Novel accelerometer-based markers for physical activity in amyotrophic lateral sclerosis based on an open-source summary metric

Abstract

Accelerometer-based physical activity monitoring has previously been shown to be an objective method for quantifying disease progression in patients with amyotrophic lateral sclerosis (ALS). Currently, accelerometer-based endpoints are based on Activity Counts (AC), a metric provided by accelerometer manufacturers to summarize raw accelerometer data into epochs. However, Activity Counts are proprietary and not calculated the same across manufacturers, making results more difficult to reproduce, less comparable, and potentially slowing down ALS research. As a result, this thesis project investigates whether markers based on the open-source summary metric Activity Index (AI) can be used to quantify disease progression and thus serve as an alternative to AC-based markers. This research made use of an existing data set from a previously conducted longitudinal cohort study. In the study, 42 ALS patients wore the ActiGraph during waking hours for seven days every 2-3 months. In addition, they provided additional information about their daily functioning as measured by the ALS functional rating score (ALSFRS-R). The raw accelerometer data was pre-processed, and 61 modified AI metrics for ten different epoch lengths were calculated. These metrics were used in the calculation of three marker types. In total, 183 different markers were defined, each of which aimed to reflect the physical activity of a patient. Finally, the associations of these markers with clinical markers of disease severity, the longitudinal rates of decline, and the associations of AI-based markers with AC-based markers were investigated. All markers were moderate to strongly associated with the ALSFRS-R total score (0.73, 95% CI 0.56 – 0.88, p<0.001), and their correlations increased with increasing epoch length. Most markers showed less between-patient variability over time than the ALSFRS-R total score (coefficient of variation 0.80 – 1.18 vs. 1.06, respectively). In comparison to some AC-based markers, AI-based markers had slightly lower correlations and a higher between-patient variability over time. AI-based markers were strongly correlated with the AC-based markers. In conclusion, this thesis project demonstrated that activity markers based on AI metrics can be used to quantify disease progression in patients with ALS and have the potential to be an alternative to markers based on Activity Counts