To perform quantitative motor analysis in X-linked dystonia parkinsonism (XDP) as a potential clinical trial endpoint.

XDP is a neurogenetic mixed movement disorder involving both parkinsonism and dystonia. For clinical trial readiness, rater-independent, quantitative assessments of motor function are urgently needed. Complex, overlapping phenotypes result in difficulties in clinical rating scale assessment.

10 symptomatic XDP patients and 4 healthy controls underwent motion sensor analysis with a standardized examination. Disease severity was assessed with the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) and Burke-Fahn-Marsden scale (BFM). We assessed specific MDS-UPDRS and BFM elements using clinical scores as the dependent variable, segmented the sensor data and derived a set of data features chosen based on prior work. Data feature selection and projection algorithms were used to inspect the data feature space and assess the suitability of the derived data features to build machine learning-based models to estimate clinical scores from sensor data.

XDP patients were of varying stages of disease (disease duration 3-28 years) and varied phenotype: parkinsonism-predominant (n=5), dystonia-predominant (n=2) and dystonia parkinsonism (n=3), with BFM scores 0-43.5 and MDS-UPDRS 13-48. Projections showed distinct clusters of data points corresponding to different MDS-UPDRS scores for different tasks (hand pronation/supination, leg agility, foot tapping, finger-to-nose), suggesting that data features derived from the sensor data are suitable to derive reliable estimates of clinical scores. We assessed upper/lower limb and cervical dystonia at rest, during  motor tasks and provocative maneuvers, compared to the BFM severity score. Projections of a dystonia-specific features set showed clear cluster separation between controls, XDP with and without dystonia.