JMIR Publications

ABSTRACT

Background:

Fitness trackers have become important instruments for health monitoring and long-term management of chronic conditions. In individuals with ulcerative colitis (UC), initial research indicates that these devices may effectively capture health-related parameters pertinent to disease monitoring and management. However, the measurement validity of widely available commercial devices, such as the Garmin Vívosmart 4 (GV4), in clinical populations remains uncertain, especially when compared to medically certified sensors.

Objective:

This study aimed to validate GV4 step count measurements by comparing them with those obtained from a clinically validated motion sensor, SENS motion®, in patients with UC under real-world conditions as part of the FilgoColitis study.

Methods:

Data were obtained from the FilgoColitis study, a prospective, observational, multicenter study conducted in Germany involving patients with UC who had recently initiated filgotinib therapy. Participants concurrently utilized both devices. For descriptive analyses, all 15-minute intervals with data from both devices were considered, regardless of whether any steps were recorded. In contrast, inferential statistical analyses included only those intervals where at least one device recorded one or more steps. Step counts were collected over several days following each study visit and aggregated into 15-minute, hourly, daily, and patient-level mean intervals. The agreement between the devices was assessed using the concordance correlation coefficient (CCC), Bland-Altman analysis, and two one-sided tests (TOST) for equivalence testing.

Results:

Data from 40 patients were analyzed, yielding 37,517 paired 15‑minute intervals of step count measurements. The CCC, based on active intervals, increased with higher aggregation levels: 0.87 for 15‑minute intervals, 0.93 for hourly intervals, 0.96 for daily intervals, and 0.93 for patient‑level mean step counts. Step count distributions were comparable across devices, with the highest agreement at the daily and patient‑level aggregations. Bland-Altman analysis showed a mean difference of 1.7 steps between the devices, with limits of agreement from -40.1 to +43.6 steps. TOST demonstrated the statistical equivalence of step counts within a margin of ±8 steps (90% confidence interval for the mean difference: -3.96 to 7.41). Discrepancies were more pronounced at shorter aggregation intervals.

Conclusions:

Commercial fitness trackers, such as the GV4, can serve as valid and widely accessible tools for capturing step-count-based proxies of physical activity in patients with UC. A high degree of agreement with a medically certified motion sensor was observed at aggregated (daily and patient‑level) intervals, supporting the use of these devices in clinical research and digital health applications. However, short-interval measurements should be interpreted with caution, as technical variability and user behavior may introduce noise. Future studies should investigate the integration of wearables capturing additional health parameters into digital health interventions and evaluate their long‑term use for telemedical monitoring and novel endpoint assessments in clinical trials. Clinical Trial: German Clinical Trials Register DRKS00027327; https://www.drks.de/DRKS00027327