[Application of directed acyclic graphs in identifying and controlling confounding bias].

Observational study has been viewed as the most convenient method in designing etiological studies. However, the presence of confounders always challenge the researchers in study design, since unadjusted confounders may lead to biased results. The traditional definition of a confounder is not intuitional in application and sometimes leading to inappropriate adjustment of nonexistent "confounders" which might induce new bias to merge. The use of directed acyclic graphs (DAGs) may identify confounders easier and more intuitional, as well as avoiding superfluous adjustment. It can also contribute to the identification of adjustment methods, and be useful in causal inference of observational studies.