Pallidal stimulation for Parkinson's disease

From the early 1940s, stereotactic surgery of the pallidum and the thalamus was performed in Parkinson's disease (PD), leading in particular to an improvement in rest tremor and rigidity.^1-6 However, side effects and a recurrence of parkinsonian signs were frequently observed. The introduction of levodopa in 1967 as a treatment for the disease⁷ largely replaced surgical treatment, except for intractable disabling tremor.

Recently, there has been resurgence of pallidotomy for advanced PD.^8-13 Pallidotomy appears to be highly effective in reducing levodopa-induced dyskinesia (LID), a troublesome complication of long-term medical treatment of PD. The effect of this surgery on cardinal signs of the disease seems to be less consistent, ranging from marked^8,9,11 or moderate^10,13 improvement to absence of efficacy.¹² Moreover, with current concepts of the functional organization of basal ganglia, it is difficult to explain how pallidotomy simultaneously improves both LID and PD signs. It is generally accepted that in PD, dopamine deficiency leads to an overactivity of the internal globus pallidus (GPi) and the subthalamic nucleus(STN).^14,15 From this model, it is therefore assumed that lesioning the GPi or the STN results in an improvement of motor parkinsonian signs. In contrast, the occurrence of LID would logically be associated with reduced GPi and STN activity. Thus, the suppression of LID by pallidotomy appears paradoxical.

Since 1985, Benabid et al.¹⁶ have been using chronic high-frequency deep brain stimulation (DBS) as an alternative nonlesioning surgical treatment for PD. DBS originally targeted the ventral intermediate thalamic nucleus and was shown to be a highly effective treatment for tremor, with results similar to those obtained with thalamotomy.¹⁶ More recently, DBS has also been applied to the STN¹⁷ and the GPi,¹⁸ with a significant reduction in parkinsonian disability.

The effect of DBS is reversible, and the stimulation parameters(principally site and voltage) can be varied to achieve the optimal therapeutic result. In addition to their therapeutic importance, postoperative data also provide an excellent means of exploring the physiology and pathophysiology of the basal ganglia. In fact, the use of quadripolar electrodes allows monopolar stimulation at four different sites covering the dorsoventral extent of the globus pallidus (GP). We studied, therefore, the effect of stimulation on parkinsonian signs and on LID in different part of the GP in PD patients treated by DBS.

Methods. Patients. We studied five patients suffering from PD who underwent bilateral stereotaxic electrode implantation into the GP for chronic high-frequency stimulation (table 1). Ages ranged from 44 to 65 years. All had a severe akinetic and rigid form of the disease and three had a mild rest tremor (patients 1, 2, and 3). Their mean baseline United Parkinson's Disease Rating Scale(UPDRS)¹⁹ score was 40 (range, 24 to 52). There was a clear response to levodopa, ranging from 50 to 80% improvement after a mean duration of the disease of 12.1 years (range, 9 to 17). Despite optimal medical treatment based on combination of levodopa (mean daily dosage, 850 mg; range, 350 to 1650 mg) and dopamine receptor agonists (pergolide in patients 1, 2, and 5; bromocriptine in patient 3; lisuride in patient 4), severe motor fluctuations and mono- and biphasic LID were seen in all patients. The mean LID score, as defined in part IV of UPDRS, was 8 (range, 6 to 10). Patients had no other neurologic or mental impairment, and their brain MRI was normal. Patients gave informed written consent. The study was approved by the French Ethical Committee.

Neurosurgical procedure. The day before surgery, MRI was performed in stereotactic conditions (ste-MR) on a 1.5-T MR unit (Sigma, General Electric, Milwaukee, WI) using an MR-compatible Leksell G stereotactic frame (Elekta Instrument, Stockholm, Sweden). Two ste-MR acquisitions were performed: 1.3-mm axial configuration slices (n = 124) with a three-dimensional spoiled gradient recalled (SPGR) acquisition and 1.3-mm contiguous slices (n = 124) with an axial three-dimensional time of flight angiographic acquisition after injection of 0.1 mmol/kg of Gadopentate-dimeglumine (Schering laboratories). The acquisitions were then transferred to an "advantages windows" work station (General Electric, Buc, France). Using the stereotaxy module of the Voxtool (General Electric) software developed at Val de Grâce Hospital (Paris, France), the targets and the trajectories for electrode implantation were calculated and visualized on both the SPGR and the angiographic acquisition. The target chosen for the tip of the electrode was the posteroventral part of the GPi. Target determination was based on the location of the anterior and posterior commissure and the midline plane and the direct visualization on MR acquisition of the GPi (see Dormont et al.²⁰ for precision achieved with the ste-MR installation and Dormont et al.²¹ for details of the calculation of targets and trajectories). Double, anteriorly and laterally, oblique trajectories were used to avoid the lateral ventricle and subependymal and cortical veins, visualized on the angiographic ste-MR acquisition.

