OBJECTIVES: The present study tests two interrelated hypotheses: (1) that bedtime ingestion of uncooked cornstarch exerts a lower and delayed nocturnal blood glucose peak compared with a conventional snack; (2) that bedtime carbohydrate supplement, administered as uncooked cornstarch, prevents nocturnal hypoglycaemia without altering metabolic control in intensively treated type 1 diabetes (IDDM) patients. DESIGN AND SUBJECTS: The above hypotheses were tested separately (1) by pooling and analysing data from two overnight studies of comparable groups of patients with non-insulin dependent diabetes mellitus (NIDDM) (14 and 10 patients, respectively), and (2) by a double-blind, randomized 4-week cross-over study in 12 intensively treated IDDM patients. SETTING: Sahlgrenska University Hospital, Göteborg. Sweden. INTERVENTIONS: (1) Ingestion of uncooked cornstarch and wholemeal bread (0.6 g of carbohydrates kg-1 body weight) and carbohydrate-free placebo at 22.00 h. (2) Intake of uncooked cornstarch (0.3 g kg-1 body weight) and carbohydrate-free placebo at 23.00 h. MAIN OUTCOME MEASURES: (1) Nocturnal glucose and insulin levels; (2) frequency of self-estimated hypoglycaemia (blood glucose [BG] levels