Objective: To assess reproducibility of published data-driven PD subtype classification systems.

Background: There is no consensus on which is the most robust PD subtype classification system for clinical research. A key feature of a valid PD subtype classification system is reproducibility in samples other than the original cohort used to define the PD subtype classification system. Longitudinal and Biomarker Studies in Parkinson’s Disease (LABS-PD) is a longitudinal cohort study with 461 PD patients enrolled in the PRECEPT randomized controlled trial within two years of diagnosis and subsequently followed at least 5 years from diagnosis. LABS-PS offers the opportunity to test reproducibility in a well-characterized cohort.

Methods: We used a Delphi-consensus process with a 4-member panel of PD experts (CMT, AEL, DG, CM) to define reproducibility of PD subtype classification systems. Using cluster analysis, eight published data-driven PD subtype classification systems were replicated in LABS-PD. For the Delphi process, the panel was provided with key demographic and clinical characteristics of the different PD subtypes generated in LABS-PD for each PD subtype classification system, and asked to judge reproducibility of subtypes in LABS-PD, as compared with the original report, and to identify sources of discrepancies. Reproducibility was determined by consensus on the similarity of subtypes identified in LABS-PD as compared to the original classification system.

Results: After two iterations using the Delphi-process, we concluded that none of the eight published PD subtype classification systems was reproduced in LABS-PD. Heterogeneity in the study populations as well as differences in study design (cross-sectional vs. longitudinal) and source of recruitment (clinical trial vs. community cohort) were identified as possible contributors to the lack of reproducibility of the published data-driven PD subtype classification systems in LABS-PD.

Conclusions: Currently published data-driven PD subtype classification systems lack reproducibility in a well-characterized cohort of patients initially recruited for a clinical trial of a disease-modifying therapy. The results call for the establishment of standards to generate a data-driven PD subtype classification system.

« Back to 2017 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/reproducibility-of-data-driven-parkinsons-disease-subtypes-for-clinical-research-using-a-delphi-process-for-expert-based-consensus/