Internal review board approval was obtained for review of medical records for patients with retinoblastoma undergoing chemosurgery between May 30, 2006, and August 11, 2009. This technique was offered as an alternative treatment in lieu of enucleation. Two of us (B.P.M. and C.H.) in a masked fashion graded 64 initial orbital angiograms from 56 patients. One of us (Y.P.G.) had performed all angiography and intra-arterial procedures in a standard manner. Prechemotherapy injection angiograms were compared with postchemotherapy injection angiograms to confirm that no variations occurred during injections and that the angiograms were consistent before and after drug delivery. The major orbital arteries were graded as present, absent, or prominent on angiograms (Figure, A). Observed were the supratrochlear, anteroethmoid, posteroethmoid, middle meningeal, supraorbital, muscular, lacrimal, posterociliary, and centroretinal arteries. Clinical photographs, drawings, and notes were evaluated for tumor response and vitreous seed response. Grading was based on the tumor volume or vitreous seed volume seen clinically at the beginning of chemosurgery and at the last follow-up examination. Low response corresponded to a 0% to 33% reduction in the initial clinical tumor volume or vitreous seed volume, medium as a 34% to 66% reduction, and high as at least a 67% reduction. Capillary filling patterns seen on angiograms were graded as present, absent, or prominent (Figure, B). Patient data were reviewed, and angiographic findings were compared between groups of patients who had and had not had tumor response, vitreous seed response, tumor recurrence, and enucleation. The association of previous (non–intra-arterial) treatment and vessel and blush data with disease response (low vs medium or high) was examined using the Fisher exact test. The association of previous treatment and vessel and blush data with time to recurrence or time to enucleation was examined using the log-rank test.