Planned Surgical Trajectory Affects Clinical Motor Outcome in Deep Brain Stimulation Targeted at Subthalamic Nucleus for Parkinson's Disease

Abstract

BACKGROUND AND OBJECTIVES:

Surgical planning is critical to achieve optimal outcome in deep brain stimulation (DBS). The relationship between clinical outcomes and DBS electrode position relative to subthalamic nucleus (STN) is well investigated, but the role of surgical trajectory remains unclear. We sought to determine whether preoperatively planned DBS lead trajectory relates to adequate motor outcome in STN-DBS for Parkinson's disease (PD).

METHODS:

In 49 participants who underwent bilateral STN-DBS for PD using a Leksell® frame, we coregistered the frame and participant MRI images to obtain participant-specific anatomical planes. We evaluated relationships between clinical data and planned trajectories relative to their midsagittal and axial planes. We computed percent change in Unified PD Rating Scale subsection 3 (Unified Parkinson's Disease Rating Scale, part III) scores before and after DBS, and performed binary logistic regression to determine whether planned trajectories affect adequate (>30% Unified Parkinson's Disease Rating Scale, part III improvement) motor outcome.

RESULTS:

Preoperatively planned left lead trajectory relative to midsagittal plane predicted likelihood of adequate right body motor outcomes (odds ratio = 0.69, P = .024), even when controlling for ventricular width through Evans index. This effect reflects that increasingly lateral angle of approach reduced odds of adequate motor outcome. Right lead trajectory lacked a similar trend.

CONCLUSION:

Left DBS lead trajectory predicts adequate right-body motor outcome after bilateral STN-DBS. Greater planned trajectory angle relative to midsagittal plane reduces motor outcomes, independent of patients' ventricular width. These data may guide patient selection, inform risk/benefit discussions, optimize surgical planning, or support evidence-based evaluation of the methodologies used to select the approach trajectory, with careful consideration of the angle of approach relative to target.

ABBREVIATIONS:

arc_post

postoperative arc

arc_pre

preoperative arc

DOS

date of surgery

EI

Evans index

FDR

false discovery rate

LEDD

levodopa equivalent daily dose

MDC

Movement Disorders Center

PD

Parkinson's disease

ring_post

postoperative ring

ring_pre

preoperative ring

STN

subthalamic nucleus

UPDRS-III

Unified Parkinson's Disease Rating Scale, part III

WUSM

Washington University School of Medicine.

Magnitude of motor benefit in those with Parkinson's disease (PD) after deep brain stimulation (DBS) of the subthalamic nucleus (STN) varies across those treated.¹ Although patient factors such as age, disease duration, and levodopa responsiveness correlate with DBS motor outcome,² there remain limited data regarding the impact of surgical planning approach on outcomes. The need to avoid anatomical structures such as ventricles, blood vessels, and “eloquent” cortical regions may influence choice of trajectory in DBS surgical planning, and trajectory or orientation of leads relative to fiber bundles and neighboring structures may affect outcome.³ Improved motor benefit occurs most often with the stimulation of the dorsolateral portion of the STN, while adverse cognitive effects, verbal fluency in particular, are associated with ventral electrode position within STN and intersection of lead trajectory with caudate nuclei.^3,4 Tamir et al. ⁵ illustrated that posterolateral trajectories yield more robust contact with the dorsolateral motor STN, but there remains lack of direct understanding regarding the impact of trajectory on motor outcome.

We investigated the relationship between preoperatively planned lead trajectory and motor outcome using binary logistic regression after transforming surgeon-recorded trajectory angles and targets from Leksell® space into participant-specific anatomical space. We hypothesized that planned trajectory angles, particularly regarding the midsagittal plane, would predict contralateral motor outcome in STN DBS for PD.

