Objective: To investigate the relationship between blood glucose-lowering therapy, glycaemia and ethnicity in urban Australians with type 2 diabetes.Design: Prospective observational community-based study of diabetes care, control and complications. Methods: We analysed cross-sectional data from 1057 patients, 238 from a southern European (SE) migrant background and 819 Anglo-Celts (AC). Follow-up data were available for 5 3 9 patients (113 SE, 426 AC) who had annual reviews over 4 years.Results: The SE patients were of similar age to the AC patients but had longer diabetes duration, were less fluent in English and had less formal education. After adjustment for diabetes duration, glycosylated haemoglobin and glutamic acid decarboxylase antibody positivity in a logistic regression model, insulin use at study entry was approximately twice as frequent amongst SE as AC patients (odds ratio (95% confidence interval); 1.90 (1.20-3.02)). In the prospective arm, progression to insulin increased in both groups, from 18.0% at baseline to 22.1% at 4 years in SE and from 7.1% to 14.4% in AC patients. beta-cell function (%B) and insulin sensitivity ('%S) using the homoeostasis model assessment in a subset of diet-treated patients at baseline showed that SE ethnicity was associated with lower %B and greater %S than in the AC group.Conclusions: SE patients with early non-antibody-mediated beta-cell failure progress to insulin requirement within the first 4-5 years of type 2 diabetes. This could reflect either a longer period of undiagnosed diabetes or a more rapid loss of beta-cell function after diagnosis.