The effects of unilateral vs. bilateral subthalamic nucleus deep brain stimulation on prosaccades and antisaccades in Parkinson’s disease

Journal: Experimental Brain Research

Lisa C. Goelz, Department of Kinesiology and Nutrition, University of Illinois; Department of Physical Therapy and Human Movement Sciences, Northwestern University, Chicago, USA

Fabian J. David, Department of Physical Therapy and Human Movement Sciences, Northwestern University, Chicago, USA

John A. Sweeney, Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati, Cincinnati, Ohio

David E. Vaillancourt, Departments of Applied Physiology and Kinesiology, Biomedical Engineering, and Neurology, University of Florida, Gainesville, USA

Howard Poizner, Institute for Neural Computation, University of California, San Diego, USA

Leonard Verhagen Metman, Department of Neurological Sciences, Section of Parkinson Disease and Movement Disorders, Rush University Medical Center, Chicago, USA

Daniel M. Corcos, Department of Physical Therapy and Human Movement Sciences, Northwestern University; Department of Neurological Sciences, Rush University Medical Center, Chicago, USA

Corresponding author: Lisa C. Goelz, Department of Kinesiology and Nutrition, University of Illinois, 901 West Roosevelt, Chicago, Illinois 60612, Tel: (312) 413-8525, Fax: (312)413-3699, Email:

Author email addresses: , , , , , ,

Statistical analyses

MDS-UPDRS: A mixed effects regression model was used to analyze the MDS-UPDRS scores. Subject was considered a random effect. The fixed effect was stimulation condition (OFF, LEFT, RIGHT, and BOTH). In the event of a significant effect of condition, pairwise comparisons between stimulation conditions were evaluated using t-tests. Critical alpha was 0.05, and uncorrected p-values are reported. A residual analysis was performed to confirm that distributional assumptions for parametric testing were not violated.

Prosaccade and Antisaccade task: For latency and gain during the prosaccade and antisaccade tasks, mixed effects regression models were used. For the prosaccade error rates in the antisaccade task, a mixed effects logistic regression model was used. For both of these mixed models, subject was the random effect, and the fixed effects were stimulation condition (OFF, LEFT, RIGHT, and BOTH), target location (Left and Right), and the condition by location interaction. In the event of a significant interaction, the following comparisons were performed to detect the effect of stimulation for each target location separately: OFF vs. LEFT, OFF vs. RIGHT, OFF vs. BOTH, LEFT vs. BOTH, RIGHT vs. BOTH, and LEFT vs. RIGHT. This resulted in 12 pairwise comparisons. All tests were two-tailed, critical alpha was 0.05, and p-values associated with pairwise comparisons were corrected using the Bonferroni method. SAS 9.4 was used for all statistical analyses. In addition, we present the mean ± standard error of the healthy age-matched control subjects in figures 2, 3 and 4. The mean ± standard error is presented as dark grey shade for left target and light grey shade for the right target. The purpose of these data is to demonstrate the extent to which STN DBS causes performance to approximate or deviate from that of healthy controls.

RESULTS

Clinical measures: MDS-UPDRS

Table 1 (main paper) presents the MDS-UPDRS scores for each individual patient for the OFF, LEFT, RIGHT, and BOTH stimulation conditions. There was a main effect of stimulation condition for MDS-UPDRS (F 3, 26= 7.95, P < 0.001). Compared to OFF, LEFT (mean difference estimate, 13.5; 95% confidence interval, 2.2 to 24.9; P = 0.02), RIGHT (11.9; 0.6 to 23.3; p = 0.04), and BOTH (26.7; 15.4 to 38.1; P < 0.001) improved MDS-UPDRS motor scores. BOTH improved scores more than LEFT (13.2; 2.2 to 24.2; P = 0.02) and RIGHT (14.8; 3.8 to 25.8; P = 0.01) stimulation.

The effect of stimulation on prosaccade latency and gain

Prosaccade latency

The stimulation condition by target location interaction was not significant. We observed a main effect of stimulation condition (F 3, 1483= 17.2, P < 0.0001) for prosaccade latency (supplementary figure A1, supplementary table 1). Pair wise comparisons between stimulation conditions revealed that latency for the BOTH condition was significantly less than for the OFF (35 ms; 15 to 54 ms; P < 0.0001), LEFT (38 ms; 20 to 56 ms; P < 0.0001), and RIGHT (46 ms; 27 to 64 ms; P < 0.0001) stimulation conditions. Supplementary figure 1A shows that this reduction in prosaccade latency brought the mean of the patients well below the mean latency of the healthy controls. Unilateral stimulation conditions did not differ from OFF. We also observed a main effect of target location (F 1, 1483=20.12, P < 0.0001). Average latency to the left target (dark grey bars, supplementary figure 1A) was greater than average latency to the right target (light grey bars, supplementary figure 1A).

