Frequency-Dependent Reciprocal Modulation of Verbal Fluency and Motor Functions in Subthalamic Deep Brain Stimulation

Background  High-frequency deep brain stimulation (DBS) of the subthalamic nucleus (STN) improves motor functions in those with Parkinson disease but may worsen frontal functions such as verbal fluency (VF). In contrast, low-frequency DBS leads to deterioration of motor functions. It is not known whether low-frequency STN DBS also has an effect on frontal functions.

Objective  To examine whether low-frequency STN DBS in contrast to high-frequency STN DBS has a positive effect on frontal functions on the basis of VF test results.

Design  A double-blind randomized crossover experiment to compare performance in 4 VF subtests and motor performance at 10 Hz, 130 Hz, and no stimulation.

Setting  University hospitals in Düsseldorf and Cologne, Germany.

Patients  Twelve patients with Parkinson disease 3 months or more after bilateral electrode implantation into the STN.

Main Outcome Measure  Mean number of words in VF at different stimulation frequencies.

Results  The VF was significantly better at 10 Hz (48.3 words) compared with 130 Hz and showed a nonsignificant trend toward worsening at 130 Hz (42.3 words) compared with no stimulation (43.8 words). These results were consistent across all subtests.

Conclusions  The study provides evidence of a beneficial effect of low-frequency (10 Hz) STN DBS on VF, which may be caused by activating neural pathways projecting to the frontal cortex. In addition, the study reproduces the negative effect of therapeutic high-frequency STN DBS on VF. The study results provide evidence for a frequency-dependent modulation of cognitive circuits involving the STN.

High-frequency (HF) deep brain stimulation (DBS) of the subthalamic nucleus (STN) at frequencies of about 130 Hz is a successful treatment of motor symptoms of Parkinson disease (PD).¹ However, some results from recent studies suggest that this therapeutic stimulation can have a subclinical negative effect on frontal functions such as verbal fluency (VF).² The involvement of the basal ganglia in nonmotor functions is supported by a model that suggests a connection to limbic and prefrontal areas.³

In contrast to HF STN DBS, low-frequency (LF) STN DBS of about 10 Hz can worsen motor functions.⁴ Furthermore, a 10-Hz pathologic oscillatory cerebral network is associated with PD tremor.⁵ This hypothesis, based on noninvasive measurements in patients with PD, is supported by intraoperative recordings in the human STN that provide evidence for pathologic synchronization at tremor frequency and around 10 Hz.⁶

There is evidence that HF STN DBS leads to improved motor functions and worsened frontal functions but that 10-Hz LF STN DBS leads to deterioration of motor functions, possibly because of enhancement of pathologic synchronization within the basal ganglia loop in patients with PD. We compared VF in patients with PD during STN stimulation at 10 Hz, at 130 Hz, and without stimulation to examine whether, in contrast to HF STN DBS, there is a positive effect of 10-Hz LF STN DBS on frontal functions suggesting opposite functional consequences of STN stimulation on the basal ganglia cognitive and motor circuit.

We used a double-blind randomized crossover method in 12 patients with PD who were selected from consecutive routine visits at the University hospitals in Düsseldorf and Cologne, Germany. They underwent examination at least 3 months (mean, 28 months; range 3-51 months) after bilateral stereotactic implantation of STN electrodes (model 3389; Medtronic, Minneapolis, Minn). All patients gave informed consent according to the Declaration of Helsinki.

To localize active poles used for long-term stimulation, we reimported intraoperative stereotactic radiographs or postoperative computed tomographic scans into the stereotactic planning system. Mean active pole position relative to the line between the anterior-posterior commissure was calculated, spatially normalized, and visualized at magnetic resonance imaging (Figure, A).

Testing was performed within 1 day after overnight withdrawal from dopaminergic medication. Stimulation was changed randomly among no stimulation (off), HF stimulation at 130 Hz or more, and LF stimulation at 10 Hz. Stimulation was kept constant with respect to the pulse width, voltage, and stimulation pole that had produced the best antiparkinsonian response during long-term stimulation (Table 1).

