Abstract
Background: Metabolic syndrome represents a cluster of cardiometabolic risk factors that significantly increase the risk of cardiovascular disease and type 2 diabetes. Traditional dietary interventions face challenges in personalization, adherence, and scalability. Artificial intelligence (AI)-based dietary systems offer promising solutions through machine learning algorithms that predict individual metabolic responses and deliver personalized nutrition interventions.
Objective: This comprehensive review synthesizes evidence from randomized controlled trials, pilot studies, and cohort programs evaluating the effectiveness of AI-based dietary systems on cardiometabolic outcomes in high-risk adult populations.
Methods: We analyzed data from multiple AI-driven dietary intervention studies including randomized controlled trials and cohort programs targeting adults with prediabetes, type 2 diabetes, obesity, and metabolic syndrome. Studies employed various AI technologies including machine learning predictive algorithms, chatbots, digital twins, and real-time monitoring platforms integrated with continuous glucose monitoring systems, wearable devices, and mobile applications.
Results: AI-mediated dietary interventions demonstrated significant improvements across multiple cardiometabolic parameters. The MedLiPal chatbot pilot (n=30, 12 weeks) achieved significant reductions in BMI (p<0.05) and waist circumference (99±28 cm to 96±25 cm, p<0.001), with Mediterranean diet adherence scores increasing from 3±3.5 to 11±3 (p<0.001). The Redicare AI-mediated program (n=320) produced average weight/BMI loss of approximately 6.5% (p<0.01), HbA1c reduction of 10.9% (p<0.01), systolic and diastolic blood pressure reductions of 7.1% and 5.9% respectively (p<0.01), and triglyceride reduction of 30% (p<0.01). The Metawell randomized controlled trial (n=220, 6 months) combining internet-based platform with meal replacement showed significantly greater reductions in waist circumference, triglycerides, fasting glucose, fasting insulin, and HOMA-IR compared to control (p<0.01). The personalized postprandial-targeting (PPT) diet RCT (n=200, 6 months) in prediabetic adults demonstrated increased microbiome alpha-diversity (p=0.007) with machine learning-predicted diet adherence mediating improvements in HbA1c, HDL-C, and triglycerides through specific microbial species.
Conclusions: AI-based dietary systems show substantial promise in preventing and managing metabolic syndrome in high-risk adult populations. These technologies enable personalized nutrition interventions that achieve clinically meaningful improvements in weight, glycemic control, blood pressure, and lipid profiles. The integration of machine learning algorithms with continuous monitoring, real-time feedback, and behavioral support represents a scalable approach to precision nutrition. Future research should focus on long-term sustainability, cost-effectiveness, and implementation strategies across diverse healthcare settings.