• Six hundred sixteen eyes with recent severe diabetic vitreous hemorrhage reducing visual acuity to 5/200 or less for at least one month were randomly assigned to either early vitrectomy or deferral of vitrectomy for one year. After two years of follow-up, 25% of the early vitrectomy group had visual acuity of 10/20 or better compared with 15% in the deferral group (P =.01). In patients with Type I diabetes, who were on the average younger and had more-severe proliferative retinopathy, there was a clear-cut advantage for early vitrectomy, as reflected in the percentage of eyes recovering visual acuity of 10/20 or better (36% vs 12% in the deferral group, P =.0001). No such advantage was found in the Type II diabetes group (16% in the early group vs 18% in the deferral group), but evidence that this advantage differed by diabetes type was of borderline significance.