Prediction of STN DBS Motor Response in PD (DBS-PREDICT)

Machine Learning Prediction of Motor Response After STN DBS in Parkinson Patients, a Retrospective Multicenter Validation Study

Despite careful patient selection for subthalamic nucleus deep brain stimulation (STN DBS), some Parkinson's disease (PD) patients show limited improvement of motor disability. Non-conclusive results and the lack of a practical implantable prediction algorithm from previous prediction studies maintain the need for a simple tool for neurologists that provides a reliable prediction on postoperative motor improvement for individual patients.

In this study, a prior developed prediction model for motor response after STN DBS in PD patients is validated. The model generates individual probabilities for becoming a weak responder one year after surgery. The model will be validated in a validation cohort collected from several international centers.

The predictive model is made public accessible before data collection on: https://github.com/jgvhabets/DBSPREDICT

Study Overview

• Status: Completed
• Conditions: Parkinson Disease
• Intervention / Treatment: Other: Prediction of motor outcome after STN DBS based on preoperative variables

Detailed Description

Predicting motor outcome after STN DBS in Parkinson Disease can be challenging for the clinician. Current prediction studies report non-conclusive results on the most important predictors and are limited by used computational methods. Traditional statistical analyses which focus on correlations are biased by predictor- and confounder-selection by the investigators. Modern computational methods like machine learning prediction models are less limited by sample size and can consider a wider range of predictors which leads to less selection-bias.

Retrospective patient data is collected from multiple international centers. This retrospective, multicenter cohort is used to validate the model which is developed based on a single-center retrospective cohort.

The goal is to develop a prediction tool that provides the clinician with a probability for weak response during the preoperative phase. This could support the clinician in including or informing the patient during preoperative counseling.

The predictive model is made public accessible before data collection on: https://github.com/jgvhabets/DBSPREDICT.

• Study Type: Observational
• Enrollment (Actual): 322

Contacts and Locations

Study Locations

• Netherlands
  • Limburg
    • Maastricht, Limburg, Netherlands, 6229 AZ
      • MaastrichtUMC

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years and older (Adult, Older Adult)
• Accepts Healthy Volunteers: No
• Genders Eligible for Study: All

• Sampling Method: Non-Probability Sample

Study Population

One cohort consisting of PD patients who underwent STN DBS in several international centers.

Description

Inclusion Criteria:

• underwent STN DBS for Parkinson's disease
• completed one year follow up after surgery

Exclusion Criteria:

- missing data in postoperative UPDRS II, III, IV

Study Plan

How is the study designed?

Design Details

• Observational Models: Cohort
• Time Perspectives: Retrospective

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
multi-center validation cohort; We collect retrospective data from several international centers containing preoperative variables (demographical and clinical) and postoperative outcome (UPDRS II, III, IV) one year postoperatively, and merge these data to one validation cohort.	Other: Prediction of motor outcome after STN DBS based on preoperative variables; Generating individual probabilities for motor response based on preoperative variables

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
area under the curve of the receiver operator curve	Motor outcome is categorised in a binary outcome variable. The model will predict to which outcome group the patient will belong one-year postoperatively. The primary outcome measure is the performance of the predicted outcome categories with the actual outcome categories. Performance of prediction models is expressed as area under the curve of the receiver operator curve, predictive accuracy, true positive prediction rate, and false positive prediction rate.	one-year postoperative
predictive accuracy	See description primary outcome 1.	one-year postoperative
true positive prediction rate	See description primary outcome 1.	one-year postoperative
false positive prediction rate	See description primary outcome 1.	one-year postoperative

Collaborators and Investigators

• Sponsor: Maastricht University Medical Center

Study record dates

Study Major Dates

• Study Start (Actual): August 1, 2019
• Primary Completion (Actual): December 17, 2019
• Study Completion (Actual): December 17, 2019

Study Registration Dates

• First Submitted: September 17, 2019
• First Submitted That Met QC Criteria: September 17, 2019
• First Posted (Actual): September 18, 2019

Study Record Updates

• Last Update Posted (Actual): September 1, 2020
• Last Update Submitted That Met QC Criteria: August 31, 2020
• Last Verified: February 1, 2020

More Information

Terms related to this study

• Keywords: deep brain stimulation; machine learning; Subthalamic nucleus; motor outcome; prediction model
• Additional Relevant MeSH Terms: Brain Diseases; Central Nervous System Diseases; Nervous System Diseases; Parkinsonian Disorders; Basal Ganglia Diseases; Movement Disorders; Synucleinopathies; Neurodegenerative Diseases; Parkinson Disease
• Other Study ID Numbers: 2019-0739-A9

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: Yes
• IPD Plan Description: Anonymized data will be shared after completing analysing
• IPD Sharing Time Frame: After data collection and analysis.
• IPD Sharing Access Criteria: Data can be made available on request.
• IPD Sharing Supporting Information Type: Study Protocol; Analytic Code

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No