Prosaccade gain

There was a significant stimulation condition by target location interaction (F 3, 1483= 3.3, P = 0.02, see supplementary table 1) for gain (supplementary figure 1B). Gain was significantly greater during BOTH stimulation compared to LEFT stimulation (left target, 0.06; 0 to 0.11; P = 0.02; right target, 0.07; 0.02 to 0.11; P = 0.001) and RIGHT stimulation (left target, 0.05; 0 to 0.1; P = 0.04; right target, 0.11; 0.06 to 0.16; P< 0.0001). Gain was significantly lower during RIGHT stimulation compared to OFF (0.08; 0.02 to 0.13; P < 0.0001) for saccades to the right target. Prosaccade gain during all stimulation conditions was less than the mean gain of the healthy control.

The effect of stimulation on antisaccade latency and gain

For the antisaccade task, rightward saccades were measured for left target presentation, and leftward saccades were measured for right target presentation.

Antisaccade latency

We observed a significant stimulation condition by target location interaction for latency (F 3, 1111 = 4.9, P = 0.002, see supplementary table 1). Pairwise comparisons showed that latency only during LEFT stimulation was significantly less than latency during OFF stimulation (37 ms; 0 to 74 ms; P= 0.02) for rightward saccades (dark grey bars supplementary figure 2A). Antisaccade latency for all stimulation conditions was longer than healthy controls.

Antisaccade gain

There was a significant stimulation condition by target location interaction for gain (F 3, 1111 = 5.4, P = 0.001, see supplementary table 1). For rightward saccades (dark grey bars supplementary figure 2B), the BOTH condition increased gain compared to the OFF (0.20; 0.08 to 0.31; P < 0.0001), LEFT (0.11; 0 to 0.22; P = 0.04), and RIGHT (0.17; 0.06 to 0.28; P < 0.0001) conditions. For leftward saccades (light grey bars supplementary figure 3B), gain was greater for RIGHT (0.15; 0.04 to 0.26; P < 0.0003) and BOTH (0.12; 0 to 0.24; P = 0.01) stimulation compared to OFF stimulation. Antisaccade gain during all stimulation conditions was well below the mean antisaccade gain of healthy controls.

The effect of stimulation on prosaccade error rates

There was a main effect of stimulation condition (F 3, 1506= 8.28, P < 0.0001; supplementary figure 3, see supplementary table 1). Prosaccade error rates were significantly higher during BOTH stimulation compared to OFF (P < 0.0001), LEFT (P = 0.003), and RIGHT (P = 0.0005) stimulation. A main effect of target location was also observed (F 1, 1506 = 10.58, P = 0.003). Prosaccade error rates were greater for the right target presentation (light grey bars supplementary figure 3) when compared to the left target presentation (dark grey bars, supplementary figure 3). Prosaccade error rate during BOTH was greater than the healthy controls.

Supplementary Table 1 Statistical comparisons for the prosaccade and antisaccade tasks

