Abstract

Background Type 2 diabetes mellitus (T2DM) is a major global health challenge. While lifestyle interventions can induce remission, their effectiveness is often constrained by inter-individual variability. Artificial intelligence (AI) and machine learning (ML) offer possibilities for precision lifestyle medicine by tailoring interventions to individual patient profiles.

Objective To synthesize evidence on AI- and ML-enhanced lifestyle interventions targeting remission of T2DM, with emphasis on algorithms, data sources, and personalization strategies.

Methods Systematic searches of PubMed, Embase, Scopus, IEEE Xplore, and Web of Science (2015–August 2025) identified studies using AI/ML in lifestyle-focused T2DM interventions. From 3,220 records, 100 studies met inclusion criteria. Eligible studies involved adults with T2DM, reported remission or glycemic outcomes, and applied AI to personalize lifestyle programs. Data were synthesized narratively according to PRISMA 2020 guidelines.

Results Studies span diverse geographies, with interventions delivered via mobile apps, conversational agents, decision support systems, and digital twins. AI methods included supervised models (random forests, gradient boosting, support vector machines) for prediction, convolutional neural networks for remission forecasting in surgical cohorts, recurrent neural networks for glucose forecasting, clustering for phenotype stratification, and reinforcement learning for adaptive insulin titration and multimorbidity management. Data inputs ranged from continuous glucose monitoring and wearables to dietary logs, electronic health records, and multi-omics. Outcomes were favorable, with remission rates exceeding 70% in some digital twin and subtype-based interventions, alongside significant HbA1c reduction, weight loss, and medication tapering.

Conclusion AI and ML can transform T2DM care by enabling adaptive, precision interventions for remission. Future work should emphasize large-scale trials, explainable algorithms, and equitable deployment to ensure sustainable real-world impact.

Statistics