Directed acyclic graphs in planning clinical and epidemiological trials

Establishing and quantifying the causal relationship between risk factors and outcomes are essential in epidemiology. Proper planning of epidemiological study makes it possible to avoid bias in estimating the effect and obtain significant results. Currently, directed acyclic graphs (causal graphs) are increasingly used to visualize the hypothesis being tested. In the article, the authors analyze directed acyclic graphs in planning epidemiological studies. Basic algorithm for the work with directed acyclic graphs is described. The last one makes it possible to avoid errors in further analysis. The main components of directed acyclic graphs are singled out. The authors present current terminology with description of each element of graphs. Graphical model shows estimation bias due to confounders and mediators of effect. The method of instrumental variables in epidemiology is presented using genetic data (Mendelian randomization method). Thus, the advantages of directed acyclic graphs are shown.