Efficacy of an app-based multimodal lifestyle intervention on body weight in persons with obesity: results from a randomized controlled trial

The study addresses whether a fully digital, app-based multimodal lifestyle intervention can induce clinically meaningful weight loss in adults with obesity. In Germany, over half of adults are overweight and about a quarter have obesity, which increases risk for multiple chronic diseases. Multimodal lifestyle interventions (diet, physical activity, behavior modification) are guideline-recommended first-line treatments, and smartphone apps are increasingly used but evidence for their effectiveness, especially for comprehensive multimodal programs, is limited. Germany’s Digital Health Applications (DiHA) framework distinguishes regulated, reimbursable medical apps from general lifestyle apps. The study’s primary objective was to evaluate the effect of a certified DiHA delivering a 12-week multimodal weight loss program on body weight after 12 weeks. Secondary objectives included body weight after 24 weeks (maintenance), quality of life, and app usability and acceptance.

Prior reviews indicate app-based interventions can improve eating behavior and BMI, though many commercial apps provide single functions rather than comprehensive programs. Evidence for app-only multimodal interventions remains heterogeneous, with modest average weight losses over months. The German DiHA framework uniquely certifies and reimburses medical apps upon proof of positive healthcare effects, distinguishing them from general lifestyle apps. Comparative trials combining apps with other tools (e.g., activity trackers, counseling) report modest short-term weight loss (~2 kg at 3 months). Longer interventions (e.g., 12 months) have shown larger losses (~8% in an RCT of another app-based multimodal program). Engagement and self-monitoring are consistently linked to better outcomes, and user acceptance and usability can influence effectiveness.

Design: Single-centre randomized controlled trial at the Institute for Nutritional Medicine, Technical University of Munich; ethics approval (45/22 S-NP); registered DRKS00025291; written informed consent obtained. Participants: Adults 18–70 years, BMI 30.0–40.0 kg/m², without severe diseases (e.g., diagnosed diabetes, cardiovascular disease, cancer), owning a smartphone. Recruitment via social media, public transport advertising, and clinic flyers (Mar 17–Aug 9, 2022). Randomization: 1:1 (ADHOC vs EXPECT) using Stat Trek Random Number Generator; allocation concealed by database assignment after data entry; study team blinded to randomization list. Interventions: DiHA “Oviva Direkt für Adipositas” app (iOS/Android) providing a 12-week multimodal program per German obesity guidelines: self-management (goal setting), self-monitoring (entries for nutrition, physical activity, weight; feedback with trajectories, reminders, notifications, dietary interpretations), and education (weekly modules via text/audio/video with knowledge checks). Initial coach safety call in week 1 and chat access for questions; otherwise no human counseling. - ADHOC: Immediate app use for 12 weeks (baseline to week 12), app available during 12-week follow-up (weeks 13–24). - EXPECT: Wait-list for first 12 weeks (no intervention), then app use for weeks 13–24. Measurements: Three study visits: baseline (V1), 12 weeks (V2), 24 weeks (V3). Anthropometrics: height (SECA 214), weight and body composition (Tanita BC-418MA) measured in light clothing, no shoes, empty bladder; BMI calculated. Quality of life: EQ-5D-5L (index using German value set) and EQ VAS (0–100) at V1, V2, V3. Usability/Acceptance: After 12 weeks of app use, Technology Acceptance Model 3 (TAM3) perceived usefulness and ease of use (8 items total) to compute TAM score; System Usability Scale (SUS). Scores categorized into tertiles (low/middle/high). App usage: Minutes per week captured by the app; summed hours over 12 weeks used as intensity proxy. Outcomes: Primary—weight change at 12 weeks. Secondary—weight at 24 weeks, body composition changes, quality of life, usability/acceptance, and association of app usage with weight change. Statistical analysis: Primary population—completers with 12-week weight. Descriptives (means ± SD, frequencies). Group comparisons at baseline: t-test or Mann–Whitney U; categorical: chi-squared or Fisher’s exact. Normality: Shapiro–Wilk; homogeneity: Levene’s test. Between-group changes at 12 weeks analyzed via multiple linear regression adjusted for gender, age, and baseline weight; standardized coefficients also reported. Within-group changes at 24 weeks described (no between-group comparison). Associations between usage (sum hours) and weight change via adjusted linear regression. TAM/SUS group comparisons: Kruskal–Wallis; pairwise as described. Significance p < 0.05. Software: RStudio 4.1.0; EQ-5D via eq5d package. Sample size: Assuming 20% dropout, 156 participants to detect a 3% weight loss difference (SD 6%) with 80% power, alpha 0.05; post-hoc power 99.9%. Missing data: Last observation carried forward (LOCF). For 12 weeks, last app-tracked weight used if clinic weight missing; for 24 weeks, last clinic and/or app-tracked self-reported weight used.

