Introduction

From experimental investigations in animals it is known that peripheral nerve injury, for example, subsequent to the amputation of fingers, leads to a change in the primary somatosensory cortex (SI), even in adult organisms (Merzenich et al. 1984; Pons et al. 1991). Pons and colleagues (Pons et al. 1991) observed massive cortical reorganization of about 1–2 cm in adult macaque monkeys that had been subjected to deafferentation of one upper extremity 12 years earlier. Stimulation of the face activated an area of SI that previously represented the finger, the ventral hand area, the upper arm and the neck. There was a point-to-point correspondence between individual stimulation sites in the face and corresponding activity in the cortical hand or arm representation.

Earlier studies in humans demonstrated the elicitation of painful and non-painful phantom sensations by touching the residual limb (Cronholm 1951). In patients with unilateral arm or hand amputation, Ramachandran et al. (1992a, 1992b) reported that they could elicit phantom phenomena upon tactile non-painful stimulation of the residual limb as well as of the mouth and chin region ipsilateral to the amputation side. There was a one-to-one correspondence between facial stimulation sites and corresponding modality-specific sensations in the amputated limb, a phenomenon they called “facial recapping” (see also Halligan et al. 1993; Aglioti et al. 1994). Studies in unilateral arm amputees (Elbert et al. 1994; Flor et al. 1995; Grüsser et al. 2001) could not find a relationship between the elicitation of referred phantom sensation and the reorganization of SI cortex but demonstrated a strong relationship between the magnitude of phantom limb pain and the amount of reorganization in SI. Knecht et al. (1995, 1996) and Grüsser et al. (2001) reported mislocalization of sensations in upper limb amputees using painful and non-painful stimulation modalities. Knecht et al. (1995, 1996) showed that mislocalization could be evoked bilaterally from the upper part of the body. They reported a strong relationship between the amount of cortical reorganization and the number of stimulated body sites that elicited sensations referred to as the phantom upon painful (but not non-painful) stimulation. Grüsser et al. (2001) showed a positive relationship between the percentage of sites from which painful referred sensations could be elicited by painful stimulation and the degree of cortical reorganization. This was not true for non-painful stimulation and non-painful referred phantom sensations. In conclusion, data from recent studies which investigated referred sensations and cortical reorganization in SI in chronic amputees suggest that painful sensations referred to upper phantom limbs as well as phantom limb pain are associated with reorganization in SI. Borsook et al. (1998) investigated a patient who underwent an amputation of the distal left humerus caused by a neuroectodermal tumor 23 h prior to the assessment using fMRI. They demonstrated activation in the contralateral SI, SII, supplementary motor area and posterior cingulate gyrus after non-painful stimulation of several trigger points on the body surface and described for the first time an elicitation of sensations referred to an upper phantom limb by stimulating the lower limb with trigger points on the dorsal surface of the foot. However, in this study the authors did not differentiate between non-painful phantom sensation, phantom pain or referred sensations with respect to the activation of brain areas. The present study examined sensations referred to the phantom arm in two long-term unilateral upper limb amputees when the lower part of the body was stimulated and analyzed concurrent cortical changes.

Materials and methods

Subjects

Both participants, one 61-year-old male (participant 1) and one 37-year-old male (participant 2), underwent an amputation of the right arm because of bone cancer disease. The time since amputation was 14.5 years for participant 1 and 22.7 years for participant 2. Participant 1 underwent a unilateral upper arm exarticulation and participant 2 an upper arm amputation with a residual limb length of 21 cm. Whereas participant 1 did not use a prosthesis, participant 2 had been using a cosmetic prosthesis daily for 23 years. Prior to his arm amputation participant 1 lost his right second digit as a result of an accident at the age of 25 years. He never felt this digit in his phantom hand.

Psychological and neurological assessment

The two amputees participated in a comprehensive neurological examination, completed a psychometric assessment including an interview about the amputation and its consequences (Flor et al. 1995; Winter et al. 2001) and the German version of the West Haven-Yale Multidimensional Pain Inventory (Kerns et al. 1985) modified to separately assess phantom limb and residual limb pain (Flor et al. 1995) and participated in an assessment of referred sensation as well as neuroelectric source imaging (both participants) and functional magnetic resonance imaging (participant 2). The study was approved by the local ethics committee and the participants gave informed consent. The study adhered to the ethical standards laid down in the 1964 Declaration of Helsinki.

