The use of molecular biomarkers in epidemiological investigations brings clear advantages of economy, speed and precision. Epidemiology--the study of the factors that control the patterns of incidence of disease--normally requires large numbers of subjects and/or long periods of time, because what is measured (the occurrence of disease) is a rare event. Biomarkers are measurable biological parameters that reflect, in some way, an individual's risk of disease-because they indicate exposure to a causative (or protective) agent, or because they represent an early stage in development of the disease, or because they allow an assessment of individual susceptibility. Biomarkers must be usable on one of the few materials available for biomonitoring of humans, i.e. blood, urine, exfoliated epithelial cells and, with some difficulty, biopsies. The approach of molecular epidemiology has a great potential is several areas of cancer research: investigating the aetiology of the disease; monitoring cancer risk in people exposed to occupational or environmental carcinogens; studying factors that protect from cancer; and assessing intrinsic factors that might predispose to cancer. The biomarkers most commonly employed in cancer epidemiology include: measurements of DNA damage--DNA breaks, altered bases, bulky adducts--in lymphocytes; the surrogate marker of chemical modifications to blood proteins, caused by agents that also damage DNA; the presence of metabolites of DNA-damaging agents (or the products of DNA damage themselves) in urine; chromosome alterations, including translocations, micronuclei and sister chromatid exchange, resulting from DNA damage; mutations in marker genes; DNA repair; and the differential expression of a variety of enzymes, involved in both activation and detoxification of carcinogens, that help to determine individual susceptibility. The molecular approach has been enthusiastically employed in several studies of occupational/environmental exposure to carcinogens. While the estimation of biological markers of exposure has certainly shown the expected effects in terms of DNA damage and adducts, the detection of the biological effects of exposure (e.g. at the level of chromosome alterations) has not been so clear-cut. This is true also when smokers are examined as a group compared with non-smokers. Several markers (especially of chromosome damage and mutation) show a strong correlation with age-indicating either an increasing susceptibility to damage with age, or an accumulation of long-lived changes. DNA repair--a crucial player in the removal of damage before it can cause mutation--may vary between individuals, and may be modulated by intrinsic or extrinsic factors, but limited data are available because of the lack of a reliable assay. Information on other enzymes determining individual susceptibility does exist, and some significant effects of phenotypic or genotypic polymorphisms have emerged, although the interactions between various enzymes make the situation very complex. The important question of whether oxidative DNA damage in normal cells is decreased by dietary antioxidants (vitamin C, carotenoids etc., from fruit and vegetables) has been tackled in antioxidant supplementation experiments. The use of poorly validated assays for base oxidation has not helped us to reach a definitive answer; it seems that, in any case, the level of oxidative damage has been greatly exaggerated. DNA-damaging agents lead to characteristic kinds of base changes (transitions, transversions, deletions). The investigation of the spectrum of mutations in cancer-related genes studied in tumour tissue should lead to a better understanding of the agents ultimately responsible for inducing the tumour. Similarly, studying mutations in a neutral marker gene (not involved in tumorigenesis) can tell us about the origins of the 'background' level of mutations. So far, interpretation of the growing databases is largely speculative. (ABSTRACT