Predicting dysglycaemia in patients under investigation for acute coronary syndrome

Abstract

Aims: To examine methods for the identification of previously undetected dysglycaemia [diabetes and impaired glucose tolerance (IGT)] in patients investigated for possible acute coronary syndrome. Specifically, we wished to examine whether the recently advocated use of glycosylated haemoglobin (HbA1c) would enhance detection rates for diabetes in these patients.

Methods: Patients (n = 200) investigated for possible acute coronary syndrome and not previously known to have diabetes were recruited and anthropometric data collected. Random plasma glucose concentrations followed by oral glucose tolerance tests, HbA1c, fasting lipids, high sensitivity C-reactive protein and homeostatic modular assessment-insulin resistance were obtained during admission. Following discharge, the fasting plasma glucose (FPG) was repeated to determine the importance of sequential fasting levels. The accuracy of individual tests, combinations and sequential testing was assessed using receiver operating characteristic curves. A predictive index (PI) was generated using stepwise logistic regression models.

Results: The prevalence of diabetes and IGT were 21 and 32%, respectively. FPG >6.0 mmol/l and HbA1c ≥ 6.0% had specificities of 94.9% and 93.6% but sensitivities of only 31.7 and 39.0%, respectively. Combination and sequential testing provided little additional benefit. Use of a PI comprising FPG, HbA1c and age provided the best overall performance (75.6% sensitivity, 77.1% specificity, negative predictive value 92.4%).

Conclusion: Our data confirm the high prevalence of dysglycaemia in this cohort. The commonly advocated screening tools have significant limitations if used in isolation, combination or sequentially. Our approach using a PI offers improved performance partly as it uses continuous data rather than arbitrary cut-off values.