Objective: To dynamically predict phenoconversion from a sporadic late-onset cerebellar ataxia (SLOCA) phenotype to MSA-C.
Background: Cerebellar multiple system atrophy (MSA-C) is diagnosed according to the consensus diagnostic criteria. Although these criteria show excellent specificity in advanced disease stages, their sensitivity remains limited inearly stages. Developing a joint survival model to predict the time to phenoconversion, defined as the transition to a fully diagnosable clinical presentation, could improve patient classification and better capture diagnostic uncertainty in atypical cases.
Method: We included 342 patients consecutively included in our SLOCA prospective registry. All patients received ataxia quantification every six months (SARA and SDFS scores). putaminal/occipital dopaminergic binding ratio (SBR) was evaluated every year through DaTscan. Brain MRI were repeated every year with pons area as well as middle cerebellar peduncles diffusivity measurement (MCP rADC). Biomarkers trajectories were compared between non-MSA-C and MSA-C using mixed-modelling. A joint survival random forest model was trained to provide dynamic prediction of time to conversion to MSA-C among SLOCA patients using the trajectories of Gaussian markers.
Results: 65 SLOCA patients converted to MSA-C while 277 did not. MSA-C patients had a shorter disease duration at inclusion (2.03 vs 3.38 years, p<0.001), more severe clinical impairment, with higher SARA (12.5 vs 10) and SDFS (3.63 vs 3.28) scores which worsened over the disease course.
MRI and DAT markers were strongly associated with MSA-C, including smaller pontine area (402 vs 535 mm²), higher MCP rADC (840.6 vs 699.9 mm²/s), and lower putaminal SBR (1.66 vs 2.06) which worsened over the disease course.
Survival was significantly shorter in MSA-C (median 7.6 years, HR 10.5, p<0.001).
A multimodal dynamic random forest model predicted phenoconversion with high accuracy (integrated Brier score 0.023, BSS 0.85), with MCP rADC as the most influential predictor.
Conclusion: Although our results warrant further validation on other SLOCA and MSA-C cohorts, the tool developped in this work may be used to identify patients at risk of converting to an MSA-C phenotype from an early stage of their disease, facilitating their recruitment in neuroprotection trials at a time when drugs may have the most significant impact on disease progression.
To cite this abstract in AMA style:
T. Wirth, A. Storck, T. Bogdan, I. Pei, IJ. Namer, S. Kremer, C. Tranchant, M. Anheim. Dynamic and Multimodal Prediction of Phenoconversion to Cerebellar Multiple Sytem Atrophy among Sporadic Late-Onset Cerebellar Ataxia [abstract]. Mov Disord. 2026; 41 (suppl 1). https://www.mdsabstracts.org/abstract/dynamic-and-multimodal-prediction-of-phenoconversion-to-cerebellar-multiple-sytem-atrophy-among-sporadic-late-onset-cerebellar-ataxia/. Accessed October 1, 2026.« Back to 2026 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/dynamic-and-multimodal-prediction-of-phenoconversion-to-cerebellar-multiple-sytem-atrophy-among-sporadic-late-onset-cerebellar-ataxia/