OUP user menu

Neurophysiological effects of stimulation through electrodes in the human subthalamic nucleus

P. Ashby, Y. J. Kim, R. Kumar, A. E. Lang, A. M. Lozano
DOI: http://dx.doi.org/10.1093/brain/122.10.1919 1919-1931 First published online: 1 October 1999


The effects of stimulation through macroelectrodes implanted in the subthalamic nucleus (STN) were studied in 14 patients with parkinsonism. Single stimuli delivered directly to the STN electrodes with an external stimulator modulated voluntary electromyography (EMG) of contralateral muscles in most patients. A short-latency facilitation (`peak') was attributed to the activation of the corticospinal system. A longer latency inhibition (`dip'), often preceded or followed by facilitations, appeared to arise from the activation of large-diameter fibres running parallel to the electrode and to be transmitted through the motor cortex. It is possible that the dip could result from the inhibition of thalamocortical neurons. With high-frequency stimulation (~100 Hz) the peaks occurred at the stimulus frequency; the dips became confluent and outlasted the duration of the stimulus train. There was no evidence that high-frequency stimulation produced `blocking'. The studies were repeated in 12 patients a mean of 5.8 months after implantation of the stimulator. The same short-latency effects were obtained. They were present on 7 out of 23 sides at the settings in use and on the majority of sides if the stimulus intensity was slightly increased. There was no clear relationship between these short-latency effects and the patients' overall clinical improvement; the effects may result from the spread of current to large-fibre systems near the STN. In five patients, high-frequency stimulation on one side immediately reduced tremor in the contralateral limbs. This effect arose from the activation of large-diameter fibres and, like the dip, had about the same threshold at each of the contacts. Frequencies as low as 70 Hz were sufficient. We conclude that the control of tremor by STN stimulation is due to the activation of a large-fibre system.

  • parkinsonism
  • subthalamic nucleus
  • deep brain stimulation
  • C–T = condition–test
  • FDI = first dorsal interosseous
  • GPE = globus pallidus externus
  • GPI = globus pallidus internus
  • MEP = motor evoked potential
  • MPTP = 1-methyl-4-phenyl-1,2,3,6 tetrahydropyridine
  • SNPR = substantia nigra pars reticulata
  • STN = subthalamic nucleus
  • UPDRS = Unified Parkinson's Disease Rating Scale


According to current concepts of the basal ganglia (De Long et al., 1990; Levy et al., 1997), the subthalamic nucleus (STN) facilitates the globus pallidus internus (GPI) and the substantia nigra pars reticulata (SNPR), augmenting the inhibition of thalamocortical neurons and, in this way, reducing motor output.

There is considerable support for this model. The firing rate of STN neurons is increased in monkeys rendered parkinsonian with 1-methyl-4-phenyl-1,2,3,6 tetrahydropyridine (MPTP) (Bergman et al., 1994). The parkinsonian features can be reduced by the injection of GABA (γ-aminobutyric acid) agonists into the STN (Aziz et al., 1991; Guridi et al., 1996), by radio-frequency lesions of the STN (Wichmann et al., 1994) or by high-frequency stimulation of this nucleus (Benazzouz et al., 1993, 1996). In human patients with parkinsonism, spontaneous haemorrhages (Yamada et al., 1992; Sellal et al., 1992) or surgical lesions of the STN (Gill and Heywood, 1997; Rodriguez et al., 1998) may alleviate the parkinsonian features, and high-frequency stimulation of the nucleus through chronically implanted electrodes has been reported to have the same effect (Limousin et al., 1995, 1998; Kumar et al., 1998; Krack et al., 1998a, b).

Because high-frequency stimulation appears to have the same clinical effect as a lesion, stimulation is assumed to `block' STN neurons. But the current model may not reflect the true complexity of the basal ganglia (Brown and Marsden, 1998; Parent and Cicchetti, 1998) and there are several other neural systems in the vicinity that could be excited or blocked by stimulation at ~70–100 Hz. What actually happens when stimuli are delivered through macroelectrodes in the region of the human STN?

To address this question we examined 14 patients with bilateral STN stimulators. Single cathodal stimuli (which can only excite neural elements) were used to see which neurons were activated within the range of the electrode. High-frequency stimulation was used to see if these effects persisted or were blocked. From the outset it was recognized that the STN has only indirect effects on motor function and that single stimuli might not have any obvious short-latency effects; also, that the nucleus is surrounded by large-fibre systems that might be recruited by low-threshold stimuli. Short-latency effects from these fibre systems would serve only to indicate the position of the electrode relative to these peri-STN pathways, perhaps to identify misplaced electrodes, or to explain certain side effects.


