Obesity is a global epidemic with substantial health and economic impacts, making scalable weight management strategies essential. A comprehensive synthesis of eHealth interventions for weight management is needed to guide clinical practice. This umbrella review evaluated mobile and web-based interventions for weight loss in adults with overweight or obesity, compared to conventional or non-intervention controls. Systematic reviews were identified across five electronic databases from inception to February 2025. Two reviewers independently selected studies and assessed methodological quality using AMSTAR 2. Pooled estimates were calculated using random-effects models. Eleven systematic reviews (261 primary studies, 62,407 participants) were included. Mobile app interventions yielded a significant reduction in body weight (MD = −1.32 kg; I² = 82%), as did long-term eHealth interventions (MD = −1.13 kg; I² = 76%). Most meta-analyses showed high heterogeneity. Web-based interventions showed no significant effect. In conclusion, mobile apps and long-term eHealth interventions resulted in modest but statistically significant reductions in body weight, body mass index, and waist circumference. The evidence for web-based approaches remains inconclusive. Further research should focus on low-resource settings, primary care, and the integration of emerging technologies such as artificial intelligence. (PROSPERO CRD42025644218).

Obesity represents a growing challenge for Primary Health Care (PHC), demanding the use of new technologies and innovative solutions. This study evaluated the adherence of Large Language Models (LLMs) to Brazilian guidelines for obesity management in PHC. An in silico experimental study was conducted in two phases. Initially, 62 LLMs underwent latency screening, with 24 meeting the usability criterion (<10s) and proceeding to content evaluation. Clinical adherence was measured by 16 questions extracted verbatim from the Clinical Protocol (PCDT), covering diagnosis, monitoring, and treatment, and was peer-reviewed (Kappa 0.68-1.00). The exclusion of 38 models (61%) due to latency highlighted infrastructure barriers. Among the 24 viable models, adherence was heterogeneous and limited, with the best performance reaching only 61.1% compliance. The correlation between model size and accuracy was moderate, but multimodality proved to be a predictor of better performance. It is concluded that current LLMs do not offer safety for autonomous operation in PHC, requiring rigorous professional curation, continuous validation, and escalation mechanisms to human care to ensure user safety.

Resumo A obesidade representa um desafio crescente para a Atenção Primária à Saúde (APS), demandando a utilização de novas tecnologias e soluções inovadoras. Este estudo avaliou a aderência de Modelos de Linguagem de Grande Escala (LLMs) às diretrizes brasileiras para o manejo da obesidade na APS. Conduziu-se estudo experimental in silico em duas fases. Inicialmente, 62 LLMs passaram por triagem de latência com 24 atingindo o critério de usabilidade (<10s) e seguindo para a avaliação de conteúdo. A aderência clínica foi mensurada por 16 questões extraídas literalmente do Protocolo Clínico (PCDT), cobrindo diagnóstico, monitoramento e tratamento, revisada por pares (Kappa 0,68-1,00). A exclusão de 38 modelos (61%) por latência evidenciou barreiras de infraestrutura. Nos 24 modelos viáveis, a aderência foi heterogênea e limitada, com o melhor desempenho atingindo apenas 61,1% de conformidade. A correlação entre tamanho do modelo e acerto foi moderada, mas a multimodalidade mostrou-se preditora de melhor desempenho. Conclui-se que os LLMs atuais não oferecem segurança para atuação autônoma na APS, exigindo curadoria profissional rigorosa, validação contínua e mecanismos de escalonamento para o cuidado humano para garantir a segurança do usuário.

Objective: To evaluate the perceived usability, acceptability, and user experience (rather than the clinical effectiveness) of Boora, an AI-assisted, human-supervised digital platform prototype for longitudinal overweight and obesity care, among users and health professionals in Brazilian primary care. Design: Convergent mixed-methods formative evaluation. Perceived usability was measured with the System Usability Scale (SUS) and summarised descriptively; semi-structured interviews conducted after hands-on use were analysed with codebook thematic analysis (Braun and Clarke); the two strands were integrated through a joint display. Qualitative reporting followed the Consolidated Criteria for Reporting Qualitative Research (COREQ). Setting: Primary health care network of Ananindeua, Pará, within the Brazilian Unified Health System (January to February 2026). Participants: Fifteen adults with overweight or obesity (BMI at least 25 kg/m2, confirmed via electronic health records) who used the patient application on their own smartphones for 24 hours, and eight primary care professionals (nurses, physicians, and a dietitian) who used the professional dashboard for approximately 20 minutes on predefined tasks with synthetic data. Main outcome measures: SUS scores and qualitative themes addressing usability, acceptability, perceived usefulness, barriers, and perceived clinical and workflow fit. Results: Boora showed good perceived usability in both cohorts (users mean 76.5, SD 10.3; professionals mean 77.5, SD 4.6; both above the SUS normative average of 68). Four themes emerged per cohort. Users valued an accessible interface and visible progress but described daily logging burden, fragile anticipated engagement, and digital-literacy and accessibility barriers. Professionals valued a clear interface and the prospect of panel-managed, proactive follow-up, while requiring training, AI governance, protected time, and interoperability with the national record. Integration indicated that the disengagement users anticipated was the risk professionals perceived the dashboard could help identify, whereas the educational AI assistant was the weakest and most ambiguous component for both groups. Conclusions: Boora was perceived as usable and acceptable, with perceived value concentrated in human-supervised, longitudinal follow-up rather than autonomous self-tracking or AI advice. These findings concern perceived usability and acceptability, not clinical effectiveness or sustained engagement. Real-world adoption would depend on accessibility refinements, electronic-record integration, and clear AI governance aligned with the principles of Brazil’s proposed risk-based AI framework and the LGPD. Key points: What is already known on this topic: Mobile and web applications can support weight management, but user engagement tends to decline over time without structured professional support, and most evidence comes from high-income countries. Autonomous AI advice in primary care can be unsafe or discordant with clinical guidelines, underscoring the need for human oversight and auditable systems. What this study adds: An AI-assisted, human-supervised platform prototype pairing a patient application with a professional dashboard was perceived as usable and acceptable by both patients and primary care professionals in the Brazilian public health system, based on brief hands-on use. Mixed-methods integration indicated that the disengagement users anticipated corresponded to the risk professionals perceived the dashboard could help identify, locating the platform’s perceived value in human-supervised longitudinal follow-up; the educational AI assistant was its weakest component. How this study might affect research, practice or policy: It supports designing AI in primary care obesity tools as auditable, supervised support rather than autonomous advice, consistent with the principles of Brazil’s proposed risk-based AI framework. Participants identified accessibility for low-literacy users, interoperability with the national electronic health record, protected time, and training as potential prerequisites for adoption, which warrant longitudinal, real-world evaluation.