Predictive Validity of ActiGraph Equations for Energy Expenditure in Older Adults in Free-Living Environments

Purpose: To compare the Freedson, Crouter and Santos-Lozano (vertical axis [VT] and vector-magnitude [VM]) equations for ActiGraph GT3X+ (AG) with indirect calorimetry for predicting energy expenditure (EE) in older adults in free-living environments.

Methods: Forty older adults (age=77.36 yr (SD+8.13; range 66 to 99 years), BMI=25.96 kg∙m-2 (SD=3.67) completed a semi-structured protocol with sitting, lying, walking and 4 self-selected household/gardening activities in their residence, while wearing an AG and a portable indirect calorimeter (Metamax 3B). EE for each activity, total physical-activity-related EE (PAEE) and total sedentary-behaviour-related EE (SBEE) predicted by the 4 equations, were compared with measured EE in terms of standard error of the estimation (SEE), mean bias, and agreement in classification of physical activity levels (PAL).

Results: Overall, the lowest SEE was found for the Santos-Lozano VM equation (0.72 METs). For PAEE, this equation had the lowest mean bias (-6.77 MET/h; 95%LoA:-62.94, 49.40), while for SBEE, the Crouter equation showed the lowest mean bias (11.04 MET/h; 95%LoA:-33.49, 55.58). All four equations had statistically significant agreement for categorising activities into PAL, with the highest overall agreement found for the Freedson equation (74.9%).

Conclusions: As none of the equations was clearly superior to the others, selection of an equation should be based on the research question. When measuring total daily EE per day, equations with more accurate estimations for sedentary behaviour may be preferred, as these behaviours account for at least 60-70% of daily activity in older adults