Studies were carried out on 14 patients with advanced Parkinson's disease who had had quadripolar macroelectrodes implanted in the STN bilaterally. The electrodes were positioned stereotactically and the final target was determined by stimulating and recording with a microelectrode (Hutchison et al., 1998). For a few days after this stereotactic procedure, the wires from each of the four contacts were led out through the scalp and connected to an external stimulator or amplifiers for testing. At a second surgical procedure the leads were connected to a totally implanted programmable stimulator (Itrel model 3625; Medtronic, Minneapolis, Minn., USA) which the patient could switch on and off with a magnet. The initial studies were carried out in the interval between the two surgical procedures, the follow-up studies several months later. The study had the approval of the Institution of the Human Experimentation Committee, Toronto, and all subjects provided informed consent.


The macroelectrodes (model 3387m DBS; Medtronic) had four platinum–iridium cylindrical surfaces 1.27 mm in diameter, 1.52 mm in length, with a 3-mm centre-to-centre separation. Contact 0 was the deepest, most caudal electrode. For the initial studies, stimuli were delivered through adjacent pairs of electrodes (cathode = –, anode = +; e.g. 0+1–, 1+2–, 2+3–) or through one electrode (e.g. 0–) with the anode over the cheek or chest. Stimuli were generated with an isolated, constant-current stimulator (model A360D-B; World Precision Instruments, Sarasota, Fla., USA). Pulse widths of 50 μs to 4 ms (usually 100 μs) and currents up to 10 mA were used. Single or paired stimuli or trains of stimuli were delivered at random intervals between 2.25 and 2.75 s controlled by a computer. For the follow-up studies the implanted stimulator was programmed to deliver stimuli at various frequencies and intensities, and the stimulus artefact, picked up with surface electrodes, was used to trigger the averaging programs. Pulse widths were usually 60 μs. The stimulus intensity of the stimulator was graduated in volts. If the electrode impedance were 1 KΩ, 1 V would generate a current of ~1 mA.

Surface EMG recordings

EMG was recorded with surface electrodes over the muscles of interest, always including the contralateral first dorsal interosseous (FDI). The signal was amplified 5000–50 000 times with a bandpass of 100 Hz to 1 kHz. The raw EMG was monitored on an oscilloscope and passed to a leaky integrator (time constant 600 ms). The level of integrated EMG was displayed on a second oscilloscope with a line representing the EMG associated with 30% maximum contraction. The subject maintained this level of contraction while stimuli were delivered through the STN electrodes. The EMG signal was digitized at 2 kHz and rectified, and 50–100 sweeps were averaged. The mean of the 50-ms prestimulus portion of the averaged rectified EMG was taken as baseline and this value was normalized to 100%. Zones in the post-stimulus period that exceeded the prestimulus mean ± 2 SD were identified as peaks or troughs. The area of a peak or trough above or below the mean was measured from the preceding and following crossings of the mean. The area was normalized by dividing it by the prestimulus mean and expressed in arbitrary units. As an example see the middle trace of the top left panel of Fig. 1, where stimulation (6 mA, 100 μs) of the right STN through contacts 0+1– produced a peak with an area of 11 units followed by a dip with an area of 9 units. To compare the effects in different muscles, the area of the peak 9 of increased EMG of each muscle was expressed as a percentage of the area of the equivalent peak in the FDI.

Fig. 1

Effects of STN stimuli on voluntary EMG of the contralateral FDI in subject 6. Single stimuli of 100 μs duration at the intensities shown on the left were delivered through electrodes to the right STN (A) or left STN (B) at time zero. Each trace is the average of 50 sweeps of rectified EMG. Dark areas in the post-stimulus period represent zones where the EMG exceeded the prestimulus mean by ± 2 SD. Note that stimuli of <5 mA had no effect in this subject. Strong stimuli produced a short-latency (<27 ms) facilitation (`peak') which was largest at the most caudal electrodes (0+1–). In this and subsequent figures, each trace has been normalized to the mean of the 50 ms prestimulus portion. This normalized mean is represented by the distance from the horizontal time scale to the mean of the prestimulus portion of the lowest trace, and the other traces are plotted at the same scale.

