We describe a randomized clinical trial of oral beta-carotene (50 mg/day) for preventing nonmelanoma skin cancer. It is a multicenter study conducted at sites in California, Minnesota, and New Hampshire. This report describes the design of the study, baseline characteristics of the 1805 randomized patients, changes in their plasma beta-carotene and retinol levels after 1 year of treatment, and plans for statistical analyses. Important features of this study are (1) a high proportion of potential subjects were found to be ineligible or chose not to enter the study, (2) the study agent is readily available over the counter and in common foods, and (3) nonmelanoma skin cancer is a relatively minor health concern for most patients. These considerations necessitated intensive efforts to encourage compliance with the study regimen. There are also some unusual statistical features of the study. One is that the study outcome is routinely assessed only at annual examinations, so the precise time of failure cannot be identified. Also, a secondary goal of the study is to determine whether beta-carotene decreases the average number of new skin cancers per patient per year, and there are no established statistical methods for analysis of data in this situation. Alternative approaches to the analysis are discussed.