Engelbrecht et al1 recently reported the visual outcomes in patients with subretinal neovascular membranes (NVMs) associated with type 2A idiopathic juxtafoveolar retinal telangiectasis (IJFRT). They concluded that the stable final visual acuity in eyes with NVMs was poor and that chorioretinal anastomosis and retinal pigment epithelial hyperplastic plaques always preceded the development of an NVM. In their citation of the staging system proposed by Blodi and me, 2 they modified stages 3 and 4 as follows: "Stage 3 is clinically characterized by the presence of right-angle–draining retinal venules or chorioretinal anastomosis. Stage 4 is characterized by the development of plaques of hyperplastic retinal pigment epithelium that are commonly associated with chorioretinal anastomosis."1(p321) These definitions together with their second conclusion mentioned above indicate that they considered right-angle venules to be indicative of chorioretinal anastomosis. This is contrary to our reported classification and to more recent evidence that the right-angle venules develop in response to changes in the outer plexus of retinal capillaries and proliferation of these capillaries into the outer retinal layers without involvement with the choroidal vessels.3 Type 2 subretinal neovascularization (SRN) (stage 5) may develop months or years after the development of right-angle venules. We concluded that SRN was primarily derived from the proliferation of retinal capillaries into the outer retina and was not derived from the choroidal circulation. Other more recent histopathological and clinical evidence supports this view.3,4 There is no clinical, fluorescein angiographic, or histopathological evidence that chorioretinal anastomosis develops during the first 4 stages of IJFRT. Chorioretinal anastomosis probably occurs infrequently in IJFRT and only during the late stages of development of large SRN complexes. Although Engelbrecht et al concluded that retinal pigment epithelial hyperplastic plaques always preceded the development of an NVM, they cite one case in which they did not, and I have seen several similar cases.