The aim of most pelvic floor disorders (PFD) research is to obtain an unbiased effect estimate and to make causal inferences. New developments in epidemiologic research, including the use of causal directed acyclic graphs (DAGs), have shown that traditional analytical strategies for research can be inadequate, leading to unintended consequences such as introducing additional bias. Although DAGs have been proven to be useful in other medical fields, their use has been limited in PFD research. The aim of this paper is to introduce DAGs and then demonstrate their application in PFD research. This paper will also illustrate how relying purely on statistical techniques can lead to pitfalls in reducing bias in research studies.
DAGs are a graphical epidemiologic tool that provide a method to select for potential confounders and minimize bias in the design and analysis of research studies. We start by providing an introduction to DAGs. We then describe six scenarios in PFD research in which DAGs can be helpful: (1) identifying appropriate confounding variables for adjustment; (2) identifying potential over-adjustment when conditioning on a mediator; (3) identifying unintended confounding due to inappropriate adjustment; (4) identifying unintended selection bias due to inappropriate adjustment; (5) planning analyses in cross-sectional studies; and (6) using DAGs as a framework to help plan data collection and analyses in PFD research.
We demonstrate how the application of DAGs as an aid to PFD research can help to decrease bias and discuss the insights and implications for study design and analytical approaches. Neurourol. Urodynam. 31:115–120, 2012. © 2011 Wiley Periodicals, Inc.