Adequate comparisons of the relative performance of different tests of beta-cell function are not available. We compared discrimination of commonly used in vivo tests of beta-cell function across a range of glucose tolerance in seven subjects with normal glucose tolerance (NGT), eight subjects with impaired glucose tolerance (IGT), and nine subjects with type 2 diabetes. In random order, each subject underwent two of each of the following tests: 1) frequently sampled 0.3-g/kg intravenous glucose tolerance test (FSIVGTT) with MinMod analysis; 2) homeostasis model assessment (HOMA) from three samples at 5-min intervals with a model incorporating immunoreactive or specific insulin measurements; and 3) continuous infusion of 180 mg x min(-1) x m(-2) glucose with model assessment (CIGMA) of three samples at 50, 55, and 60 min (1-h CIGMA) and at 110, 115, and 120 min (2-h CIGMA). The discrimination of each test was assessed by the ratio of the within-subject SD to the underlying between-subject SD, the discriminant ratio (DR). The degree to which tests measured the same physiological variable was assessed using Pearson's correlation coefficient adjusted for attenuation due to test imprecision. An unbiased line of equivalence, taking into account the imprecision of both tests, was used to compare results. Beta-cell function assessed from HOMA and beta-cell function assessed from CIGMA (CIGMA%beta) (using immunoreactive insulin) had higher DRs than first-phase intravenous glucose tolerance test-derived incremental insulin peak, area, insulin-to-glucose index, and acute insulin response to glucose from FSIVGTT-MinMod. CIGMA%beta (immunoreactive insulin) had the highest DR. FSIVGTT-derived first-phase insulin response tests correlated only moderately with HOMA and CIGMA. Using specific rather than immunoreactive insulin for HOMA and CIGMA did not improve discriminatory power. Simple tests such as HOMA and CIGMA, using immunoreactive insulin, offer better beta-cell function discrimination across subjects with NGT, IGT, and type 2 diabetes than measurements derived from FSIVGTT first-phase insulin response.