Modeling a Composite Score in Parkinson's Disease Using Item Response Theory

Area of Science:

• Neurology
• Biostatistics
• Psychometrics

Background:

• Parkinson's disease (PD) progression is complex and challenging to quantify longitudinally.
• The Movement Disorder Society (sponsored revision) of Unified Parkinson's Disease Rating Scale (MDS-UPDRS) is a key clinical endpoint.
• Existing methods may not fully capture the nuanced longitudinal changes in MDS-UPDRS scores.

Purpose of the Study:

• To develop and validate an Item Response Theory (IRT) model within a non-linear mixed effects framework.
• To characterize longitudinal changes in MDS-UPDRS scores in de novo Parkinson's disease patients.
• To provide a robust analytical framework for PD clinical trials.

Main Methods:

• Utilized data from the Parkinson's Progression Markers Initiative (PPMI) database (163,070 observations, 430 subjects).
• Developed a three-latent variable IRT model with mixture implementation.
• Modeled the probability of item scores as a function of subject disability and time.

Main Results:

• The three-latent variable model adequately described longitudinal changes at both total and item levels.
• Patient-reported and non-sided items showed similar linear progression rates (approx. 50 months/SD).
• Sided items exhibited differential progression, with the better side deteriorating faster than the disabled side.

Conclusions:

• The proposed IRT framework effectively characterizes MDS-UPDRS longitudinal changes in Parkinson's disease.
• The model provides insights into differential progression rates across symptom domains.
• This methodology can enhance the design and analysis of Parkinson's disease clinical trials.