Our knowledge about the risk of hypoglycaemia associated with diabetes treatment is derived from studies that often exclude elderly people. Aim of this study was to determine the incidence and risk factors for developing severe hypoglycaemia among persons aged 80 yr or older, with Type 2 diabetes mellitus (T2DM). During a 2-yr period, all episodes of severe hypoglycaemia occurred in T2DM patients aged 80 yr or older were identified. Hypoglycaemia was defined as a symptomatic event requiring treatment with i.v. glucose and confirmed by a blood glucose determination of less than 50 mg/dl. A detailed history and blood laboratory profile were obtained for each patient. During the period of the survey a total of 124 diabetic subjects aged 80 yr or older were hospitalised and severe hypoglycaemia was reported in 31 patients (25%). This group of patients had a marked comorbidity and was found to have HbA1c values of 5.1% indicating that their diabetes was well controlled. Of these hypoglycaemic episodes, 23 (74%) occurred in patients taking glibenclamide. Diabetes therapy was prescribed by general practitioners in 24 of these patients. Seventeen subjects concomitantly received drugs that potentiated hypoglycaemia. Only 10 patients performed regular blood glucose self-monitoring. In conclusion, severe hypoglycaemia is a serious and not uncommon problem among elderly patients with T2DM; it is more frequent in patients undergoing aggressive diabetes management and in users of a long-acting sulphonylurea (eg, glibenclamide). A normal HbA1c level in this age group appears to be a powerful indicator of the risk of severe hypoglycaemia and should alert clinicians to change therapy. Finally, each patient's risk for hypoglycaemia should be considered and therapy should be individualised accordingly; in our opinion, a great number of episodes of serious hypoglycaemia may be prevented by teaching the principles of blood glucose monitoring and involving general practitioners in outpatient management of diabetes mellitus in the elderly.