Background: Insulin analogues have been developed to reduce the risk of hypoglycaemia in patients with diabetes who require insulin-based treatment, but their effect on this endpoint in patients with type 1 diabetes complicated by recurrent severe hypoglycaemia is unknown. We compared the occurrence of severe hypoglycaemic episodes in such patients during treatment with insulin analogues or human insulin. Methods: In this investigator-initiated, prospective, randomised, open-label, blinded-endpoint crossover trial at seven medical centres in Denmark, we recruited patients (aged ≥18 years) with type 1 diabetes (diagnosed for >5 years) who had reported two or more episodes of severe hypoglycaemia in the preceding year. Patients were randomly assigned (1:1) using computer-generated site-specific randomisation lists in blocks of four to treatment with basal-bolus therapy with either analogue insulin (detemir and aspart) or human insulin (human neutral protamine Hagedorn and human regular) in a balanced crossover design. A 1-year plus 1-year treatment period was specified, consisting of two 3-month run-in periods, each followed by a 9-month maintenance period. The primary endpoint was the number of validated episodes of severe hypoglycaemia (defined by need for treatment assistance from others) reported during the maintenance periods, analysed by intention to treat. The study is registered with ClinicalTrials.gov, number NCT00346996. Findings: Between May 9, 2007, and Oct 30, 2009, 159 patients were randomly assigned. 18 patients discontinued during the first run-in period, leaving 141 patients in the intention-to-treat population. 136 severe hypoglycaemic episodes were reported during treatment with human insulin and 105 episodes were reported during treatment with insulin analogues, resulting in an absolute rate reduction of 0·51 episodes (95% CI 0·19-0·84) per patient-year with insulin analogues. This result corresponds to a relative rate reduction of 29% (95% CI 11-48; p=0·010). Interpretation: Treatment with insulin detemir and aspart in patients with type 1 diabetes and recurrent severe hypoglycaemia resulted in a clinically significant reduced rate of severe hypoglycaemia compared with human insulin. Patients with the greatest chance of benefitting from improved insulin therapy should be offered treatment with insulin analogues and be included in future trials of new insulins. Funding: Novo Nordisk A/S.