A retrospective analysis of the University of Minnesota (Minneapolis) experience with retinoblastoma is presented. Seventy-five patients were diagnosed with retinoblastoma between 1958 and 1983, of which 53 (71%) had at least one Reese-Ellsworth group V eye. Nineteen group V patients and one group II patient developed extraocular disease recurrence. The cumulative actuarial rate of recurrence at 12 years was 36% for patients with group V disease. The median time from diagnosis to recurrence for unilateral patients was seven months and for bilateral patients 28 months (P = .001). Patients developing extraocular disease had a 10-year actuarial survival rate postrecurrence of 34%. The four long-term survivors of extraocular recurrences had had isolated orbital or local soft tissue recurrences only. Features of group V patients associated with extraocular recurrences were identified by univariate life table analyses. Clinical poor-risk factors included the nongenetic form of the disease (P = .03) and male sex (P = 0.2). Pathologic poor risk factors included rubeosis (P = .01), undifferentiated histology (P = .03), large tumor size (P = .05), and intraocular extension to the anterior segment (P = .02), retinal pigment epithelium (P = .03), choroid (P < .001), and optic nerve beyond the lamina cribrosa (P = .02). Treatment-associated poor-risk factors included an optic nerve length of < 5 mm removed at enucleation (P = .003). Multivariate life table analyses demonstrated the following parameters to be independent poor-prognostic factors: optic nerve length of < 5 mm removed at enucleation (P = .001), optic nerve involvement (P = .004), and large tumor size (P = .01). These results will help to identify patients with retinoblastoma who are at greatest risk for extraocular recurrence.