Fasting plasma glucose and Type 2 diabetes risk: a non-linear relationship

The current definition of impaired fasting glucose (IFG, >or=100 mg/dl) has been criticized as being too low for selective identification of individuals at risk for Type 2 diabetes. Furthermore, it is unclear whether any cut-off is justifiable from the shape of association between fasting plasma glucose (FPG) and diabetes. We therefore evaluated the association between FPG and incidence of Type 2 diabetes in a prospective, population-based study.

A case-cohort study within the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam cohort study involved 589 fasted participants in a randomly selected subcohort and 153 incident cases. Restricted cubic spline regression was used to examine non-linearity of associations and we calculated pairs of sensitivities and specificities for different cut-offs of FPG.

Spline regression with adjustment for age, sex, body mass index, waist circumference, education, physical activity, alcohol consumption, and plasma levels of triglycerides, high-density lipoprotein cholesterol and gamma-glutamyltransferase indicated that FPG was associated with risk in a non-linear fashion. Risk with higher FPG increased only above approximately 84 mg/dl. FPG>or=84 mg/dl yielded a sensitivity of 95.4% at a false-positive rate of 86.8%. In comparison, FPG>or=100 and>or=110 mg/dl yielded sensitivities of 78.4 and 42.5% and false-positive rates of 27.8 and 6.8%, respectively. The optimal cut-off of FPG was at approximately 102 mg/dl (sensitivity: 75.8%, false-positive rate: 20.7%).

Although our study suggests a threshold for increasing diabetes risk at 84 mg/dl, this cut-off would classify the vast majority of the population as at risk. The statistically optimal cut-off supports the current definition of IFG (>or=100 mg/dl).