2313-P: AI-Based Meal Logging Reveals Nutritional Differences by Type 2 Diabetes Status in a Real-World Smart-Ring User Cohort

Introduction and Objective: AI-based meal logging in wearables offers a scalable way to capture daily dietary intake for research, but the quality and utility of these data are uncertain. This study sought to characterize users’ nutrient intake distributions with and without Type 2 Diabetes (T2D). Methods: We analyzed de-identified data from 10,378 Oura Ring users (74.4% female; age 18-90 years; BMI: 25.6 ± 5.7 kg/m²; 1.7% self-reported T2D) who logged ≥1 breakfast, lunch, and dinner on ≥5 days during Jan-Nov 2025 using the Oura App. Dietary outcomes (Fig. 1), provided by AI, included carbohydrates, added sugar, fiber, protein, and fat. Differences in dietary patterns between individuals with and without T2D were assessed using Mann-Whitney U tests. Results: AI-based nutritional data produced normally distributed intakes across the main macronutrient categories. T2D status was associated with lower carbohydrate (−37.4 g/day; P < .001) and added sugar intakes (−10.1 g/day; P < .001), higher energy shares from fat (2.8%; P < .001) and protein (1.4%; P < .001), and lower total fiber intake (−3.7 g/day; P < .001). All values represent mean differences (T2D - non-T2D). Conclusion: In this study, individuals with T2D logged meals consistent with a lower carbohydrate and added sugar, higher fat dietary pattern, but also with less fiber. This difference suggests that AI-based meal logging can capture meaningful variation in everyday dietary behavior. Disclosure T. Kinnunen: Employee; Current; Oura Health Oy. P. Ohukainen: Employee; Current; Oura Health oy. M. Karsikas: Employee; Current; Oura Health.