Psychometric evaluation of MDS-UPDRS Part I and II to support patient-focused drug development in early Parkinson’s clinical trials

Abstract Background The MDS-UPDRS is the gold-standard clinical outcome assessment (COA) used in Parkinson’s disease (PD) clinical trials. Given the rich pipeline of therapeutics with a focus on earlier intervention, evaluation of the measure for an early PD target population is needed. The aim of this work is to better understand the psychometric/statistical support of the MDS-UPDRS items in an early PD population. Multiple-group latent variable models were used to examine the performance of the patient-reported outcome portions (Parts I and II) of the MDS-UPDRS. Methodology Data from multiple studies, both observational and clinical trials, included in the Critical Path for Parkinson’s Integrated Database were pooled and used in the reported analyses. The latent variable analysis framework was used to explore the dimensionality of the noted parts of the MDS-UPDRS using multiple-group item factor analysis and item response theory models, with groups defined by Hoehn & Yahr (H&Y) stage. Specifically, multiple-group (H&Y Stages 1, 2, and 3) item factor analyses were used to explore the dimensionality of the examined MDS-UPDRS parts and, after achieving acceptable model, the final factor analysis model was reparameterized as a multiple-group item response theory model to provide a more detailed evaluation of the performance of the measure and the individual items contributing to the measure. Results The unidimensionality of Part II was supported by latent variable modeling results. For Part I, a multidimensional model that included method factors for reporter (i.e., clinician-reported versus patient-reported items) provided good fit to the data. The scores from both examined parts of the MDS-UPDRS were found to have acceptable reliability for continued use in PD research. Conclusions Based on the multiple-group analyses of Part II of the MDS-UPDRS, summed scores from these items may be supportable. In contrast, the best fitting model applied to Part I of the MDS-UPDRS suggests that summed scores may not be appropriate for summarizing these items, as accounting for reporter in scoring appears to be important. These results provide further guidance for the use of this tool in early PD for urgently needed disease-modifying treatments.