Vascular Abnormalities in Acute Reflex Sympathetic Dystrophy (CRPS I): Title and subTitle BreakComplete Inhibition of Sympathetic Nerve Activity With Recovery FREE

REFLEX SYMPATHETIC dystrophy (RSD) is a painful neuropathic disorder that may develop as a disproportionate consequence of a minor trauma affecting the limbs, a bone fracture, or a remote process like stroke or myocardial infarction. The characteristic clinical features are spontaneous pain, hyperalgesia, impairment of motor function, swelling, changes in sweating, and vascular abnormalities in a single extremity. An overt nerve lesion is not detectable.^{1 - 2} Regardless of the site of the precipitating event, the abnormalities show a spreading tendency with a generalized distal distribution that is not confined to innervation territories of peripheral nerves or roots. In many cases, interruption of the efferent sympathetic supply to the affected extremity can dramatically relieve the pain. According to Classification of Chronic Pain,³ RSD is now called complex regional pain syndrome type I (CRPS I). This syndrome is distinguished from causalgia (CRPS II), in which a partial lesion of a peripheral nerve is necessary for the diagnosis.

Characteristically, patients with CRPS I demonstrate a warm and vasodilated affected extremity in the early stages and cold and pale skin in the late stages.^{4 - 8} The cause of these vascular abnormalities is unknown, and it is still under debate whether the sympathetic nervous system is involved in the generation of these changes.⁹

Animal studies have addressed the cause of vascular disturbances in neuropathic pain syndromes using experimental partial nerve lesions. Within the innervation zone of the affected nerve, changes in neurovascular transmission (ie, sympathetic denervation of cutaneous vessels resulting in vasodilation and subsequent development of supersensitivity to circulating catecholamines leading to vasoconstriction) were demonstrated as the cause of secondary blood flow abnormalities.^{10 - 12} Moreover, after a partial nerve lesion, excessive antidromic activation of undamaged afferent C fibers and neuropeptide release leading to acute vasodilation within the innervation zone of the affected nerve were shown.¹³

These models may apply to CRPS II (causalgia), in which a peripheral nerve lesion is by definition present. However, CRPS I (RSD) characteristically develops after a bone fracture or even a minor tissue lesion without any overt damage of peripheral nerves or major nerve branches. The autonomic symptoms spread to skin territories away from the originating injury. Therefore, interruption of sympathetic fibers and denervation of blood vessels cannot account for vascular changes in these patients. The question arises, "What alternative mechanisms might be responsible for autonomic abnormalities in CRPS I?"

Cutaneous sympathetic nerves mainly consist of vasoconstrictor fibers involved in thermoregulation. To assess their function quantitatively, sympathetic thermoregulatory and respiratory reflexes were induced under controlled conditions. Using these neurophysiological techniques, we demonstrated that acute RSD/CRPS I is characterized by complete functional inhibition of cutaneous sympathetic vasoconstrictor activity on the affected extremity. This abnormality is capable of recovering within weeks.

The patient was a 52-year-old woman seen at the Pain Clinic of the Neurological, Anaesthesiological, and Orthopedic Clinics in Kiel, Germany, with acute RSD (CRPS I) after having fractured the right distal radius. Two healthy volunteers served as control subjects. The patient and the controls were asked to refrain from coffee and food intake and from smoking for at least 3 hours before being tested. They were not taking drugs affecting vascular function. The aim of the study and the nature of the tests were explained to the patient and the controls according to the Helsinki Declaration. The study was approved by the local ethics committee. The patient and the controls gave informed consent after the nature of the procedures had been fully explained. The procedures followed were in accordance with institutional guidelines.

All neurophysiological tests were performed between 3 and 6 PM. The patient and the controls were tested while in the supine position (room temperature of 24°C). In the patient, the first set of experiments was performed during the early phase of the disease, ie, 24 days after the fracture and 2 weeks after the onset of typical CRPS I symptoms (week 2). Five weeks later the experiments were repeated (week 7).

Cutaneous blood flow in the glabrous skin (of the index fingertip) was measured bilaterally using a laser Doppler flowmeter (Periflux pf 4001 and integrating probe pf 413; Perimed, Stockholm, Sweden). The fingertips were selected for investigation because the abundant arteriovenous anastomoses of this area are under strict sympathetic vasoconstrictor control; thus, changes mediated by cutaneous vasoconstrictor activity are prominent, and vasomotor reflexes are extensive. Simultaneously, skin temperature was measured bilaterally in the dorsum of the hand (continuously) and in all fingers at 5-minute intervals with infrared thermometers.

