Molecular epidemiology combines biological markers and epidemiological observations in the study of the environmental and genetic determinants of cancer and other diseases. The potential advantages associated with biomarkers are manifold and include: (a) increased sensitivity and specificity to carcinogenic exposures; (b) more precise evaluation of the interplay between genetic and environmental determinants of cancer; (c) earlier detection of carcinogenic effects of exposure; (d) characterization of disease subtypes-etiologies patterns; (e) evaluation of primary prevention measures. These, in turn, may translate into better tools for etiologic research, individual risk assessment, and, ultimately, primary and secondary prevention. An area that has not received sufficient attention concerns the validation of these biomarkers as surrogate endpoints for cancer risk. Validation of a candidate biomarker's surrogacy is the demonstration that it possesses the properties required for its use as a substitute for a true endpoint. The principles underlying the validation process underwent remarkable developments and discussion in therapeutic research. However, the challenges posed by the application of these principles to epidemiological research, where the basic tool for this validation (i.e., the randomized study) is seldom possible, have not been thoroughly explored. The validation process of surrogacy must be applied rigorously to intermediate biomarkers of cancer risk before using them as risk predictors at the individual as well as at the population level.