more important thalamic changes appear to concern the motor thalamus, parallel or secondary activation of medial thalamic regions (either by direct connection from motor cortex (Powell and Cowan, 1967) or via the reticularis and VA nuclei) may trigger a cascade of synaptic events in¯uencing activity in other pain-related structures, including the anterior cingulate gyrus, insula/MTL, subthalamic areas and upper brainstem. It is thus conceivable that thalamic functional changes may need to reach a threshold in order to activate other structures; the lack of clinical effect might result in some instances from a failure to attain such threshold. In a very recent study, Katayama et al. (1998) have suggested that a high degree of corticospinal impairment may be a predictor of poor MCS ef®cacy. Our patients are scarcely comparable to those of Katayama's series since only two of them had severe weakness in the painful limbs. On the other hand, three of our patients with plexus lesions had totally preserved corticospinal tracts, yet two of them exhibited a very poor clinical response to MCS. The cortical and brainstem structures activated by MCS may modify the pain experience at different levels. The anterior cingulate cortex appears to be implicated in both the cognitive and the affective integration of pain stimuli (review in Devinsky et al., 1995), and a number of neurosurgical case reports suggest that lesions involving this structure may speci®cally reduce the emotional component of chronic pain (Foltz and White, 1962, and see Devinsky et al., 1995). Among the cingulate subdivisions, the perigenual region bordering orbitofrontal cortex (BA32, i.e. the region with CBF changes in this study) is considered to subserve in part the affective components of pain (Vogt et al., 1992; Devinsky et al., 1995), and generally the processing of emotional stimuli (Lane et al., 1997). Thus, the analgesic effects of MCS might partly derive from a transient blunting of the distressful reaction to pain, rather than to an actual decrease of its intensity. On the other hand, MCS may also lead to descending inhibition of pain impulses at the dorsal horn level. This second putative mechanism of MCSinduced analgesia was already postulated by Adams et al. (1974), who considered that activation of corticospinal axons was able to inhibit nociceptive neurons at the spinal level. In our study, changes in spinal re¯exes during cortical stimulation also supported the implication of descending mechanisms (Fig. 7). Descending inhibition triggered by cortical stimulation might explain the ef®cacy of the procedure upon the `evoked components' of pain (i.e. allodynia, hyperalgesia and hyperpathia). This notion is further supported by data from patient #6 (Table 1), who, although unsatis®ed with the results of MCS on spontaneous pain, experienced a decrease of provoked pain during MCS, which was concomitant to a decrease of spinal re¯exes. Early investigators noted that stimulation of central motor ®bers may inhibit afferent transmission in the dorsal horn (Lindblom and Ottosson, 1957; Andersen et al., 1962), an L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 effect that might be at the basis of ¯exion re¯ex attenuation in our patients. 