The occurrence of referred sensations was assessed twice in each participant. Fifty-seven standardized sites spread over the entire body were stimulated. Ten sites were located in the face, 23 sites were located on the upper body part and the remaining sites covered the lower body (see Fig. 1).

Fig. 1
figure 1

Stimulated sites in participants 1 and 2 (open circles) and sites that elicited referred sensations (closed circles)

Full size image

Stimulation was performed in randomized order for the two modalities pain and touch, by using an algesiometer with an exchangeable taper for the elicitation of painful and cotton tips for non-painful referred sensations. The duration of the pain and touch stimuli at each site was 2 s per site in the first and 20 s in the second assessment. In both sessions the participants had to localize and evaluate the quality and intensity of stimulation they felt at the site of stimulation and to report and evaluate sensations elsewhere (e.g., in the phantom). The localization of sensations at the stimulation site and the referred site was drawn in a body template. The intensity was evaluated on a 7-point rating scale ranging from 0 = no sensation to 6 = very strong sensation and the quality was described by adjective lists.

Neuroelectric source imaging

Neuroelectric source imaging (ESI) was used by applying 1,000 pneumatic stimuli to the first and the fifth digits of the intact hand and the left and the right corner of the mouth. A 61-channel electrode matrix was employed with AgAgCl electrodes mounted in an electrocap where the corners of the 10×6 matrix were F7, T5, F8 and T6 according to the 10–20 EEG system and the reference electrode was placed at Cz. The individual matrix distribution was digitized with an infrared camera system (Optotrak). Two Synamps amplifiers with Scan software (Neuroscan) were used for data acquisition. The signals were continuously sampled with a 1-kHz sampling rate and filter settings from d.c. to 200 Hz. The data were then digitally filtered from 2 to 20 Hz and were analyzed with the MUSIC algorithm, which fits probe dipoles to potential dipole locations in the region of interest (Mosher et al. 1992). The region of interest here was a grid covering the individual surface of the cortex of the patients. The time range was 0–100 ms with respect to the trigger onset. The signal subspace contained the largest four components of the data singular value decomposition. The generator in SI was determined by selecting those dipole locations with maximum probability within the somatosensory cortex (Mühlnickel et al. 1999). Anatomical data gathered by magnetic resonance imaging (Siemens Vision MR 1.5 T scanner, T1-weighted, TR = 22 ms, TE = 10 ms, α = 30 degrees, slice thickness = 1 mm) were used to construct the individual head geometry and were utilized to overlay the dipole reconstructions on the anatomical data (Mühlnickel et al. 1999). The coordinates of the localization of the fingers and the mouth were used to compute the Euclidean distance as a measure of cortical reorganization (Elbert et al. 1994; Flor et al. 1995). The measure of cortical reorganization was determined in the following manner: the Euclidean distance between the fingers (mean of the first digit and the fifth digit) and the mouth representation on the intact side was computed for the anterior-posterior, medial-lateral and superior-inferior coordinates. This yielded a mouth-digit distance measure in millimeters. Based on the symmetry assumption of the two hemispheres (cf. Gallen et al. 1993), the locations of the fingers of the intact side were transposed along the midsagittal plane to the hemisphere that represents the amputation side and mirror locations of the digits were determined. The same distance measure was now calculated for the amputated side, and the difference between the Euclidean distances on the intact and the amputation side was then used as an indicator of the amount of cortical reorganization (Elbert et al. 1994; Schaefer et al. 2002a, 2002b). This procedure yielded a measure of the shift of the cortical representation of the mouth into the representation of the deafferented area in millimeters.

Functional magnetic resonance imaging

Functional magnetic resonance imaging was used in one patient to determine the cortical representation of the feet, which were difficult to image with the EEG. A foil ring electrode was placed around the left and the right first toe with a proximal cathode. Electric monophasic square-wave pulses with a duration of 200 µs were delivered with a randomized interstimulus interval of 70–140 ms (Digitimer DS7A). The stimulus intensity was adjusted to be at 75% between sensory and pain threshold (6.9–11.8 mA) and was characterized by the subject as non-painful. The subject laid supine in a 1.5 T Siemens Vision scanner. For functional MRI an echoplanar imaging (EPI) sequence (TR = 3.3 ms, TE = 66 ms, α = 90°, matrix dimension 64×64, slice thickness 4 mm, gap 1 mm) was used encompassing 24 axial slices that covered the entire cerebrum and as much as possible of the cerebellum. For anatomical reference a 3D MPRAGE (magnetization prepared rapid gradient echo, 1×1×1 mm3) image data set was acquired. fMRI slices were oriented axially to the AC-PC plane. The toes and the finger were stimulated in successive sessions that consisted of five stimulation blocks (duration 16.5 s) interrupted by rest blocks of varying duration (26.4–33 s). The total acquisition time was 5 min for each session. The entire fMRI experiment lasted 60 min. After motion correction and temporal and spatial smoothing a hemodynamic response function based on a statistical comparison between rest and activation was calculated (Brainvoyager 2000).

