Individual Calibration of a Multiinput Wearable Device Improves Exercise Energy Expenditure Estimation During Intermittent High-Intensity but Not Resistance Exercise in an Athletic Female Population

This is the authors' abstract. We don't add key points for this paper.

This study aimed to determine the accuracy of a multiinput wearable device (Actiheart 5) and commercially available accelerometer (GENEActiv) in the estimation of exercise energy expenditure (EEE) during intermittent high-intensity and resistance exercise sessions among 31 trained females. Data were collected over two testing sessions involving (a) measurement of resting metabolic rate, physique and a graded exercise test on a treadmill to enable individualized calibration of the wearables, and (b) A modified Bath University Rugby Union Shuttle Test followed by a resistance exercise session. GENEActiv and Actiheart EEE-derived measures, using both generic and individually calibrated algorithms, were compared against (a) indirect calorimetry for Bath University Rugby Union Shuttle Test and (b) indirect calorimetry plus the estimated anaerobic energy contribution based on blood lactate responses for resistance exercise. Equivalence was evaluated using predefined bounds of ±10% mean indirect calorimetry EEE. The use of individually-calibrated Actiheart algorithms improved estimation of EEE during Bath University Rugby Union Shuttle Test (mean bias 4.5 ± 60.4 kcal) compared with individually-calibrated GENEActiv algorithms (mean bias 69.5 ± 86.6 kcal), and equivalence with indirect calorimetry was achieved, although substantial individual error remained (95% limits of agreement -113.9 to 122.8 kcal and intraclass correlation coefficients, .62 [.34 to .80]). Individualized Actiheart algorithms provided equivalent group-level estimates of EEE during intermittent high-intensity exercise, whereas neither Actiheart nor GENEActiv demonstrated acceptable validity or equivalence against the reference method for resistance training, suggesting Metabolic Equivalents of Task may be a more practical method of estimating EEE.