Quantitation of the earliest changes in abnormal retinal morphology using fluorescein angiography is potentially superior to retinal photography. However, the critical importance of flawless technique, limitations in the size of the field available for detailed study, and observer variability constitute major disadvantages. A protocol describing standards of photography and of injection was developed. Methodology for counting microaneurysms (Ma) was developed at a central laboratory and applied in suitable photographs obtained at 0 (baseline), 4, and 8 mo in 68 patients. Counts of "definite" or "possible" Ma were made on films projected under standard conditions by two observers known to achieve consistently reproducible results. Semiquantitative assessment of diffusibility of fluorescein reflecting capillary leakage was performed in 61 patients. Leakage was graded according to three degrees of severity permitting study of observer variation, concordance of change in pairs of eyes, and treatment effects. The average number of Ma at baseline was slightly higher in the continuous subcutaneous insulin infusion (CSII) group than in the conventional insulin treatment (CIT) group, but the difference was not statistically significant at the 0.05 level. At both 4 and 8 mo, definite Ma were more prevalent in the CSII group, and the difference was statistically significant at the 0.05 level using both parametric and nonparametric tests. In 27 CSII and 23 CIT patients having complete sets of 0, 4, and 8-mo photographs, Ma counts increased during the 0-4-mo interval. During the 4-8 mo interval, a further increase occurred in the CSII group but contrasted with a decrement observed in the CIT group, which showed no net change from baseline.(ABSTRACT TRUNCATED AT 250 WORDS)