The importance of glycaemic control for the development of proliferative retinopathy and nephropathy was assessed by monitoring glycated haemoglobin for 5 years or more before the diagnosis of these complications. The study comprised Type 1 (insulin-dependent) diabetic patients diagnosed at an age less than 31 years, and with diabetes duration 25 years or less. They were followed for an average of 7.9 years with 3.3 measurements per year. Of 172 patients screened for retinopathy 60 had no retinopathy, 104 had background retinopathy, and 8 had proliferative retinopathy The mean HbAlc (95% confidence intervals) of the groups was 6.4% (6.2-6.7%), 7.3% (7.1-7.5%) and 8.9% (8.1-9.6%), respectively (p < 0.0001); the mean duration of diabetes was 12, 18, and 17 years. Of 186 patients 7 had nephropathy (albuminuria > 200 mg/l). Mean HbAlc in patients without nephropathy was 7.0% (6.8-7.1%) and in patients with nephropathy 8.8% (7.8-9.9%, p < 0.001). Mean diabetes duration was 16 years in both groups. Multiple logistic regression including mean HbAlc, age at onset, duration, sex, and hypertension, was for both proliferative retinopathy and nephropathy significant only for mean HbAlc. In all cases, proliferative retinopathy and nephropathy were preceded by poor glycaemic control over several years, suggesting that these complications are caused by poor glycaemic control.