Single motor unit recordings

The action potentials of single motor units, activated by a gentle voluntary contraction, were recorded with an intramuscular concentric needle electrode (13150; Dantek, Skovlunde, Denmark) and isolated with a window discriminator. Peristimulus time histograms of the discharge times of the motor unit relative to stimuli delivered through the STN electrodes were generated with a computer.

Evoked potentials

Evoked potentials were recorded from adjacent pairs of STN electrodes in response to stimuli delivered to the median nerve at the contralateral wrist at the threshold of the alpha motor neuron axons or to flashes from a strobe light placed 0.5 m in front of the subject. The signals were amplified 500 000 times using a bandpass of 2 Hz to 1 kHz, and 100 sweeps were averaged.

Transcranial magnetic stimulation

Following safety studies (Kumar et al., 1999), a Magstim 200 (Magstim, Dyford, UK) connected to a circular coil 13 cm in diameter was used to deliver transcranial stimuli to the motor cortex. The coil was positioned so as to generate a motor evoked potential (MEP) in the contralateral FDI with the lowest threshold. Magnetic stimuli were given alone or at various intervals after a stimulus to the ipsilateral STN in random order and at random intervals. Twenty motor evoked potentials were averaged at each condition–test (C–T) interval. The averaged MEP amplitude with the subject at rest was used to measure the excitability of the corticospinal system following an STN stimulus. Suppose that STN stimuli caused facilitation of voluntary EMG at 26 ms and inhibition at 60 ms and that the MEP latency was 25 ms; if these effects were transmitted via the fast-conducting corticospinal pathway they would be leaving the cortex 1 and 35 ms after the STN stimulus. Thus, 1 and 35 ms would be the appropriate C–T intervals to test the excitability of the cortex.


Initial neurophysiological studies

The 14 patients were aged between 40 and 76 years, had had parkinsonism for 6–21 years and had preoperation total motor unified Parkinson's disease rating scale (UPDRS) scores (Part III), off medication, of 37–82 (possible scores, 0–108) (Table 1).

View this table:
Table 1

Data on the 14 patients with bilateral STN stimulators

PatientAge (years)SexDuration of disease (years)Total UPDRSRight 0+1−Right 1+2−Right 2+3−Left 0+1−Left 1+2−Left 2+3−Overall scoreGood tremor control
F = female; M = male; Total UPDRS = score on preoperative Unified Parkinson's Disease Rating Scale, Part III (possible scores0–108) off medication. The neurophysiological effects of single stimuli given through the right and left STN electrodes include: 0 = no effect; P, p = large (e.g. seen with stimuli >5 mA peak) or small peak; D, d = large or small dip. Overall score at follow-up: 0 = not improved; 4 = much improved. Patients with good tremor control are indicated by +; NA = not applicable (i.e. no tremor or tremor not immediately controlled when stimulator turned on).
440M 639.25DDDPD0NA
1455F 682.5dddPDDd4+

Motor effects of single stimuli delivered through STN electrodes

We looked first for short-latency motor effects by delivering single stimuli (~1 Hz) through adjacent pairs of STN electrodes (0+1–, 1+2–, 2+3–) while averaging rectified EMG from the contralateral FDI. With low-intensity (100 μs, <5 mA) stimulation, three types of response were seen: (i) no response (Figs 1 and 2, right STN); (ii) short-latency facilitation (<27 ms), which we will call the `peak' (Fig. 3, left STN); or (iii) a longer-latency (~50 ms) inhibition, which we will call the `dip'. This dip was frequently followed by facilitation and was sometimes preceded by a period of facilitation distinct from and later than the short-latency peak (Fig. 2, left STN; Fig. 3, right STN). The thresholds for these effects were as low as 1 mA. As the stimulus intensity or pulse width was increased, these motor effects, especially the dip, were more likely to occur (Figs 1–3 and Table 2). Monopolar stimulation produced much larger peaks and dips for a given voltage than bipolar stimulation in the three patients in whom these forms of stimulation were compared.