During all the experiments described below, laser Doppler and continuous thermometer signals were recorded online (IBM-compatible computer) with an analog digital converter and counter interface (Burr-Brown PCI-20000; Burr-Brown Co, Tucson, Ariz, data acquisition software CARDS by Stefan Tiedemann, Department of Physiology, University of Kiel) for subsequent analysis.

To assess phasic modulation in sympathetic activity, respiratory reflexes were examined. Deep inspiration induces a phasic activation of cutaneous vasoconstrictor neurons that is followed by a marked short-lasting decrease of blood flow through the skin.¹⁴ Degeneration or dysfunction of these neurons results in an attenuation of the vasoconstrictor response.^{15 - 16}

Maximal phasic activation of sympathetic neurons was achieved by performing forced breathing. The desired respiration frequencies and inspiration-expiration ratios were visualized graphically by a moving vertical mark on a computer screen. The actual breathing parameters were monitored by a specially designed electronic device: a small plastic tube (14 cm long) with 1 end in 1 nostril. Air movement through the tube was detected by a sensitive thermistor within the tube.¹⁷

The persons were asked to breathe at a low frequency of 5/min with a high tidal volume. The inspiration-expiration ratio was 1:3. These parameters are known to induce a massive rhythmic activation of sympathetic neurons.¹⁷

Controlled thermoregulatory reflexes were performed to induce a physiologic tonic change of sympathetic skin nerve activity. This was achieved by changing environmental temperature. The subject was lying in a thermal suit supplied by tubes in which circulating water of 12°C and 50°C (inflow temperature), respectively, was used to cool or warm the whole body. Both hands were not covered by the suit. Whole-body cooling is the most effective stimulus to induce a massive tonic activation of cutaneous vasoconstrictor neurons, as demonstrated in microneurographical recordings,¹⁴ and warming leads to a considerable decrease of this activity. Degeneration or dysfunction of these neurons results in an attenuation of the cooling response. Alteration of sympathetic activity was controlled by simultaneously measuring skin blood flow and skin temperature in the hands as described above. To assess the central effects of whole-body temperature change, tympanic membrane temperature (close to core body temperature) was measured with an infrared thermometer at 10-minute intervals and blood pressure was documented online with a noninvasive finapress device (Ohmeda; Englewood, Colo).

The assessment of blood flow reactions evoked by sympathetic reflexes is an indirect measure of sympathetic nerve activity. Simultaneously occurring intense vasodilation may interfere with sympathetic vasoconstrictor activity. Theoretically, it is possible that a profound antidromic vasodilation caused by afferent C fiber axon reflex activation and neuropeptide release (neurogenic inflammation) may be present in this patient and therefore mimic a loss of sympathetic vasoconstrictor reflexes.

To address this problem, the following control experiment was performed. Histamine, a potent stimulator of afferent C fibers, was iontophoresed into the glabrous skin (thenar) of 2 controls. Cutaneous application of histamine induces an intense axon reflex vasodilation (antidromic vasodilation or flare reaction) within an area of several centimeters around the application site caused by the release of calcitonin gene-related peptide and substance P from axon collaterals in the skin.^{18 - 19} Within the area of axon reflex vasodilation (1 cm from the application site), laser Doppler measurements were performed during simultaneous activation of cutaneous sympathetic vasoconstrictor neurons using the same protocol as described above (forced breathing and controlled thermoregulation).

Histamine was applied 3 times during each experiment. After acclimatization in the laboratory (low sympathetic activity), the first application was performed to induce neurogenic inflammation and axon reflex vasodilation. To maximize the vasodilatory effect, a second application was performed at the same site 10 minutes later. Five minutes after this stimulus, whole-body cooling was started to induce activation of cutaneous vasoconstrictor fibers. During whole-body cooling (high sympathetic activity), the third application of histamine was performed at the same site as previously.

To further assess sympathetic function, plasma levels of norepinephrine from the venous effluent were examined. About 80% of this value reflects secretion by sympathetic postganglionic vasoconstrictor terminals to muscle and mainly to skin. Two weeks after the onset of CRPS I symptoms (week 2), venous blood samples were taken from veins bilaterally at the dorsum of the hands under resting conditions. Norepinephrine was measured by high-pressure liquid chromatography with electrochemical detection (Biorad Laboratories, Hercules, Calif).

