Objective To assess the adequacy of the first-phase insulin response for predicting development of insulin-dependent diabetes.

Research Design and Methods Determinations were made of 1- and 3-min insulin responses to glucose (0.5 g/kg i.v.), islet cell antibodies (ICAs), insulin autoantibodies (IAAs), and HLA. We studied 220 first-degree relatives (aged 3–29 yr) of diabetic patients; 75 underwent two or more tests.

Results At the first test, insulin responses correlated with age in ICA⁻ children ≤ 11 yr old (r = 0.46, P < 0.001). Individual responses varied widely in all ages, and low values were common (5th percentile: 108 pM in children < 5 yr old, 180 pM thereafter). No correlation was found between insulin responses and IAAs or HLA. The responses of 15 ICA⁺ subjects were not significantly different from those of ICA⁻ subjects after excluding the influence of age. At subsequent tests, ICA⁺ and ICA⁻ subjects displayed distinct changes; the mean ± SE insulin response increased in ICA⁻ subjects from 619.2 ± 40.8 to 716.4 ± 50.4 pM (P < 0.001) but declined in ICA⁺ subjects from 403.2 ± 91.8 to 313.8 ± 67.2 pM (P < 0.02). During follow-up, 5 of 9 (56%) consistently ICA⁺ siblings developed diabetes or impaired glucose tolerance compared with 1 of 54 (2%) consistently ICA⁻ subjects. The sensitivity and specificity of two or more low insulin responses (300 pM) for predicting progression to diabetes were 60 and 96%, respectively; the predictive value was 43%. The highest predictive value (75%) was achieved by the combination consistently ICA⁺, consistently low insulin response, and HLA-DR3/4. However, in no subject could the time of onset of diabetes be deduced from the decline of the insulin response.

Conclusions Consecutive intravenous glucose tolerance tests are a useful complement for predicting progression to diabetes but not its onset.