Monitoring Steps and Heart Rate Using a Withings Smartwatch in Children and Adolescents With Cancer: Validation Study

Background: Consumer smartwatches offer potential for continuous monitoring of physical activity during childhood cancer treatment. However, their validity for measuring step count and heart rate in this population has not been established. Objective: This study examined the validity of a Withings smartwatch for monitoring step count and heart rate in children and adolescents undergoing cancer treatment. Methods: Children and adolescents aged 8 to 18 years under active cancer treatment wore a Withings smartwatch for 12 weeks. For validation purposes, children and adolescents simultaneously wore an ActiGraph wGT3X-BT accelerometer for 1 week under free-living conditions to assess step count agreement and were connected to a clinical heart rate monitor under resting conditions to assess heart rate agreement. Linear mixed models with participant ID as a random effect were used to estimate mean differences and derive 95% limits of agreement for repeated measurements. Results: A total of 14 children and adolescents participated in each of the validation analyses. The smartwatch recorded significantly fewer steps than the ActiGraph accelerometer for both steps per hour (mean difference 174 steps per hour; 95% limits of agreement=−283 to 633; P<.001) and steps per day (mean difference 3154 steps per day; 95% limits of agreement=−394 to 6702; P<.001), with increasing discrepancy at higher activity levels, suggesting proportional bias. No statistically significant difference in heart rate was observed between the smartwatch and the clinical heart rate monitor under resting conditions (mean difference −1.07 beats per minute; 95% limits of agreement=−17.55 to 15.41; P=.24). Conclusions: In children and adolescents undergoing cancer treatment, the Withings smartwatch showed minimal systematic bias for resting heart rate assessment, although individual variability was considerable. In contrast, step counts demonstrated systematic underestimation and wide limits of agreement compared with a research-grade accelerometer. The use of smartwatch-derived step counts for clinical benchmarking or goal setting in this population cannot currently be recommended.