Pilot Sequential Multiple Assignment Randomised Trial of LvL UP: an Adaptive Holistic mHealth Coaching Intervention Integrating Physical Activity, Diet, and Mental Health

Background Mobile Health (mHealth) interventions are promising for addressing due burden of noncommunicable diseases and common mental disorders but often focus on single domains and lack adaptability. LvL UP contributes novel evidence by operationalising a holistic mHealth coaching intervention that integrates physical activity, diet, and emotional regulation, with adaptive human support. Methods The eight-week trial (April–July 2024) recruited adults in Singapore aged 21–59 at risk of chronic conditions. Participants were randomised 2:1 to the intervention (LvL UP app with a peer supporter–LvL UP Buddy) or comparison (control app with educational resources). After four weeks, non-responders (completed < 6 digital coaching sessions or rated session usefulness < 4/5) were re-randomised 1:1 to continue or receive three additional motivational interviewing (MI)-informed sessions with a human coach; responders remained on their original allocation. Primary outcomes included feasibility indicators: recruitment, LvL UP Buddy enrolment, non-responder rate, retention, data completion, and engagement. Secondary outcomes measured changes from baseline to eight weeks in mental well-being, psychological distress, physical activity, sleep duration, and fruit and vegetable intake. Six progression criteria were prespecified to guide advancement to a full SMART trial. Results Of the 458 individuals screened, 394 were eligible, and 123 were enrolled (82 interventions; 41 controls). Most intervention participants (95.1%) were paired with a LvL UP Buddy. Thirty-eight participants (46.3%) were non-responders; of those receiving MI sessions, 52.6% (10/19) completed all three. Eight-week retention was high (91.5% intervention; 92.7% control), with 12.2% missing data. Positive trends were observed in mental well-being (2.12, 95% CI [-0.58, 4.82]), psychological distress (-0.94 [-2.08, 0.20]), and sleep duration (0.49 hours/week [0.17, 0.82]). The study met five of six prespecified progression criteria: recruiting ≥ 60 participants within six weeks, achieving ≥ 75% retention, maintaining ≤ 20% missing data, obtaining a 40–60% non-responder rate, and showing a positive change in ≥ 1 health-related outcome. The digital coaching session adherence fell below the target (39.5% vs 70%). Conclusions LvL UP was feasible for delivery and evaluation using a SMART design. The results provide strong operational guidance and a solid foundation for the refinement and implementation of a fully powered trial.