The procedure started with the nondominant hemisphere. A 14-mm burr hole was made in accordance with the predetermined trajectory. Both intraoperative electrophysiologic and clinical exploration were performed with an array of five simultaneous concentric semimicroelectrodes (no. 14-30-08, FHC Inc., Bowdoiham, ME) in a cross-pattern arrangement, the four outer electrodes being separated by 2 mm from the central one. All five were advanced in parallel with a hydraulic microdrive and exploration was performed millimeter by millimeter starting 10 mm before the target established by MR. Stereotactic x-ray controls were regularly performed to check that the electrodes were following the calculated trajectories. This was done using the short x-ray radiologic device of the Leksell stereotactic frame. The definitive site chosen for the target was determined by the location of single unit characteristic GPi activity recordings²² and by a maximal improvement in contralateral upper limb rigidity and finger tapping with minimal side effects during stimulation. A chronic DBS quadripolar electrode (3387, Medtronic, Minneapolis, MN) was implanted to replace the microelectrode directed toward the optimal target. The electrode permits stimulation at four contacts, contact 3 being the uppermost(proximal) and contact 0 the lowest (distal) (figure 1). Contacts are 1.3 mm in diameter, 1.5 mm in length, and 1.5 mm apart; thus, the total length covered is 10.5 mm from the inferior border of the lowest contact to the superior border of the uppermost contact. The final depth of the lowest contact exceeded that of the optimal target while always remaining above the optical tract, the latter defined by imaging and by induction of phosphenes during stimulation. During the same session, an electrode was similarly implanted in the dominant hemisphere, but the exploration was done in a more simplified manner, using single electrode on the central trajectory.

For the purposes of this pathophysiologic study, one electrode (patient 2, right electrode) was excluded on the grounds that the lowest contact was not located in the posteroventral part of the GPi, as determined on control MR performed the day after surgery. Therefore, only nine electrodes were studied.

Several days after surgery, programmable stimulators (ITREL II, Medtronic) were implanted in the subclavicular area and connected to the electrode by a tunneled subcutaneous extension cable. The stimulator delivered a rectangular current with four variable parameters: location of the stimulating contact(contacts 0, 1, 2, and 3), frequency (0 to 185 Hz), pulse width (60 to 420µsec), and amplitude (0 to 10.5 V). Postoperatively, parameters could be changed by an electromagnetic programmer (7532 neurologic programmer, Medtronic).

Postoperative evaluation of pallidal electrical stimulation. Ten days after electrode implantation, we assessed the effects of electrical stimulation via each of the four contacts of the electrodes on parkinsonian disability to determine the therapeutically optimal contact and parameters of GPi stimulation for each hemisphere. Clinical evaluations were performed, after a night without stimulation, in two states: "off" state, as defined by the Core Assessment Program Intracerebral Transplantation,²³ that is, after interruption of antiparkinsonian medication for at least 12 hours, and stable "on" state throughout the day after usual drug intake, as clinically assessed by the investigator.

Parkinsonian motor signs were evaluated using four subscales of the UPDRS part III: gait (UPDRS, item 20; maximal score = 4); contralateral upper and lower limb rigidity (UPDRS, item 22; maximal score = 8); contralateral akinesia, defined as the total score for thumb-index tapping, opening and closing hand, repetitive and alternating pronation-supination movements of the hands, and leg agility (UPDRS, items 23, 24, 25, and 26; maximal score = 16); and contralateral upper and lower limb rest tremor(UPDRS, item 20; maximal score = 8). For each item, additional half points were used to increase sensitivity of the testing.