4.5. Non-speci®c rCBF changes during MCS Finally, a bilateral decrease of perfusion was observed in both occipital areas during MCS, involving BA 18/19 bilaterally but sparing primary visual cortex (BA17), and thus unrelated to visual inputs during the PET recording. Bilateral decrease of perfusion in occipital cortex is a common ®nding in PET activation studies that use non-visual stimuli, and has been described during vestibular stimulation (Wenzel et al., 1996), semantic tasks (Warburton et al., 1996), selective attention (Carter et al., 1995) and allodynic stimulation (Peyron et al., 1998). Since visual association cortex has a privileged share of CBF distribution during control conditions (at rest), in relation to the richness of visual imagery, diversion of ¯ow from the occipital areas to regions functionally activated may occur during any kind of extra-visual stimulation, thus explaining the relative occipital hypoperfusion during activation relative to rest conditions. 5. Conclusion Changes in regional CBF assessed by PET during motor cortex stimulation highlighted the thalamus as the key structure mediating functional effects. The primary site of CBF increase was the motor thalamus (VL, VA) suggesting a direct effect mediated by connections from motor (and possibly premotor) areas. Either in parallel or secondarily, a restricted network of anatomical regions including medial thalamic and subthalamic areas, insula, brainstem and anterior cingulate also underwent increases in blood ¯ow during MCS. These regions are implicated in pain process and control, and could therefore participate as mediators of MCS clinical effects. Changes in polysynaptic spinal re¯exes during MCS further suggest that descending inhibitory mechanisms might be set up by this neurosurgical procedure, and participate to its analgesic effects. Of course, this model is not exhaustive and does not exclude other possible mechanisms of action. Further studies using last generation PET or fMRI cameras with greater sensitivity and spatial resolution will help to delineate the spectrum of MCS-related neurophysiological activation, and therefore to optimize both patient selection and procedure ef®cacy. References Adams JE, Hosobuchi Y, Fields HL. Stimulation of internal capsule for relief of chronic pain. J Neurosurg 1974;41:740±744. Amassian VE, Stewart M, Quirk GJ, Rosenthal JL. Physiological basis of motor effects of transient stimulus to cerebral cortex. Neurosurgery 1987;20:74±93. Andersen P, Eccles JC, Sears TA. Presynaptic inhibitory action of cerebral cortex on the spinal cord. Nature 1962;194:740±741. 271 Baleydier C, MauguieÁre F. The duality of cingulate gyrus in monkey: neuroanatomical study and functional hypothesis. Brain 1980;103:525±554. Boivie J. An anatomical reinvestigation of the termination of the spinothyalamic tract in the monkey. J Comp Neurol 1979;186:343±370. Carmel PW. Efferent projections of the ventral anterior nucleus of the thalamus in the monkey. Am J Anat 1970;128:159±184. Carpenter MB. The diencephalon. Core text of Neuroanatomy, Williams and Wilkins, 1995 pp. 251-296. Carstens E, Campbell IG. Parametric and pharmacological studies of midbrain suppression of the hind limb ¯exion withdrawal re¯ex in the rat. Pain 1988;33:201±213. Carter CS, Mintun M, Cohen JD. Interference and facilitation effects during selective attention: an (H2O)- O-15 PET study of Stroop task performance. Neuroimage 1995;2:264±272. Casey KL, Minoshima S, Berger KL, Koeppe RA, Morrow TJ, Frey KA. Positron emission tomographic analysis of cerebral structures activated speci®cally by repetitive noxious heat stimuli. J Neurophysiol 1994;71:802±807. Cesaro P, Ollat H. Pain and its treatments. Eur Neurol 1997;38:209±215. Cesaro P, Mann MW, Moretti JL, Defer G, RoualdeÁs B, Nguyen JP, Degos JD. Central pain and thalamic hyperactivity: a single photon emission computerized tomographic study. Pain 1991;329±336. Coghill RC, Talbot JD, Evans AC, Meyer E, Gjedde A, Bushnell C, Duncan GH. Distributed processing of pain and vibration in the human brain. J Neurosci 1994;14:4095±4108. Craig AD, Reiman EM, Evans A, Bushnell