In this multicentre double-blind study, 100 insulin-treated diabetics with background retinopathy were randomly assigned to treatment with either 250 mg ticlopidine b.i.d. (49 patients) or placebo (51 patients). The primary aim of the study was to assess the evolution of retinopathy by fluorescein angiography performed annually for at least 3 years. The metabolic control of diabetes was evaluated by quarterly assessment of the haemoglobin A1 level. Safety parameters, especially haematologic variables were closely monitored. The proportion of patients with favourable results was higher in the ticlopidine group (55.2%) than in the placebo group (36.6%). However, this difference did not reach statistical significance (p = 0.123), most probably because of the too limited number of patients studied. In the placebo group, a significant relationship was found between the unfavourable evolution of diabetic retinopathy and more fluctuating haemoglobin A1 levels. Such a relationship was not observed in the ticlopidine group. This could be regarded as an expression of the therapeutic effect of the compound. The number and nature of side-effects was similar in both treatment groups. However, the proportion of patients who prematurely discontinued study treatment because of side-effects was significantly higher in the ticlopidine group. Mean levels of biological, and in particular haematologic, parameters did not change significantly except for a significant increase in the cholesterol level in the ticlopidine group, which, however, remained within normal limits.