Dyskinesia was assessed qualitatively and quantitatively in the contralateral limbs, using a global score ranging from 0 to 4 (0 = absence of dyskinesia; 1 = mild, occasional; 2 = moderate, intermittent; 3 = severe, permanent; 4 = extremely severe dyskinesia).

Electrical stimulation procedure. The contacts of the electrodes implanted in each pallidum were tested separately, one by one. The order in which the various contacts were tested was randomized. The time needed to test one contact in the "off" state was 2 to 3 hours. At least 30 minutes without stimulation was mandatory between testing two different contacts on the same electrode to ensure a return to the baseline state; no more than two contacts were evaluated per day. A rest day was allowed between two evaluations; during that day, evaluation in the "on" state was done. The total time needed to test the four contacts of the two electrodes in each patient was about 10 days.

Stimulation was unipolar; the frequency and pulse width of the stimulation current were kept at 130 Hz and 60 µsec, respectively. The voltage amplitude was increased by 0.1 V and varied from 0 to 5 V. At every 0.5-V step, a delay was respected before clinical evaluation was performed. The delay was 5 minutes for evaluation during the "off" state. It was reduced to 2 minutes during the "on" state because of the short duration of a stable"on" state after levodopa intake.

We recorded the scores observed for the different clinical signs at each level of stimulation. The magnitude of the motor score increased with increase in the stimulation intensity. However, after a certain level of stimulation, the scores reached a plateau (figure 2). For each patient, to evaluate the effect of stimulation, we used two values for each contact. The first or "baseline score" was the score obtained before stimulation. The second or "optimal score" was the maximum score observed during the plateau phase.

The differences between the clinical effects observed using contact 3("upper contact") and contact 0 ("lower contact") were strikingly clear and reproducible for the nine electrodes tested. When stimulation was applied through contacts 1 or 2, the effects were intermediate; contact 1 tended to give results similar to those of contact 0 and contact 2 to those of contact 3. We therefore elected to study only the upper and the lower contacts for the analysis.

A control blind evaluation of clinical effects of stimulation of the upper and the lower contacts was performed by a neurologist (P.D.) for all patients.

Statistical comparison of the means was done using a Wilcoxon signed-rank test for nonparametric data.

Results. Postoperative evaluations of the patients' motor disability (UPDRS) in the "off" state and without stimulation (39 ± 6) were not different from those obtained before surgery (40 ± 6).

Parkinsonian disability score changed after a sigmoid curve as voltage was increased (figure 2). No clinical effect was observed when the voltage was low. Above a given threshold, clinical effects became significant. The amplitude of these effects increased progressively with voltage, reaching a plateau at which motor signs remained unchanged. The plateau always started at a stimulation level less than 5 V ("optimal voltage") with values varying widely (range, 1 to 4.5 V) from one contact to another and from one subscore to another. Voltages in excess of 5 V were sometimes found to produce side effects, such as nausea, paresthesia, and dystonia, and thus were not systematically tested in all patients. Stimulation effects were reproducible and reversible, as assessed by the return to baseline after stimulation withdrawal.

Effect of pallidal electrical stimulation in the "off" state. When stimulation was applied through the lower contact, rigidity progressively improved, whereas akinesia and gait worsened with increasing voltage (seefigure 2A). At the optimal voltage, the rigidity score improved by 59% (p < 0.005), whereas gait and akinesia scores worsened by 26% (p < 0.01) and 35% (p < 0.005), respectively (table 2). Tremor was completely suppressed in patients 1 and 3 but was unchanged in patient 2. No dyskinesia was provided by stimulation on the lower contact, except a foot dystonia in patient 4 starting from 0.5 V and remaining stable when voltage increased. Pre-existing "off" dystonia in patients 1, 2, and 3 were relieved, however, by stimulation at this contact.