METHODS

Participants

We retrospectively selected participants with PD who underwent bilateral STN-DBS implantation at the Movement Disorders Center (MDC) in Washington University School of Medicine (WUSM) and consented to DBS outcomes research. We sampled 50 total participants in a pseudo-random fashion that completed the procedure between 2007 and 2017, a timeline where institutional surgical methods and microelectrode localization was similar. Power analysis estimated that 38 participants would be necessary to achieve 80% power to detect a medium effect at α = 0.05, thus our sample size was sufficient to test the study hypothesis. We selected participants from a pool of which a single surgeon (JLD) preoperatively defined the lead trajectory (see “Neurosurgical procedure”). One participant was excluded from all analyses due to unavailable clinical computed tomography (CT) images, and one was excluded only from postoperative angle analyses due to inaccurate contact localization. We evaluated demographic data for all participants. These were compared with overall demographic and outcome data from the entire STN-DBS cohort to ensure that the sample was representative.

Ethical Adherence

The Washington University Human Research Protection Office approved this study of DBS outcomes, and all participants provided written consent.

Neurosurgical Procedure

One neurosurgeon performed all bilateral STN-DBS implantation surgeries as previously described, including the use of microelectrode recordings and intraoperative testing.⁶ All participants were prescreened for comorbidities that could preclude safe lead placement, including neuropsychological testing to exclude dementia. The surgeon relied on preoperative MRI to define the lead trajectory aimed to avoid deep cortical sulci, superficial vessels, or ventricles while targeting dorsolateral STN. This trajectory was visualized and determined on a StealthStation® (Medtronic), defining trajectory angles to be set on the Leksell® (Elekta Instrument AB) stereotaxic frame.

Computation of Trajectory Angles

Given the Leksell defined space represents variability in frame placement relative to target was critical to transform planned trajectories into individual participant-specific neuroanatomical space. Full methods for computing transforms, preoperative planned trajectories, postoperative measured trajectories, and lead displacement from intended target are contained in Supplemental Digital Content 1, http://links.lww.com/ONS/B150. Our transformations yielded the planned trajectory's angle against the participant's midsagittal plane, which we referred to as the preoperative arc (arc_pre), as well as the angle against an axial plane, the preoperative ring (ring_pre). A greater arc would indicate more lateral trajectory with an entry point farther from the midsagittal plane, and a greater ring would suggest a more posterior entry point.

Postoperative measured trajectories were computed based on CT using a validated method for contact localization.⁷ Electrode contacts and point of entry into the skull were used to reconstruct measured ring and arc (ring_post and arc_post, respectively). The mean electrode curvature (radian/mm) and maximum deviation from a straight line (mm) were also computed, collectively referred to as electrode deformation. We also computed Cartesian displacement of electrode from intended target position in 2 dimensions (along the transverse plane) and 3 dimensions in atlas space to minimize the effect of MER-guided depth adjustment.

Clinical Evaluation

Motor parkinsonism was assessed using the Unified Parkinson's Disease Rating Scale, part III (UPDRS-III) at the WUSM MDC. All UPDRS-III ratings were obtained after overnight withdrawal of antiparkinsonian medication (OFF), while postoperative ratings were obtained ON-stimulation. To mitigate the effect of fluctuations in scores within individuals, we averaged all qualifying scores for the 12-month period before surgery for the preoperative scores, and 15 months after surgery (minimum 2 weeks postoperative) for the postoperative scores, and computed percent change after DBS. Adequate outcome was defined as percent improvement of at least 30%, which has been established as the clinical standard for DBS outcome in PD.⁸ We calculated levodopa equivalent daily dose (LEDD) at the time of surgery according to standard criteria.⁹ Additional demographic, clinical, and neuropsychological information was obtained through chart review including sex, handedness, disease duration at time of surgery, and preoperative Mattis Dementia Rating Scale.

Ventricular Width

To account for surgical planning that potentially favored wider trajectories to avoid the lateral ventricle, we calculated Evans index (EI) for all 49 participants analyzed. EI captures ventricular width in comparison with total intracranial width and is thus an ideal measure to account for the dimension of ventricle size primarily contributing to laterality of electrodes. We measured maximum width of frontal horns of lateral ventricle and maximum internal diameter of the skull at the same level in the axial T2-weighted magnetic resonance scan using an open source Digital Imaging and Communications in Medicine viewer (https://ivmartel.github.io/dwv-jqui/demo/stable/index.html).

