Calibration and validation of the ActiGraph GT3X+ in 2-3 year olds

Objective To calibrate and validate the ActiGraph GT3X+ to measure sedentary behaviour (SB) and physical activity (PA) in 2-3 year olds, using 5-second epochs; and to compare the predictive validity of the resulting cut-points with that of NHANES’, Trost’s, and Pate’s 15-second cut-points.

Design Cross-sectional study

Method Eighteen children (2.86 ± 0.60years) wore an ActiGraph GT3X+ during video-recorded semi-structured calibration activity sessions. Activity was coded following Children’s Activity Rating Scale (CARS). Receiver-Operating Characteristic analysis was used to derive Axis1 and Vector Magnitude (VM) cut-points for SB and MVPA at 5-second epochs. Agreement with CARS was assessed with Cohen’s kappa, Lin’s concordance, and Bland-Altman plots. Predictive validity of all cut-points was assessed in an independent sample of 20 children (2.99 ± 0.48years) video-recorded during free-play, using the same procedures as the calibration phase.

Results During calibration, VM cut-points (SB≤96.12 counts; MVPA≥361.94 counts) showed slightly better classification agreement with CARS than Axis1 cut-points (SB≤5 counts; MVPA≥165 counts), but the latter showed the lowest bias in estimated SB and MVPA time. In the validation sample, 5-second Axis1 cut-points showed the best predictive validity and lowest mean differences of all cut-points between predicted and observed SB (-2.31%), light PA (-24.40%), and total PA time (-0.95%). MVPA time was significantly overestimated by all cut-points (128.33-184.17%).

Conclusions Because MVPA was highly overestimated, using only the 5-second Axis1 SB cut-point to distinguish SB from total PA is advised. The high accuracy indicates that these cut-points are useful for epidemiological studies involving the SB and PA of 2-3 year olds.