PURPOSE: The purpose of this study was to investigate (i) the effect of diabetes on retinal ganglion cell death in diabetic dogs and mice, (ii) the effect of prolonged glycemic control on diabetes-induced death of retinal ganglion cells, (iii) whether retinal ganglion cell death in diabetes is associated with degeneration of retinal capillaries, and (iv) the effect of diet on diabetes-induced degeneration of retinal ganglion cells in mice. METHODS: Diabetes was induced in dogs using streptozotocin, and levels of glycemic control (good, moderate, and poor) were maintained for 5 years. Diabetes was studied in two mouse models (diabetes induced in C57Bl/6J mice using streptozotocin and spontaneously diabetic Ins2Akita mice). Retinal ganglion cell death was investigated by counting the number of axons from the ganglion cells in the optic nerve and with terminal transferase deoxyuridine triphosphate nick-end labeling and annexin V staining in mice. RESULTS: As reported previously, the development and severity of vascular lesions of diabetic retinopathy in diabetic dogs were strongly associated with glycemic control. Loss of retinal ganglion cells was extensive in dogs kept in poor glycemic control, and was essentially prevented in diabetic dogs kept in good glycemic control for the 5 years of study. In contrast, "moderate" glycemic control (intermediate between poor and good glycemic control) caused a significant increase in vascular pathology, but did not cause loss of retinal axons in the optic nerve. Using this validated optic nerve axon counting method, the two mouse models of diabetic retinopathy were studied to assess ganglion cell death. Despite 10 months of diabetes (a duration that has been shown to cause retinal capillary degeneration in both models), neither mouse model showed loss of optic nerve axons (thus suggesting no loss of retinal ganglion cells). Likewise, other parameters of cell death (terminal transferase deoxyuridine triphosphate nick-end labeling and annexin V labeling) did not suggest ganglion cell death in diabetic C57Bl/6J mice, and ganglion cell death was not increased by a different commercial diet. CONCLUSIONS: Retinal ganglion cell death in diabetic dogs is significantly inhibited by good or even moderate glycemic control. The finding that diabetic dogs in moderate glycemic control had appreciable vascular disease without apparent retinal ganglion cell degeneration does not support the postulate that neural degeneration causes the vascular pathology. Studies of diabetic mice in our colony again fail to find evidence of ganglion cell death due to prolonged diabetes in this species.