A 52-year-old woman was seen with a fracture of the right distal radius in January 1997. No other injuries were detectable, in particular no peripheral nerve lesions. Ten days after immediate reduction of the fracture under plexus anesthesia and plaster, the patient reported a change in symptoms. She complained of a marked generalized swelling of the hand, a feeling of heat, and increasing pain in the right forearm and hand that was now of burning character.

Two weeks after the onset of these symptoms (week 2), the patient was clinically examined and the neurophysiological tests were performed. At that time she complained of spontaneous pain that she rated from 5 to 8 on a numerical rating scale (NRS: 0 indicates no pain; 10, the maximum of imaginable pain). Local warming and orthostatic load increased the pain, whereas moderate cooling and lifting reduced it. The right hand was warmer, but no side difference in sweating was observed. The hand was swollen, but there were no trophic changes. Voluntary movements of all fingers were markedly reduced, and handgrip force was extremely weak. The patient reported an increase of these motor disturbances during the previous 2 weeks. A distally generalized dynamic mechanical hyperalgesia (allodynia, pain caused by lightly touching the skin) was present in the affected forearm and hand without being confined to an innervation zone of any individual nerve or nerve root. No other neurologic abnormalities were observed. Results of the 3-phase bone scan demonstrated characteristic scintigraphic findings in phase 3, ie, a diffuse increase in the uptake of tracer around distal joints on the affected side.

The patient was a nonsmoker and had never had any other severe disease. Before the first examination, no sympatholytic treatment had been performed. Because of the clinical symptoms without any definable nerve lesion, the history of a distal radial fracture, and the typical scintigraphic findings, CRPS type I (RSD) was diagnosed.

Three weeks later (week 5), the patient was reexamined clinically after a series of sympatholytic procedures had been performed for diagnostic and therapeutic reasons. These interventions repeatedly relieved the pain, indicating a sympathetically maintained pain component. For example, use of a diagnostic regional guanethidine sulfate block led to a reduction of spontaneous pain from NRS 7.5 before the block to NRS 5 at 10 minutes, NRS 4.5 at 30 minutes, and NRS 0 at 60 minutes after the block. In addition to the sympatholytic interventions, nonsteroidal anti-inflammatory drugs were given and the extremity was immobilized. At week 3, therapy with corticosteroids was started. On examination at week 5, the typical symptoms of CRPS I were still present but of less severity. (1) The spontaneous pain was rated NRS 2. (2) The right hand was still warmer. (3) Swelling was diminished, and the skin of the right hand was shiny and dry. (4) Voluntary movements and handgrip force were improved. (5) Dynamic mechanical hyperalgesia was less intense.

At week 6, active physiotherapy could be started. After another week (week 7), during which there was no further invasive treatment, the third clinical examination was performed and the neurophysiological tests were repeated. At this time, pain was present only during active and passive movements of the hand. No swelling or temperature difference was detectable. All fingers of the right hand had small hyperkeratoses at their distal ends. Voluntary movements and strength showed further improvements, and the allodynia was further reduced.

Half a year later, the patient was reexamined clinically. At that time she had completely recovered from pain and allodynia, and range of movement was only minimally reduced. One year after the fracture she had completely recovered from all symptoms.

The first set of experiments was performed 2 weeks after the onset of CRPS I symptoms. Under forced breathing conditions (5/min), a rhythmical variation in cutaneous blood flow occurred in the finger of the unaffected side, indicating intact vasoconstrictor activity.¹⁷ The inspiratory phase was followed by considerable vasoconstriction, with a latency of several seconds between inspiration and decrease in blood flow (Figure 1). The mean relative decrease in blood flow was 58%.

On the affected side, almost no variation in blood flow was detectable (Figure 1). A small oscillation that is phase shifted in comparison with the respiratory rhythm on the unaffected side is generally thought to be passively induced by changes in blood pressure and venous tone during respiration.¹⁷ These data indicate that, on the affected side, no vasoconstriction could be induced by forced breathing.

