AIMS/HYPOTHESIS: The Diabetes Reduction Assessment with Ramipril and Rosiglitazone Medication (DREAM) trial reported that 3 years of therapy with rosiglitazone reduced the primary outcome of diabetes or death by 60%. Here we investigated whether an effect on diabetes prevention persists more than 1.5 years after therapy has been discontinued. METHODS: The DREAM On passive follow-up study was conducted at 49 of the 191 DREAM sites. Consenting participants were invited to have a repeat OGTT 1-2 years after active therapy ended. A diagnosis of diabetes at that time was based on either a fasting or 2 h plasma glucose level of >/=7.0 mmol/l or >/=11.1 mmol/l, respectively, or a confirmed diagnosis by a non-study physician. Regression to normoglycaemia was defined as a fasting and 2 h plasma glucose level of <6.1 mmol/l and <7.8 mmol/l, respectively. RESULTS: After a median of 1.6 years after the end of the trial and 4.3 years after randomisation, rosiglitazone participants had a 39% lower incidence of the primary outcome (hazard ratio [HR] 0.61, 95% CI 0.53-0.70; p < 0.0001) and 17% more regression to normoglycaemia (95% CI 1.01-1.34; p = 0.034). When the analysis was restricted to the passive follow-up period, a similar incidence of both the primary outcome and regression was observed in people from both treatment groups (HR 1.00, 95% CI 0.81-1.24 and HR 1.14, 95% CI 0.97-1.32, respectively). Similar effects were noted when new diabetes was analysed separately from death. Ramipril did not have any significant long-term effect. CONCLUSIONS/INTERPRETATION: Time-limited exposure to rosiglitazone reduces the longer term incidence of diabetes by delaying but not reversing the underlying disease process.