There is still no consensus on the natural history and optimal management of optic pathway gliomas (OPG) in children. In order to tackle optimal management issues, we need to clearly understand the prognostic and confounding factors affecting relapse of OPG. We propose the use of the Cox proportional hazards (PH) model in a retrospective study of childhood OPG of 69 children seen from 1977 to 1994. We have developed a comprehensive model capable of multivariate analyses and handling time-dependencies. Our studies showed that relapse-free survival improves with increasing age, the presence of neurofibromatosis 1 (NF1), and chemotherapy and radiotherapy (p < 0.0005). Sex, tumor position and surgery do not significantly affect survival. Older children with NF1 have extremely good prognosis. We noted behavior that departs from the strictly proportional hazards model, but results were inconclusive.