The incidence of low nocturnal blood glucose values (i.e., less than 65 mg/dl) was assessed in 20 insulin-dependent diabetic patients treated with continuous subcutaneous insulin infusion supported by capillary blood glucose monitoring before each meal and the evening snack. Patients were randomly assigned to a control or experimental group. Both groups followed an identical protocol for the first part of the study (baseline). Patients were instructed to determine capillary blood glucose measurements five times during the night for three consecutive nights. The same procedure was repeated one week later, but this time the subjects in the experimental group were instructed to have an extra snack if capillary blood glucose levels at 10:30 P.M. were 120 mg/dl or less. The control group continued with the usual routine of one evening snack at 9 P.M. At baseline, the incidence of capillary blood glucose values of less than 65 mg/dl was 13 percent. The ingestion of an extra snack at bedtime resulted in the absence of capillary blood glucose values of less than 65 mg/dl in the experimental group, whereas the incidence of capillary blood glucose values of less than 65 mg/dl in the control group remained 13 percent (p = 0.038). The capillary blood glucose concentration at 10:30 P.M. was highly predictive of the risk of nocturnal blood glucose values below 65 mg/dl (p = 0.015) and fasting capillary blood glucose values above 140 mg/dl (p = 0.0001). These data show that nocturnal hypoglycemia may be a considerable problem during continuous subcutaneous insulin infusion therapy even if the basal infusion rate is adjusted in the hospital on the basis of nocturnal blood glucose concentrations. The ingestion of an extra snack at bedtime for capillary blood glucose values below 120 mg/dl has the potential to minimize this risk. The capillary blood glucose concentration at 10:30 P.M. is a significant predictor of nocturnal hypoglycemia.