Results

Phantom sensations and phantom pain

Participant 1 reported retrospectively preamputation pain. He described pricking and cutting pain sensations in the shoulder area before amputation. He furthermore described attacks of phantom limb pain with a duration of 2–3 s for several hours 1–4 times a month. The quality of his pain sensations was mainly pricking and cutting, localized in the articulation of the phantom hand, the edge of the phantom hand and digits 1 and 5 with an intensity of 1.3 on the MPI. Non-painful phantom sensations were described as a position of the phantom arm in a 90° angular position and a quality of sensation in all phantom digits such as pressure, tingling, numbing and feeling of warmth, with an intensity of 14 on a visual analogue scale (0–100). He localized his non-painful phantom phenomena mainly in the articulation of the phantom hand, the edge of the phantom hand, digits 1 and 5 and sometimes the lower arm. A combination of painful and non-painful phantom sensations was reported for about 60% of the time. The patient did not report sensations in the residual limb whether painful or not.

Referred sensations were elicited only in the second assessment due to the longer duration when non-painful stimulation but not when painful stimulation was applied to two sites at the lower limb (calf, first toe) ipsilateral to the amputation (see Fig. 1). The localization of referred sensations in participant 1 was the articulation of the phantom hand, the edge of the phantom hand, and digits 1 and 5. The quality and intensity of the referred sensations were similar to the painful and non-painful phantom sensations the patient habitually experienced.

Participant 2 did not report preamputation pain. After amputation he described cramping and knocking phantom pain with an intensity of 4 (MPI) and a duration of 4–6 h daily localized in all five digits. The quality of his non-painful phantom sensations was mainly described as pressure, itching, tingling and the feeling of cold on the hand surface and in all digits with an intensity of 99 (VAS). In addition, he reported a telescoped phantom. His phantom hand was localized in the residual limb and it was shrunk in size by about 60%. The fingertips of digits 1, 2 and 3 were connected to one another and digits 4 and 5 were bent. The non-painful phantom phenomena occurred in all digits and there was no combination of painful and non-painful phantom sensations at the same time. Crawling non-painful residual limb sensations and burning, throbbing and piercing residual limb pain were reported.

In both assessments participant 2 demonstrated several trigger points where painful and non-painful stimulation elicited referred sensations in his phantom. The sites eliciting referred sensations were the chest, shoulder and calf ipsilateral and the first toe and the dorsal and ventral surface of the foot ipsi- and contralateral to the amputation side as well as several sites on the residual limb (see Fig. 1). The referred sensations were localized on the thenar eminence and all digits (mainly digits 1–3). The quality and intensity of referred sensations were similar to the painful and non-painful phantom sensations the patient habitually experienced. He reported a point-to-point correspondence between the stimulation sites and the sensations referred to the phantom digits when painful and non-painful stimulation was applied to the chest.

Cortical reorganization

The neuroelectric source imaging data demonstrated for participant 1 a reorganization of the mouth representation into the cortical arm area. The amount of cortical reorganization was 11.6 mm (see Fig. 2). Participant 2 also showed a reorganization of the mouth representation into the cortical arm area amounting to 15.2 mm (see Fig. 2). In contrast, the fMRI data showed a symmetric representation of both toes (see Fig. 3). When the Euclidean distance of each of the toes from Cz was calculated, it amounted to 13.93 mm on the right and 13.38 mm on the left side.

