To analyse the results of diabetic vitrectomy according to the severity of proliferation [severe (SPG) vs. less-severe proliferation group (LSPG)], and methods of the operation, which was complete removal of anteroposterior vitreous traction with or without complete removal of preretinal memebrane, we compared both groups by using anatomic success rate and postoperative visual acuities (VA). The results were as follows: The anatomic success rate and postoperative VA were significantly better in LSPG than in SPG. In SPG, anatomic success rate and postoperative VA tended to be better when complete removal of anteroposterior traction was possible than when impossible. In SPG, postoperative VA tended to be better when complete removal of preretinal membrane was possible, but the anatomic success rate was the same for each group. So, when severe proliferation (including table-top elevation of posterior retina), complete removal of anteroposterior traction only can improve the anatomic success rate of the surgery.