Carcinogen-DNA adducts are thought to be a useful biomarker in epidemiologic studies seeking to show that environmental exposures to xenobiotics cause cancer. This paper reviews the literature in this field from an epidemiologic perspective and identifies several common problems in the epidemiologic design and analysis of these studies. Carcinogen-DNA adducts have been used in studies attempting to link xenobiotic exposures to hepatocellular carcinoma, smoking related cancers and breast cancer. Adduct measurements have been useful in further implicating aflatoxin exposure in the etiology of hepatocellular carcinoma. For smoking related cancers, associations with carcinogen-DNA adducts are commonly seen in current smokers but less so in ex- or non-smokers. In breast cancer the associations have been inconsistent and weak and there is little evidence that carcinogen-DNA adducts implicate xenobiotic exposures in the etiology of breast cancer. Methodological issues common to these studies are the use of target versus surrogate tissues and how this choice impacts control selection, disease effects on adduct levels, the time period reflected by adduct levels, the use of inappropriate statistical analyses and small sample sizes. It is unclear whether the lack of association between carcinogen-DNA adducts and cancer reflects a lack of association between the xenobiotic exposure of interest and cancer or the effects of these methodological issues. A greater focus needs to be placed on designs that allow measurements of adduct levels in tissues collected years prior to cancer diagnosis, there is little need for further hospital based case-control studies in which adducts are measured at the time of or after diagnosis. New designs that address these issues are suggested in the paper.