Post-load glucose spike: a better predictor of post-MI prognosis than post-load glucose in patients without known diabetes

Abstract

Abstract Funding Acknowledgements Type of funding sources: None. Background In non-diabetic population, number of deaths increase as the post-challenge glucose spike (PGS) i.e. difference between the fasting and 2h-PG increases in the DECODE study. Purpose To assess whether PGS is a better predictor of post-ACS prognosis in patients without known diabetes. Methods Retrospective cohort analysis of post-MI survivors (n=1056) without known diabetes undergoing pre-discharge oral glucose tolerance test. A standardised dataset was collected for the Myocardial Infarction National Audit Project. MACE (death or non-fatal re-infarction) over 5 (median 3.4) years was collected from notes review. Predictors of MACE were determined from Cox proportional hazard regression (stepwise) models adjusted for several variables (tested for collinearity) including GRACE score (GRS), fasting (FG) and 2-hour post-load glucose (2hG) and PGS as % of the 2hG (%PGS) (VIF=3.12). Incremental predictive value of PGS over 2hG was tested using probabilities of MACE calculated from logistic regression models using the same covariates and FG, 2hG and PGS individually and in combination applying c-statistics, continuous net reclassification improvement (cNRI) and integrated discrimination improvement (IDI). Results Of the patients 469 (44.4%) had normal glucose tolerance, 378(35.8%) preDM and 209 (19.8%) had new DM. In the whole cohort, MACE was predicted by %PGS (HR 1.02 (1.01-1.03), p<0.0001), GRS (HR 1.01, (1.01-1.02), p<0.0001), discharge with betablockers (HR 1.50 (1.14-1.98), p=0.0037), previous history of MI (HR 1.44 (1.02-2.03), p=0.039) but neither FG nor 2hG were included in the final model. MACE was predicted by %PGS (HR 1.04 (1.02-1.06), p<0.0001) in the non-diabetic group but not in the new DM patients (HR 1.02 (0.99-1.05), p=0.296). Addition of PGS to models containing FG and 2hG improved reclassification by 22.6% for events and 5.0% for non-events (NRI>0 0.275, p<0.001) and discrimination (IDI 0.0076, p=0.033). However the c-statistics did not change significantly (ΔAUC 0.0094, p=0.099). Conclusion PGS is a better predictor of post-MI prognosis than FG or 2hG in patients without known diabetes.