Preliminary Investigations of the Validity and Interinstrument Reliability for Classification of Accelerometer Physical Activity Cut-Points Against Indirect Caliometry in Healthy Adults

Wrist and waist-worn accelerometers are frequently used in human participant studies to quantify physical activity; however, cut-points to process raw accelerometry data into activity intensity categories require validation. This study aimed to investigate wrist and waist-based cut-points for sedentary behavior, light, moderate and vigorous intensity physical activity in healthy adults for three research-grade wearable sensors. Healthy adults (n=30) completed a six-phase treadmill protocol (0.0, 2.4, 5.6, 6.4, 7.2 and 8.0 km/h) while wearing GENEActiv, ActiGraph wGT3X-BT and Verisense sensors simultaneously on the waist and nondominant wrist. Metabolic equivalent (MET) levels were assessed by energy expenditure derived indirect calorimetry. Correlations between accelerometer values and METs were generated along with Receiver operating characteristics (ROC) curves to define accelerometer values maximizing sensitivity and specificity for classification of sedentary behavior ( $\leq 1.5$ METs) and moderate to vigorous activity ( $> 6$ METs). For all devices, correlations ranged from $\mathrm{r}=0.91-0.93$ for wrist placement and $\mathrm{r}=0.96-0.97$ for waist placements, demonstrating excellent correlations against METs. The area under the ROC curves were investigated to discriminate sedentary activity, with values ranging from 0.844-0.896 for wrist placements and 0.699-0.923 for waist placements, and moderate to vigorous activity ranging from 0.991-0.993 for wrist placements and 0.989-0.993 for waist placements (95% CI, $\mathrm{P} < 0.001)$ . Further analyses are required before novel cut-points can be generated for each wearable device; however, the initial findings and ROC curves show promising correlations.