Agreement Between TDK Silmee W22 and ActiGraph wGT3X-BT for Estimating Daily Step Counts and Moderate to Vigorous Physical Activity in Free-Living Adults: Comparative Study.

Background: Wearable Internet of Things (IoT) devices are powerful tools for remotely collecting intensive longitudinal data. The TDK Silmee W22, a wristband-type wearable IoT device with a built-in 3-axis acceleration sensor, provides minute-by-minute physical activity data such as estimated metabolic equivalents (METs) and step counts. These measurements can be aggregated to daily estimates; however, their accuracies have not been fully explored in adults under free-living settings. Objective: This study aims to assess the agreement between the TDK Silmee W22 and the research-grade activity monitor, ActiGraph wGT3X-BT, in estimating daily steps and time spent in moderate to vigorous physical activity (MVPA ≥3 METs) in adults under free-living settings. Methods: A convenience sample of young to older adults was recruited from communities across several prefectures in Japan. Participants concurrently wore a TDK Silmee W22 on their nondominant wrist and an ActiGraph wGT3X-BT on the left side of the waist during waking hours for 7 consecutive days. Data were aggregated to daily steps and time spent in MVPA (≥1952 vertical axis counts/minute for ActiGraph) for each participant. A valid day was defined as having ≥ 10 hours of accumulated ActiGraph wear time with ≥100 and <50,000 accumulated steps from both devices. Each valid day was classified as either an active day (≥10,000 steps/day or ≥21.4 minutes MVPA per day) or an inactive day. Bland-Altman plots combined with multilevel analysis and κ statistics were used to assess the agreement between physical activity estimates from the devices. Results: A total of 129 participants (n=66, 51.2% women) aged 23-89 years provided the final dataset of 884 observations (5 to 7 daily observations/participant). The TDK Silmee W22 estimated an overall mean of 6369 (SE 242) steps/day and 40.3 (SE 1.9) minutes/day spent in MVPA. Although Bland-Altman plots suggested no obvious proportional bias, fixed biases were observed; the TDK Silmee W22 estimated -1203 steps/day (95% limits of agreement [LoA] -4202 to 1796) and +5 minutes/day (LoA-23 to 34) spent in MVPA compared with those estimated by the ActiGraph wGT3X-BT. TDK Silmee W22 and ActiGraph wGT3X-BT, respectively, classified 14% and 23.1% as active by daily step counts (κ=0.65, 95% CI 0.59-0.72), and 70.4% and 60.9% as active by daily time spent in MVPA (κ=0.64, 95% CI 0.59-0.69), both indicating moderate agreement. Conclusions: TDK Silmee W22 underestimated step counts and overestimated time spent in MVPA compared with the research-grade ActiGraph wGT3X-BT, which may lead to misclassification of active and inactive days. Caution is warranted when using TDK Silmee W22 data over relatively short periods, as discrepancies-particularly when compared with research-grade monitors-may affect feedback or goal setting.