Aims/hypothesis: The value of diagnostic categories of glucose intolerance for predicting type 2 diabetes is much debated. We therefore sought to estimate relative and population-attributable risk of different definitions based on fasting (impaired fasting glucose [IFG]) or 2 h plasma glucose concentrations (impaired glucose tolerance [IGT]) and to describe the associated clinical phenotypes.
Methods: We prospectively observed a population-based cohort of 1,963 non-diabetic participants (mean age 47 years), in whom an OGTT was performed at baseline and 7 years later.
Results: IGT was fivefold more prevalent (13.5%) than IFG. In both categories, participants were older, heavier, hyperinsulinaemic, hyperproinsulinaemic and dyslipidaemic compared with participants with normal glucose tolerance. Relative risk of incident diabetes was similar for IFG and IGT categories (3.73 [95% CI: 2.18-6.39] and 4.01 [95% CI: 3.12-5.14], respectively), but the population-attributable risk was fivefold higher for IGT (29% [95% CI: 26-32]) than for IFG (6% [95% CI: 5-7]). Isolated IFG carried no increase in risk. Lowering the threshold to 5.6 mmol/l raised the population-attributable risk of IFG to 23% (95% CI: 20-25); its contribution to diabetes progression, however, was largely due to co-existent IGT. In multivariate analysis adjusting for sex, age, familial diabetes and BMI, fasting and 2 h glucose were independent predictors.
Conclusions/interpretation: Fasting and 2 h glucose values are independent predictors of incident diabetes. Isolated IFG is not a high-risk condition; lowering the diagnostic threshold increases the population-attributable risk of IFG fourfold, but performing an OGTT captures additional diabetes progressors compared with the number identified by IFG.