Performance of Glycated Hemoglobin and a Risk Model for Detection of Unknown Diabetes in Coronary Patients

Introduction and objectives Traditionally, the oral glucose tolerance test has been useful to diagnose unknown diabetes. Recently, the American Diabetes Association committee has accepted glycated hemoglobin ≥6.5% as a criterion for unknown diabetes. The aim was to determine the benefit of glycated hemoglobin for diagnosing unknown diabetes and also create a predictive model that adjusts the indication for oral glucose tolerance test in coronary patients. Methods We examined the glycemic profile of 338 coronary patients without previous diagnosis of diabetes, applying 2010 American Diabetes Association criteria. A unknown diabetes risk predictive model was developed using logistic regression analysis, and then validated in another cohort. Results Using the glycated hemoglobin criteria and/or fasting plasma glucose, unknown diabetes was diagnosed in 26 patients. The remaining patients were classified according to oral glucose tolerance test as follows: unknown diabetes 53 (17%), prediabetes 144 (46.2%), and normoglycemic 115 (36.8%). The diagnostic method for unknown diabetes was fasting plasma glucose in 25.3%, glycated hemoglobin in 7.6%, and oral glucose tolerance test in 67.1%. A risk model including fasting plasma glucose, glycated hemoglobin, left ventricular ejection fraction, age, and noncoronary vascular disease was shown to effectively predict unknown diabetes after oral glucose tolerance test: area under the ROC curve 0.8 (95% interval confidence: 0.74-0.87). When the oral glucose tolerance test is restricted to patients with a risk score >6 (31% of our sample) we properly identify 83% of unknown diabetes cases (sensitivity: 75%, specificity: 73%, positive predictive value: 40%, negative predictive value: 93%). The model was adequately validated in another cohort of 115 patients (area under the ROC curve 0.84 [95% interval confidence: 0.74-0.95]). Conclusions In coronary patients, glycated hemoglobin alone failed to detect many cases of unknown diabetes. However, its inclusion in a risk prediction model leads to optimizing the usefulness of oral glucose tolerance test.