Abstract P053: Diabetes Risk Prediction and Sickle Cell Trait in African Americans From CARDIA and the Jackson Heart Study

Abstract

Listen

Translate article

Introduction: Existing models to predict incident diabetes mellitus (DM) perform better in Whites than African Americans. In models that incorporate hemoglobin A1c (A1C) as a predictor of DM, the difference in model performance by race is more pronounced. In a recent study, we found that A1C was systematically underestimating glycemia in African Americans with sickle cell trait (SCT). Hypothesis: Given the poorer performance of DM prediction models in African Americans than Whites and the impact of SCT on the A1C-glycemia association, we hypothesized that incorporating sickle cell trait into DM prediction models would improve the ability of the model to predict future risk of DM. Methods: We pooled data collected from 2000-2012 on 3,122 African Americans (8.6% with SCT) from the Jackson Heart Study (JHS; n=2,065; mean age=54.71 years) and CARDIA (n=1,057; mean age=44.53). Over 5 years of follow-up in CARDIA and 10 years of follow-up in JHS, 85 CARDIA participants (8.1%) and 342 JHS participants (16.6%) developed DM. Using generalized estimating equations to account for correlation of repeated measures, we compared the discriminative ability and net reclassification improvement (NRI) resulting from the addition of SCT for a series of prediction models. Results: Overall, the addition of SCT to prediction models did not result in significant improvement in the discriminative ability. However, by the NRI index, the addition of SCT to measures of glycemia and to a fuller risk prediction model did improve prediction of DM. In the full model, adding SCT*A1C as a predictor resulted in 2% of events being reclassified as higher risk and 45% of non-events being reclassified as lower risk. Conclusion: Our results suggest that incorporating SCT into DM prediction for African Americans may result in modest improvement in model performance.