View this table:
Table 2

Frequency of the various effects (peaks and dips) on voluntary EMG of the contralateral FDI produced by bipolar stimulation through adjacent pairs of electrodes at various intensities and pulse widths

StimulusNo effectEarly peakDip
Data are from 77 electrode pairs in 14 subjects. Stronger stimuli were more likely to give rise to a peak or dip.
100 ms, <5 mA411224
100 ms, >5 mA181841
1 ms, <5 mA121946
1 ms, >5 mA 32351
Fig. 2

Effect of single STN stimuli on voluntary EMG of the contralateral FDI in subject 13. Traces as for Fig. 1. Note that stimulation through the right STN electrodes (A) at <5 mA had very little effect. Stimulation through the left STN electrodes (B) resulted in inhibition of voluntary EMG (dip) preceded and followed by facilitation. The dip had about the same magnitude at all three sites. At higher stimulus intensities there was an additional short-latency facilitation (peak) which had the lowest threshold and highest amplitude at the lower electrodes (0+1–).

Fig. 3

Effect of single STN stimuli on voluntary EMG of the contralateral FDI in subject 11. Traces as for Fig. 1. Note that stimulation through the right STN electrodes (A) resulted in inhibition (dip) which was about the magnitude same at each site. Stimulation through the left STN electrodes (B) resulted in a short-latency facilitation (peak) maximum at the lowest electrode (0+1–).


The mean latency of the early peaks was 21.5 ms (SE = 0.7 ms, range 18–26, n = 15). This mean latency was obtained from runs using a 100 μs pulse and the maximum stimulus intensity used in that subject, and where the peaks or dips were >10 units (see Methods). The peaks were significantly larger with stimulation through the deepest, most caudal electrode pair (see Methods) (0+1–) than at the other two sites (0+1– versus 1+2–, t = 2.65, P = 0.014; 0+1– versus 2+3–, t = 2.9, P = 0.009) (Fig. 4). For these calculations we used those data for which the identical stimulus (100 μs, 10 mA) had been used for all three of the electrode pairs in that subject (23 sides). The distribution of the short-latency facilitation (studied in detail in one subject) was strictly contralateral; it was largest in the distal upper limb muscles although quite substantial in the proximal upper limb muscles. The latency of the early peak was on average 3.8 ms shorter than the motor potential evoked by transcranial magnetic stimulation on the same side (n = 15). Peristimulus time histograms of single units in FDI were studied in two subjects. The increase in firing probability had a mean duration of 2 ms (range 0.6–3 ms, n = 5).

Fig. 4

Comparison of the amplitudes of the peaks (top) and dips (bottom) produced by single stimuli given through electrodes 0+1–, 1+2– and 2+3–. Note that the peaks (n = 23) were largest at the most caudal electrodes (0+1–) and the dips (n = 23) were about the same amplitude at the three sites (slightly larger at the most caudal electrodes).


The dips had a mean onset latency of 41.7 ms (SE = 0.9 ms; range 31–55 ms, n = 48). These data were obtained from runs using a 100 μs pulse and the maximum stimulus intensity used in that subject and where the dips were >10 units. The dip occurred alone at 13 out of 48 sites, was followed by facilitation at 23 out of 48 sites and was preceded by a facilitation (later than the early peak) at 12 out of 48 sites. Examples are shown in Fig. 2 (left STN) and Fig. 3 (right STN). The dips were of similar magnitude at each of the electrodes (Fig. 4). They were slightly larger at the deepest electrode pair but the difference was not as pronounced as for the peaks (0+1– versus 1+2–, t = 1.8, P = 0.08; 0+1– versus 2+3–, t = 2.5, P = 0.02). For these calculations we used only those data for which the same stimulus (100 μs, 10 mA) had been tested at all three electrode pairs in that subject (23 sides). The result was the same if the runs with a large early peak, which might have contributed to the dip, were excluded. When single STN stimuli produced inhibition of voluntary EMG, the same stimuli depressed the motor evoked potential produced by transcranial magnetic stimulation with the subject at rest at corresponding latencies in the four subjects tested (Fig. 5).

Fig. 5

Effect of single stimuli delivered through the STN electrodes on the MEP in the contralateral FDI produced by transcranial magnetic stimulation with the subject at rest. Pooled data from four subjects (4, 9, 12 and 13) in whom single stimuli produced inhibition of EMG (as in Fig. 3, right STN). The condition–test intervals have been made comparable to the latency of the inhibition of voluntary EMG by adding the latency of the MEP (see Methods).

Relationship of the peak and dip to the position of the electrode

Postoperative MRI scans were available on 10 out of 14 patients. It was difficult to judge the relationship of the electrodes to the STN, but the electrode tip generally extended down to the brainstem at the level of the red nucleus, and at this level the relationship to the peduncle and red nucleus could be measured from the films. The tip of the electrode at this level was, on average, 9.7 (SE = 0.5) mm from the midline, 10.2 (SE = 0.4) mm posterior to the anterior border of the peduncle, 4.5 (SE = 0.3) mm from the estimated position of the corticospinal tract in the middle of the peduncle and 5.4 (SE = 0.6) mm from the centre of the red nucleus (n = 20).