Five minutes after each change of the stimulation condition, motor functions were scored by using the motor score of the Unified Parkinson's Disease Rating Scale under double-blinded conditions. A further 10 minutes later, 4 different VF tests were each performed for 1 minute.

In the formal lexical test, patients were asked to produce words beginning with a particular letter. In a second test, patients had to produce words of a certain semantic category, such as animals. These tests for divergent thinking emphasize the creativity of search strategies. Two other tests were categorical change subtests of each of the first 2 tests with stronger emphasis on flexibility functions. In the formal lexical category change test, patients were asked to switch between 2 different letters (eg, a word beginning with H and then a word beginning with T) and finally were given a semantic category change test in which they were asked to switch between semantic categories (eg, furniture and tools). Three parallel test versions were used for each stimulation condition to avoid learning effects.

Parallel test versions of the 4 subtests, formal lexical, semantic category, formal lexical category change, and semantic category change, were A: words with P, animals, words with H, then words with T, clothes and flowers; B: words with M, food, words with D, then words with U, furniture and tools; and C: words with S, first name, words with G, then words with R, sports and fruit. Two patients were tested in the following sequence: parallel test A at 10 Hz, parallel test B at 130 Hz, and parallel test C with no stimulation. Two patients were tested in the following sequence: A at 10 Hz, B in the off condition, C at 130 Hz, and so on.

After testing was completed in all 12 patients, every combination of stimulation frequency and test version had been randomly used twice. This randomly changed testing sequence helped to avoid potential systematic carryover effects of stimulation on fluency results. Because the raw number of words can neither interindividually nor intraindividually be compared because of different word frequencies of the parallel tests in everyday language, mean values were computed after testing all subjects and all conditions. Motor scores and VF scores were analyzed statistically by using the distribution-free Wilcoxon test for nonnormal samples.

The active pole was located in the STN in most cases; however, 2 left-sided poles were located in the zona incerta. The spatially normalized mean (SD) position of the active stimulation pole at the right hemisphere was as follows: 11.1 mm (1.2 mm) lateral to the anterior-posterior commissure line (x-axis), 1.0 mm (1.9 mm) behind the middle of the anterior-posterior commissure line (y-axis), and 0.7 mm (3.0 mm) below the anterior-posterior commissure line (z-axis). Coordinates for the left hemisphere correspondingly were as follows: x-axis, 11.2 mm (1.4 mm); y-axis, 0.8 mm (1.9 mm); and z-axis, 0.5 mm (2.4 mm). Altogether, the mean stimulation position was in the anterior-dorsal part of the subthalamic area (Figure, A).

All patients had mean (SD) Unified Parkinson's Disease Rating Scale motor scores in the pathologic range with no stimulation (43 [13.3]) and mean (SD) Unified Parkinson's Disease Rating Scale scores that were highly improved during HF stimulation (22.1 [7.5]); the worst scores were during LF stimulation, (44.9 [11.6]). The difference between HF and LF was highly significant (P<.001) (Figure, B). Reciprocally, across all subjects and subtests, mean VF (mean number of words [SD]) was significantly better (P = .04) during LF stimulation (48.3 [9.7]) compared with HF stimulation (42.3 [11.1]) (Figure, B). Furthermore, VF showed a nonsignificant trend to improve during LF stimulation and worsen during HF stimulation compared with no stimulation: 43.8 (9.6). Each subtest taken alone showed the same trend: VF was best during LF stimulation followed by no stimulation; HF stimulation showed the worst results in all subtests. However, these differences in the subtests were not significant (Table 2).

