Leukoplakia is associated with increased risk of oral cancer and is considered a premalignant lesion. Retinoids, particularly 13-cis retinoic acid, can frequently reverse leukoplakia. However, these drugs have considerable toxicity and are not suitable for large-scale use in the prevention of oral cancer. Beta-carotene is a naturally occurring, nontoxic carotenoid with biologic properties that suggest that it might be efficacious against oral leukoplakia. In 1986, we began a randomized study of 13-cis retinoic acid (1 mg/kg/d) versus beta-carotene (30 mg/d) in leukoplakia. However, owing to the marked differences in toxicity between the two compounds outlined in the consent form, 11 of the initial 16 eligible patients refused to participate unless they were "guaranteed" beta-carotene. Therefore, the study design was changed to a phase II trial of beta-carotene in which the compound was given daily for 3 months. Responding patients were continued for another 3 months of treatment. All lesions were examined histologically at entry. Responses were monitored by bidimensional measurements and photography done at entry, then monthly while on treatment and at study completion. Twenty-four evaluable patients were treated, and 17 had major responses (two complete, 15 partial), a response rate of 71% (95% confidence limits, 53% to 89%). There was no significant toxicity requiring drug discontinuation or dose reduction. These results indicate that beta-carotene has substantial activity in oral premalignancy. Because of its lack of toxicity, it is an excellent candidate for a preventive agent for oral cancer.