Statistical Analysis

Statistical analyses were performed using IBM SPSS 28.0 (IBM Corp). The primary relationship of interest was whether laterality of trajectory (arc_pre) predicted an adequate motor outcome, with exploratory analyses of anterior trajectory (ring_pre), ventricular width, and accuracy of placement (Cartesian electrode displacement) as predictors. We used the minimum UPDRS-III reduction with levodopa for DBS candidacy ⁸ to set the threshold for adequate motor outcome at 30% UPDRS-III reduction from the preoperative OFF-medication score to postoperative OFF-medication/ON-DBS score (see Clinical Evaluation, above). A binary outcome with a logistic regression approach was used to improve reliability of the predictive model and maximize clinical interpretability. We performed binary logistic regression to predict adequate motor outcome of the right and left hemibody from the respective left and right arc_pre and ring_pre alone, followed by backward stepwise regression using the following exploratory predictors and covariates to account for confounding: ring_pre, EI (to control for effect of ventricular width), sex, date of surgery (DOS), age at DOS, handedness, disease duration, LEDD at time of surgery, Mattis Dementia Rating Scale, and both 2-dimensional and 3-dimensional Cartesian electrode displacement. Overall model significance was adjusted using Benjamini-Hochberg false discovery rate (FDR) for 2 models with nonindependent variables, and significance of exploratory variables was also adjusted within model using FDR. To further explore these relationships, we performed Pearson correlations between variables of interest, including surgeon-recorded angles, arc_pre and ring_pre, arc_post and ring_post, electrode deformation, UPDRS-III percent change, age at DOS, LEDD at time of surgery, EI, disease duration, and Cartesian displacement of intended vs actual targeted lead location (Supplemental Digital Contents 2, http://links.lww.com/ONS/B151 SDC 3, http://links.lww.com/ONS/B152, SDC 4, http://links.lww.com/ONS/B153, SDC 5, http://links.lww.com/ONS/B154, SDC 6, http://links.lww.com/ONS/B155, SDC 7, http://links.lww.com/ONS/B156). For bivariate correlations, multiple comparisons were controlled using FDR.

RESULTS

Participant demographic, anatomical, and clinical data are summarized in Table 1. Our sample represents the total DBS population at WUSM MDC, with gender being the single discrepancy (51% vs 39% female, respectively). Average corrected electrode laterality (arc_pre) was 11.5° (SD = 3.1) on the left and 10.2° (SD = 2.9) on the right (Table 2).

TABLE 1.

Participant Characteristics

N	49
Age at date of surgery (y)	64.0 ± 9.0
Gender	24 male, 25 female
Handedness	43 right-handed, 6 left-handed
Disease duration (y)	11.8 ± 5.1
Evans index	0.27 ± 0.03
Levodopa equivalent daily dose (mg)	1513.6 ± 657.1
Mattis Dementia Rating Scale	137.1 ± 5.1
UPDRS-III overall percent change (%)	−35.9 ± 13.7
UPDRS-III left lateralized percent change (%)	−36.3 ± 17.9
UPDRS-III right lateralized percent change (%)	−38.9 ± 18.8

UPDRS-III, Unified Parkinson's Disease Rating Scale, part III.

Relevant values reported as mean ± SD.

TABLE 2.

Preoperative (n = 49) and Postoperative (n = 48) Trajectory Angles

Angle	Left	Right
Arcpre	11.5 ± 3.1	10.2 ± 2.9
Ringpre	61.2 ± 6.0	61.2 ± 5.8
Arcpost	10.1 ± 3.0	7.6 ± 2.7
Ringpost	58.2 ± 4.5	59.4 ± 4.3

arc_pre, preoperative arc; arc_post, postoperative arc; ring_pre, preoperative ring; ring_post, postoperative ring.

Relevant values reported as mean ± SD.

