Quantifiable, well-characterized cancer risk factors demonstrate the need for chemoprevention and define cohorts for chemopreventive intervention. For chemoprevention, the important cancer risk factors are those that can be measured quantitatively in the subject at risk. These factors, called risk biomarkers, can be used to identify cohorts for chemoprevention. Those modulated by chemopreventive agents may also be used as endpoints in chemoprevention studies. Generally, the risk biomarkers fit into categories based on those previously defined by Hulka: 1) carcinogen exposure, 2) carcinogen exposure/effect, 3) genetic predisposition, 4) intermediate biomarkers of cancer, and 5) previous cancers. Besides their use in characterizing cohorts for chemoprevention trials, some risk biomarkers can be modulated by chemopreventive agents. These biomarkers may be suitable surrogate endpoints for cancer incidence in chemoprevention intervention trials. The criteria for risk biomarkers defining cohorts and serving as endpoints are the same, except that those defining cohorts are not necessarily modulated by chemopreventive agents. A primary criterion is that the biomarkers fit expected biological mechanisms of early carcinogenesis-i.e., differential expression in normal and high-risk tissue, on or closely linked to the causal pathway for the cancer, and short latency compared with cancer. They must occur in sufficient number to allow their biological and statistical evaluation. Further, the biomarkers should be assayed reliably and quantitatively, measured easily, and correlated to cancer incidence. Particularly important for cancer risk screening in normal subjects is the ability to use noninvasive techniques that are highly specific, sensitive, and quantitative. Since carcinogenesis is a multipath process, single biomarkers are difficult to correlate to cancer, as they may appear on only one or a few of the many possible causal pathways. As shown in colorectal carcinogenesis, the risks associated with the presence of biomarkers may be additive or synergistic. That is, the accumulation of genetic lesions is the more important determinant of colorectal cancer compared with the presence of any single lesion. Thus, batteries of biomarker abnormalities, particularly those representing the range of carcinogenesis pathways, may prove more useful than single biomarkers both in characterizing cohorts at risk and defining modulatable risks. Risk biomarkers are already being integrated into many chemoprevention intervention trials. One example is the phase II trial of oltipraz inhibition of carcinogen-DNA adducts in a Chinese population exposed to aflatoxin B1. Also, urine samples from subjects in this trial will be screened for the effect of oltipraz on urinary mutagens. A second example is a chemoprevention protocol developed for patients at high risk for breast cancer; the cohort is defined both by hereditary risk and the presence of biomarker abnormalities. Modulation of the biomarker abnormalities is a proposed endpoint. Also, dysplastic lesions, such as prostatic intraepithelial neoplasia, oral leukoplakia and colorectal adenomas, have been used to define high-risk cohorts and as potential modulatable surrogate endpoints in chemoprevention trials.