1299-P: Individualized Type 2 Diabetes Risk Prediction

Introduction and Objective: Current T2D risk prediction models do not include individualized preventive intervention effects. We developed an individualized T2D prediction model with differential treatment effects based on an individual’s fasting glucose (FG) level and body mass index (BMI). Methods: We included 2640 participants from the Diabetes Prevention Program (DPP) randomized to placebo, metformin, or lifestyle arms. Using a 50/50 train/test split, we developed and validated an individualized Cox model predicting T2D risk at 3 years (3yr) with adjustment for sex, HbA1c, triglycerides, FG, BMI, intervention, and interactions between intervention and age, FG, and BMI. We repeated this process 100 times and took the mean concordance (C)-statistic of the model. We then fit the model to all available data and calculated counterfactual T2D risk for each intervention. Results: Mean (standard deviation) age was 51 years (11) and 54% of participants were women. The mean C-statistic was 0.78 (95% confidence interval: 0.74, 0.81). The optimal intervention (lowest 3yr risk) was lifestyle for 86% of participants and metformin for the remaining 14% (table). If assigned to lifestyle, 3yr risk was 10% for those where lifestyle was optimal and 21% when metformin was optimal. When metformin was optimal, it was associated with 44% lower T2D risk than placebo and 29% lower than lifestyle. Conclusion: Individualized T2D risk prediction may improve preventive strategies. Disclosure B.C. Jaeger: None. B.J. Wells: None. J.M. Stafford: None. R. Casanova: None. Y. Demesie: None. M. Bancks: None. Funding American Diabetes Association (11-22-ICTSPM-18)