Estimation of Caloric Expenditure Using Triaxial Accelerometers

Accelerometers-based activity monitors are commonly used to measure physical activity energy expenditure (PAEE). Newly designed wrist and hip-worn triaxial accelerometers claim to accurately predict PAEE across a wide range of activities.

Purpose To determine if the Nike FuelBand (NFB), Fitbit (FB) and ActiGraph GT3X+ (AG) accurately estimate PAEE in various activities.

Methods 21 healthy, college-aged adults wore a NFB on the right wrist, a FB on the left hip, and AG on the right hip, while performing 17 activities. AG data were analyzed using Freedson’s kcal regression equation. PAEE was measured using the Cosmed K4b2 (K4). Repeated measures ANOVAs were used to compare mean differences in PAEE (kcal/min). Paired sample t-tests with Bonferroni adjustments were used to locate significant differences.

Results For each device, the overall mean difference in PAEE was significantly different from the K4 (NFB, -0.45 + 2.8, FB, 0.48 + 2.27, AG, 0.64 + 2.59 kcal/ min, p = 0.01). The NFB significantly overestimated most walking activities (e.g., regular walking; K4, 3.1 + 0.2 vs. NFB, 4.6 + 0.2 kcal/min) and activities with arm movements (e.g., sweeping; K4, 3.0 + 0.8 vs. NFB, 4.7 + 0.4 kcal/min, p < 0.05). The NFB showed a trend towards overestimating sport activities (basketball, K4; 10.8 + 0.8 vs. NFB, 12.2 + 0.5 kcal/min) (racquetball, K4, 9.6 + 0.8 vs. 11.1 + 0.5 kcal/min). The FB and the AG significantly overestimated walking (FB, 5.4 + 0.3, AG, 5.8 + 0.4 kcal/min, p = 0.01) and underestimated PAEE of most activities with arm movements (e.g., Air Dyne, K4 5.6 + 0.2; Fitbit, 0.3 + 0.2; AG, 0.2 + 0.1 kcal/min, p < 0.05) (racquetball, K4, 9.6 + 0.8 kcal/minute vs. FB, 7.4 + 0.6 kcal/minute, vs. AG, 6.5 + 0.4 kcal/minute, p < 0.05).

Conclusions The NFB overestimated PAEE during most activities with arm movements and tended to overestimate sport activities, while the hip-worn AG and FB overestimated walking and underestimated activities with arm movements. Overall, the wrist-worn NFB had similar accuracy to the waist-worn tri-axial accelerometers; however, none of the devices were able to estimate PAEE across a wide range of activities.