Objective: To systematically evaluate data-driven Parkinson’s disease (PD) subtyping studies, focusing on their methodological quality and clinical applicability.

Background: The heterogeneity of PD suggests distinct etiological subtypes [1]. While traditional subtyping has been hypothesis-driven [2], a recent focus on data-driven approaches has emerged. However most data-driven studies exhibited significant methodological limitations, limiting their clinical applicability [3]. Given the rapid advancement of research in the field, an updated evaluation of existing studies is warranted.

Method: We systematically evaluated PD subtyping studies using data-driven methodologies published in PubMed/Medline, Web of Science, and Scopus from inception to May 2024. At each screening and review stage, studies were assessed by reviewers with clinical and machine learning expertise. Study quality was measured using a methodological instrument for data-driven studies that evaluates aspects such as population description, replication, longitudinal follow-up and if the study provided a classification algorithm for clinical use, with scores ranging from 0 to 17, where higher values indicate better quality.

Results: A total of 82 studies were included in our review. Of note, the most recent studies had  the highest quality and largest sample sizes, and included more non-clinical features in their analyses.

Most studies were cross-sectional orhad a short follow-up (73.2%), and lacked replication (75.6%). Only 1.2% provided an individual classification algorithm. While most studies were rated as low quality (6.1 ± 3.1), a slight improvement in quality was noted in studies published 2021 onward (6.6 ± 3.2 vs. 5.6 ± 3.0). 

Clinical features were most used as input features for clustering (85.4%), followed by neuroimaging (22%) and omics and fluid biomarkers data approaches (7.3% and 3.7%, respectively). Only 15 (18.3%) studies made their analysis codes available.

Based on the evaluation of eight proposed parameters for predicting disease progression [1], no subtyping approach was deemed predictive, and only three studies evaluated therapeutic response across subtypes.

Conclusion: Several aspects of study design, methodological rigor, and comprehensive data inclusion remain lacking. Closing these gaps is key to developing reproducible PD subtypes that advance precision medicine, clinical trials, and treatment strategies.

Figure 1

Table 1

Table 2

Figure 2

References: 1. Marras C, Fereshtehnejad SM, Berg D, et al. Transitioning from Subtyping to Precision Medicine in Parkinson’s Disease: A Purpose-Driven Approach. Mov Disord. 2024;39(3):462-471.
2. Zetusky WJ, Jankovic J, Pirozzolo FJ. The heterogeneity of Parkinson’s disease: clinical and prognostic implications. Neurology. 1985;35(4):522-526.
3. Mestre TA, Fereshtehnejad SM, Berg D, et al. Parkinson’s Disease Subtypes: Critical Appraisal and Recommendations. Journal of Parkinson’s Disease. 2021;11(2):395-404. doi:10.3233/jpd-202472

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/a-systematic-review-on-the-reliability-of-data-driven-parkinsons-disease-subtypes/