The longer acting basal insulin analogs glargine and detemir have shown a lower incidence of hypoglycemia compared to insulin NPH in clinical studies. We evaluated the real-life risk of severe hypoglycemia among new users of insulins in the working-age population in Finland.All persons aged 18-65 years with diabetes mellitus who were newly prescribed with insulins NPH, glargine, or detemir during 2006-2009, were identified from national registers. Risk of severe hypoglycemia requiring hospital care was compared between insulin types.A total of 16,985 persons initiated basal insulin treatment (5586, 7499, and 3900 patients started NPH, glargine, and detemir, respectively) during follow-up. Five hundred and thirty-six persons were hospitalized because of severe hypoglycemia. Absolute rate (per 1000 patient-years) was 20.6 (95% CI 17.9, 23.8), 17.8 (15.6, 20.3), and 12.4 (9.9, 15.5) for NPH, glargine, and detemir initiators, respectively. With NPH as reference, the adjusted hazard ratio (HR) was 0.92 (95% CI 0.74, 1.15, p = 0.47) for glargine, and 0.70 (0.51, 0.94, p= 0.018) for detemir. The HR for detemir compared to glargine was 0.76 (0.58, 0.99, p = 0.040).Initiating insulin treatment with detemir, but not with glargine, was associated with a significantly lower risk of severe hypoglycemia compared to NPH, among working-age adults. KEY MESSAGES The comparative safety of modern basal insulins regarding hypoglycemia among the working-age population is unclear. Large reductions in the incidence of severe hypoglycemia were seen among real-life patients who started insulin detemir, as compared to patients who initiated glargine or especially NPH insulin. Given the large amount of patients using insulin, these findings may have considerable clinical consequences at the population level.