OBJECTIVE: To determine the impact of four bedtime (HS) snack compositions on nocturnal glycemic control, including frequency of hypoglycemia (<4 mmol/l) and morning hyperglycemia (>10 mmol/l), in adults with type 1 diabetes using lispro insulin before meals and NPH insulin at bedtime. RESEARCH DESIGN AND METHODS: Substitutions of 15 g carbohydrate (one starch exchange) for an equivalent amount of uncooked cornstarch or pure protein were compared to a standard snack (control: two starch + one protein exchange) and to no snack (placebo) in 15 adults using a randomized, cross-over design. All snacks were equivalent in kcal, fat, and total available glucose. An intravenous facilitated hourly blood glucose sampling during the night (11:00 P.M. to 7:00 A.M.). RESULTS: The glycemic level at bedtime (<7, 7-10, and >10 mmol/l) mediated the effects observed. A total of 14 hypoglycemic episodes, in 60% of patients, and 23 morning hyperglycemic episodes occurred over 50 nights. Most hypoglycemic episodes (10 of 14, 71%) occurred with no snack compared to any snack (P < 0.001) and at HS levels of <7 mmol/l (P = 0.05). The standard and protein snacks resulted in no nocturnal hypoglycemia at all HS glucose levels (P < 0.001). Only HS glucose >10 mmol/l was protective against hypoglycemia, even in the absence of a snack (P = 0.05); 46% of morning hyperglycemic episodes were associated (r = 0.37, P = 0.07) with this HS glucose level. CONCLUSIONS: The need for and composition of an HS snack depends on the HS glucose such that no snack is necessary at levels >10 mmol/l. At levels between 7 and 10 mmol/l, any snack is advised, and at <7 mmol/l, a standard or protein snack is recommended.