RADIATION-INDUCED INJURY TO THE VISUAL PATHWAY

Radiation-induced injury to the visual pathway was reviewed in a cohor t of patients treated for various cancers of the nasal cavity and para nasal sinuses between 1969 and 1985. The study subjects consisted of 2 19 patients, 137 men and 82 women, in whom detailed records were avail able on the extent of orbital shielding, treatment plan for estimating doses delivered to various optic structures, and visual acuity follow -up information. There was a wide range in doses administered to vario us optic structures because patients with different primary lesion typ es were included and the radiotherapy techniques used varied during th is era. The endpoint of the study was visual acuity <20/100. The Cox p roportional hazard model was used to assess the influence of various f actors on the latent time to visual impairment and to fit the LQ model to the failure-time data. Actuarial curves showing the proportion of patients with visual impairment as a function of dose and time were ge nerated. Corneal injury occurred in 24 of the 49 patients treated with the 3-field technique without any orbital shielding; variables affect ing the incidence of cornea injury were total radiation dose and chemo therapy. Symptomatic retinopathy was diagnosed in 7 of 77 patients who received irradiation to a relatively large retinal surface, but no va riables were found to correlate with this complication. Eight patients developed ipsilateral blindness due to optic neuropathy and 11 patien ts had bilateral visual impairment secondary to chiasm injury. The tot al radiation dose was identified as the predominant determinant. None of patients receiving a dose of <50 Gy developed optic neuropathy or c hiasm injury. In contrast, the 10-year actuarial incidences of optic n erve-chiasm injury were similar to 5% and similar to 30% for patients receiving 50-60 Gy and 61-78 Gy, respectively. A fit of the LQ model y ielded an alpha/beta estimate of 1.6 Gy for optic neuropathy. Actuaria l dose-incidence curves were generated for different types of visual c omplications. This information would be useful in guiding rational sel ection of radiation dose in various subgroups of patients based on the extent of disease and risk of recurrence.