AIMS: Risk of Type 2 diabetes varies by ethnicity, but whether ethnicity remains important among those who have impaired glucose tolerance or impaired fasting glucose is uncertain. Whether the effect of thiazolidinedione treatment on diabetes prevention in persons with non-diabetic dysglycaemia varies by ethnicity is also not known. We addressed these questions using data collected in the DREAM trial. METHODS: A 2-by-2 factorial double-blind randomized controlled trial to compare the effects of rosiglitazone and ramipril on the primary outcome of diabetes or death in persons meeting criteria for impaired glucose tolerance or impaired fasting glucose. The effect of these interventions by ethnicity was estimated using Cox regression analysis. RESULTS: Of 5269 adults, 2365 were randomly assigned to rosiglitzone and 2634 to placebo. South Asians showed a higher hazard for the primary outcome compared with Europeans (hazard ratio, 95% confidence interval 2.21, 1.41-3.47) adjusted for age, gender, BMI, waist-hip ratio and geographic region. A lesser increase in risk was seen in Black people (1.37, 1.04-1.81). A significant reduction in risk of the primary outcome with rosiglitazone treatment assignment was seen in all ethnic groups, but the treatment effect significantly differed by ethnicity (P=0.0242), with South Asians experiencing a smaller, and Latinos a larger preventive effect. CONCLUSIONS: Ethnicity is an important risk factor for Type 2 diabetes in dysglycaemic persons. All ethnic groups experienced a large significant reduction in diabetes risk because of rosiglitazone. The magnitude of this reduction differed by ethnicity. Given the post hoc nature of this analysis, further confirmation of these findings is needed.