Prognostic impact of admission blood glucose for all-cause mortality in patients with acute coronary syndromes: added value on top of GRACE risk score

Background: Abnormal glucose metabolism is a predictor of worse outcome after acute coronary syndrome (ACS). However, this parameter is not included in risk prediction scores, including GRACE risk score. We sought to evaluate whether the inclusion of blood glucose at admission in a model with GRACE risk score improves risk stratification. Methods: Study of consecutive patients included in a single centre registry of ACS. Our primary endpoint was the occurrence of all-cause mortality at one-year follow-up. The ability of the two logistic regression models (GRACE risk score alone and in combination with blood glucose) to predict death was analysed. Continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI), with corresponding 95% confidence intervals (CIs), were also calculated. Results: We included 2099 patients, with a mean age of 64 (SD=13) years, 69% males. In our sample, 55.1% presented with ST-segment elevation ACS and 13.1% in Killip class >= 2. Only 25% were known diabetic at admission. In-hospital mortality was 5.8% and 9.7% at one-year follow-up. The best cut-point for blood glucose was 160 mg/dl (sensitivity 62% and specificity 68%), and 35.2% of the patients had increased levels. This group was elderly, had more prevalence of cardiovascular risk factors, worse renal function and GRACE score as well as more frequently Killip class >= 2. Treatment was similar in both groups besides less frequent use of clopidogrel in high glycaemic patients. The hyperglycaemia group had higher one-year mortality (17.2% vs. 5.6%, p<0.001). Moreover, binary blood glucose remained a predictor of death independently of the GRACE risk score and the presence of diabetes (odds ratio (OR) 1.99, 95% CI 1.40-2.84, p<0.001). The inclusion of blood glucose, as a continuous variable, in a logistic regression model with GRACE score, increased the area under the ROC curve from 0.80 to 0.82 (p=0.018) as well as the goodness-of-fit and was associated with an improvement in both the NRI (37%) and the IDI (0.021), suggesting effective reclassification. Conclusions: A blood glucose level on admission >= 160 mg/dl is an independent predictor of mortality in mediumterm follow-up. It offers an incremental predictive value when added to the GRACE risk score, although with a modest magnitude of improvement, probably due to the high predictive performance of the GRACE risk score alone.