Assessment of feasibility of actigraphy as a measure of clinical change in response to an experimental interventional treatment in adolescents and adults with autism spectrum disorder

Abstract

Objective: The use of actigraphy as a continuous experimental measure of clinical change was explored through a comparison of two clinical studies in autism spectrum disorder (ASD). The data quality, implementation ease, wear compliance, and clinical outcome correlation of actigraphy as a measure were assessed.

Methods: Two clinical studies were conducted and used as a basis of comparison: (1) AUT2001, a Phase 2A interventional study in ASD (N=63), and (2) AUT0002, a Phase 0 non-interventional study in typically-developing (TD) participants (N=53). Participants in both studies wore a wrist-based actigraph throughout enrollment. Actigraphy features were identified based on potential clinical relevance and calculated as weekly averages for each participant's study timepoints. Expert review was used to confirm validity of automated sleep/wake period detection. Feature differences were then assessed using t tests/ANCOVA. Spearman rank correlations between actigraphy features and caregiver reported outcome measures were also examined.

Results: Results from both clinical studies were combined during analysis. Actigraphy was shown to be feasible as a measure of longitudinal change in ASD, but with notable challenges in adherence: participant exclusions due to poor wear compliance substantially reduced the size of the final dataset. Despite this limitation, several findings were noted. Significant differences in sleep disturbance were observed at baseline between the ASD and TD populations as measured by physical activity occurring within the defined sleep period. No significant between-group differences were noted in changes from baseline to endpoint in key sleep variables. Caregiver reported sleep quality significantly correlated with actigraphy measures of sleep disturbance. Additional significant correlations were observed between caregiver reported outcomes of self-regulation and actigraphy features measuring daytime physical activity. Finally, potentially relevant correlations with anxiety, social responsiveness, and restricted and repetitive behaviors are reported.

Conclusions: The observed correlations suggest there may be alignment between some generalized features of actigraphy and core and associated domains of ASD. The clinical utility of actigraphy as a biomarker of clinically relevant outcomes in ASD requires further study. Actigraphy may provide supportive evidence of treatment outcome, providing clinical context, or as a objective behavioral measure (e.g., of sleep or activity level) when combined with more traditional clinical outcome measures.

Keywords: actigraphy; autism spectrum disorder; biomarker; clinical trials; feasibility; outcome measurement.

Figure 1

Comparison of an HDCZA detected sleep period and one detected using the Cole-Kripke sleep state algorithm (37) with Tudor-Locke period boundary detection (38), a standard offering available through the ActiGraph™ CentrePoint system.

Figure 2

Sleep period detection correction example: a visual inspection/correction procedure was implemented to address issues arising from GGIR’s HDCZA method of detecting sleep and sleep period boundaries: suspected non-wear periods that occurred near a detected sleep period were occasionally combined, resulting in an inaccurate sleep period start time that affected downstream features. The inspector attempted to identify these occurrences, updated the sleep period start time, and provided these values to GGIR for a second run of feature extraction.

Figure 3

Visual inspection workflow used to determine final feature set used in analyses.

Figure 4

Relationship between ABI Self-Regulation subdomain and number of moderate-or-vigorous activity (MVPA) fragments (A) and number of sustained inactivity bouts (SIBs) during waking hours (B).

Figure 5

Relationship between “Duration of Sleep During Sleep Period” and ABI Self-Regulation (A) and the SRS-2 Total Score (B).