The safety, reproducibility, and reliability of an insulin infusion test for assessment of adequate glucose counterregulation were evaluated in 18 patients with type I (insulin-dependent) diabetes mellitus. When the test (a 60-min, 30-mU/m2/min insulin infusion) was administered on three separate occasions at 3-4-wk intervals, coefficients of variation for plasma glucose and counterregulatory hormone (glucagon, epinephrine, cortisol, and growth hormone) responses averaged less than 8%. No patient experienced symptoms requiring discontinuation of the test and plasma glucose concentrations increased spontaneously after stopping the insulin infusion. Using objective criteria based on plasma glucose nadirs or postnadir rates of plasma glucose recovery, no patient judged to have adequate glucose counterregulation by the test (postnadir rates of plasma glucose recovery or plasma glucose nadir above 0.4 mg/dl/min and 45 mg/dl) developed severe hypoglycemia (plasma glucose less than 40 mg/dl) during up to 7 mo of intensive insulin therapy, whereas nearly all patients with inadequate counterregulation did. We conclude that this test, when performed in standardized conditions, is safe and reproducible and can reliably predict those patients with type I diabetes who are at risk of developing severe hypoglycemia during intensive insulin therapy.