Impact of cut-point methods on classification of physical activity and sedentary behaviour of toddlers.

Background: Classification of physical activity (PA) depends on the cut-point method used to allocate PA counts from accelerometer measurements. This study investigates how three validated cut-point methods affect the time spent in various levels of PA and sedentary behaviour (SB), and how they impact toddlers estimated adherence to PA guidelines.

Methods: PA was assessed using an ActiGraph wGT3X-BT accelerometer in a cohort of 653 two-year-old children participating in the Toddler Milk Intervention study. Children wearing the ActiGraph for at least four days, with a minimum of six hours wear-time per day, were included. Time spent in SB and different activity levels were estimated according to three cut-point methods and were standardized to individual mean wear-time. We used one cut-point method based on the vertical axis (VA) (Trost VA), with an epoch length of 15 s and two cut-point methods based on either the VA (Costa VA) or on the vector magnitude (VM) (Costa VM) with an epoch length of five seconds. Estimates of SB and PA for each method were compared with repeated measures ANOVA.

Results: The time toddlers spent in PA was significantly different depending on the cut-point methods. Costa VM classified on average 62 min (95% CI 61, 64] more per day as SB and 57 min (95% CI -58, -56] less per day as LPA compared to Trost VA (both p < 0.0001). For MVPA, the mean difference between Costa VA and Trost VA was 6.8 min (95% CI -7, -6; p < 0.0001). Concurrently, the proportion of children meeting the WHO recommendation of 180 min of total PA differed between cut-point methods, with 86% according to Costa VM and 97% according to Trost VA.

Conclusions: The time toddlers engage in different intensities of PA is significantly determined by the selection of cut-point method. Notably, the use of a different cut-point method leads up to a 10% difference in the estimated time spent in LPA and SB, but only a 1% difference of moderate-vigorous PA. These differences change the estimated adherence to recommendations. Future research is needed to standardize the data processing methods for better comparability between studies analysing toddlers' PA.

Registry: ClinicalTrials.gov, TRN: NCT02907502, Registration Date: 31 August 2016.