Abstract: Introduction: The ocular lesions of Behcet's disease (BD) progress usually toward severe loss of vision or blindness, and therefore need aggressive treatment. However, even such a treatment is not always successful. We used classification and regression tree (cart) method to see if some factors could predict the result of the future treatment. Methods: Eight hundreds and thirty patients who fulfilled the Iran criteria for BD were selected for this study. They had all active posterior uveitis and/or retinal vasculitis. They were treated with an association of one cytotoxic drug (cyclophosphamide, methotrexate, chlorambucil, or cyclosporine A) and prednisolone for at least 12 months. A good outcome was defined as an eye with a visual acuity (VA) of 0.4 or better on the Snellen chart, in a patient who improved his Total Adjusted Disease Activity Index (VA, anterior and posterior uveitis, retinal vasculitis) by 20% or more from the base line. A bad outcome was the opposite, and a null result was between the two. Patients classified as having a good outcome were 420, and a bad outcome 125. Eighty six parameters were selected for the CART analysis. The tree was grown to13 terminal nodes and then pruned back to 3 terminal nodes. Results: The (CART) has 3 terminal nods (1 good, 2 bad). Specificity: 89% ↓————VA<6<10————↓ Sensitivity: 32% Accuracy: 61% ↓——Genital aphthosis——↓ Bad Confidence interval: 4.04 Good Bad The tree predicting the outcome of ocular lesions, after systemic treatment with cytotoxic drugs and prednisolone. Us very dimple. If before the treatment an eye has a VA inferior to 0.6 (6/10) on the Snellen chart the outcome will be bad. For the eyes with a VA better than 0.6 the presence of genital aphthosis may predict a bad outcome. Conclusion: With a 61% accuracy (6 times on 10) the tree may predict the outcome of ocular lesions. The power to predict a bad outcome is much higher (89%) than a food outcome (32%).