Artificial Intelligence-Generated Diet Plans for Hypertension and Dyslipidemia: Adherence and Nutritional Insights

Abstract

Background: We evaluated diet plans generated by ChatGPT for hypertension and dyslipidaemia. Methods: In October 2024, ChatGPT was used to generate meal plans for 24 simulated patients with different cardiovascular health problems. Data were used from men (n=12) and women (n=12), aged 56 yr, with mean heights of 176 cm and 161 cm respectively. Weight categories were based on BMI: normal, overweight, and obese, using weights of 56, 71, and 84 kg for women and 67, 85, and 101 kg for men. Four health conditions were assessed: hypertension stages 1 and 2 (systolic BP 130-139 mm Hg and >= 140 mm Hg; diastolic BP 80-89 mm Hg and >= 90 mm Hg), and elevated LDL levels (>= 130 mg/dL and >= 160 mg/dL). Menus were evaluated for adherence to Mediterranean and DASH diets, including recommendations. Results: Adherence to the Mediterranean and DASH diets was low across all groups, with median scores below 9 and 4.5, respectively. Common recommendations included weight loss, physical activity, reduced salt intake, stress management, and omega-3s for both hypertension and LDL reduction. Plant sterols/stanols were suggested only for LDL. No advice was given on smoking or alcohol use. Nutrient content did not differ significantly between hypertension and LDL menus (P>0.05). Conclusion: This pioneering study found that AI-generated dietary models had low adherence to DASH and Mediterranean diets, though most recommendations were generally appropriate. Since the prompts only requested basic nutrition plans, future research should use more specific, personalized prompts to better assess AI's role in managing chronic diseases.