Introduction: In the struggle for tight glycemic control many type 1 diabetes patients daily experience large blood glucose fluctuations. It has been demonstrated that increased glucose variability is a predictor of hypoglycemia, which is the main limiting factor for achieving the recommended HbA1c target. The relationship between glucose variability and hypoglycemia might seem intuitive, but the exact nature of the relationship is difficult to elucidate. In this study we wanted to clarify the relationship between occurrence of non-severe hypoglycemia and glycemic variability in type 1 diabetic patients at high risk of severe hypoglycemia. Material and Methods: 72 patients participated for 4 x 3 days of blinded CGM recordings during a 2-year randomized balanced cross-over study with human insulins (regular/NPH) and insulin analogs (aspart/detemir). CGM recordings were used to calculate the coefficient of variation (CV) in glucose. Results: Treatment with insulin analogs resulted in a significant lower nocturnal CV (mean 29.2%, SD 9.1) than treatment with human insulin (mean 36.9%, SD 9.1) (p<0.0001). CV was a significant risk marker of nocturnal asymptomatic hypoglycemia ?3.9 mmol/l in a multivariate analysis. A 1% increase in CV corresponds to a 5% (p<0.0001) increase in asymptomatic hypoglycemia. CV did not predict hypoglycemia during daytime at this threshold. Conclusion: Treatment with insulin analogs provides less glycemic variability during night-time. The coefficient of variation in patients suffering from recurrent severe hypoglycemia predicts the occurrence of especially nocturnal hypoglycemia.