In 1982 we proposed the concept and a framework for implementing molecular cancer epidemiology. Here, we review progress during the past 17 years in validating and applying this approach to cancer prevention. There have been major advances, notably in the understanding of environment-susceptibility interactions in human cancer. However, a review of major findings to date reveals several urgent research needs to keep pace with the rapid evolution in knowledge of mechanisms in carcinogenesis. Although much valuable progress continues to be made in the study of carcinogens that cause direct DNA damage and are mutagenic, exogenous and endogenous carcinogens can also act by altering gene expression, cell proliferation and differentiation. The mechanisms include aberrant DNA methylation, oxidative damage, effects on metabolism of nitrogen oxide and nitrites, activation of receptors and transcription factors, cyclins and other cell cycle proteins. Sensitive, validated biomarkers are needed to detect these mechanisms in small numbers of cells, tissues or fluids. There is also increasing recognition that individual risk from carcinogen exposure varies as a function of both inherited and acquired factors. Recent advances in genomics, microassay technologies and informatics hold promise for rapid identification of polymorphic variants or changes in expression of genes influencing both response and susceptibility to carcinogens. Another emerging area of molecular epidemiology concerns the role of nutrition and specific dietary factors (including studies on antioxidants, energy metabolism, insulin and various growth factors) and the modulating effect of genetic polymorphisms. Finally, molecular epidemiology has enormous potential in cancer prevention through the early identification of 'at risk' populations and the rapid assessment of intervention efficacy. Its success in fully reaching this potential will depend on the application of validated biomarkers, with adherence to sound epidemiologic and ethical principles.