In several subjects the motor effects were quite different on the two sides (Table 1). This provided an opportunity to make anatomical correlations. For example, in subject 11 (Fig. 3), stimulation on the left produced a peak. The electrode tip on this side was 3.1 mm from the peduncle and 6.3 mm from the centre of the red nucleus. Stimulation on the right produced a dip. The electrode tip on this side was 7.5 mm from the peduncle and 1.3 mm from the centre of the red nucleus. Figure 6 is a reconstruction of the positions of the electrode tips relative to the anterior border of the peduncle and to the midline for the 10 patients for whom adequate MRI data were available. The peaks (open circles) arose from more lateral sites than the dips (filled circles). We also measured the distance of the electrode from the centre of the red nucleus and the estimated position of the corticospinal tract in the middle of the peduncle. When stimulation resulted in a peak (n = 6) the electrode was closer to the peduncle. When stimulation resulted in a dip (n = 10) the electrode was closer to the red nucleus (Table 3).

View this table:
Table 3

Mean distance (mm, SE in brackets) of the electrode tip from the corticospinal tract in the cerebral peduncle and from the centre of the red nucleus when stimulation produced a peak or a dip

Peak (n = 6)Dip (n = 10)
When stimulation produced a peak, the electrodes were closer to the peduncle. When stimulation produced inhibition, the electrodes were closer to the red nucleus. *Significant difference between means in that column; **significant difference between means in that row.
Peduncle4.3* (0.6)5.5* (0.6)
Red nucleus6.9** (0.78)2.3** (0.5)
Fig. 6

Reconstruction of the positions of the most caudal electrode in relation to the midline and to the anterior border of the peduncle. The sites where short-latency facilitation was obtained (open spheres) are more lateral than those where inhibition occurred (filled spheres). The zig-zag line represents the anterior border of the peduncle. The two circles represent the mean diameter of the red nuclei positioned at the mean distance from the anterior border of the peduncle. These landmarks were measured at the 5 mm vertical plane, where the red nucleus had its greatest diameter, but are represented at the zero plane for clarity. Vertical lines have been dropped from the electrode positions to the zero plane to show their x, y coordinates.

Recordings from STN electrodes

Small evoked potentials could be recorded from the STN electrodes following stimulation of the contralateral median nerve at the wrist. The mean onset latency was 14.9 ms (SE = 1.5, n = 84) and the mean amplitude 1.1 μV (SE = 0.9, n = 84). Neither the amplitude nor the latency differed significantly between electrodes (0+1–, 1+2–, 2+3–). There was a weak correlation between the amplitude of the evoked potential and the magnitude of the dip (r = 0.36, P < 0.001) but no correlation with the amplitude of the peak. No visual evoked potentials could be recorded in any of the seven patients tested.

Trains of stimuli

When single stimuli produced a peak, continuous stimulation at up to 100 Hz produced multiple peaks without evidence of attenuation (Fig. 7). In three patients, in whom single stimuli gave inhibition with or without later facilitation, trains of stimuli produced prolonged inhibition of voluntary EMG which could outlast the stimulus (Fig. 8). Train rates of >100 Hz were required (n = 1).

Fig. 7

Effect of continuous left STN stimulation at 100 Hz (120 μs, 2+3–; voltages shown on the left) on voluntary EMG of the right FDI in subject 11. Single stimuli (100 μs) to the left STN in this subject produced short-latency facilitation (Fig. 3, left STN). With 100 Hz stimulation, the peaks occurred at 100 Hz.

Fig. 8

Effect of trains of stimuli (200 Hz, 100 μs, 0+1–) to the right STN on voluntary EMG of the left FDI in subject 11. Single stimuli to the right STN in this subject produced inhibition (Fig. 3, right STN). Trains of stimuli of duration 20–1000 ms produced inhibition which outlasted the stimulus train.

Follow-up neurophysiological studies

The 14 patients were reviewed at a mean follow-up time of 5.8 months. The patients were given a score according to their overall response to stimulation by the clinician most familiar with their programming and management (0 = no effect, 4 = marked improvement) (see Overall score in Table 1). Ten of the patients had improved (three of them quite dramatically); four had not improved, and two of these were no longer using their stimulators. The clinical results of a double-blind assessment of the first consecutive eight patients retaining bilateral stimulators has been reported (Kumar et al., 1998).

