IMPORTANCE Threshold of retinal nonperfusion for the development of proliferative diabetic retinopathy (PDR) is unclear.OBJECTIVES To identify a threshold of retinal nonperfusion for the presence of retinal neovascularization and the distribution and area of retinal nonperfusion in eyes with severe nonproliferative diabetic retinopathy (NPDR), PDR, neovascularization of the optic disc (NVD), and retinal neovascularization elsewhere (NVE). DESIGN, SETTING, AND PARTICIPANTS This cross-sectional image analysis study was performed between September 24, 2018, and October 24, 2018, at a multicenter national study in the United Kingdom. Baseline images were obtained from 2 completed randomized clinical trials (Ranibizumab for Diabetic Macular Edema Panretinal Photocoagulation [RDP] study and Clinical Efficacy of Intravitreal Aflibercept vs Panretinal Photocoagulation for Best Corrected Visual Acuity in Patients With Proliferative Diabetic Retinopathy at 52 Weeks [CLARITY] study). The RDP study recruited eyes with severe NPDR between April 1, 2014, and December 31, 2015, and the CLARITY study recruited eyes with PDR between August 22, 2014, and November 20, 2015. Ultra-widefield angiography images of eyes with no prior panretinal photocoagulation treatment were included. MAIN OUTCOMES AND MEASURES The total area of retinal nonperfusion, the area of posterior pole retinal nonperfusion, and the area of peripheral retinal nonperfusion were measured. RESULTS A total of 92 patients (92 eyes) were included in the study: 59 in the PDR group (mean [SD] age, 42 [15] years; 20 female [33.9%]) and 33 in the NPDR group (mean [SD] age, 63 [10] years; 3 female [9.1%]). Forty eyes had NVE and 19 had NVD with or without NVE.We identified a retinal nonperfusion threshold of 118.3 disc areas (DA) with a specificity of 84.9% (95% CI, 68.1% to 94.9%) for PDR. The median area of retinal nonperfusion was 67.8 DA (95% CI, 44.2 to 107.3 DA) in the NPDR eyes and 147.9 DA (95% CI, 127.4 to 173.5 DA) for eyes with proliferative changes, with a difference of 69.0 DA (95% CI, 42.2 to 97.7 DA; P < .001). No difference was found in the median area of posterior nonperfusion between NPDR and PDR, with a difference of 0 DA (95% CI, −6.7 to 5.2 DA; P = .56). As for peripheral nonperfusion, NPDR eyes measured 64.1 DA and PDR eyes measured 130.6 DA, with a difference of 70.8 DA (95% CI, 48.4 to 94.9 DA; P < .001). Eyes with NVD had the largest total area of retinal nonperfusion, with a difference of 65.1 DA (95% CI, 28.6 to 95.8 DA; P < .001) compared with eyes with only NVE.CONCLUSIONS AND RELEVANCE These findings suggest eyes with at least 107.3 DA of nonperfusion are at risk of proliferative disease, and eyes with NVD have the largest area of retinal nonperfusion.