Fig. 2
figure 2

Localization of the representation of the mouth (white square) and the thumb (white circle) of the intact body side and the representation of the mouth of the lesioned body site (black square) mirrored on the intact hemisphere. Data of participants P1 and P2

Full size image
Fig. 3
figure 3

fMRI activation sites of participant 2 for the stimulation of the left and right first toe. Due to the specific anatomy of P2’s brain, visual inspection of the localization may suggest asymmetry; however, actually calculated side differences between the centers of activation were <3 mm, which is within the normal interhemispheric variation (Gallen et al. 1993)

Full size image

Discussion

Most studies on referred sensation and phantoms found mislocalization of sensation elicited at points on the body surface adjacent to the deafferented area (Ramachandran et al. 1992a, 1992b; Knecht et al. 1996) in accordance with the notion that referred sensations might be the perceptual correlate of cortical reorganization subsequent to amputation (Ramachandran et al. 1992b). Here the elicitation of phantom sensation by stimulation in the lower body part and contralateral to the amputation was found. In SI, the lower body part (e.g., the first toe) is represented rostral from the deafferented area and separated by several representation zones of the body surface, i.e., the hip and the trunk. In both patients sensations referred to the phantom arm were elicited from the first toe and from the calf ipsilateral to the amputation site, in one patient also contralateral. In one of the patients sensations referred to the phantom were also elicited by non-painful stimulation of the lower limb ipsilateral to the amputation but not at trigger points elsewhere on the body surface.

The present data showed a reorganization of the mouth area into the deafferented hand area in SI that exceeded the normal interhemispheric difference of about 3–6 mm in healthy humans as well as amputees without phantom pain (Gallen et al. 1993; Schaefer et al. 2002a, 2002b) in both subjects. We also observed the elicitation of sensation referred to the upper phantom limb by non-painful stimulation of points on the intact lower limb in both participants. At the same time there was no reorganization of the foot representation in the one patient who was tested with fMRI.

Reorganization in SI could only underlie the phenomenon of referred sensation from the toe if reorganization of the foot representation into the now absent arm and hand representation were present. The data of this study make it unlikely that the phenomenon of referred sensation is related to changes in SI. Rather, reorganization in SI seems to be related to the presence of habitual phantom limb pain (cf. Flor et al. 1995; Grüsser et al. 2001). There are several potential neural correlates of referred phantom sensation: (1) subcortical mechanisms might contribute to the phenomenon of referred sensations. Somatotopic reorganization in the brainstem and thalamus (Churchill et al. 2001; Jones and Pons 1998) following peripheral nerve injury in adult primates has been described. It is known that at the level of the thalamus the upper (fasciculus cuneatus Burdach) and the lower body part (fasciculus gracilis Goll) are connected by interneurons (Zilles and Rehkämper 1993). In contrast to the somatotopy of SI, the somatotopy of the ventral posterolateral nucleus (VPL) of the thalamus (Mountcastle and Henneman 1952) shows the hand area adjacent to the leg area. However, since the thalamic afferents reach both SI and SII cortex, it is unlikely that thalamic reorganization would only manifest itself in SII but not in SI. (2) A further candidate for reorganizational change is SII. In SII, the foot and hand representation are adjacent just as in the thalamus. In macaques, Pons et al. (1988) reported massive somatotopic reorganization 6–8 weeks after hand amputation, involving more than half the areal extent of SII, which may reflect a greater capacity for reorganizational changes in higher order than in primary sensory cortical areas. Nearly all of the region that had been represented by the hand was now found to be occupied by an expanded foot representation. The role of SII in phantom sensation is supported by the report of decreased ipsilateral SII activation in amputees with experimentally elicited referred phantom sensation (Flor et al. 2000) and a patient with a supernumerary limb (Hari et al. 1999). (3) A further region of potential interest is the posterior parietal cortex. It was previously shown that the occurrence of referred phantom sensation in the laboratory was accompanied by significantly enhanced activation in posterior parietal cortex (Flor et al. 2000). The posterior parietal cortex—and specifically area 5—is involved in the maintenance of the body image and associated with neglect as well as lack of recognition of body parts as one’s own (Berlucchi and Aglioti 1997). Its involvement in phantom phenomena is therefore highly likely. This assumption was confirmed by PET data (e.g., Bonda et al. 1995; Kew et al. 1994, 1997), indicating that the PPC is involved in the perception of body image as well as phantom sensation. (4) Finally, referred phantom sensation could also be related to activation in right dorsolateral prefrontal cortex, a region that was found to be activated by incongruent visual feedback from motor movements (Fink et al. 1999). Harris (1999) postulated that this region might also be involved in pathological pain such as phantom limb pain. To what extent non-painful referred sensation might be mediated by activation in this region remains to be determined.

Conclusion

Taken together the data of the present study suggest involvement of areas other than SI in the generation of referred phantom sensations. This assumption is in accordance with recent findings suggesting that non-painful phantom sensations are associated with alterations in area 5 and SII whereas painful sensation is strongly associated with reorganization in SI (Flor et al. 2000).