Laterality of Planned Trajectory Predicts Adequacy of Motor Outcome

The logistic regression model using left-sided trajectory data predicted adequate motor outcome of the right hemibody (χ² = 11.6, P = .009, R² = 0.31) and correctly classified 79.2% of cases. In the optimized model after stepwise parameter selection, laterality of left trajectory (left arc_pre) significantly predicted adequate motor outcome (P = .024, OR = 0.69, 95% CI 0.50-0.95). Exploratory predictors (ring_pre, ventricular width, Cartesian electrode displacement) were nonsignificant within this model. The model using right-sided trajectory data also predicted adequate motor outcome of the left hemibody (χ² = 12.0, P = .007, R² = 0.30), although only correctly classifying 66.2% of cases. In this model, laterality of right trajectory (right arc_pre) did not significantly predict adequate motor outcome (P = .056, OR = 0.78, 95% CI 0.60-1.0). Ventricular width showed trend-level significance (P = .046) but was not significant after multiple comparisons, while other exploratory predictors were nonsignificant. In both models, the direction of this relationship suggested that increasingly large lateral trajectory angles were associated with decreased odds of adequate motor outcome.

Laterality of left planned trajectory (left arc_pre) correlated moderately with right body UPDRS-III percent change, but this was not significant after adjustment for multiple comparisons (r = 0.282, P = .049, Figure). Laterality of right planned trajectory (right arc_pre) and left body UPDRS-III percent change did not significantly correlate (r = 0.140, P = .337, Figure).

FIGURE.

A, Right body UPDRS-III percent change vs left preoperatively planned arc angle (n = 49) which positively correlate (r = 0.282, P = .049). B Left body UPDRS III percent change vs right preoperative arc angle (n = 49) without significant correlation (r = 0.140, P = .337). UPDRS-III, Unified Parkinson's Disease Rating Scale, part III.

Relationship of Planned Trajectories With Postoperative Measured Trajectories, Electrode Placements, and Motor Outcome

To verify that intended trajectory reflected actual trajectory, we correlated planned trajectory angles with those measured postoperatively. Arc_pre strongly correlated with arc_post (L: r = 0.902, P < .001; R: r = 0.850, P < .001), while ring_pre more moderately correlated with ring_post (left: r = 0.672, P < .001; right: r = 0.449, P = .001), particularly on the right (Supplemental Digital Content 8, http://links.lww.com/ONS/B157 and Supplemental Digital Content 9. Figure 6a and 6b, http://links.lww.com/ONS/B158). Mean 2-dimensional (axial) Cartesian displacement of the planned target from actual postoperative location was 2.3 ± 1.6 mm on the left and 1.8 ± 1.1 mm on the right. Arc_pre and ring_pre did not correlate with Cartesian displacement between actual and planned electrode placement (Supplemental Digital Content 10. Table 1, http://links.lww.com/ONS/B159). We found no significant correlations between lead deformation measures and trajectories or UPDRS-III percent change, after Bonferroni correction (Supplemental Digital Content 10. Table 1, http://links.lww.com/ONS/B159). Cartesian displacement did not predict lateralized adequate motor outcome in an optimized logistic regression model on either left or right.

DISCUSSION

We demonstrate a predictive relationship between trajectory data and adequacy of motor outcome in STN DBS for PD, although this relationship was stronger for left electrode data on right hemibody outcome. Most notably, laterality of planned left electrode trajectory (left arc_pre) predicted corresponding motor outcome on the right hemibody. A similar effect was observed for laterality of right electrode trajectory (right arc_pre) on left hemibody outcome, although with a more variable odds ratio and lack of statistical significance. The strength of the relationship observed supports a modest effect of electrode trajectory angle on motor outcome, although this effect is independent of likely confounding factors such as age, ventricular width, and inaccuracy in surgical placement from intended target.

Many factors may affect DBS lead placement accuracy. In our data, there is no relationship between electrode deformation, trajectory angles, or motor outcome. Our planned and measured electrode angles highly correlate, supporting precision of surgical planning. We evaluated CT scans obtained immediately after surgery, so are unable to account for electrode displacement occurring after the immediate postoperative period.⁷ Approaches at greater angles from midsagittal plane may aggravate such deviations, resulting in less effective stimulation of STN motor domain and surrounding fiber tracts, or unwanted nontarget stimulation.⁴

Asymmetric observations here may reflect lead placement sequence, where the first implanted side is always the left lead. Electrode locations on the second implanted side have been described as more dispersed.¹⁰ Implantation of the first lead could cause brain shift or pneumocephalus, which reduces accuracy or exacerbates effects of greater ventricular width for the second implantation.¹⁰ Asymmetries in motor circuitry related to handedness or more affected side in PD may also underlie the observed asymmetry.¹¹ However, it should be noted that right electrode trajectory showed a trend toward the same predictive value as the left.

