Abstract
Researchers using the GT3X+ accelerometer have the option of utilising a Low-Frequency Extension (LFE) to the data filtering algorithm to increase sensitivity to low intensity movement. Recent studies show that the LFE results in significantly higher estimates of physical activity in free-living populations although no study has yet provided comparisons to a criterion measure.
PURPOSE: To determine the effects of the LFE and sample frequency on the prediction accuracy of energy expenditure (EE) during structured light and moderate intensity activity.
METHODS: Eighteen active adults (14 males, age 28 ± 8 yr, weight 76.1 ± 12.4 kg) completed a continuous incremental protocol on a treadmill comprising 5 min walking at 4 km/hr, 3 min walking at 5 and 6 km/hr and then 3 min running at 10 and 12 km/hr. Respiratory variables were measured throughout via indirect calorimetry (IC) and EE for each stage was calculated using the Weir equation. Accelerometer data was collected using two GT3X+ devices, one of which was set to a high sample frequency (100 Hz) and the other to a low sample frequency (30Hz). The data was downloaded using both the normal filter (NF) and the LFE for each device and EE estimated in all four conditions (30NF, 30LFE, 100NF, 100LFE) using the VM3 equation. Differences between predicted and the IC criterion method of EE measurement were assessed using paired t-tests. Agreement between methods was established by Pearson’s correlation coefficient and Bland and Altman analysis following log transformation of the data.
RESULTS: Predictions of EE were significantly correlated with IC (R=0.92, R=0.92, R=0.91, R=0.91 all P<0.001 for 30NF, 30LFE, 100NF and 100LFE respectively). Mean EE was significantly lower with 30NF (8.0 ± 3.5 kcal/min) compared to IC (8.5 ± 4.8 kcal/min, P=0.02). Mean EE in the other conditions was not different to IC (P>0.05). Mean bias and error ratios were 1.02 x/÷ 1.50, 0.99 x/÷ 1.50, 1.00 x/÷ 1.54 and 0.97 x/÷ 1.53 for 30NF, 30LFE, 100NF and 100LFE, respectively.
CONCLUSIONS: Utilisation of the LFE at lower sampling frequencies appears to prevent statistically significant differences in predicted EE during structured light and moderate intensity activity. However, a large error ratio limits the capacity of accelerometers to predict EE at the individual level regardless of processing decisions.
PURPOSE: To determine the effects of the LFE and sample frequency on the prediction accuracy of energy expenditure (EE) during structured light and moderate intensity activity.
METHODS: Eighteen active adults (14 males, age 28 ± 8 yr, weight 76.1 ± 12.4 kg) completed a continuous incremental protocol on a treadmill comprising 5 min walking at 4 km/hr, 3 min walking at 5 and 6 km/hr and then 3 min running at 10 and 12 km/hr. Respiratory variables were measured throughout via indirect calorimetry (IC) and EE for each stage was calculated using the Weir equation. Accelerometer data was collected using two GT3X+ devices, one of which was set to a high sample frequency (100 Hz) and the other to a low sample frequency (30Hz). The data was downloaded using both the normal filter (NF) and the LFE for each device and EE estimated in all four conditions (30NF, 30LFE, 100NF, 100LFE) using the VM3 equation. Differences between predicted and the IC criterion method of EE measurement were assessed using paired t-tests. Agreement between methods was established by Pearson’s correlation coefficient and Bland and Altman analysis following log transformation of the data.
RESULTS: Predictions of EE were significantly correlated with IC (R=0.92, R=0.92, R=0.91, R=0.91 all P<0.001 for 30NF, 30LFE, 100NF and 100LFE respectively). Mean EE was significantly lower with 30NF (8.0 ± 3.5 kcal/min) compared to IC (8.5 ± 4.8 kcal/min, P=0.02). Mean EE in the other conditions was not different to IC (P>0.05). Mean bias and error ratios were 1.02 x/÷ 1.50, 0.99 x/÷ 1.50, 1.00 x/÷ 1.54 and 0.97 x/÷ 1.53 for 30NF, 30LFE, 100NF and 100LFE, respectively.
CONCLUSIONS: Utilisation of the LFE at lower sampling frequencies appears to prevent statistically significant differences in predicted EE during structured light and moderate intensity activity. However, a large error ratio limits the capacity of accelerometers to predict EE at the individual level regardless of processing decisions.
| Original language | English |
|---|---|
| Pages (from-to) | 492 |
| Number of pages | 1 |
| Journal | Medicine and Science in Sports and Exercise |
| Volume | 46 |
| Issue number | 5S |
| DOIs | |
| Publication status | Published - May 2014 |