Performance Of The Wrist-worn Actigraph GT3X+ In Measuring Physical Activity In Older Women

Abstract

The ActiGraph GT3X+ (ActiGraph LLC, Pensacola, FL) has been worn on the wrist in studies with the purpose of measuring sleep. However, its performance in assessing physical activity (PA) when worn on the wrist remains unclear. PURPOSE: This study examined the performance of the wrist-worn GT3X+ accelerometer in measuring energy expenditure of PA (PAEE) and classifying PA intensity in a healthy, non-obese population of older women. METHODS: Women [n = 89, age = 65.6 (4.3)] wore the GT3X+ and the SenseWear Armband Mini monitor (SWAM, BodyMedia Inc. Pittsburgh, PA) concurrently for 2 weeks when then doubly labeled water method (DLW) was used to determine total daily energy expenditure (EE). Indirect calorimetry was performed on the last day of the 2 weeks to determine resting EE. Manufacturer-provided software was used to process GT3X+ (2 equations were selected) and SWAM data. Bivariate correlations, Intra Class Correlations, and Bland-Altman plots were performed to evaluate agreement between the GT3X+ and criterion measures for time spent in sedentary, light and moderate-to-vigorous PA (determined by SWAM) and PAEE (determined by SWAM and by DLW and resting EE). An epoch-by-epoch (EBE) analysis was performed to evaluate discrepancy and agreement of PA intensity classification between the GT3X+ and the SWAM. RESULTS: For PAEE, the GT3X+ showed a moderate correlation with criterion measures (r = 0.413 and 0.400 with the SWAM and r = 0.564 and 0.501 with DLW and resting EE), but Bland-Altman plots showed large variability. When estimating time spent in PA intensity, the GT3X+ underestimated sedentary time and overestimated PA intensity when compared to the SWAM. During the EBE analysis, the GT3X+ misclassified light intensity PA as moderate-to-vigorous PA 72% of the time. However, vector magnitude counts per minute (CPM) showed a strong correlation with criterion measures (r = 0.68 and 0.625 for SWAM and DLW and resting EE respectively). CONCLUSIONS: The currently available PAEE estimation equations from the manufacturer-provided software do not allow us to accurately measure PAEE or PA intensity with the wrist-worn GT3X+ in a population of older women. A high correlation of CPM from the GT3X+ with criterion measures indicates that the problem may lie in the available equations and cut points, not in the device itself.