JMIR Publications

ABSTRACT

Background:

Prader-Willi Syndrome (PWS), the most common syndromic cause of obesity, is characterized by hyperphagia and behavioral outbursts for which there are limited FDA-approved treatments. Objective outcomes to assess change in behaviors with treatment are needed for future clinical trials.

Objective:

The objectives of this pilot study were to collect accelerometry, heart rate (HR), and heart rate variability (HRV) data from adults with PWS living in a group home setting to (1) develop an algorithm using accelerometry data for detecting eating in this population, (2) explore the association of HR and HRV with questionnaires commonly used as outcomes in PWS trials, and (3) assess the acceptability and feasibility of using wrist-worn wearable devices to collect data from people with PWS.

Methods:

Residents of a PWS group home wore a LEAPTM accelerometer (Ametris, formerly ActiGraph, Pensacola, FL) on their dominant hand and a Vivoactive® 5 smartwatch (Garmin, Olathe, KS) on their non-dominant hand simultaneously for two weeks to collect activity, HR, interbeat interval, and 3-axis accelerometer data. Resident schedules were recorded by group home staff. An eating-detection algorithm was developed from accelerometry data using a random forest model. HR and HRV were assessed for correlation with behavioral outcomes from caregiver-reported questionnaires (Hyperphagia Questionnaire for Clinical Trials [HQ-CT], Prader-Willi Syndrome Anxiousness and Distress Behaviors Questionnaire [PADQ], and PWS Acuity Scale) and used in linear regression models to predict questionnaire scores. Post-study surveys and interviews with residents and staff were used to assess acceptability and feasibility.

Results:

Six group home residents and three staff members were enrolled. Half the residents were male, four had a deletion genetic subtype, median age was 29.5 years, and median body mass index was 27. All residents were white, and one self-identified as Hispanic. Eating behavior was predicted for each resident with a median accuracy of 76%. Most HR and HRV features were moderately to strongly correlated with questionnaire scores, with the strongest correlation observed between pre-meal HR (difference from resting HR) and the HQ-CT (r=-0.96). HR and HRV accurately predicted scores for HQ-CT but not the other questionnaires. Both residents and staff endorsed the acceptability and feasibility of using smartwatches for data collection.

Conclusions:

Wrist-worn wearable devices have the potential to provide objective outcome measures for detecting change in behaviors, including hyperphagia, in PWS clinical trials. Use of these devices is acceptable and feasible in the PWS population, supporting further research to develop and validate wearables outcomes.