Methodological issues in pooled analysis of biomarker studies.

The number of epidemiological studies involving biological markers has dramatically increased in the last years. These studies are generally small sized and this feature has called attention to the need to summarize the individual results, while waiting for the completion of larger studies, designed to answer questions that have been raised by preliminary studies. We describe here some of the methodological issues related to pooling data of biomarker studies, taking advantage of the experience accumulated by conducting two pooled analyses, one of studies of metabolic gene polymorphisms and cancer, the other on cytogenetic damage. Topics that are considered are: data standardization, population selection and bias, statistical analysis, ethical issues. Pooled analysis seems to provide a relevant improvement over meta-analysis in molecular epidemiology studies, though more research on methodology is needed.