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

Abstract

Abstract.In the current work, we present the methodology for development of an Item Response Theory model within a non-linear mixed effects framework to characterize the longitudinal changes of the Movement Disorder Society (sponsored revision) of Unified Parkinson's Disease Rating Scale (MDS-UPDRS) endpoint in Parkinson's disease (PD). The data were obtained from Parkinson's Progression Markers Initiative database and included 163,070 observations up to 48 months from 430 subjects belonging to De Novo PD cohort. The probability of obtaining a score, reported for each of the items in the questionnaire, was modeled as a function of the subject's disability. Initially, a single latent variable model was explored to characterize the disease progression over time. However, based on the understanding of the questionnaire set-up and the results of a residuals-based diagnostic tool, a three latent variable model with a mixture implementation was able to adequately describe longitudinal changes not only at the total score level but also at each individual item level. The linear progression rates obtained for the patient-reported items and the non-sided items were similar, each of which roughly take about 50 months for a typical subject to progress linearly from the baseline by one standard deviation. However for the sided items, it was found that the better side deteriorates quicker than the disabled side. This study presents a framework for analyzing MDS-UPDRS data, which can be adapted to more traditional UPDRS data collected in PD clinical trials and result in more efficient designs and analyses of such studies.

Progression of MDS‐UPDRS Scores Over Five Years in De Novo Parkinson Disease from the Parkinson's Progression Markers Initiative Cohort

“…Future evaluation of the PPMI cohort after more time has passed could help determine whether symptom severity progression continues in a linear manner beyond the first 5 years after diagnosis. However, in support of our results, another recent study reported a linear progression in MDS‐UPDRS scores over time . Gottipati and colleagues studied the PPMI cohort and utilized an Item Response Theory model to determine that a typical participant's MDS‐UPDRS total score (excluding Part IV) increases linearly by 1 standard deviation (estimated at 13.1 points) in 50 months .…”

“…However, in support of our results, another recent study MDS-UPDRS Part III Score (SE) reported a linear progression in MDS-UPDRS scores over time. 19 Gottipati and colleagues studied the PPMI cohort and utilized an Item Response Theory model to determine that a typical participant's MDS-UPDRS total score (excluding Part IV) increases linearly by 1 standard deviation (estimated at 13.1 points) in 50 months. 19 This equates to a change of 3.1 points per year for the MDS-UPDRS total score, which lies close to our estimation using alternate methodology.…”

Item Response Theory as an Efficient Tool to Describe a Heterogeneous Clinical Rating Scale in De Novo Idiopathic Parkinson’s Disease Patients

“…the neuronal disability) and therefore, captures more precisely the heterogeneity of the scale (18). Recent publications in the field of central nervous system diseases (19,20) have demonstrated the use of IRT to leverage the information provided by clinical rating scales.…”

Item Response Theory Analysis of the MDS‐UPDRS Motor Examination: Tremor vs. Nontremor Items

“…14 The use of IRT as an approach to examine clinimetric properties of the MDS-UPDRS is recent, with methodological variability in the IRT models used, characteristics of PD samples studied, and portions of the scale used in the analyses. Most of these studies use longitudinal data from the Parkinson's Progression Marker initiative, [15][16][17][18] a cohort of early-stage PD patients who do not require treatment at baseline and are followed longitudinally. The majority of these studies applied IRT methods to examine longitudinal change of either total scale scores or unique combinations of different items of the MDS-UPDRS.…”