Objective: The Scale for the Assessment and Rating of Ataxia (SARA) is the most widely applied clinical outcome assessment (COA) for genetic ataxias, but presents metrological and regulatory challenges. To facilitate trial planning, we characterize its responsiveness (including subitem-level relations to ataxia severity and patient-focused outcomes) across a large number of ataxias, and provide first natural history data for several of them.
Methods: Subitem-level correlation- and distribution-based analysis of 1637 SARA assessments in 884 patients with autosomal-recessive/early-onset ataxia (370 with 2-8 longitudinal assessments), complemented by linear mixed-effects modeling to estimate progression and sample sizes.
Results: While SARA subitem responsiveness varied between ataxia severities, gait/stance showed a robust granular linear scaling across the broadest range (SARA<25). Responsiveness was diminished by incomplete sub-scale use at intermediate or upper levels, non-transitions ('static periods'), and fluctuating decreases/increases. All subitems -except nose-finger- showed moderate-to-strong correlations to activities of daily living, indicating that metric properties -not content validity- limit SARA responsiveness. SARA captured mild-to-moderate progression in many genotypes, e.g., SYNE1-ataxia: 0.55 points/year, AOA2: 1.14, POLG-ataxia: 1.56; but no change in others (ARSACS, COQ8A-ataxia). While sensitivity to change was optimal in mild ataxia (SARA≤10), it substantially deteriorated in advanced ataxia (SARA>25; 2.7-fold sample size). Use of a novel rank-optimized SARA without subitems finger-chase and nose-finger reduces sample sizes by 20-25%.
Interpretation: This study comprehensively characterizes COA properties and annualized changes of the SARA across and within a large number of ataxias. It suggests specific approaches for optimizing its responsiveness that might facilitate regulatory qualification and trial design.