Nutrients, Vol. 18, Pages 1340: Implementation and Applications of Artificial Intelligence in Nutrition: A Systematic Review of Use in Practice and Research
Nutrients doi: 10.3390/nu18091340
Authors:
Celia Fabiola Vásquez-García
María Elizabeth Tejero
Marlen Naranjo-Martínez
Alexa Zagorin-Djaddah
Background: Artificial intelligence (AI) is increasingly incorporated into nutrition research and practice; however, the extent of its clinical integration and impact on health outcomes remains unclear. This systematic review evaluated how AI-based systems have been implemented in human nutritional interventions and summarized reported outcomes. Methods: PubMed, Scopus, Google Scholar, SpringerLink, JMIR, and MDPI were searched from January 2020 to March 2025 (search completed in March 2025). Randomized controlled trials and prospective or retrospective cohort studies published in English or Spanish were included if they evaluated AI-driven nutritional interventions in human populations and reported health-related outcomes. Risk of bias was assessed using RoB 2 and ROBINS-I. A qualitative synthesis was performed. Results: Sixteen studies involving 10,863 participants were included. Most were randomized controlled trials targeting metabolic disorders, particularly type 2 diabetes and obesity. Eleven studies evaluated metabolic outcomes, including HbA1c, body weight, fat mass, lipid levels, and insulin resistance indices. Six studies assessed gastrointestinal symptom severity scores, and two examined quality-of-life or patient-reported outcomes. Several trials reported short-term improvements favoring AI-supported interventions in glycemic control, weight reduction, and symptom severity. However, effects were heterogeneous and often observed within multimodal programs, limiting attribution of outcomes solely to the AI component. Conclusions: AI integration in nutrition remains in an early phase of clinical implementation. Although preliminary findings suggest potential benefits, interpretation should be cautious given methodological heterogeneity and moderate-to-high risk of bias across studies. Larger, rigorously designed investigations are required to determine sustained clinical effectiveness.
Background: Artificial intelligence (AI) is increasingly incorporated into nutrition research and practice; however, the extent of its clinical integration and impact on health outcomes remains unclear. This systematic review evaluated how AI-based systems have been implemented in human nutritional interventions and summarized reported outcomes. Methods: PubMed, Scopus, Google Scholar, SpringerLink, JMIR, and MDPI were searched from January 2020 to March 2025 (search completed in March 2025). Randomized controlled trials and prospective or retrospective cohort studies published in English or Spanish were included if they evaluated AI-driven nutritional interventions in human populations and reported health-related outcomes. Risk of bias was assessed using RoB 2 and ROBINS-I. A qualitative synthesis was performed. Results: Sixteen studies involving 10,863 participants were included. Most were randomized controlled trials targeting metabolic disorders, particularly type 2 diabetes and obesity. Eleven studies evaluated metabolic outcomes, including HbA1c, body weight, fat mass, lipid levels, and insulin resistance indices. Six studies assessed gastrointestinal symptom severity scores, and two examined quality-of-life or patient-reported outcomes. Several trials reported short-term improvements favoring AI-supported interventions in glycemic control, weight reduction, and symptom severity. However, effects were heterogeneous and often observed within multimodal programs, limiting attribution of outcomes solely to the AI component. Conclusions: AI integration in nutrition remains in an early phase of clinical implementation. Although preliminary findings suggest potential benefits, interpretation should be cautious given methodological heterogeneity and moderate-to-high risk of bias across studies. Larger, rigorously designed investigations are required to determine sustained clinical effectiveness. Read More