Comparing therapeutic HF STN DBS with no stimulation, the present study reproduces the negative trend of HF STN DBS on VF while improving motor functions. The effect of HF STN DBS on VF has been shown in several studies, whereas some studies failed to do so.² Results of imaging studies suggest that HF STN DBS may worsen VF by deactivating the left inferior frontal gyrus⁷ by affecting the striatothalamocortical circuits. A current basal ganglia model supports this view, showing pathways from the basal ganglia through the ventral dorsomedial thalamus to different prefrontal areas.³ Imaging data also show that motor improvement by HF STN DBS is connected with motor-related facilitation of premotor areas.⁸

Comparing LF (10 Hz) STN DBS with no stimulation, the present study reveals the positive influence of LF stimulation on VF while worsening motor functions. Although this effect was not significant, the direct comparison of LF stimulation at 10 Hz with HF stimulation provides evidence for a reciprocal, highly significant, frequency-dependent modulation of VF and motor functions. The positive VF effect of LF STN DBS may be caused by activating neural pathways projecting to the inferior frontal cortex, whereas motor dysfunction is supposed to be caused by aggravation of pathologic synchronization in an oscillatory network,⁴ which was described in a magnetoencephalography study in patients with PD tremor.⁵

Results from a recent positron emission tomography study of thalamic DBS for essential tremor⁹ showed that DBS has more than a blocking effect and also allows a gradual tuning of neuronal activity within functional circuits. To our knowledge, the present study results provide the first evidence of the possibility of frequency-dependent tuning of cognitive circuits interconnected with the STN. The functional relevance of neural activity in the 10-Hz domain for cognitive functions is supported by findings of enhancing cognitive performance by means of repetitive transcranial magnetic stimulation at α frequency.¹⁰ Deep brain stimulation of the STN at 10 Hz obviously has an inhibitory effect on the motor circuit and a facilitatory effect on the cognitive circuit.

The exact mechanisms and structures underlying the network of frontal functions as tested by means of VF remain to be determined. For motor control in this study, the STN was not stimulated mainly in the medial-ventral cognitive part; however, more ventral stimulation might lead to a more significant VF effect and might also help to explain differential aspects of the VF subtests, which showed a homogenous trend toward improvement with LF DBS but failed to reach significance.

Besides electrode localization, other factors might influence the results. We controlled for training effects on VF by using different parallel test versions. In addition, we balanced the order of tests across subjects. From the results of this study, we cannot answer the question of how the implantation procedure per se affected VF because patients were tested only after surgery. Generally, systematic investigations on the implantation effect comparing preoperative with postoperative (within 3 months) VF with no stimulation are rare. A comparison in 3 cases was inconclusive.¹¹ Study results in patients undergoing subthalamotomy have demonstrated no significant decrease in VF 3 to 12 months postoperatively.¹² Thus, it is most likely that VF effects 3 or more months after STN electrode implantation are attributable to the electrical stimulation.

In summary, this study provides the first evidence of frequency-dependent differential effects of STN stimulation on frontal functions such as VF. This finding supports the hypothesis that the STN is part of a segregated cognitive network that can be modulated in a way opposite to that of the motor network.

Correspondence: Alfons Schnitzler, MD, Department of Neurology, Heinrich Heine University, Moorenstr 5, 40225 Düsseldorf, Germany (schnitza@uni-duesseldorf.de).

Accepted for Publication: April 4, 2006.

Author Contributions:Study concept and design: Wojtecki, Timmermann, Jörgens, Maarouf, Gross, and Schnitzler. Acquisition of data: Wojtecki, Timmermann, Jörgens, Treuer, Gross, Lehrke, and Sturm. Analysis and interpretation of data: Wojtecki, Timmermann, Südmeyer, Maarouf, Gross, Koulousakis, Voges, Sturm, and Schnitzler. Drafting of the manuscript: Wojtecki and Timmermann. Critical revision of the manuscript for important intellectual content: Wojtecki, Jörgens, Südmeyer, Maarouf, Treuer, Gross, Lehrke, Koulousakis, Voges, Sturm, and Schnitzler. Statistical analysis: Wojtecki and Jörgens. Obtained funding: Schnitzler. Administrative, technical, and material support: Wojtecki, Timmermann, Südmeyer, Maarouf, Treuer, and Gross. Study supervision: Timmermann, Südmeyer, Maarouf, Gross, Lehrke, Koulousakis, Voges, Sturm, and Schnitzler.

Funding/Support: This study was supported by grant J/73240 from the Volkswagen Stiftung.

Acknowledgment: We thank the patients for their excellent cooperation during the study.