After acclimatization in the laboratory (room temperature of 24°C, supine position, and thermal suit on), the skin temperature at the dorsum of the hand was 36°C on both sides (Figure 2). The finger skin temperature on the affected side was slightly higher than on the unaffected side (35.4°C vs 34.6°C, Figure 3). Thereafter, controlled whole-body cooling was performed to achieve tonic activation of sympathetic vasoconstrictor neurons innervating the skin. Skin blood flow and skin temperature on the unaffected side showed a normal pattern of regulation (Figure 2 and Figure 3). Whole-body cooling immediately produced a massive activation of vasoconstrictor neurons leading to a considerable and prolonged drop in skin blood flow that reached a minimum of 19% of baseline flux before cooling. After a short latency, skin temperature slowly decreased, reaching a minimum of 24.2°C in the fingers and 29.0°C in the dorsum of the hand (Figure 2 and Figure 3). After having switched to whole-body warming, the thermoregulatory cycle reversed (Figure 3). At the end of warming, considerable vasodilation was present because of nearly complete inhibition of vasoconstrictor activity.

On the affected side (CRPS I), however, whole-body cooling was not capable of inducing a decrease in skin blood flow and temperature. Throughout the thermoregulatory cycle the skin temperature was continuously high, with values that were close to core body temperature (approximately 36°C) indicating maximal vasodilation (Figure 2 and Figure 3). As a result of this constant vasodilation, differences in skin temperature between the affected and unaffected sides measured up to 14.0°C at the end of cooling. At the end of warming no difference in skin temperature was present. These results indicate that on the affected side no vasoconstriction could be induced by tonic thermoregulatory reflexes.

Venous blood samples were taken bilaterally from veins on the dorsum of the hand under resting conditions. These veins are draining a considerable amount of cutaneous blood flow from the hand. The affected side had a lower norepinephrine level (1856 nmol/L) compared with the unaffected side (3115 nmol/L), indicating that norepinephrine release from cutaneous vasoconstrictor terminals is decreased in the affected hand.

At week 7, vasoconstrictor responses induced by forced breathing were the same in both hands and not significantly different from the controls. The mean relative decrease in blood flow was 70% on the affected side and 63% on the unaffected side (Figure 4). The shape and time course of the respiratory reflexes were normal. These results indicate that vasoconstriction induced by repiratory reflexes had completely recovered within 7 weeks of disease onset. Tonic sympathetic reflexes were not performed at week 7.

To test whether an intense antidromic vasodilation may be capable of mimicking the loss of sympathetic vasoconstrictor reflexes observed in the patient, control experiments were performed. Histamine iontophoresis was used to induce an axon reflex vasodilation in the glabrous skin of the hand. During this vasodilation, phasic and tonic activation of cutaneous sympathetic vasoconstrictor neurons was performed (forced breathing and controlled thermoregulation) using the same protocol as in the patient.

After acclimatization in the laboratory (room temperature of 24°C, supine position, thermal suit on, and low sympathetic activity), histamine application at the thenar skin induced an intense antidromic vasodilation caused by afferent C fiber axon reflex activation and neuropeptide release. The increase in blood flow was 60% (Figure 5). Furthermore, the blood flow signal was stabilized so that transient sympathetic reflexes induced by arousal stimuli extensively present in 1 control before histamine application were absent after histamine application (Figure 5). This observation indicates that antidromic vasodilation can override sympathetic vasoconstriction caused by phasic arousal reflexes. Moreover, vasoconstriction evoked by forced breathing that was extensive before histamine application was nearly abolished after histamine application (Figure 6). Therefore, antidromic vasodilation can also mimic the loss of sympathetic vasoconstriction caused by phasic respiratory reflexes.

Using tonic, long-lasting sympathetic vasoconstrictor reflexes, the situation was different. Sympathetic vasoconstrictor activity induced by whole-body cooling performed during the presence of antidromic vasodilation in the hand induced a sustained vasoconstriction. Additional application of histamine produced only a small increase in blood flow, indicating that axon reflex vasodilation was sufficiently depressed by ongoing sympathetic vasoconstrictor activity (Figure 5). These results show that sympathetic vasoconstriction caused by tonic sympathetic reflexes can override antidromic vasodilation.

Vascular abnormalities, often abnormal vasodilation and skin warming in the early phase and vasoconstriction in later stages, are characteristic symptoms of RSD/CRPS I. We present for the first time evidence that a complete inhibition of sympathetic activity may be reponsible for the skin warming and vasodilation observed during the early phase of CRPS I. Several findings can be summarized: (1) Ten days after the inciting trauma, ie, a distal radius fracture, typical clinical symptoms of CRPS I developed at the distal extremity, ie, swelling, pain, impairment of movement, and vascular abnormalities. (2) Skin temperature was higher on the affected side compared with the unaffected side in normal room temperature and during controlled thermoregulation. (3) Phasic and tonic stimulation of cutaneous vasoconstrictor neurons evoked by sympathetic respiratory and thermoregulatory reflexes did not induce a decrease in skin blood flow and temperature on the affected side (Figure 1, Figure 2, and Figure 3). (4) Venous norepinephrine levels were lower on the affected side. (5) In relation to clinical improvement, loss in vasoconstrictor function completely recovered within weeks (Figure 4).