MC. Functional imaging of an illusion of pain. Nature 1996;384:258±260. Davis KD, Wood ML, Crawley AP, Mikulis DJ. fMRI of human somatosensory and cingulate cortex during painful electrical nerve stimulation. NeuroReport 1995;7:321±325. Decety J, Perani D, Jeannerod M, Bettinardi V, Tadary B, Woods R, Mazziota JC, Fazio F. Mapping motor representations with positron emission tomography. Nature 1994;371:600±602. Devinsky O, Morrell MJ, Vogt BA. Contributions of the anterior cingulate cortex to behaviour. Brain 1995;118:279±306. Di Piero V, Jones AKP, Iannotti F, Powell M, Perani D, Lenzi GL, Frakowiak RSJ. Chronic pain: a PET study of the central effects of percutaneous high cervical cordotomy. Pain 1991;46:9±12. Dostrovsky JO, Hutchison WD, Davis KD, Lozano A. Potential role of orbital and cingulate cortices in nociception. In: Besson JM, Guilbaud G, Ollat H, editors. Forebrain areas involved in pain processing, Paris: John Libbey Eurotext, 1995. pp. 171±181. Foltz EL, White LE. Pain relief by frontal cingulumotomy. J Neurosurg 1962;19:89±100. Friston KJ, Frith CD, Liddle PF, Lammertsma AA, Dolan RD, Frackowiak RSJ. The relationship between local and global changes in PET scans. J Cer Blood Flow Metab 1990;10:458±466. Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ. Plastic transformation of PET images. J Comp Ass Tomogr 1991a;15:634±639. Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ. Comparing functional (PET) images: the assessment of signi®cant change. J Cer Blood Flow Metabol 1991b;11:690±699. GarcõÂa-Larrea L, Sindou M, MauguieÁre F. Nociceptive ¯exion re¯exes during analgesic neurostimulation in man. Pain 1989;39:145±156. GarcõÂa-Larrea L. Multimodal approaches to generators of laser evoked potentials: with a little help from our friends. Pain Forum 1998;7:216±220. Gerhart KD, Yezierski RP, Fang ZR, Willis WD. Inhibition of primate spinothalamic tract neurons by stimulation in ventral posterior lateral thalamic nucleus: possible mechanisms. J Neurophysiol 1983;49:406± 423. Gybels J, Kuypers R. Subcortical stimulation in humans and pain. In: Desmedt JE, Bromm B, editors. Pain and the brain. Series: advances in pain research and therapy. New York: Raven Press, 1995. pp. 187± 199. Handwerker HO, Iggo A, Zimmermann M. Segmental and supraspinal 272 L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 actions on dorsal horn neurons responding to noxious and non-noxious skin stimuli. Pain 1975;1:147±165. Hardy SGP, Leichnetz GR. Cortical projections to the periaqueductal gray in the monkey: a retrograde and anterograde horserdish peroxydase study. Neurosci Lett 1981;22:97±101. Herregodts P, Stadnik T, Deridder F, Dhaens J. Cortical stimulation for central neuropathic pain: 3-D surface MRI for easy determination of the motor cortex. In: Advances in Stereotactic and Functional Neurosurgery 1995;11:132±135. Herscovitch P, Markham J, Raichle ME. Brain blood ¯ow measured with intravenous H215O. I. Theory and error analysis, J Nucl Med 1983;24:782±789. Hirayama T, Tsubokawa T, Katayama Y, Maejima S, Koyama S, Yamamoto S. Tonic changes in activity of thalamic lemniscal relay neurons following spino-thalamic tracheotomy in cats: effects of motor cortex stimulation. Pain, suppl. 