Effect is independent of ventricular width, which may otherwise influence surgical targeting. Wider lateral ventricles tend to result in wider lateral trajectories regarding the midsagittal plane.¹² We observed significant correlations between ventricular width and arc_pre but none between trajectory and lead displacement. Although increasing laterality does not seem to affect accuracy, it is plausible that this orientation places viable contacts at less ideal orientations relative to “sweet spots” within STN and nearby structures.

Limitations

Our study is limited as a single-site retrospective study, with all demographic and clinical data prospectively collected during 2007 to 2017. As surgical date was not a significant covariate in our analyses, the effect of learning does not appear to be a confounder. Although the correlations between ring_pre and surgeon recorded ring were relatively low, it likely represents the variability in stereotactic frame placement and highlights the necessity of angle transformation into anatomical space.

CONCLUSION

We conclude that laterality of electrode trajectory has a modest effect on motor outcome, with more lateral trajectories being associated with a reduced odds of a good motor outcome. Clinicians and surgeons should be aware of this relationship when selecting patients, discussing potential risk/benefit of the procedure, and planning surgical trajectories within a practically defined approach for STN DBS in PD. These data may inform future evidence-based evaluations of the methodologies used in selecting the angle of approach.

Funding

This work was supported by NIH NS124789 (SAN) NIH NS121630 (JRY) and by Barnes-Jewish Hospital Foundation (Elliot Stein Family Fund and Parkinson disease research fund), American Parkinson Disease Association (APDA), Advanced Research Center at Washington University, Greater St. Louis Chapter of the APDA, Paula and Rodger Riney Fund, and the Jo Oertli Fund.

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

SUPPLEMENTAL DIGITAL CONTENT

Supplemental Digital Content 1. Methods, 2 Figures. Methods. Full explanation on the computation of trajectory angles described in the Methods section of the manuscript. Figure 1. 3D model of the Leksell frame and sample slices (axial, coronal, sagittal) from the “standard” Leksell frame 3D raster image. Figure 2. Graph depicting planes and vectors as defined by the 3 anatomical points and our mathematical equations (not to scale).

Supplemental Digital Content 2. Figure 3a. Age at DOS vs EI values for all included participants (n = 49) with positive correlation (r = 0.520, P < .001).

Supplemental Digital Content 3. Figure 3b. EI vs preoperatively defined arc angle (n = 49) which correlate positively for both left (r = 0.400, P = .004) and right (r = 0.407, P = .004) brain electrode trajectories. Least-squares fit trendlines for each side (left: dotted, right: dashed) displayed for clarity.

Supplemental Digital Content 4. Figure 4a. Right body UPDRS-III percent change vs left preoperative ring angle (n = 49) without significant correlation (r = -0.184, P = .205).

Supplemental Digital Content 5. Figure 4b. Left body UPDRS-III percent change vs right preoperative ring angle (n = 49) without significant correlation (r = -0.015, P = .917).

Supplemental Digital Content 6. Figure 5a. Calculated preoperative arc values vs surgeon recorded arc values (L: r = 0.867, P < .001; R: r = 0.874, P < .001).

Supplemental Digital Content 7. Figure 5b. Calculated preoperative ring values vs surgeon-recorded ring values (L: r = 0.262, P = .069; R: r = 0.214, P = .140).

Supplemental Digital Content 8. Figure 6a. Graph showing preoperative and postoperative arc values (L: r = 0.902, P < .001; R: r = 0.850, P < .001).

Supplemental Digital Content 9. Figure 6b. Graph showing preoperative and postoperative ring values (L: r = 0.672, P < .001; R: r = 0.449, P = .001).

Supplemental Digital Content 10. Table 1. Correlation coefficients (P-value) for image-guided approach angles and electrode deformation measures (n = 48).