Background: A significant proportion of individuals with diabetes or impaired glucose tolerance have fasting plasma glucose less than 6.1 mmol/L and so are not identified with fasting plasma glucose measurements. In this study, we sought to evaluate the utility of plasma lipids to improve on fasting plasma glucose and other standard risk factors for the identification of type 2 diabetes or those at increased risk (impaired glucose tolerance).
Methods and findings: Our diabetes risk classification model was trained and cross-validated on a cohort 76 individuals with undiagnosed diabetes or impaired glucose tolerance and 170 gender and body mass index matched individuals with normal glucose tolerance, all with fasting plasma glucose less than 6.1 mmol/L. The inclusion of 21 individual plasma lipid species to triglycerides and HbA1c as predictors in the diabetes risk classification model resulted in a statistically significant gain in area under the receiver operator characteristic curve of 0.049 (p<0.001) and a net reclassification improvement of 10.5% (p<0.001). The gain in area under the curve and net reclassification improvement were subsequently validated on a separate cohort of 485 subjects.
Conclusions: Plasma lipid species can improve the performance of classification models based on standard lipid and non-lipid risk factors.