Evaluating the Accuracy of Sex-Dependent Algorithms in Commercial Wrist-Worn Wearables for Exercise Energy Expenditure

Authors

Olivia R. Perez, University of Nevada, Las VegasFollow
Dustin W. Davis, University of Nevada, Las VegasFollow
Michael W. H. Wong, University of Nevada, Las VegasFollow
James W. Navalta FACSM, University of Nevada, Las VegasFollow

Abstract

Commercial wrist-worn wearable devices are popular among sports and health enthusiasts. Energy expenditure (EE) is a valuable measurement returned from these devices, but the accuracy is variable. A gap exists in sport and exercise science concerning sex-specific accuracy and input algorithms. Purpose: This study aimed to evaluate the sex-specific validity of five commercial wrist-worn wearable devices in estimating exercise EE, with indirect calorimetry as the criterion. Methods: 97 adult participants were outfitted with two devices (n = 20 per device), one set to the "female" sex option on the right wrist and the other set to the "male" sex option on the left. Devices included the Garmin Forerunner 265S, Fitbit Charge 6, Polar Vantage M3, Samsung Galaxy Watch7, and Apple Watch Ultra 2. Participants were fitted for the ParvoMedics TrueOne 2400 facemask to measure EE during maximal exercise testing (V̇O2max). Validity thresholds included mean absolute percentage error (MAPE) of ≤10% and Lin’s concordance correlation (CCC) of ≥0.9. Results: The Samsung Galaxy Watch7 met MAPE and CCC thresholds for both sex-aligned (9.42%, 0.96) and sex-not aligned (9.66%, 0.96) conditions with all participants included and when gender-diverse individuals were subtracted from the total for sex-aligned (9.42%, 0.96) and sex-not aligned (9.66%,0.96) conditions. The other wrist worn experimental devices did not meet thresholds with MAPE ≤10% and/or CCC ≥0.9 with sex-aligned (Garmin = 14.65%, 0.89; Fitbit = 13.97%, 0.94; Polar = 21.10%, 0.85, Apple = 25.95%, 0.82; 27.43%, 0.87) and sex-not aligned (Garmin = 29.01%, 0.61; Fitbit = 13.16%, 0.93; Polar = 23.89%, 0.81; Apple = 21.35%, 0.85; 29.59%, 0.86) conditions with all participants included, and when gender-diverse individuals were subtracted for sex-aligned (Garmin = 13.25%, 0.86; Fitbit = 13.13%, 0.94; Polar = 21.65%, 0.84; Apple = 25.85%, 0.81; 27.43%, 0.87) and sex-not aligned (Garmin = 26.14%, 0.48; Fitbit = 14.21%, 0.93; Polar = 23.26%, 0.81; Apple = 21.32%, 0.84; 29.59%, 0.86) conditions. Discussion: Only the Samsung device met accuracy thresholds. These findings suggest the built-in requirement to input sex information may not contribute to EE estimation inaccuracy in the majority (4/5) of device algorithms. Manufacturers are encouraged to adopt inclusive practices while improving EE estimation algorithms.

Recommended Citation

Perez, Olivia R.; Davis, Dustin W.; Wong, Michael W. H.; and Navalta, James W. FACSM (2025) "Evaluating the Accuracy of Sex-Dependent Algorithms in Commercial Wrist-Worn Wearables for Exercise Energy Expenditure," International Journal of Exercise Science: Conference Proceedings: Vol. 14: Iss. 5, Article 102.
Available at: https://digitalcommons.wku.edu/ijesab/vol14/iss5/102