The 12 patients using their implanted stimulators (23 sides) were restudied. At the time of the study 13 sides had been set for monopolar stimulation and 10 sides for bipolar stimulation. The mean electrode impedance was 1594 Ω (range 729–2000), slightly higher than that obtained at the initial study (1156 Ω, range 645–1957; t = 4.26, P < 0.001, n = 36). The pulse width was usually 60 μs (90 μs in five, 120 μs in one), the rate usually 130 Hz (160 Hz in one, 185 Hz in five) and the mean voltage 3.4 V (monopolar) and 4.0 V (bipolar) (range 2.3–8 V).

In order to observe the effects of individual stimuli at the electrode intensities and pulse widths being used for clinical benefit, the implanted stimulator rate was programmed to 2 Hz and the stimulus artefact was used to trigger the computer. On seven out of 23 sides, short-latency modulations of the contralateral EMG were seen (a peak on three sides and a dip on four, no effect on 16). If the stimulus intensity was increased, short-latency effects occurred on all except two sides (a peak on six sides and a dip on 15 and no effect on two). The threshold for these effects was an average of 2.1 V higher than the patients' current therapeutic settings. Curiously, the three patients with the greatest clinical improvement (overall score = 4) all had short-latency effects on one or other side at the settings used (a peak in two and a dip in one) and the mean threshold for a short-latency effect (five sides) was 0.1 V lower than the voltage currently in use (range –2.5 to +2.1 V). In each subject the short-latency effects were the same as those recorded at the initial examination.

Relationship between neurophysiological findings and overall benefit

We looked for a relationship between initial and follow-up motor effects and the overall score but found none. For example, the seven subjects (13 sides) who did well (those with overall score 3 or 4) had, at the settings used clinically but with the rate reduced to 2 Hz, two peaks, two dips and nine `no effects'. When the stimulus current was increased, there were four peaks, seven dips and two `no effects'. Of the seven subjects who did less well (those with an overall score of 0, 1 or 2), five were using their stimulators. They had one peak, two dips and seven `no effects' at the clinical settings being used and, when the current was increased, two peaks and eight dips.

Control of tremor

In five patients, high-frequency stimulation on one side (bipolar in three, monopolar in two) produced an immediate reduction in contralateral tremor. Short-latency motor effects to single stimuli were seen at these settings in all five (three peaks, two dips). We plotted strength–duration curves with pulse widths 60, 130, 210 and 450 μs for the thresholds for tremor control (at 130 Hz) and for the appearance of the peak or dip (at 2 Hz). The strength–duration curves (Fig. 9) were similar, suggesting that all three effects arose from the activation of neural elements with similar characteristics. To see if these effects came from the same neural elements, we compared the thresholds for tremor control and for the peak or dip at each electrode in turn (Fig. 10). The thresholds for the peak and for tremor control separated progressively at the more rostral electrodes but the thresholds for the dip and for tremor control appeared to be closer. In three subjects we altered the stimulus frequency systematically. In all three subjects tremor control began when the stimulus frequency was as low as 70 Hz.

Fig. 9

Strength–duration curves for the control of tremor at 130 Hz (triangles, subjects 5, 6, 7, 12 and 13) and for the thresholds of the peak (open circles, subjects 5 and 13) and for the dip (filled circles, subjects 7 and 12) produced by stimulation at 2 Hz (pulse widths 60, 120, 210 and 450 μs). Note that the peaks and dips and the control of tremor could all be obtained with narrow pulse widths, suggesting that they arose from neural elements with similar low thresholds.

Fig. 10

Thresholds at various electrodes for the control of tremor with stimulation at 130 Hz (triangles) and for the appearance of the peak (open circles) or dip (filled circles) at 2 Hz in five subjects. Bipolar stimulation was used in the subjects on the left side of figure (from the top down, subjects 6, 5 and 7) and monopolar stimulation in the subjects on the right side (top, subject 13; bottom, subject 12). Note that the threshold for the peak was lower at more caudal contacts (0, 1) but that the thresholds for tremor control and for the dip were similar for each electrode.


