To study insulin dependent diabetic patients for change in non-proliferative retinopathy and its relation to glycaemic control and to various clinical background data. Prospective study with follow up for seven years. Outpatient departments of university hospitals. Glycated haemoglobin concentration; degree of retinopathy. Retinopathy worsened by an overall increase in counts of microaneurysms and haemorrhages from 17 (SD 25) to 45 (58) (p = 0.005). Intensified insulin treatment and home blood glucose monitoring improved concentrations of glycated haemoglobin (HbA1) from 11.2% (2.2%) at the start of the study to a mean of 9.5% (1.5%) over the seven years of the study (p less than 0.0001). A mean value for HbA1 greater than 10% was associated with an increased risk of progression of retinopathy and a mean value less than 8.7% was associated with a diminished risk. Multiple regression analysis identified four independent variables as indicative of outcome of retinopathy after seven years: HbA1 value at baseline; the change in HbA1 from start to the mean level through the seven years; duration of diabetes; and retinopathy at start. Age, blood pressure, and urinary albumin excretion were not related to the presence or progression of retinopathy. Secondary intervention by long term lowering of glycated haemoglobin has a beneficial impact on non-proliferative retinopathy. A four factor regression model can determine patients at high risk of severe retinopathy.