Besides the bone fracture, no skin or deep tissue lesions, in particular no nerve lesions, were present. At the time of the accident, results of a thorough neurologic examination did not reveal any neurologic abnormalities. Therefore, the loss of vasoconstrictor responses observed 2 weeks after the onset of CRPS I is unlikely to be explained as a consequence of a peripheral lesion of sympathetic fibers. These findings are supported by results of histological examinations of skin biopsy samples in patients with CRPS I. No differences in distribution of cutaneous sympathetic or nociceptor fibers was demonstrated.²⁰ Another observation that challenges peripheral nerve damage as a causative factor is that the sympathetic reflex abnormalities observed in the present study were reversible within 5 weeks (Figure 4).

Alternatively, changes in the neurovascular transmission may lead to a lack of vessel responsiveness to sympathetic stimulation. However, it seems unlikely that such changes occur without the presence of a structural lesion of sympathetic postganglionic fibers.

Furthermore, an ongoing C nociceptor barrage and profound antidromic vasodilation within the symptomatic skin may interfere with sympathetic outflow and therefore mimic a loss of vasoconstrictor response. Such neurogenic inflammation has been suggested to be the source of skin warming and vasodilation in CRPS.^{13 ,21 - 22} In fact, results of the control experiments performed in this study show that intense antidromic vasodilation in the glabrous skin induced by histamine iontophoresis overrides vasoconstriction evoked by phasic sympathetic reflexes such as short-lasting arousal stimuli or respiratory reflexes (Figure 6). However, tonic sympathetic thermoregulatory reflexes, ie, sympathetic activation caused by whole-body cooling, can overcome antidromic vasodilation (Figure 5). Comparable results have been obtained by Hornyak et al,²³ who used transcutaneous electrical stimulation to induce antidromic vasodilation in the glabrous skin and whole-body cooling to change sympathetic activity. During high sympathetic activity, vasodilation was markedly diminished and in some experiments even totally abolished. Accordingly, vasoconstriction in the hand achieved by intraneural microstimulation was found to override the antidromic vasodilator effect induced by intraneural stimulation of C nociceptors.²⁴ This was confirmed in animal experiments investigating the interaction between sympathetic vasoconstriction and antidromic vasodilation. Electrical stimulation of the sympathetic chain with high frequencies significantly reduced axon reflex vasodilation induced by dorsal root stimulation.²⁵ Furthermore, in a patient with a neuropathic pain syndrome after burn injury, abnormal C nociceptor sensitization was identified microneurographically, suggesting that antidromic vasodilation was the source of local skin warming. In this case, sympathetic vasoconstrictor reflexes were normal.²⁶ Taking these results together, it is unlikely that afferent antidromic mechanisms are involved in the skin warming, vasodilation, and loss of vasoconstrictor responses in the patient with CRPS I described herein.

However, other vasodilatory mechanisms not tested in the present investigation may be more powerful than histamine-evoked antidromic vasodilation. Endothelium-derived nitric oxide and prostacyclins are known to induce a profound relaxation of blood vessels. Therefore, vasodilation caused by an abnormally high release of these substances might interfere with sympathetic vasoconstriction.

In summary, anatomic damage of sympathetic fibers and excessive antidromic vasodilation caused by neurogenic inflammation are not responsible for the loss of vasoconstrictor responses and the skin warming observed in our patient. Therefore, it seems reasonable to suggest that the loss of vasoconstrictor responses is related to a functional inhibition of sympathetic neuronal activity. The sympathetic inhibition is so intense that respiratory and thermoregulatory vasoconstrictor reflexes are completely abolished. This inhibition of sympathetic outflow is confined to the extremity where the inciting trauma occurred.

In accordance with the idea of an inhibition of sympathetic activity, the norepinephrine level in the venous blood samples from the affected side was considerably lower compared with that from the healthy side, indicating a substantial decrease of transmitter release from postganglionic sympathetic vasoconstrictor fibers. In similar studies,^{27 - 29} norepinephrine, its intracellular metabolite 3,4-dihydroxyphenylethyleneglycol, and neuropeptide Y, which coexist with norepinephrine in sympathetic vasoconstrictor neurons, were shown to be reduced in venous blood samples from affected limbs of patients with CRPS.