At present it is not clear which neural elements are affected by high-frequency stimulation of the STN or whether they are activated or blocked. In general, electrical stimulation within the neuropil excites axons before cell bodies, large axons before small, those near the cathode before those near the anode, and those running parallel to the flow of current before those running transversely to it (Ranck, 1975, 1981; Nowak and Bullier, 1998a, b). Large-diameter axons have the shortest refractory periods. Thus, the neural elements most likely to be recruited at low threshold are the least likely to be blocked. Stimulation at 70 Hz is unlikely to block large axons, although blocking could occur at their fine nerve terminals or synapses (O'Mara et al., 1988). STN neurons in the rat can discharge at up to 500 Hz (Nakanishi et al., 1987); presumably their efferents are expected to carry impulses at that rate.

Neural systems that might be activated by STN stimulation

As the electrode contacts have a centre-to-centre separation of 3 mm, only one or two contacts of the macroelectrode will actually be in the STN, the others being in adjacent structures (Hutchison et al., 1998). The dorsolateral STN is the zone that is considered to be most involved with the motor function of the limbs. The largest number of STN efferents arises from cells in the lateral two-thirds of the STN and project to the globus pallidus externus (GPE). A much smaller number of cells in the medial region of the nucleus project to the GPI (Parent, 1995). The axons of STN neurons branch within the STN, sending collaterals to the GPE, the GPI and the SNPR. The fibres pass caudoventrally to the SNPR and rostrolaterally to the GPI (Sato et al., 1998). There are also projections to other nuclei (Parent and Cicchetti, 1998).

According to current models of the basal ganglia (DeLong, 1990; Levy et al., 1998), blocking the neurons which facilitate the GPI and SNPR would enhance motor output, and this, of course, is the rationale for high-frequency stimulation of the STN. But increased motor output could also result from activation of the efferent fibres from the STN to the GPE or from activation of the afferents from the GPE to the STN; orthodromic activation of these afferents would inhibit the STN; antidromic activation, if there were collaterals, would inhibit the GPI (Fig. 11). There are also a number of fibre systems which pass close to the STN. The lenticular fasciculus lies just rostral and dorsal to the STN. Blocking these fibres from the GPI to the thalamus would enhance motor output and blocking the SNPR, which lies just below the STN, would have the same effect.

Fig. 11

Diagram of some of the pathways in the vicinity of the STN relevant to the effects of STN stimulation. Facilitatory neurons are white; inhibitory neurons are black. Ansa = lenticularis; f = fasciculus; other abbreviations are defined below the Summary.

The relative sizes of nerve fibres in the region of the STN, and thus their recruitment order, have not been studied in detail. Many of the fibre systems connecting the STN to other structures have fine branches or a variable, beaded configuration that would make generalizations about average size difficult (Sato et al., 1998). The largest fibre system in the vicinity is probably the corticospinal tract (~4–5 μm). The fibres from the GPI to thalamus may be ~2–3 μm and those from the STN to the GPE and GPI are ~1–2 μm (A. Parent, personal communication, 1998). In the rat the axons from the STN to the SNPR conduct at 0.05–2 m/s (~0.3 μm) and those from the cortex to the STN at 2–8 m/s (~1 μm) (Kitai and Deniau, 1981).

STN stimulation in animals

In normal rats, stimulation of STN can increase or decrease the firing rate of neurons in the endopeduncular nucleus (the rat equivalent of the GPI) and in the SNPR (Benazzouz et al., 1995). Single stimuli (60 μs, 0.5 mA) facilitated SNPR neurons. So did trains of stimuli lasting 5 s if the rate of stimulation was <50 Hz. But with trains of more rapid stimulation (130 Hz, 60 μs, 0.3 mA) the firing rates of neurons in the GPE were increased (for the subsequent 50–160 s) and the firing rates of SNPR and EP neurons were depressed for the subsequent 50–120 s. The authors postulated that the stimuli either activated GPE axons retrogradely inhibiting GPI or blocked STN axons. Gao and colleagues (Gao et al., 1997) showed that STN stimulation in the rat increased the firing rate of ventrolateral thalamic neurons. Narrow pulses and slow stimulus rates (20–50 Hz) were sufficient, implying that the effect arose from neural activation rather than blocking.

We used single stimuli to determine which elements were within range of the stimulating electrode and which might be excited or blocked by high-frequency stimulation. We found two main effects from single stimuli.

