2432-PUB: Predictors of Imminent Glycemic Worsening in Adults with Screen-Detected Diabetes

Reliable clinical indicators of near-term risk of worsening glycemia in type 2 diabetes (T2D) could help guide individual-level interventions. Here we analyzed data from the Early Diabetes Intervention Program (EDIP) to identify predictors of near-term glycemic progression. EDIP evaluated progression of fasting blood glucose (FBG) above 140 mg/dl on consecutive quarterly measurements in individuals with screen-detected T2D and FBG below 140 at study entry (age 52.6 + 11.5 years, 41.9% male, mean entry FBG 121±14 mg/dL; mean 2-hour OGTT glucose 236±30; mean HbA1C 6.4+0.57%). We investigated accessible clinical parameters (weight, glycemia, lipids, and insulin levels) as potential near-term predictors of progression, evaluating absolute levels and yearly changes preceding progression. Variables were evaluated individually in age/sex/race-adjusted analyses and in combination in multivariable Cox progression models. Change over the preceding 1 year in fasting glucose, but not in weight or insulin levels, was predictive of glycemic progression in the subsequent year in adjusted univariate analyses (p<0.03). In multivariable modeling, the FBG concentration (hazard ratio HR 1.14) and the antecedent 1-year change in FBG (HR 0.90) were the strongest predictors of progression, and both retained significance in multivariable models including all specified variables (p< 0.0001). Adding changes in fasting insulin, HOMA-IR, or oral disposition index to the modeling did not change this result. There was a significant correlation between current FBG and the antecedent 1-year change (r=0.587; p<0.01), highlighting that change in FBG was contributing in part via its effect on current FBG. In the EDIP study population with screen-detected T2D, current FBG and antecedent 1-year change in FBG were the best predictors of upcoming glycemic worsening. These data underscore the value of serially monitoring fasting glucose, and suggest a similar monitoring paradigm should be evaluated in prediabetes.