ObjectiveThe purpose of this study was to investigate if the levonorgestrel-impregnated intrauterine device (LNG-IUS, Mirena®) is safe and effective as therapy for low-risk and medium-risk endometrial hyperplasia compared with oral medroxyprogesterone (MPA).DesignA multicentre randomised trial.SettingNorway.PopulationIn all, 170 women aged 30–70 years with low- or medium-risk endometrial hyperplasia who met inclusion criteria.MethodsPatients were randomly assigned to one of three treatment arms: LNG-IUS; oral MPA 10 mg administered for 10 days per cycle, or continuous oral MPA 10 mg daily, for 6 months.Main outcome measuresThe primary outcome measure was normalisation or persisting hyperplasia.ResultsAfter 6 months all three treatment regimens showed significant effect when the outcome was evaluated as therapy response or not (P < 0.001). Responses were obtained for all the women in the LNG-IUS group (53/53, 95% CI 0.93–1.0) and for 96% of the women in the continuous oral group (46/48, 95% CI 0.86–0.99). Only 69% of the women in the cyclic oral group were responders (36/52, 95% CI 0.55–0.81). Adverse effects were relatively common with minimal differences between therapy groups.ConclusionIn the first trial of its kind, women treated with the LNG-IUS showed histologically normal endometrium after 6 months of therapy for endometrial hyperplasia. Cyclical progestogens are found to be less effective compared with continuous oral therapy and LNG-IUS and should not be used for this purpose.