OBJECTIVES--To evaluate integrated care for diabetes in clinical, psychosocial, and economic terms. DESIGN--Pragmatic randomised trial. SETTING--Hospital diabetic clinic and three general practice groups in Grampian. PATIENTS--274 adult diabetic patients attending a hospital clinic and registered with one of three general practices. INTERVENTION--Random allocation to conventional hospital clinic care or integrated care. Integrated care patients seen in general practice every three or four months and in the hospital clinic annually. General practitioners were given written guidelines for integrated care. MAIN OUTCOME MEASURES--Metabolic control, psychosocial status, knowledge of diabetes, beliefs about control of diabetes, satisfaction with treatment, disruption of normal activities, numbers of consultations and admissions, frequency of metabolic monitoring, costs to patients and NHS. RESULTS--A higher proportion of patients defaulted from conventional care (14 (10%)) than from integrated care (4 (3%), 95% confidence interval of difference 2% to 13%). After two years no significant differences were found between the groups in metabolic control, psychosocial status, knowledge, beliefs about control, satisfaction with treatment, unscheduled admissions, or disruption of normal activities. Integrated care was as effective for insulin dependent as non-insulin dependent patients. Patients in integrated care had more visits and higher frequencies of examination. Costs to patients were lower in integrated care (mean 1.70 pounds) than in conventional care (8 pounds). 88% of patients who experienced integrated care wished to continue with it. CONCLUSIONS--This model of integrated care for diabetes was at least as effective as conventional hospital clinic care.