Capturing longitudinal change in cerebellar ataxia: Context-sensitive analysis of real-life walking increases patient relevance and effect size

OBJECTIVESWith disease-modifying drugs for degenerative ataxias on the horizon, ecologically valid measures of motor performance that can detect patient-relevant changes in short, trial-like time frames are highly warranted. In this 2-year longitudinal study, we aimed to unravel and evaluate measures of ataxic gait which are sensitive to longitudinal changes in patients real life by using wearable sensors. METHODSWe assessed longitudinal gait changes of 26 participants with degenerative cerebellar disease (SARA:9.4{+/-}4.1) at baseline, 1-year and 2-year follow-up assessment using 3 body-worn inertial sensors in two conditions: (1) laboratory-based walking (LBW); (2) real-life walking (RLW) during everyday living. In the RLW condition, a context-sensitive analysis was performed by selecting comparable walking bouts according to macroscopic gait characteristics, namely bout length and number of turns within a two-minute time interval. Movement analysis focussed on measures of spatio-temporal variability, in particular stride length variability, lateral step deviation, and a compound measure of spatial variability (SPCmp). RESULTSGait variability measures showed high test-retest reliability in both walking conditions (ICC > 0.82). Cross-sectional analyses revealed high correlations of gait measures with ataxia severity (SARA, effect size {rho}[≥]0.75); and in particular with patients subjective balance confidence (ABC score, {rho}[≥]0.71), here achieving higher effect sizes for real-life than lab-based gait measures (e.g. SPCmp: RLW {rho}=0.81 vs LBW {rho}=0.71). While the clinician-reported outcome SARA showed longitudinal changes only after two years, the gait measure SPCmp revealed changes already after one year with high effect size (rprb=0.80). In the subgroup with spinocerebellar ataxia type 1, 2 or 3 (SCA1/2/3), the effect size was even higher (rprb=0.86). Based on these effect sizes, sample size estimation for the gait measure SPCmp showed a required cohort size of n=42 participants (n=38 for SCA1/2/3 subgroup) for detecting a 50% reduction of natural progression after one year by a hypothetical intervention, compared to n=254 for the SARA. CONCLUSIONSGait variability measures revealed high reliability and sensitivity to longitudinal change in both laboratory-based constrained walking as well as in real-life walking. Due to their ecological validity and larger effect sizes, characteristics of real-life gait recordings are promising motor performance measures as outcomes for future treatment trials.