- Participants (N=168): 64.3% women, mean age 46.8 ± 11.0 years, mean BMI 34.2 ± 2.8 kg/m²; baseline characteristics balanced between groups. - Adherence: 12-week visit completed by 82.7% (139/168): ADHOC 75.0% (63/84), EXPECT 90.5% (76/84). 24-week completion: ADHOC 70.2% (59/84), EXPECT 64.3% (54/84). - Primary outcome (12 weeks, completers): ADHOC lost −3.2 ± 3.2 kg (−3.2% ± 3.0%); EXPECT lost −0.4 ± 2.6 kg (−0.3% ± 2.6%). Adjusted between-group difference: −2.9 kg (95% CI −3.8 to −1.9), p < 0.001; percentage difference −2.9% (95% CI −3.8 to −1.9), p < 0.001; standardized β ≈ 0.45–0.46. - Body composition (12 weeks, completers): Fat mass change: ADHOC −2.0 ± 2.7 kg vs EXPECT −0.1 ± 2.2 kg (p < 0.001); fat mass %: ADHOC −0.8 ± 1.9 vs EXPECT +0.04 ± 1.7 (p = 0.008); fat-free mass: ADHOC −1.3 ± 1.7 kg vs EXPECT −0.3 ± 2.0 kg (p = 0.002). - Quality of life (12 weeks, completers): No significant between-group differences (EQ-5D-5L index mean change: −0.02 ± 0.1 vs −0.001 ± 0.1; p = 0.32). EQ VAS favored ADHOC numerically (+4.7 ± 14.8 vs +0.6 ± 11.0; p = 0.06). - LOCF analysis (12 weeks, ITT N=168): ADHOC −2.6 ± 3.1 kg (−2.6% ± 3.0%) vs EXPECT −0.3 ± 2.5 kg (−0.3% ± 2.5%); adjusted difference −2.3 kg (95% CI −3.1 to −1.4), p < 0.001; percentage difference −2.3% (95% CI −3.2 to −1.5), p < 0.001; effect size ~0.38–0.39. - 24 weeks (completers within-group): ADHOC maintained weight loss: −3.3 ± 4.8 kg (−3.1% ± 4.5%) from baseline; EXPECT (who used the app in weeks 13–24) −2.1 ± 3.3 kg (−2.0% ± 3.3%). - Sensitivity analyses: Greater 12-week weight loss in ADHOC vs EXPECT across subgroups by gender, age (<50 vs ≥50), and BMI (30.0–34.9 vs 35.0–40.0), all p < 0.001. Men lost more than women; ≥50 years lost more than <50; higher baseline BMI lost more than lower BMI within ADHOC. - App usage and acceptance: In ADHOC, higher TAM score associated with greater weight loss (β = −1.7 kg per TAM unit [95% CI −2.7; −0.7], p = 0.002). SUS score not associated (p = 0.11). Weight loss differed across TAM tertiles in ADHOC (p = 0.003): middle/high > low. EXPECT showed a TAM-tertile difference (p = 0.03), but pairwise not significant. - App engagement: Minutes spent on app decreased over 12 weeks in both groups. Greater cumulative app use associated with greater weight loss: ADHOC β = −0.10 kg per hour [95% CI −0.18; −0.01], p = 0.03; EXPECT β = −0.12 [95% CI −0.22; −0.02], p = 0.02.

The trial demonstrates that a certified, fully digital multimodal app can achieve statistically significant and clinically meaningful short-term weight loss in adults with obesity compared with a wait-list control. The findings address the central question of whether an app-only, guideline-based intervention yields measurable benefits: the app produced greater reductions in body weight and fat mass at 12 weeks and maintained weight loss at 24 weeks among immediate users. Quality of life improvements did not differ significantly between groups at 12 weeks, though trends favored the intervention. Engagement and acceptance were key moderators: higher technology acceptance (TAM) and greater time spent using the app were linked to larger weight losses, underscoring the importance of user engagement and behavioral self-monitoring. The magnitude of weight loss aligns with other app-based or mobile health interventions over similar timeframes and is modest compared to longer, more intensive programs, suggesting that extending duration may enhance outcomes. Given the scalability of DiHA in the German healthcare system, such app-based programs can serve as accessible initial interventions for a broad population, with the potential to triage individuals needing more intensive care. The multimodal design, integrating self-management, self-monitoring, and education with behavior change techniques, likely contributes to effectiveness and aligns with evidence that multi-component, behaviorally informed interventions outperform single-feature apps.

A fully digital, multimodal app-based lifestyle program produced moderate, clinically relevant weight loss after 12 weeks that was maintained at 24 weeks among immediate users. Effectiveness was influenced by user acceptance and engagement. These findings support the use of certified, reimbursable digital health applications as initial weight management tools in routine care for adults with obesity. Future research should evaluate longer intervention durations, strategies to sustain engagement, identification of likely responders using acceptance/usability measures, integration with stepped-care models for higher-risk patients, and generalizability across diverse populations and settings.

The study was single-centre and conducted in the Munich region, limiting generalizability and representativeness for the broader German population. The primary intervention period was 12 weeks, which may be too short to realize maximal weight loss typically observed at 4–6 months or later. App engagement declined over time, potentially attenuating effects. LOCF imputation incorporated app-tracked self-reported weights when clinic measurements were missing, which could introduce bias.