The objective of this study was to characterise test properties and agreement for fasting glucose cut-offs used for screening diabetes in Indigenous Australian communities, across a range of diabetes prevalence. The oral glucose tolerance test (OGTT) was administered to adult volunteers (n=3249) for community-based diabetes screening in isolated settlements (n=25). Two-hour post-load glucose concentration was the 'gold standard' against which test properties were characterised for fasting plasma glucose cut-offs at concentrations of > or = 7.0 mmol/l (ADA criterion) and > or = 7.8 mmol/l (WHO criterion). Test sensitivity (95% CI) was greater for the ADA (72.5%, 67.4-77.1%) versus the WHO criterion (61.5%, 56.1-66.6%). Specificity was high (>98%) using either criterion. The post-test likelihood of diabetes, given a population prevalence of 14.8%, was 89.1% using the ADA criterion, and 94.9% using the WHO criterion. The ADA criterion gave better agreement than did the WHO criterion for diabetes prevalence > or = 8.6%, and the probability of false results was lower using the ADA criterion, when diabetes prevalence was > or = 12.8%. According to the ADA criterion most individuals ascribed false negative results were aged > or = 35 years, overweight/obese, or had impaired fasting glucose. The fasting glucose criterion of 7.0 mmol/l was superior to 7.8 mmol/l in screening for diabetes.