PD650 - SAFETY AND CLINICAL ACCURACY OF ARTIFICIAL INTELLIGENCE–GENERATED RECOMMENDATIONS ON DIETARY SUPPLEMENT USE IN CHRONIC DISEASE: A SCENARIO-BASED COMPARATIVE STUDY

Linked sessions

PD650

SAFETY AND CLINICAL ACCURACY OF ARTIFICIAL INTELLIGENCE–GENERATED RECOMMENDATIONS ON DIETARY SUPPLEMENT USE IN CHRONIC DISEASE: A SCENARIO-BASED COMPARATIVE STUDY

F. A. Bozoklu1, H. Çorakçı2, Y. Aykemat1,*

1Nutrition and Dietetics, Nuh Naci Yazgan University, 2Nutrition and Dietetics, Kayseri Metropolitan Municipality Sports Inc., Kayseri, Türkiye

 

Rationale: The increasing use of artificial intelligence (AI) for health-related information includes guidance on dietary supplement use, particularly among individuals with chronic diseases. In these populations, inappropriate recommendations may lead to clinically significant adverse effects and drug–supplement interactions, highlighting the need for systematic evaluation. This study aimed to assess the accuracy, safety, and clinical appropriateness of AI-generated dietary supplement recommendations in chronic disease contexts.

Methods: A cross-sectional, scenario-based content analysis was conducted using standardized cases representing common chronic conditions, including type 2 DM, hypertension, cardiovascular disease, depression, and polypharmacy. Each scenario was submitted to multiple AI systems using identical prompts under controlled conditions. Responses were independently evaluated by experts in nutrition and pharmacology using a predefined scoring rubric. Evaluation domains included scientific accuracy, safety, identification of drug–supplement interactions, consideration of contraindications, disease-specific appropriateness, and referral to healthcare professionals. Scores were assigned on an ordinal scale and composite scores were calculated.

Results: The analysis revealed notable variability in accuracy and safety across AI systems, with inconsistent identification of clinically relevant interactions and contraindications. Recognition of high-risk conditions was often incomplete, and referral to healthcare professionals remained limited.

Conclusion: AI systems may provide general information on dietary supplements; however, their clinical reliability in chronic disease management remains uncertain. Standardization and improved safety frameworks are required before their use in high-risk populations can be considered appropriate.

Disclosure of Interest: None declared