Sleep-specific DBS Therapy in Parkinson's Disease

Overview

Sleep-wake disturbances are a major factor associated with reduced quality of life of individuals with Parkinson's disease (PD), a progressive neurological disorder affecting millions of people in the U.S and worldwide. The brain mechanisms underlying these sleep disorders, and the effects of therapeutic interventions such as deep brain stimulation on sleep-related neuronal activity and sleep behavior, are not well understood. Results from this study will provide a better understanding of the brain circuitry involved in disordered sleep in PD and inform the development of targeted therapeutic interventions to treat sleep disorders in people with neurodegenerative disease.

Sleep-wake disturbances are a major factor associated with reduced quality of life of individuals with Parkinson's disease (PD), a progressive neurological disorder affecting millions of people in the U.S and worldwide. This study will examine the sleep of people who are receiving Deep Brain Stimulation (DBS) therapy for PD. DBS uses electrical fields to change the activity in a few structures deep in the brain, to improve motor (movement) symptoms of Parkinson's disease (PD). The DBS system includes a wire (the lead) that goes deep in the brain to deliver electricity and a battery pack (IPG) that controls the electricity that is delivered by the lead. Patients receiving this treatment may have a lead implanted on just one side of their brain (unilateral) or both sides (bilateral). In this proposal, we will recruit people who have DBS systems with the electrical lead implanted in the internal segment of the globus pallidus (GPi) or the subthalamic nucleus (STN). The GPi and STN are two parts of the brain that are commonly targeted in DBS therapy to treat PD symptoms. The site of lead implantation (GPi or STN) will be decided by their clinical team, as is standard-of-care. Participation in this study will have no impact on site selection (and in fact, site selection will take place before recruitment and enrollment in this study). There are two separate studies in this proposal: an Aims 1 \& 2 study, and an Aim 3 study. All participants will be 21 years or older. All participants in Aims 1 \& 2 may potentially be recruited for Aim 3, however enrollment in Aim 3 is not contingent upon enrollment in Aims 1 \& 2. Participants in both studies will have received a 7 Tesla MRI scan (either as part of their regular clinical treatment, or as part of a different study, run by Dr. Noam Harel), which is a very detailed image of the brain. Participants will be asked for their permission to use these images as research data for these studies. All patients will undergo standard-of-care pre-surgery UPDRS-III exams, sleep-specific exams including the Pittsburgh Sleep Quality Index (PSQI), Epworth Sleepiness Scale (ESS), and REM Sleep Behavior Disorder Questionnaire (RBDSQ). AIMS 1 \& 2 STUDY For Aims 1 and 2, we will recruit 24 patients who have been approved for GPi or STN (12 in each group) DBS surgery. Participants must be approved for DBS surgery to treat PD at the University of Minnesota. DBS surgery is done in two stages: the first stage surgery is to implant the lead(s), and the second stage surgery is to implant the IPG and connect it to the wires. Patients who choose to participate in this study will have two research-only procedures done during their lead implantation surgery, which will otherwise be as standard of care: ï Recording strips (ECoG strips) will be placed in two spots on the surface of the brain, and their location will be confirmed using fluoroscopy ï The ends of the DBS leads will be left outside of the body (ìexternalizedî, under a sterile bandage) so that they can be hooked up to research recording equipment After the first stage lead implantation surgery is complete, participants will be discharged to the M Health Clinical Research Unit (CRU), where they will spend 2 nights having research recordings done before having their second stage IPG implantation surgery. In the CRU, participants will have standard, clinical DBS programming done. This is when a clinician figures out the best settings for their DBS system. The DBS leads that were left externalized will be hooked up to equipment that can record the signal and control DBS stimulation. Participants will be able to take their regular PD medications during the daytime, but will be asked not to take any PD meds after 6pm. Reaching Task: The signals from the implanted lead(s) and strips will be recorded while the participant does a simple touchscreen reaching task. They will be asked to reach out and touch the screen with the arm on the opposite side of the side of the brain in which the DBS lead is implanted. Sleep Recording and Staging: For one of the nights in the CRU, participants will have their DBS stimulation turned ON, and for the other night, they will have it turned OFF. The order of the ON vs OFF nights will be picked by chance, but also making sure that the number of participants is split evenly over the two possible orders. For the ON stimulation night, the DBS stimulation will be turned on at 6pm so the stimulation can have time to work up to its full effect. Video-polysomnography (v-PSG) will be collected according to American Academy of Sleep Medicine (AASM) standards and include EEG (recording from the surface of the scalp), measurement of eye movement, measurement of activity in certain muscles. This will be set up by a qualified technician, following the directions of sleep neurologist Dr. Michael Howell. ECoG strip removal: The morning after the second night of CRU data collection, participants will return to the surgical suite for placement of the IPG, as would be done even if they werenít in this research study, and removal of the ECoG strips that were used for research recordings. AIM 3 STUDY For Aim 3, we will recruit forty subjects who have been approved for bilateral STN or GPi (20 in each group) DBS surgery at UMN, or who have already received their bilateral STN or GPi DBS system implanted. Baseline Sleep Assessments Baseline sleep data will be collected in patients after DBS implantation but prior to their regular, clinical initial programming (which usually happens 4-6 weeks after DBS surgery). Participants will be sent home with an actigraphy ìwatchî which is worn on the wrist, and is commonly used for sleep assessments in our sleep clinic. Participants will be instructed to wear the watch for the two weeks prior to their standard-of-care initial programming session. Sleep rating scales Sleep rating scales will be collected as part of their standard-of-care initial programming visit (typically 4-6 weeks post-surgery), which combined take only 15-20 minutes to complete. Follow-up Sleep rating scales will