Performance of homeostasis model assessment and serum high-sensitivity C-reactive protein for prediction of isolated post-load hyperglycaemia.

ABSTRACT

AIMS: To evaluate whether homeostasis model assessment and high-sensitivity C-reactive protein improve the prediction of isolated post-load hyperglycaemia. METHODS: The subjects were 1458 adults without self-reported diabetes recruited between 2006 and 2010. Isolated post-load hyperglycaemia was defined as fasting plasma glucose < 7 mmol/l and 2-h post-load plasma glucose ≥ 11.1 mmol/l. Risk scores of isolated post-load hyperglycaemia were constructed by multivariate logistic regression. An independent group (n = 154) was enrolled from 2010 to 2011 to validate the models' performance. RESULTS: One hundred and twenty-three subjects (8.28%) were newly diagnosed as having diabetes mellitus. Among those with undiagnosed diabetes, 64 subjects (52%) had isolated post-load hyperglycaemia. Subjects with isolated post-load hyperglycaemia were older, more centrally obese and had higher blood pressure, HbA(1c), fasting plasma glucose, triglycerides, LDL cholesterol, high-sensitivity C-reactive protein and homeostasis model assessment of insulin resistance and lower homeostasis model assessment of ß-cell function than those without diabetes. The risk scores included age, gender, BMI, homeostasis model assessment, high-sensitivity C-reactive protein and HbA(1c). The full model had high sensitivity (84%) and specificity (87%) and area under the receiver operating characteristic curve (0.91), with a cut-off point of 23.81; validation in an independent data set showed 88% sensitivity, 77% specificity and an area under curve of 0.89. CONCLUSIONS: Over half of those with undiagnosed diabetes had isolated post-load hyperglycaemia. Homeostasis model assessment and high-sensitivity C-reactive protein are useful to identify subjects with isolated post-load hyperglycaemia, with improved performance over fasting plasma glucose or HbA(1c) alone.