Adjustment of the GRACE score by 2-hour post-load glucose improves prediction of long-term major adverse cardiac events in acute coronary syndrome in patients without known diabetes

Abstract

Global Registry of Acute Coronary Events (GRACE) risk score (GRS), a powerful predictor of prognosis after acute coronary event (ACE), does not include a glucometabolic measure. We investigate whether 2 h post-load plasma glucose (2h-PG) could improve GRS based prognostic models in ACE patients without known diabetes mellitus (DM). A retrospective cohort study of 1056 ACE survivors without known DM who had fasting plasma glucose (FPG) and 2h-PG measured pre-discharge. Death and non-fatal myocardial infarction were recorded as major adverse cardiac events (MACE) during follow-up. GRS for discharge to 6 months was calculated. Cox proportional-hazards regression was used to identify predictors of event free survival. The predictive value of 2h-PG alone and combined with GRS was estimated using likelihood ratio test, Akaike's information criteria, continuous net reclassification improvement (NRI>0), and integrated discrimination improvement (IDI). During 40.8 months follow-up 235 MACEs (22.3%) occurred, more frequently in the upper 2h-PG quartiles. Two-hour PG, but not FPG, adjusted for GRS independently predicted MACE (hazard ratio 1.091, 95% confidence interval 1.043-1.142; P = 0.0002). likelihood ratio test showed that 2h-PG significantly improved the prognostic models including GRS (χ2 = 20.56, 1 df; P = 0.000). Models containing GRS and 2h-PG yielded lowest corrected Akaike's information criteria, compared to that with only GRS. 2h-PG, when added to GRS, improved net reclassification significantly (NRIe>0 6.4%, NRIne>0 24%, NRI>0 0.176; P = 0.017 at final follow-up). Two-hour PG, improved integrated discrimination of models containing GRS (IDI of 0.87%, P = 0.008 at final follow-up). Two-hour PG, but not FPG, is an independent predictor of adverse outcome after ACE even after adjusting for the GRS. Two-hour PG, but not FPG, improves the predictability of prognostic models containing GRS.