The relationship between the progression of diabetic retinopathy and consecutive blood sugar control measured by HbA1 levels was statistically analyzed based on 422 eyes of 211 diabetic patients who were followed up 2 years or more. Cox regression (proportional hazard model) with HbA1 as a time-dependent covariate was applied and the relative risk of HbA1 vis-à-vis the progression of retinopathy was evaluated. For a given value of HbA1 at each time point during progression, the average for x preceding months was calculated and the value of x was changed from 1 to 30 to assess the cumulative effect of HbA1. The cumulative effect of HbA1 was statistically significant and the absolute value of the coefficient changed little if the periods for calculating the average are equal to or more than 6 months. Relative risks corresponding to the increase of HbA1 (averaged at 6 months) by 1, 2 and 3% were 1.18, 1.39 and 1.63, respectively. When the HbA1 level was adjusted to its average value (9.56%), estimated cumulative progression rates for 6, 12, 18 and 24 follow-up months were 3.6, 14.5, 22.5 and 30.5%, respectively. Difference of cumulative progression rate between a well-controlled group with a low (3% less than the average) HbA1 value and a poorly-controlled group with a high (3% higher than the average) HbA1 value was estimated as 3% (6 follow-up months) and 23% (24 months).