Longitudinal Analysis of Natural History Progression of Rare and Ultra‐Rare Cerebellar Ataxias Using Item Response Theory

Degenerative cerebellar ataxias comprise a heterogeneous group of rare and ultra-rare genetic diseases. While disease-modifying treatments are now on the horizon for many ataxias, robust trial designs and analysis methods are lacking. To better inform trial designs, we applied item response theory (IRT) modeling to evaluate the natural history progression of several ataxias, assessed with the widely used scale for assessment and rating of ataxia (SARA). A longitudinal IRT model was built utilizing real-world data from the large autosomal recessive cerebellar ataxia (ARCA) registry. Disease progression was evaluated for the overall cohort as well as for the 10 most common ARCA genotypes. Sample sizes were calculated for simulated trials with autosomal recessive spastic ataxia Charlevoix-Saguenay (ARSACS) and polymerase gamma (POLG) ataxia, as showcased, across multiple design and analysis scenarios. Longitudinal IRT models were able to describe the changes in the latent variable underlying SARA as a function of time since ataxia onset for both the overall ARCA cohort and the common genotypes. The typical progression rates varied across genotypes between relatively high in POLG (~ 0.98 SARA points/year at SARA = 20) and very low in COQ8A ataxia (~ 0.003 SARA points/year at SARA = 20). Smaller trial sizes were required in case of faster progression, longer trials (~ 75-90% less with 5 years vs. 2 years), and larger drug effects (~ 70-80% less with 100% vs. 50% inhibition). Simulating under the developed IRT model, the longitudinal IRT model had the highest power, with a well-controlled type I error, compared to total score models or end-of-treatment analyses. The established longitudinal IRT framework allows efficient utilization of natural history data and ultimately facilitates the design and analysis of treatment trials in rare and ultra-rare genetic ataxias.