1990;5:S273. Holsheimer J. Effectiveness of spinal cord stimulation in the management of chronic pain: analysis of technical drawbacks and solutions. Neurosurgery 1997;40:990±999. Hsieh JC, Belfrage M, Stoneelander S, Hansson P, Ingvar M. Central representation of chronic ongoing neuropathic pain studied with positron emission tomography. Pain 1995;63:225±236. Hsieh JC, Stahlebackdahl M, Hagermark O, Stoneelander S, Rosenquist G, Ingvar M. Traumatic nociceptive pain activates the hypothalamus and the periaqueductal gray: a positron emission tomography study. Pain 1996;64:303±314. Iadarola MJ, Max MB, Berman KF, Byassmith MG, Coghill RC, Gracely RH, Bennett GJ. Unilateral decrease in thalamic activity observed with positron emission tomography in patients with chronic neuropathic pain. Pain 1995;63:55±64. Iadarola MJ, Berman KF, Zef®ro TA, Byassmith MG, Gracely RH, Max MB, Bennett GJ. Neural activation during acute capsaicin-evoked pain and allodynia assessed with PET. Brain 1998;121:931±947. Ilinsky IA, Jouandet ML, Goldman-Rakic PS. Organization of the nigrothalamo-cortical system in the rhesus monkey. J Comp Neurol 1985;236:315±330. Jasper HH. The ten-twenty electrode system of the International Federation. Electroencephalogr clin Neurophysiol 1958;10:371±375. Jeanmonod D, Magnin M, Morel A. Thalamus and neurogenic pain: physiological, anatomical and clinical data. NeuroReport 1993;4:475±478. Jessurun GAJ, Dejongste MJL, Blanksma PK. Current views on neurostimulation in the treatment of cardiac ischemic syndromes. Pain 1996;66:109±116. Jones AKP, Brown WD, Friston KJ, Qi LY, Frackowiak RSJ. Cortical and subcortical localization of response to pain in man using positron emission tomography. Proc R Soc Lond B 1991;244:39±44. Katayama Y, Tsubokawa T, Maejima S, Hirayama T, Yamamoto T. Corticospinal direct responses in humans: identi®cation of the motor cortex during intracranial surgery under general anesthesia. J Neurol Neurosurg Psychiatry 1988;51:50±59. Katayama Y, Tsubokawa T, Yamamoto T. Chronic motor cortex stimulation for central deafferentation pain: experience with bulbar pain secondary to Wallenberg syndrome. Stereotactic Funct Neurosurg 1994;62:295±299. Katayama Y, Fukaya C, Yamamoto T. Poststroke pain control by chronic motor stimulation: neurological characteristics predicting a favorable response. J Neurosurg 1998;89:585±591. Kim R, Nakano K, Jayaraman A, Carpenter MB. Projections of the globus pallidus and adjacent structures: an autoradiographic study in the monkey. J Comp Neurol 1976;169:263±289. Lane RD, Fink GR, Chau PML, Dolan RJ. Neural activation during selective attention to subjective emotional responses. NeuroReport 1997;8:3969±3972. Lindblom U, Ottosson JO. In¯uence of pyramidal stimulation upon the relay of coarse cutaneous afferents in the dorsal horn. Acta Physiol Scand 1957;38:309±318. Lindblom U, Tapper N, Wiesenfeld Z. The effect of dorsal column stimula- tion on the nociceptive response of dorsal horn cells and its relevance for pain suppression. Pain 1977;4:133±144. Magnusson JE, Vaccarino AL. Reduction of autotomy following peripheral neurectomy by a single injection of bupivacaine into the cingulum bundle of rats. Brain Res 1996;723:214±217. Melzack R, Wall P. Pain mechanisms: a new theory. Science 1965;150: 971±979. Merlet I, GarcõÂa-Larrea L, Froment JC, MauguieÁre F. A simpli®ed method to project dipolar sources onto brain anatomy. Neurophysiol Clin 1999;29:39±52. Meyerson BA, Lindblom U, Linderoth B, Lind G, Herregodts P. Motor cortex stimulation as a treatment of trigeminal neuropathic pain. Acta Neurochirurgica, suppl. 