Using "Big Data" Methods to Better Plan Clinical Trials and Predict Outcomes of Parkinson's Disease

Study Rationale:
Parkinson’s disease (PD) progresses differently in every case, which makes reliable prediction of future outcomes challenging. Looking for better ways to predict disease progression, researchers have been trying to identify endpoints, or specific clinical signs, symptoms or test results that people living with PD would have at some point in the future. To do this, researchers first need to identify sub-types of PD by grouping together people whose disease has been progressing similarly. However, most of these studies did not use the “big data” approach, which is a new, more powerful research method yielding more informative findings.

Hypothesis:
We will compare the abilities of traditional and “big data” research methods to predict future endpoints and identity sub-types of PD. We are planning to use unbiased recursive partitioning, a statistical method for analysis that allows us to classify members of a population such as Parkinson’s patients into sub-populations based on independent variables. We hypothesize that this method will help us better understand Parkinson’s progression.

Study Design:
Using unbiased recursive partitioning, we will analyze the data from the Parkinson’s Progression Markers Initiative (PPMI), a landmark study to find biomarkers – disease indicators that are critical missing links in the search for better PD treatments.

Impact on Diagnosis/Treatment of Parkinson’s disease:
Because unbiased recursive partitioning allows us to easily group people with PD into sub-groups using their data, it may improve the prediction of future endpoints and planning of future clinical trials.

Next Steps for Development:
Unbiased recursive partitioning could be a useful clinical tool for predicting outcomes and planning clinical trials.