Cytomegalovirus (CMV) retinitis is the most common intraocular infection in patients with AIDS and affects an estimated 20% of these patients. Two drugs, ganciclovir and foscarnet, are currently available for the treatment of CMV retinitis. The Foscarnet-Ganciclovir CMV Retinitis Trial was designed: (1) to evaluate the relative efficacy and safety of foscarnet and ganciclovir for the treatment of CMV retinitis in patients with AIDS; and (2) to compare the relative benefits of immediate treatment versus deferral of the treatment for patients with disease not involving the posterior pole. Prior to randomization, patients were assigned to one of two strata based upon the location and extent of retinitis in the more severely involved eye. Patients with retinitis in zone 1 (posterior pole) or extensive retinitis (extent greater than or equal to 25% retina) in zones 2 and/or 3 (peripheral retina) were randomly assigned to immediate treatment with either ganciclovir or foscarnet. Patients with CMV retinitis confined to less than or equal to 25% of the retina and in zones 2 and/or 3 were offered the option of participating in the comparison of immediate treatment versus deferral of treatment. Patients opting to participate in this comparison were randomly assigned to immediate treatment versus deferral. Patients preferring to choose either immediate treatment or deferral (treatment preference design) were randomized only to foscarnet or ganciclovir for treatment of CMV retinitis. Patients in the deferral group were started on drug treatment when the retinitis became more immediately sight-threatening either by virtue of location (involvement of zone 1) or size (extent greater than 25% of the retina). Outcome measures included: survival, retinitis progression, visual function (visual acuity and visual field), drug side effects, and morbidity. Enrollment in the trial began in March 1990 and was completed in October 1991.