When stimulation was applied through the upper contact, all scores progressively improved with increasing voltage. At the optimal voltage, all scores improved: gait by 36% (p < 0.005), akinesia by 38%(p < 0.005), and rigidity by 50% (p < 0.01) (seetable 2). A bilateral improvement, mainly in the lower limbs, was noted with three (patients 1 and 3) of nine electrodes. Tremor present in patients 1, 2, and 3 was improved by 60, 40, and 100%, respectively. For seven (all patients) of nine electrodes, stimulation applied through the upper contact induced contralateral and sometimes axial dyskinesia. The mean score of stimulation-induced dyskinesia was 1.7 ± 0.3 (figure 3A). All types of dyskinesia were observed(chorea, ballism, dystonia, pedaling, and repetitive limb movements).

No side effects were noted, except a transient visual flash sometimes produced by stimulation at the lower contact above a certain voltage (range, 2 to 3 V).

Effect of pallidal stimulation in the "on" state. When stimulation voltage through the lower contact was increased, contralateral LID decreased in intensity until complete relief, in contrast to rigidity, which was not affected, and akinesia and gait, which progressively worsened(see figure 2B). Without stimulation, the mean contralateral LID score was 2.8 ± 0.3. At optimal voltage stimulation on the lower contact, the mean score decreased by 97% (p < 0.005) (figure 3B). At 5 V, contralateral LID was completely suppressed in eight of nine electrodes (a lower limb dystonia was partially persistent in patient 4). A bilateral complete suppression of dyskinesia was obtained in four electrodes tested. The voltage required for complete suppression of contralateral LID was always below the threshold voltage that suppressed dyskinesia bilaterally.

An increase in stimulation voltage applied through the lower contact resulted in a progressive worsening of akinesia for seven of nine electrodes(no effect with the other two). This worsening led to a complete suppression of the levodopa effect on gait and contralateral akinesia in three electrodes, thus rendering patients hemiparkinsonian and in two cases led to a more severe akinesia than in the patients' baseline "off" state. The voltage required for worsening gait and akinesia was always above the voltage needed for complete suppression of contralateral LID. Rigidity was still improved in all patients.

Stimulation applied through the upper contact had no immediate effect on parkinsonian motor signs, but it resulted in a slight, although not significant, increase in LID (see figure 3B).

Discussion. Using DBS to explore different sites within the GP, we were able to identify two targets where stimulation had strikingly distinct effects. In the first target, the upper GP, stimulation dramatically improved parkinsonian signs (akinesia, rigidity, and gait) and could result in abnormal involuntary movements in the absence of levodopa (i.e., when the patient was in the "off" state); under levodopa treatment ("on" state), stimulation of the upper part of the GP did not significantly worsen LID. In the second target, the lower GP, in contrast, stimulation worsened akinesia and gait while still improving rigidity, when the patient was in the "off" state; in the "on" state, stimulation of the lower part of the GP was able to suppress LID, but, at higher voltages, significantly worsened akinesia and gait, thus canceling out the effect of levodopa.

Although distribution of the neural networks underlying the upper and the lower parts of the GP are unknown, their location can be inferred using post-operative MR, which allows the contacts to be localized within the structure. The lower contact was localized in the posteroventral segment of the GPi in each patient. The upper contact was situated in the dorsal segment of the GP in all patients, but its exact localization was difficult to ascertain because of slight variations in the trajectories of the electrodes(to avoid the cortical veins and ventricle) from one patient to another. In fact, even with the 1.5-T MR resolution, we were unable to discern clearly the limit between the external GP (GPe) and GPi on all the axial slices. We cannot, therefore, rule out the possibility that the upper contact was confined to the intermediate zone between GPi and GPe, or to the GPe.

Even though the physiologic effects of stimulation are still largely unknown, it is assumed to act by neural inactivation, and the volume of neural tissue affected is probably in the order of cubic millimeters.^16,17 We used these hypothetical mechanisms to try to understand our results.