The short-latency peak

The short-latency peak is probably generated by the spread of current to the corticospinal fibres in the peduncle. The strength–duration curves indicate that the peak arises from the activation of large myelinated axons. The fibres project to the motor neurons of contralateral muscles, particularly those of the distal upper limb, and the single motor unit studies suggest that this projection is monosynaptic. The latencies of the peaks were a few milliseconds shorter than similar peaks induced by transcranial magnetic stimulation in the same subjects, implying that the corticospinal system is activated deep in the brain. Larger peaks occurred with stimulation through the most caudal electrodes and with electrodes placed more laterally, closer to the peduncle, so the corticospinal system is probably activated at that level. It is unlikely that the peaks arise from the antidromic activation of the projections from the motor cortex to the STN because they are small (Kitai and Deniau, 1981; Parent, 1995). The corticospinal fibres responsible for the peak are capable of transmitting at 100 Hz without any evidence of blocking. In clinical practice the recruitment of this fibre system produces a tonic contraction of contralateral muscles that limits the stimulus intensity that can be used. It is unlikely that subclinical activation of the corticospinal system has any therapeutic effect, although the antidromic effects of high-frequency activation of efferents from the motor or prefrontal cortex are unknown (Ashby et al., 1998).

The dip

The dip and its flanking facilitations also appear to arise from the activation of large axons. The dip was seen more often with medial electrode placements and was obtained at much the same threshold at caudal and rostral electrodes, implying that the responsible fibre system is aligned with the longitudinal axis of the electrode. The modulations appear to be transmitted through the motor cortex as there are parallel fluctuations in cortical excitability.

The dips bear some resemblance to the modulations produced by stimulating through electrodes targeted at the cerebellar thalamus [cf. Fig. 2, left STN, above (present paper) with the middle panel of Fig. 1 in Strafella et al., 1997]. But the small somatosensory evoked potentials recorded with STN electrodes compared with those recorded with electrodes targeted to VIM (ventro intermedius nucleus of the thalamus) indicate that the STN electrodes are further from the sensory nucleus of the thalamus. Nevertheless, the cerebellothalamic fibres lie just posterior to the thalamic fasciculus and might be within range of the STN electrodes. Alternatively, the dip could be produced by direct activation of the lenticulate or thalamic fasciculi or the SNPR, or by the indirect activation of the GPI or SNPR via the facilitatory projections to them from the STN. In each case the result would be a transient inhibition of thalamocortical cells. The dip was often followed by a facilitation and sometimes by a further dip. This oscillatory activity was too fast to represent entrainment of motor neurons and its period varied with stimulus intensity (Fig. 2) so it was probably generated centrally.

When single stimuli gave rise to a dip, trains of stimuli could produce prolonged inhibition that outlasted the stimulus train. There was no evidence that this inhibitory mechanism was `blocked'.

Relationship between neurophysiological findings and overall therapeutic benefit

The peaks and dips, although constant in a given individual, varied considerably between patients and between the two sides of a single patient (Figs 1–3 and Table 1). These differences must reflect the `physiological position' of the contacts. We had hoped to find some relationship (positive or negative) between these short-latency motor effects and the ultimate therapeutic benefit that could be used to identify an appropriate electrode position in the operating room and to shorten the time required for programming. We were unable to do so. This may be because the number of patients was small and clinical benefit multifactorial. But it is possible that there are no short-latency effects from single-pulse activation of STN neurons and that the peak and dip reflect the activation of neighbouring fibre systems that have no effects on bradykinesia.

System responsible for the control of tremor

In five patients tremor of the contralateral limbs could be immediately curtailed by high-frequency stimulation, confirming recent reports (Rodriguez et al., 1998; Krack et al., 1998a). Strength–duration curves indicated that tremor control arose from the activation of a low-threshold (presumably large-fibre) system. Frequencies as low as 70 Hz were sufficient. Since large fibres should be able to transmit at 14 ms intervals, the effect may have been due to activation rather than `blocking' of neural elements. The threshold for tremor control was about the same at each of the contracts, implying that the site of activation of the responsible neural elements was not confined to the STN. By comparing the relative thresholds for tremor control, the peak and the dip at various electrodes (Fig. 10), it was clear that tremor control and the short-latency peak arose from separate fibre systems, while tremor control and the dip could not be distinguished on this basis. Stimulation through electrodes implanted in the ventrolateral thalamus for the control of tremor also give rise to a dip (Strafella et al., 1997). Thus it is possible that the neural system responsible for the dip is also involved in the control of tremor.


View Abstract