196-LB: Heterogeneity in Elevated Glucose and A1C as Predictors of the Prediabetes-to-Diabetes Transition—Framingham Heart Study, Multiethnic Study on Atherosclerosis, Jackson Heart Study, and Atherosclerosis Risk In Communities

Abstract

The clinical utility of the elevated altered glucose tolerance and prediabetes, and novel phenotypes of diabetes risk is a hotly debated. The heterogeneity of risk and incidence across diverse cohorts needs to be considered when developing new approaches for screening and when communicating the uncertainty in the transition rate. How the definition of prediabetes relates to cohort-level sources of variation and predicts the transition from prediabetes to overt diabetes has not been well characterized. Here, we study the range and heterogeneity in the absolute incidence and the relative risk for diabetes across competing definitions of prediabetes in participants from four large US cohorts, the Framingham Heart Study (FHS), Multi-Ethnic Study on Atherosclerosis (MESA), Jackson Heart Study (JHS), and Atherosclerosis Risk in Communities (ARIC). We computed the incidence rates across prediabetes (fasting glucose, A1C%; ADA or WHO), outcome definition (elevated glucose or overt T2D), body mass index, blood pressure, age, race, sex and cohort. The cross-cohort meta-analytic estimate rate for transition to diabetes for FBG 100-125 mg/dL, 110-125 mg/dL, and A1C 5.7-6.4% were 4.2, 8.2, 2.8 cases per 100 person-years, respectively. This contrasts with incidence rates of 0.6, 1.0, and 0.6 for the individuals with FBG less than 100, FBG less than 110, and A1C% less than 5.7% respectively. All the meta-analytic estimates had I2 greater than 93%. The between-cohort range FBG 110-125 mg/dL was 4 (FHS Generation 3), 5 (FHS Generation 2), 5.5 (ARIC), 12.5 (JHS), and 14.8 (MESA). In contrast, the range of the incidence rate for A1C 5.7-6.4% was 1.6 (ARIC) to 5.2 (JHS). We found that these inter-cohort differences cannot be explained by body mass index or ethnic differences and reporting of rates need to consider the wide prediction interval between cohorts. Disclosure C. Patel: None. Funding National Institutes of Health (R01ES032470)