Epidemiologic observations and preclinical experimental investigations suggest that the prevention or reversal of precancers should be an effective strategy in humans to control cancer. Although "proof of principle" has been established in humans, the results of randomized trials have not been confirmatory in most cases. Toxicity in normal or near-normal populations has also been greater than anticipated. We examine the problems associated with testing chemoprevention agents in humans and offer a process and guidelines that may better inform the logical development of this relatively young clinical field.