Inactivation of the dorsal part of the GP allowed an improvement of all cardinal signs of PD and induced dyskinesia, even if the patient was in the"off" state. Dyskinesia was either similar to that previously experienced by the patient after administration of levodopa or slightly different. In view of the location of the upper contact in the dorsal part of the GPi or the GPe, inactivation via this contact probably modulated the output of the GPe to the GPi alone or to both the GPi and the STN. A similar improvement in akinesia has been previously obtained using intraoperative stimulation of areas anterior to the posteroventral GPi, which could correspond to the GPe.²⁴ These effects are hard to explain with the classic functional model of the basal ganglia,^14,15 where the GPe is viewed almost exclusively as a relay structure to the GPi through its projection to the STN. In PD, dopamine deficiency is generally held to decrease the activity of the GPe, resulting in an overactivity in the GPi, thus leading to parkinsonian motor signs. Conversely, in this model, dyskinesia appears to be a consequence of the reverse imbalance between GPe and GPi, with hyperactivity in the GPe and hypoactivity in the GPi.²⁵ The recent suggestion that the GPe, and particularly GPe efferents to the GPi, have a greater integrative function in the basal ganglia than previously thought^26,27 might help to provide an explanation for these apparent contradictions.

Inactivation of the posteroventral GP dramatically relieved LID, eventually leading to the recurrence of PD signs when stimulation voltage was high. A similar and consistent effect on LID is obtained with pallidotomy, targeting the same area of the GP. These observations are also difficult to explain with the classic functional model of basal ganglia organization,^14,15 because this would suggest that dyskinesia is the consequence of hypoactivity of the GPi. A possible explanation comes from the work of Matsumara et al.,²⁸ suggesting that dyskinesia is not the consequence of a simple imbalance between GPe and GPi but resulted from a mixture of decreased and increased activity in the two segments of the GP. Thus, as is probably the case with lesioning, inactivation of the posteroventral GP would suppress this abnormal mixture of activities causing LID. As reported with unilateral pallidotomy,¹³ we also observed a bilateral suppression of LID after a unilateral inactivation of the posteroventral GP.

Inactivation of the posteroventral GP provided a differential effect on rigidity and akinesia. The improvement in rigidity, also reported after posteroventral pallidotomy,^8-11,13 could result from decreased GPi overactivity, considered to result in parkinsonian symptomatology. This effect is difficult to reconcile with the simultaneous worsening of gait and akinesia observed in our study. Gait impairment and akinesia were rarely reported after bilateral lesioning of the GPi in nonparkinsonian patients²⁹ and after pallidotomy in PD patients.¹² However, the increase in the levodopa dose needed to obtain an optimal motor response after pallidotomy could suggest a worsening of akinesia.³⁰ The differential effect provided by the inactivation of the posteroventral GP suggests that the pathophysiology of rigidity and akinesia is different. One explanation could be that the two signs involve the dysfunction of different pallidal outflows; whether the segregated outputs to the pedunculopontine nucleus and to the thalamus are good candidates remains to be demonstrated.

The clearly different clinical effects obtained with inactivation of the dorsal and the posteroventral part of the GP underline the complexity of the functioning of this nucleus in the basal ganglia circuitry. The evidence of two distinct anatomofunctional sites in the GP may explain the contrasting results of pallidotomy on parkinsonian signs.^8-13 The consistent effect of pallidotomy on LID might be due to the fact that such lesioning invariably includes the posteroventral part of the GP. One might also suppose that the effect on akinesia and gait varies in relation to the size of the lesion and its extension to the dorsal GP. Our results emphasize that the choice of optimal target site in the GP is not straightforward and has to be adapted individually to each patient. The dorsal GP would appear to be the prime target for patients with severe akinesia and rigidity, whereas the posteroventral GP might be more appropriate for patients with disabling LID. When chronic DBS is chosen to treat PD patients, we usually aim for a combination between the effects on both targets. To this end, we use stimulation on an intermediate contact or stimulation on the upper contact on one side and stimulation on the lower contact on the other side, depending on the severity of the parkinsonism and LID on each part of the patient's body.

In conclusion, our study demonstrates that two different targets are present in the GP. Inactivation of the dorsal GP provides an improvement of all cardinal signs of PD and can even generate dyskinesia when the patient is in the "off" state; in contrast, inactivation of the posteroventral GP suppresses LID when the patient is in the "on" state and can worsen akinesia and gait when the stimulation voltage is increased. The reasons for the different effects of stimulation in the dorsal and posteroventral GP are not explained by the current model of basal ganglia circuitry. Further physiologic and anatomic studies are needed to explain their basis and also to improve the choice of the optimal target for pallidal stimulation and pallidotomy.