Prosaccade / Antisaccade
Latency / Gain / Latency / Gain / Error Rate
Main Effect of
Target Location
(Left v. Right) / Diff = 22 s
t=4.49
p < 0.0001 / Diff = 0.05
t = 5.8
p < 0.0001 / Diff= 0.007s
t = 1.09
p = 2.8 / Diff = -0.02
t = -1.3
p = 0.19 / Diff = 7.7 %
t = 3.5
p = 0.003
Main Effect of
DBS Condition / F = 17.2
p < 0.0001 / F = 17.3
p < 0.0001 / F = 0.62
p = 0.6 / F = 12
p < 0.0001 / F = 8.2
p < 0.0001
Pairwise comparisons
OFF v. LEFT / Diff=-0.004 s
t=-0.50 p=1.00 / Diff=0.023
t = 1.88
p = 0.363 / Diff=0.007
t = 0.86
p = 1.00 / Diff = 0.09
t = 3.6
p = 0.002 / NS
OFF v. RIGHT / Diff =-0.01s
t = -1.53
p = 0.76 / Diff = -0.04
t = 3.4
p = 0.005 / Diff = 0.007
t = 0.80
p = 1.00 / Diff = 0.09
t = 3.6
p = 0.002 / NS
OFF v. BOTH / Diff = -35 s
t = 4.7
p < 0.0001 / Diff = 0.04
t = 3.1
p = 0.01 / Diff = 0.013
t = 1.34
p = 1.00 / Diff = 0.16
t = 5.9
p < 0.0001 / Diff = 15.7 %
t = 4.4
p < 0.0001
LEFT v. RIGHT / Diff=-0.008s
t = -1.14
p = 1.00 / Diff=0.02
t = 1.69
p = 0.552 / Diff=-0.0004s
t = -0.o5
p = 1.00 / Diff=-0.002
t = 0.09
p = 1.00 / NS
LEFT v. BOTH / Diff = -38 s
t = 5.7
p < 0.0001 / Diff = 0.06
t = 5.4
p < 0.0001 / Diff = 0.005s
t = 0.59
p = 1.00 / Diff = 0.07
t = 2.8
p = 0.03 / Diff = 10.7 %
t = 3.5
p = 0.003
RIGHT v. BOTH / Diff = -46 s
t = 6.6
p < 0.0001 / Diff = 0.08
t = 6.8
p < 0.0001 / Diff=0.006s
t = 0.63
p = 1.00 / Diff = 0.07
t = 2.7
p = 0.047 / Diff = 13.7 %
t = 4
p = 0.0005
Interaction Effect of
DBS Condition x Target Location / F = 1.00
p = 0.39 / F = 3.3
p = 0.02 / F = 4.9
p = 0.002 / F = 5.4
p = 0.001 / F = 1.23
p = 0.30
Pairwise comparisons
Left Target
OFF v. LEFT / NA* / Diff = 0.015
t = 0.84
p = 1.00 / Diff = -37 s
t = 3.4
p = 0.02 / Diff=-0.08
t = -2.48
p = 0.16 / NA*
OFF v. RIGHT / NA* / Diff = 0.009
t = 0.46
p = 1.00 / Diff = 0.26s
t = 2.13
p = 0.39 / Diff =-0.28
t = -0.80
p = 0.42 / NA*
OFF v. BOTH / NA* / Diff = -0.05
t = -2.31
p = 0.25 / Diff =0.021s
t = 1.60
p = 1.00 / Diff = 0.2
t = 5.2
p < 0.0001 / NA*
LEFT v. RIGHT / NA* / Diff =-0.006
t = -0.37
p = 1.00 / Diff = -0.01s
t = -1.06
p = 0.29 / Diff = 0.06
t = 1.75
p = 0.97 / NA*
LEFT v. BOTH / NA* / Diff = 0.06
t = 3.4
p = 0.02 / Diff=-0.02s
t = -1.34
p = 1.00 / Diff = 0.11
t = 3.2
p = 0.04 / NA*
RIGHT v. BOTH / NA* / Diff = -0.05
t = -2.92
p = 0.04 / Diff=-0.005s
t = -0.38
p = 1.00 / Diff = 0.17
t = 4.7
p < 0.0001 / NA*
Right Target
OFF v. LEFT / NA* / Diff = 0.03
t = 1.91
p = 0.68 / Diff=- 0.02s
t = -1.92
p = 0.66 / Diff = 0.15
t = 4.4
p = 0.004 / NA*
OFF v. RIGHT / NA* / Diff = -0.08
t = 4.6
p = 0.0001 / Diff = -0.012s
t = -0.99
p = 1.00 / Diff = 0.15
t = 4.4
p = 0.0003 / NA*
OFF v. BOTH / NA* / Diff = -0.04
t = -2.16
p = 0.37 / Diff=0.004s
t = 0.32
p = 1.00 / Diff = 0.12
t = 3.2
p = 0.01 / NA*
LEFT v. RIGHT / NA* / Diff = 0.05
t = -2.92
p = 0.04 / Diff = 0.01s
t = 0.95
p = 1.00 / Diff= -0.06
t = -1.81
p = 0.07 / NA*
LEFT v. BOTH / NA* / Diff = 0.07
t = 4.3
p = 0.001 / Diff=0.03s
t = 2.10
p = 0.43 / Diff= -0.03
t = -0.82
p = 1.00 / NA*
RIGHT v. BOTH / NA* / Diff = 0.11
t = 7
p < 0.0001 / Diff = 0.16s
t = 1.25
p = 1.00 / Diff = 0.03
t = 0.80
p = 1.00 / NA*

*Not applicable- pairwise comparisons were not performed if interaction effect was not significant

Supplementary Figure 1.Prosaccade latency and gain

Supplementary Figure 1. These plots show the mean (± standard error) prosaccade latency (A) and gain (B) for all stimulation conditions. The average latency and gain for the healthy control group is indicated by the shaded horizontal band (mean ± standard error). Bilateral (BOTH) STN DBS significantly reduced prosaccade latency compared to OFF, LEFT, and RIGHT (A). This reduction brought the mean latency well below the mean values for healthy controls (A). Stimulation did not improve prosaccade gain (B).

Supplementary Figure 2. Antisaccade latency and gain

Supplementary Figure 2.These plots show the mean (± standard error) antisaccade latency (A) and gain (B) for all stimulation conditions. The average latency and gain for the healthy control group is indicated by the shaded bands (mean± standard error). Stimulation did not improve antisaccade latency (A). With respect to gain (B), for rightward saccades (dark grey bars), the BOTH condition increased gain compared to the OFF, LEFT, and RIGHT stimulation conditions. For leftward saccades (light grey bars), gain was greater for RIGHT and BOTH stimulation compared to OFF stimulation. The mean gain for all conditions remained well below the mean gain of healthy controls.

Supplementary Figure 3.Prosaccade error rate

Supplementary Figure 3. This plot shows the mean (± standard error) percentage of prosaccade errors in the antisaccade task for all stimulation condition. The average percentage of prosaccade errors for the healthy control group is indicated by the shaded bands (mean± standard error). Bilateral (BOTH) stimulation significantly increased the percentage of prosaccade errors compared to OFF, LEFT, and RIGHT stimulation conditions