Wearables for Health Monitoring: Body Composition Estimates of Commercial Smartwatch and Clinical Bioelectrical Impedance Device

Bryson Carrier¹Amanda Melvin¹Jacob R Outwin²Marni Wasserman³Adam Audet¹Katherine Soldes⁴Kenn Kozloff¹Adam Lepley^1*

• ¹University of Michigan, Ann Arbor, United States
• ²The George Washington University, Washington, United States
• ³Indiana University, Bloomington, United States
• ⁴Washington University in St Louis, St. Louis, United States

Abstract Introduction Body composition is a critical health measure. Accurate monitoring of body composition, such as body fat percentage (BF%) and skeletal muscle mass percentage (SM%), enables individuals to make informed decisions regarding nutrition, exercise, health status and management. Recent advancements have integrated bioelectrical impedance analysis (BIA) into wearable technology, presenting accessible options for tracking body composition measures. However, the validity of wearable BIA devices in comparison to criterion methods remains underexplored. Therefore, this study aimed to assess the validity of a wrist-worn consumer device and a clinical BIA device against the criterion measure of dual-energy X-ray absorptiometry (DXA). Methods This study included 108 physically active participants (56 females, 52 males). Participants underwent assessments using DXA, a wearable smartwatch BIA device (wearable-BIA; Samsung Galaxy Watch5), and a clinical standing hand-to-foot BIA analyzer (clinical-BIA; InBody 770). Measures of interest included BF% and SM%. Data were analyzed for accuracy using tests of error (mean absolute error [MAE], mean absolute percentage error [MAPE]), linearity (Pearson's r, Deming regression), agreement (Lin's CCC), and equivalence, complemented by Bland-Altman plots to visually represent bias. Results When assessing BF%, both the wearable-BIA (r = 0.93; CCC = 0.91) and clinical-BIA (r = 0.96; CCC = 0.86), demonstrated very strong correlations and agreement compared to DXA, with MAPEs of 14.3% and 21.1%, respectively. Female participants using the wearable-BIA device showed the greatest accuracy for BF% (CCC=0.91, MAPE=9.19%, equivalence supported). Bland-Altman analysis revealed proportional bias, particularly in individuals with higher BF%. Although correlations were considered strong for SM%, agreement was classified as weak (wearable-BIA: r = 0.92, CCC = 0.45; MAPE = 20.3%; clinical-BIA, r = 0.89; CCC = 0.25; MAPE = 36.1%). Discussion Both the wearable-and clinical-BIA device revealed mixed validity, demonstrating strong correlations for both BF% and SM%, and high levels of agreement and low error for BF%. Additionally, the wearable-BIA demonstrated acceptable accuracy for estimating BF% in females. However, wider limits of agreement and variability suggest limitations in validity, particularly for skeletal muscle mass and in individuals with higher body fat percentages. Continued validation efforts are recommended to enhance accuracy and consistency across diverse populations.