Effects of filter choice in GT3X accelerometer assessments of free-living activity

ActiGraph accelerometers are widely used devices to objectively assess physical activity. The GT3X version has two filter options to be selected before data assessment (normal and low-frequency extension filter option). It is not clear whether the resulting physical activity levels differ depending on the choice of the filter. The aims were to compare GT3X data collected using the different filter options during free-living activities and to establish correction factors if the results were not comparable.

Sixty-five participants of the population-based SAPALDIA-cohort (50.8% women, age range = 40-80 yr) wore two GT3X accelerometers with different filter selections simultaneously during 8 d. Spearman correlations, Wilcoxon rank sum tests, McNemar tests, scatter plots, and Bland-Altman plots were used to compare the data. Correction factors were established using linear regression models.

Although Spearman correlations were high (r ≥ 0.93), there were significant differences in minutes per day between filter options for nonwearing time and time spent in sedentary, light, and moderate-to-vigorous physical activity (all P < 0.001), with more remarkable differences in the lower range of activity (sedentary and light activities). Mean counts per minute and steps per day were significantly higher using the low-frequency extension filter (P < 0.001). Most differences could be resolved using the correction factors.

The observed differences are especially important when research is focusing on sedentary and light activities. In future studies, it is important to carefully evaluate the suitable filter option and to specify the filter choice in publications. The correction factors can be used to make data assessed using the low-frequency extension filter comparable to data assessed using the normal filter option.