Sources of bias, effect of confounding in the application of biomarkers to epidemiological studies

This article addresses some methodological aspects of the application of biomarkers of exposure, effect and susceptibility to cancer epidemiology. The application of biomarkers to cancer epidemiology should enhance the validity of exposure and outcome measurement, and strengthen the statistical association between exposure and diseases, thus reducing the possibility of random and systematic error. However, the use of biomarkers provides in turn new opportunities for bias and confounding. Small sample size is a limitation of many molecular epidemiology studies. Three major types of bias are recognized: selection bias, information bias (measurement error), and confounding. An important aspect of confounding is that biomarkers can be seen both as exposure and outcome. A second aspect of confounding lies in the role of the biological marker in the causal pathway between exposure and disease. Molecular epidemiology offers better opportunity for the elucidation of interactions between genetic and environmental factors. Genetic susceptibility studies should demonstrate the highest disease risk for exposed susceptible groups and the lowest risk for non-susceptible unexposed groups.