Perioperative control of blood glucose in patients with diabetes has been difficult due to frequent occurrence of hypoglycemia or hyperglycemia. We developed a two-step protocol for management of insulin-treated patients during general anesthesia. Regular insulin was given intravenously before anesthetic induction according to the step I formula (initial): initial blood glucose - 150/10 = U. Regular insulin was then given during surgery according to the step II formula (hourly): blood glucose 150-250 mg/dl = 2 U; blood glucose >250 mg/dl = 4 U. Fluid replacement of 5% dextrose with lactated Ringer's solution, 3 ml/kg estimated ideal body weight, was given hourly and additional lactated Ringer's solution was given as indicated. Thirty patients received preoperative (long- or intermediate-acting) insulin and were managed by the two-step protocol. Ten patients received preoperative (long- or intermediate-acting) insulin and were managed by a standard method used in the same institution. Thirty patients did not receive preoperative insulin and were managed by the two-step protocol. All patients underwent retinal surgery under general anesthesia. In comparing the two groups pretreated with insulin, patients treated by the standard method had significantly higher (mean) blood glucose levels (360.2 ± 100.4 mg/dl) than those treated with the two-step protocol (181.2 ± 50.8 mg/dl) (P = 0.0001) at the end of surgery. Of the two-step protocol patients, those pretreated with one-half the usual morning dose of long- or intermediate-acting insulin had lower (mean) blood glucose levels (225 ± 87 mg/dl) than patients not pretreated (310 ± 130.8 mg/dl) (P = 0.0069) the morning after surgery. Hypoglycemia was not evident. We conclude that use of the two-step protocol can provide a safe, effective method of perioperative control of blood glucose and that pretreatment by one-half the usual morning dose of long- or intermediate-acting insulin preoperatively is effective in improving control during the first 24 h postoperatively.