Purpose: To examine and illustrate the properties of recently developed statistical models for clinical trials having time to event outcomes in both eyes. Methods: Clinical trials with different study designs were assessed by statistical models that formulate the marginal distribution of the event times with the familiar Cox proportional hazards model. Designs and clinical trials included: Case 1: Two eyes enrolled for all patients, different treatments to each eye; Diabetic Retinopathy Study, Glaucoma Laser Trial; Case 2: Two eyes enrolled for each patient, same treatment to both eyes, patients randomly assigned to treatments; Studies of the Ocular Complications of AIDS; Case 3: One or two eyes enrolled for each patient, different treatments to each eye within two-eye patients, one-eye patients randomly assigned to treatments; Choroidal Neovascularization Prevention Trial. Results: These methods offer several desirable features: 1) estimation of the magnitude of risk associated with both treatment variables and other risk factors; 2) avoidance of underestimating (two eyes, different treatments) or overestimating (two eyes same treatment) the statistical significance of the treatment effect; 3) ability to adjust for other risk factors that might influence the treatment effect; 4) ablility to include subjects who have one or two eyes included in the study. Software to perform these analyses (SAS,S) is available and relatively easy to implement. Conclusion: Recently developed methods allow greater flexibility in analysing data survival data involving two eyes.
|Original language||English (US)|
|Journal||Investigative Ophthalmology and Visual Science|
|State||Published - Feb 15 1996|
ASJC Scopus subject areas
- Sensory Systems
- Cellular and Molecular Neuroscience