**Background:** Childhood obesity affects 340 million children and adolescents globally and is a major risk factor for future adverse health conditions. Multicomponent parent-child interventions are considered first-line treatment, with clinically significant weight reduction defined as a BMI z-score drop of 0.2 units. Digital health interventions offer potential solutions but face challenges in engagement and adherence. The ENDORSE platform was developed to integrate m-health technologies, artificial intelligence, and serious games into a coordinated software ecosystem connecting healthcare professionals, children, and parents.
**Methods:** This 12-week feasibility pilot study was conducted from March 2021 to May 2022 at the Obesity outpatient Clinic of the First Department of Pediatrics, National and Kapodistrian University of Athens at "Aghia Sophia" Children's Hospital. Fifty overweight and obese children (mean age 10.5 years, 52% girls, 58% pubertal, median baseline BMI z-score 2.85) and their mothers were enrolled. The study included three consecutive groups: a pre-pilot group (n=20, 60% boys, mean age 11 years, mean BMI z-score 2.85), a pilot active control group (n=15, mean BMI z-score 2.71), and a pilot intervention group (n=15, mean BMI z-score 2.89). The pre-pilot group tested an early platform version; the control group received weekly personalized messages from the clinical team without the serious game or recommendation system; the intervention group received automated AI-generated messages with the serious game. Adherence was measured using objective usage metrics for each module (parental app, activity tracker, serious game). Anthropometrics, dietary habits, physical activity, screen time, and sleep were assessed pre- and post-intervention.
**Key Results:** Overall, a clinically and statistically significant BMI z-score reduction was achieved across all participants (mean BMI z-score reduction −0.21 ± 0.26, p < 0.001). In the pre-pilot group, mean BMI z-score decreased by −0.24 (p = 0.001). In the pilot phase, the intervention group showed a −0.16 reduction (p = 0.002) and the control group −0.21 (borderline significant, p = 0.068), with no statistically significant difference between groups. A statistically significant correlation was found between activity tracker average usage and BMI z-score improvement (−0.355, p = 0.017). The overall adherence score also correlated significantly with BMI z-score change (−0.299, p = 0.046). The attrition rate was low: 10% in pre-pilot, 13.33% in control, and 6.66% in intervention groups. Most participants (75%) achieved medium to high overall adherence levels. Health behavior improvements across all participants (n=45) included: increased fruit intake (mean +0.62 servings/day, p < 0.001), increased vegetable intake (mean +0.80 servings/day, p < 0.001), decreased fast food consumption (mean −0.22 servings/week, p = 0.042), increased physical activity (mean +24.33 min/day, p < 0.001), reduced weekday screen time (mean −0.47 h/day, p = 0.005), and increased sleep duration (mean +0.54 h/day, p = 0.005). No adverse psychological events were reported.
**Clinical Implications:** This feasibility study demonstrates that an integrated digital platform combining m-health, AI, and gamification can support clinically meaningful weight reduction in children with obesity. The significant correlation between activity tracker adherence and BMI z-score improvement supports the inclusion of wearables in pediatric weight management programs. The low attrition rate and positive acceptability ratings suggest the platform is feasible for clinical implementation. However, the low engagement with the serious game highlights the challenge of maintaining user interest in gamified interventions. The study was limited by its nonrandomized design, small sample size, and conduct during the COVID-19 pandemic, which may have influenced lifestyle behaviors. Larger-scale randomized controlled trials are needed to establish effectiveness and sustainability.