These findings support the idea that vascular abnormalities of acute CRPS I are associated with a disturbed sympathetic innervation of the affected limb. An abnormal unilateral reflex pattern of sympathetic vasoconstrictor neurons evoked by respiratory and thermoregulatory stimuli is present. The pathophysiological mechanisms underlying such disturbed sympathetic reflex activity must be located in the central nervous system.³⁰ This interpretation is consistent with experimental findings, which show that the centrally generated reflex pattern in cutaneous vasoconstrictor neurons changes in neuropathic animals.^{11 ,31}

There are several other symptoms of CRPS I that favor a central origin of the disorder: (1) Hyperhidrosis, a typical feature of many patients with CRPS I, cannot be explained by a peripheral mechanism because, in contrast to blood vessels, sweat glands do not develop denervation supersensitivity.³² Therefore, an increase in sweating must be explained by an increase in activity in sympathetic sudomotor neurons that is of central origin.^{33 - 34} (2) Impairment of muscle strength involving all muscles of the affected distal extremity that cannot be explained by pain, edema, or severance of peripheral nerves also are the result of a centrally mediated impulse abnormality in the motor neuron pool projecting to the distal extremity. Also, a neglectlike syndrome responsible for severe motor dysfunctions described recently³⁵ points to a central mechanism. (3) Moreover, an increased physiologic tremor, present in approximately 50% of patients with CRPS I, is caused by central changes.³⁶

Interruption of the efferent sympathetic nerve supply to the affected extremity may relieve the pain in patients with CRPS I. The pain is therefore called "sympathetically maintained pain." Also, in the patient presented herein, use of regional guanethidine blocks relieved the pain. This observation seems to contradict the finding that the sympathetic outflow to the affected limb is already inhibited or abolished. However, the sympathetic tests used in this study exclusively assess the function of cutaneous vasoconstrictor neurons. It is possible that although the cutaneous sympathetic outflow is inhibited, the sympathetic innervation of deeper tissues such as muscle or bone is normal or even enhanced. Results of recent animal experimental studies³⁷ indicate that separate functional channels of the sympathetic nervous system exist that can be activated selectively and independently, eg, the cutaneous vs the muscle vasoconstrictor outflow. The pathological interaction of sympathetic fibers and nociceptive afferents that builds the basis for sympathetically maintained pain might therefore involve sympathetic neurons innervating deep tissues rather than cutaneous sympathetic neurons. Evidence for the involvement of deep tissues includes abnormalities in the 3-phase bone scan³⁸ and the appearance of the spontaneous pain that characteristically is located deep in the affected extremity.

Several conclusions can be summarized from this case of CRPS I (RSD). (1) In the early stages of CRPS I, cutaneous sympathetic vasoconstrictor reflexes may be completely abolished on the affected side. (2) This loss of response is not caused by anatomic damage of sympathetic nerve fibers. (3) Although an intense vasodilatory mechanism that mimics the loss of vasoconstrictor response cannot be ruled out completely, it is reasonable to consider a functional unilateral inhibition of sympathetic activity to be the likely cause. (4) The pathophysiological mechanism of these sympathetic abnormalities is located in the central nervous system. (5) The functional sympathetic inhibition may be reversible within weeks of the disease course, leading to complete recovery of symptoms. (6) Sympathetically maintained pain is unlikely mediated by cutaneous vasoconstrictor fibers in the early stage of CRPS I. The underlying sympathetic-afferent interaction might be located in deep tissue, ie, bone or muscle.

Accepted for publication June 26, 1998.

This work was supported by the Wilhelm Sander-Stiftung, Munich, and the Alexander von Humboldt-Stiftung, Bonn, Germany.

We thank G. Deuschl, MD, director of the Neurology Clinic, University of Kiel, Kiel, Germany, for providing facilities for the study; H. L. Fields, MD, PhD, and P. Green, PhD, for their constructive criticism of the manuscript; and U. Ertl and B. Luig for their excellent technical assistance.

Reprints: Ralf Baron, MD, Klinik für Neurologie, Christian-Albrechts-Universität Kiel, Niemannsweg 147, 24105 Kiel, Germany (e-mail: r.baron@neurologie.uni-kiel.de).