Comparability of children’s sedentary time estimates derived from wrist worn GENEActiv and hip worn ActiGraph accelerometer thresholds

Objectives: to examine the comparability of children’s free-living sedentary time (ST) derived from raw 37 acceleration thresholds for wrist mounted GENEActiv accelerometer data, with ST estimated using the 38 waist mounted ActiGraph 100 count∙min-1 threshold. 39 Design: Secondary data analysis 40 Method: 108 10-11-year-old children (n=43 boys) from Liverpool, UK wore one ActiGraph GT3X+ 41 and one GENEActiv accelerometer on their right hip and left wrist, respectively for seven days. Signal 42 vector magnitude (SVM; mg) was calculated using the ENMO approach for GENEActiv data. ST was 43 estimated from hip-worn ActiGraph data, applying the widely used 100 count∙min-1 threshold. ROC 44 analysis using 10-fold hold-out cross-validation was conducted to establish a wrist-worn GENEActiv 45 threshold comparable to the hip ActiGraph 100 count∙min-1 threshold. GENEActiv data were also 46 classified using three empirical wrist thresholds and equivalence testing was completed. 47 Results: Analysis indicated that a GENEActiv SVM value of 51mg demonstrated fair to moderate 48 agreement (Kappa: 0.32-0.41) with the 100 count∙min-1 threshold. However, the generated and empirical 49 thresholds for GENEActiv devices were not significantly equivalent to ActiGraph 100 count∙min-1. 50 GENEActiv data classified using the 35.6 mg threshold intended for ActiGraph devices generated 51 significantly equivalent ST estimates as the ActiGraph 100 count∙min-1. 52 Conclusions: The newly generated and empirical GENEActiv wrist thresholds do not provide equivalent 53 estimates of ST to the ActiGraph 100 count∙min-1 approach. More investigation is required to assess the 54 validity of applying ActiGraph cutpoints to GENEActiv data. Future studies are needed to examine the 55 backward compatibility of ST data and to produce a robust method of classifying SVM-derived ST.