From innovation to implementation: translating adaptive deep brain stimulation in Parkinson's disease into clinical practice

Abstract

Introduction
Parkinson’s disease (PD) is the second most common neurodegenerative disorder worldwide, leading to a wide range of motor and non-motor symptoms that significantly affect quality of life. Motor symptoms include, among others, rigidity, slowness of movement (bradykinesia), tremor and gait/balance problems. As the disease progresses, response to dopaminergic medication may become increasingly variable, resulting in motor fluctuations, including OFF periods and dyskinesia—involuntary hyperkinetic movements as a side effect of high dopaminergic medication state.

For these motor response fluctuations, continuous Deep Brain Stimulation (cDBS) is an effective treatment that delivers continuous high-frequency electrical stimulation via implanted electrodes to brain nuclei in the basal ganglia. However, continuous DBS (cDBS) has several limitations in some patients, including stimulation-induced side effects, such as dysarthria or balance problems, and persisting symptom fluctuations related to concurrent medication treatment.

Adaptive Deep Brain Stimulation (aDBS) was designed to overcome these limitations by automatically adjusting the stimulation parameters according to a biomarker that correlates with symptom severity to provide personalized stimulation. A commonly used biomarker is activity in the β-frequency band in the local field potentials (LFPs), local compound neuronal activity measured via the implanted electrodes. Despite promising results in early studies, no standardized workflow for aDBS has been defined yet, while the process is inherently complex with multiple different parameters that should be selected and optimized. Therefore, the primary aim of this thesis was to develop a workflow algorithm for aDBS.

According to a previously reported systematic review, an initial protocol was proposed consisting of four phases: (1) pre-aDBS patient selection, (2) biomarker selection, (3) initial aDBS setup and (4) iterative optimization. In the optimization phase, a decision tree was proposed to decide on which parameter to adjust based on clinical symptoms and results of home monitoring analysis.

Methods
In a retrospective case series, the setup process of all patients treated with aDBS was described and compared to the initial protocol. PD symptoms and home-monitoring data from the last clinical visit before aDBS initiation and from all subsequent visits during aDBS optimization were retrieved from the EHR and from the programming tablets. For the optimization phase, all adjustments were grouped by clinical state and adherence to protocol. Furthermore, an existing Toolbox was extended to allow for visualization of home monitoring data to aid in parameter adjustment decision-making.

Results
Eleven patients were assessed for aDBS, of whom nine patients eventually initiated aDBS. Initial setup varied greatly among patients and did not include biomarker selection OFF-medication. Optimization took median 2 (range: 1–13) sessions. Four patients discontinued aDBS due to lack of clinical benefit after a brief optimization period. Retrospectively, 13 parameter adjustments were in accordance with the protocol, 19 deviated from the protocol, and 14 were not covered by the protocol. Descriptively, protocol adherence led more often to beneficial outcomes than protocol deviations (46% vs 21%). Within these case series, different phenomena were identified. These included changes in baseline biomarker value, resulting in changes in overall stimulation. Moreover, not all patients showed a relationship between biomarker and events or medication intake, but still benefited from aDBS.

Conclusion
In this work, an extended protocol for aDBS setup and optimization was proposed to account for real-world scenarios encountered during clinical practice. This protocol provides a basis for further refinement and validation of standardized aDBS workflows.