1993;58:150±153. Nguyen JP, Pollin B, Feve A, Geny C, Cesaro P. Improvement of action tremor by chronic cortical stimulation. Movement Disord 1998;13:84± 88. Pagni CA, Canavero S. Functional thalamic depression in a case of reversible central pain due to a spinal intramedullary cyst: case report. J Neurosurg 1995;83:163±165. Percheron G, Francois C, Talbi B, Yelnik J, Fenelon G. The primate motor thalamus. Brain Res Rev 1996;22:93±181. Peyron R, GarcõÂa-Larrea L, Deiber MP, Cinotti L, Convers P, Sindou M, MauguieÁre F, Laurent B. Electrical stimulation of precentral cortical area in the treatment of central pain: electrophysiological and PET study. Pain 1995;62:275±286. Peyron R, GarcõÂa-Larrea L, GreÂgoire MC, Convers P, Lavenne F, Bonnefoi F, Michel D, MauguieÁre F, Laurent B. Allodynia after lateral-medullary (Wallenberg) infarct: a PET study. Brain 1998;121:345±356. Powell TPS, Cowan WM. The interpretation of the degenerative changes in the intralaminar nuclei of the thalamus. J Neurol Neurosurg Psychiatry 1967;30:140±153. Rinaldi PC, Young RF, Albe-Fessard DC, et al. Spontaneous neural hyperactivity in the medial and intralaminar thalamic nuclei of patients with deafferentation pain. J Neurosurg 1991;74:415±424. Sokoloff L, Porter A, Roland P, Wise O, Frakowiack RH, Jones T, Raichle ME, Fox P, Plum F. General discussion. In: Chadwick C, Derek J, Whelan J, editors. Exploring brain functional anatomy with positron emission, Ciba Foundation Symposium, London: Wiley and Sons, 1991. pp. 43±56. Talairach J, Tournoux P. Coplanar stereotaxic atlas of the human brain, Stuttgart: Thieme, 1988. Talbot JD, Marrett S, Evans AC, Meyer E, Bushnell MC, Duncan GH. Multiple representations of pain in human cerebral cortex. Science 1991;251:1355±1358. Trebossen R, Mazoyer BM. Some physical characteristics of a timeof-¯ight positron tomograph (CEA-LETI-TTV03) obtained with the EEC emission phantom. Med Prog through Technol 1991;17:165± 171. Treede RD, Lorenz J, Kunze J, Bromm B. Assessment of nociceptive pathways with laser-evoked potentials in normal subjects and patients. Pain and the Brain. Series: advances in pain research and therapy no. 22. New York: Raven Press, 1995. pp. 377±392. Tsubokawa T, Katayama Y, Yamamoto T, Hirayama T, Koyama S. Chronic motor cortex stimulation for the treatment of central pain. Acta Neurochirurgica, suppl. 1991;52:137±139. Tsubokawa T, Katayama Y, Yamamoto T, Hirayama T, Koyama S. Chronic motor cortex stimulation in patients with thalamic pain. J Neurosurg 1993;78:393±401. Tsubokawa T, Katayama Y, Yamamoto T, Hirayama T, Maejima S, Koyama S. Fujii M. Treatment of deafferentation pain with thalamic and motor cortex stimulation: possible role of reorganization of neural circuits. VIIth World Congress on Pain. Book of Abstracts, IASP Publications, Seattle, 1993, pp. 504-505. Valeriani M, Rambaud L, MauguieÁre F. Scalp topography and dipolar source modelling of potentials evoked by CO2 laser stimulation of the hand. Electroencephalogr clin Neurophysiol 1996;100:343±353. Vogt B, Finch DM, Olson C. Functional heterogeneity in cingulate cortex: L. GarcõÂa-Larrea et al. / Pain 83 (1999) 259±273 the anterior executive and posterior evaluative regions. Cerebral Cortex 1992;2:435±443. Warburton E, Wise RJS, Price CJ, Weiller C, Hadar U, Ramsay S, Frackowiak RSJ. Noun and verb retrieval by normal subjects; studies with PET. Brain 1996;119:159±179. Wenzel R, Bartenstein P, Dieterich M, Danek A, Weindl A, Minoshima S. Deactivation of human visual cortex during involuntary oscillations. A PET activation study. Brain 1996;119:101±110. Willer JC. Comparative study of perceived pain and nociceptive ¯exion re¯ex in man. Pain 1977;3:69±80. 273 Willer JC. Nociceptive ¯exion re¯ex as a physiological correlate of pain sensation in humans, Handbook of Physiology. Amsterdam: Elsevier S.C, 1984 pp. 87-110. Wood CC, Spencer DD, Allison T, McCarthy G, Williamson PD, Goff WR. Localization of human sensorimotor cortex during surgery by cortical surface recordings of somatosensory evoked potentials. J Neurosurg 1988;68:99±111. Worsley KJ, Evans AC, Marrett S, Neelin P. A three dimensional statistical analysis for CBF activation studies in human brain. J Cereb Blood Flow Metab 1992;12:900±918.