Comparison Between the Health Effects of an AI-driven Model With Those of Human Professional Guidelines for the Treatment of Obesity

Overview

Many non-communicable diseases are diet-related and have a significant impact on public health. It is stated that global dietary shifts are needed to change disease patterns, highlighting the importance of nutrition in addressing public health issues, such as obesity. The field of nutrition has been dependent on clinical and observational studies; however, the emergence of Artificial Intelligence (AI) is transforming these approaches. ChatGPT may be used to provide dietary recommendations due to its high speed, extensive access to a variety of meal data, and low complexity. However, initial evaluations have shown that ChatGPT may be inaccurate in terms of safety and reliability, and traditional nutrition approaches are highly reliant on experts' knowledge and the validity of nutritional guidelines. It has been suggested that AI in nutrition may be beneficial; however, further investigations are needed. Our proposal aims to fill the represented critical evidence gaps. The study aims to compare the health effects of an AI-driven model with those of human professional guidelines for the treatment of obesity. Furthermore, investigators seek better strategies to utilize AI, if appropriate, for weight loss and other health benefits.

An 8-week parallel-randomized clinical trial will be conducted. Participants will be recruited through printed flyers posted at approved locations, including Texas Tech University campus buildings and community centers in Lubbock, Texas. In addition to physical flyers, recruitment may include unpaid postings on social media platforms. No paid or targeted advertisements will be used. Flyers will include a QR code directing interested individuals an online survey. Based on their completed survey, they will be screened according to inclusion and exclusion criteria. If they meet the criteria, they will be invited for the baseline visit. Eligible participants will be asked to come fast for 8 hours to the Nutrition and Metabolic Health Initiative (NMHI). Upon arrival, the written consent form will be provided, and they will be enrolled in the study. They will be randomly allocated into one of the three groups in a 1:1:1 ratio. The REDCap will be used for electronic data capture for randomization and data management. The experimental groups will include: * An AI-only group using ChatGPT Premium. * A dietitian-led group (human), in which dietary plans are developed by a licensed dietitian using standard professional resources and clinical judgment, without the use of generative AI tools. * An AI-assisted dietitian group (combination), in which the same dietitian is permitted to use ChatGPT Premium as a supportive tool during the consultation. This design reflects real-world clinical practice, where dietitians commonly use professional judgment and non-AI resources. After allocation, blood pressure will be measured. Resting metabolic rate, fasting blood glucose, lipid profile, and HbA1 will be recorded. Participants will be blinded to their group assignment; however, the study will be unable to blind the personnel due to the nature of the study. Additionally, the participants will be asked to complete the short form of the International Physical Activity Questionnaire (IPAQ). The AI group participants will be in touch only with a trained study staff member (operator) who will facilitate communication with ChatGPT Premium so that the participants cannot see the operator's screen. The operator will not be a licensed dietitian or healthcare provider and will not provide any independent dietary advice. The operator will receive standardized training prior to study initiation to ensure consistency across participants. The results of the BIA and finger prick tests will be given to the AI and asked to provide a dietary meal plan based on a standard prompt. The human group participants will sit in the same room, but instead of the operator, they will have 60-minute consultations with the dietitian. The results of baseline assessments would be provided; however, the dietitian does not have any AI access. Participants in the combination group will sit in the same room and be consulted for 60 minutes by the same dietitian as the human group. However, in this group, the dietitian has access to ChatGPT premium and can use it. After 8 weeks, all participants will be asked to attend the clinic for their final visit and glycemic index, lipid profile, HbA1c, and body composition will be measured. In addition, the participants will be asked about their self-report compliance assessment.