To determine if culturally appropriate health education is more effective than 'usual' health education for people with diabetes from ethnic minority groups living in high- and upper-middle-income countries. A systematic review with meta-analysis, following the methodology of the Cochrane Collaboration. Electronic literature searches of nine databases were made, with hand searching of three journals and 16 author contacts. The criteria for inclusion into the analysis were randomized controlled trials of a specified diabetes health education intervention, and a named ethnic minority group with Type 2 diabetes. Data were collected on HbA(1c), blood pressure, and quality-of-life measures. A narrative review was also performed. Few studies fitted the selection criteria, and were heterogeneous in methodologies and outcome measures, making meta-analysis difficult. HbA(1c) showed an improvement at 3 months [weighted mean difference (WMD) -0.32%, 95% confidence interval (CI) -0.63, -0.01] and 6 months post intervention (WMD -0.60%, 95% CI -0.85, -0.35). Knowledge scores also improved in the intervention groups at 6 months (standardized mean difference 0.46, 95% CI 0.27, 0.65). There was only one longer-term follow-up study, and one formal cost-effectiveness analysis. Culturally appropriate health education was more effective than 'usual' health education in improving HbA(1c) and knowledge in the short to medium term. Due to poor standardization between studies, the data did not allow determination of the key elements of interventions across countries, ethnic groups and health systems, or a broad view of their cost-effectiveness. The narrative review identifies learning points to direct future research.