be collected at their standard of care 6 and 12 month visits. Participants will be sent actigraphy watches again, 3 weeks prior to these appointments. Participants will be instructed to again wear the watch for the two weeks prior to their upcoming clinic visit, and will return the watch at their appointment SAMPLE SIZE Power analysis: In the following, for each aim we will compute the minimum detectable effect size at 80% power and Type I error rate of 5% using our proposed sample size of N=24 for SA1 and SA2 and N=40 for SA3. For SA1, using a paired t-test, the minimum detectable Cohenís d for the difference in the mean sleep-related neural activity and two sleep stages (e.g., wake vs light sleep) is d=0.60 standard deviations; here, ìstandard deviationî describes the variation in the sleep-related neural activity between the 24 subjects. For comparison, our preliminary data from n=5 patients yielded a much larger effect size of d=1.48 when comparing beta band power during wake vs NREM sleep stages. For SA2, we used Monte Carlo simulations to compute the power to the interaction between DBS and the indicator for the sleep stage while respecting the within-subject design. In the simulations, we assumed two sleep stages whose difference in neural activity was d=0.60 (i.e., the same as the minimum detectable effect calculated for SA1). Using these specifications, we calculated 80.5% power to detect an interaction effect with a magnitude of ?2=0.29. This number is the proportion of variation in the sleep-related neural activity explained by the interaction between DBS and sleep stage. For instance, an interaction effect of this magnitude is present when the difference in neural activity, calculated by wake minus light sleep, is d=0.67 for off-DBS and d=1.51 for on-DBS. Our preliminary data comparing beta power in the wake sleep stage when on-DBS vs off-DBS was d=1.93. For SA3, the minimum detectable correlation between a sleep outcome and a pathway activation is r=0.43. DATA ANALYSIS Statistical analysis of all clinical and physiological data will be done under the direction of M. Fiecas (Co-I), Division of Biostatistics. General approach. For all aims, linear mixed models (LMMs) will be used to estimate associations between neurophysiological measures and clinical outcomes, and to estimate mean comparisons in these outcomes across sleep stages (wake, light sleep, SWS, REM sleep) and across experimental conditions (e.g., ON vs OFF DBS). LMMs use random effects to account for the repeated measurements within a subject across nights or visits, across experimental conditions, and across topography (e.g., within STN/GPi). LMMs contain the repeated measures analysis of variance (ANOVA) as a special case, but has greater modeling flexibility and can handle missing data should they arise. If an outcome is not normally distributed, generalized linear mixed models (GLMMs) will be used instead. The method of restricted maximum likelihood (REML) will be used to estimate the parameters of the LMMs. Likelihood ratio tests (LRTs) will be used to test hypotheses about model parameters, using a significance level of ?=0.05. Biological variables. All models will include fixed effects for biological variables important in PD, such as sex and age, determined in consultation with the Dr. Fiecas (Co-I), so that we can examine whether primary effects differ by sex. Type I error control. In all statistical analyses, the False Discovery Rate procedure will be used to adjust the p-values for multiple testing. Assessment of model fit. We will use residual analyses to assess how well each model fits the data, compute objective measures such as the Akaike information criterion177 to compare model fit across models and use predictive cross-validation178 to assess the predictive capability. Based on residual diagnostics, we will consider potential data transformations or other functional forms (e.g. logarithmic) to improve model fit, interpretability, and predictive power. Missing data. Different amounts of neurophysiological data may be collected from different people (e.g., different amount of data in specific sleep stages) and this is not considered missingness. Missing neurophysiological data could arise if an experimental condition (e.g., DBS ON) was missed. We will use Multiple Imputation using Chained Equations (MICE)179 because of its flexibility in accommodating multiple variable types (e.g., continuous, binary). All consented participants will be used in analyses. In SA1, each of the oscillatory activity, coupling, and connectivity measures will be the dependent variable in an LMM. First, we model the association between the sleep-related neural activity and the sleep stages. The primary predictors in the LMM will be indicator variables associated with the sleep stages. The LMM parameter associated with each indicator variable captures the mean sleep-related neural activity within a particular sleep stage. Estimating and testing for differences in the sleep-related neural activity between sleep stages can be accomplished by setting up appropriate contrasts in the LMM. Next, to model the changes in neural activity related to sleep fragmentation, we restrict our analysis to data collected before and after microarousals, and in the LMM we use only one indicator variable that marks the wake stage. The LMM parameter associated with this indicator variable thus captures the change in the neural activity associated with nighttime awakenings. The LMMs for SA2 build on the LMMs from SA1 by including a main effect for DBS and its interaction with the other primary predictors. For instance, to model the effect of DBS on the sleep-related neural activity across sleep stages, as described above, the LMM will have indicator variables for the sleep stages, and we now include a main effect for DBS and an interaction term between each indicator variable with DBS. In this model, the main effect of DBS quantifies the effect of DBS averaged across the sleep stages, and the interaction terms quantify the difference between off-DBS and on-DBS in their associations between the neural activity and the sleep stages. The interaction terms will be of primary interest. The other LMMs for SA2 will be constructed similarly. In SA3, each sleep outcome (e.g., sleep fragmentation, excessive daytime sleepiness, PSQI, ESS), will be the dependent variable in an LMM. The primary predictors will be the axonal pathway activations, a main effect for DBS, and an interaction term between DBS and the pathway activations. In this model, the main effect for the pathway activations quantifies the association between pathway activation and the sleep outcome averaged across off-DBS and on-DBS, and the interaction term quantifies the difference in this association between off-DBS and on-DBS.