관찰 연구에서 선택 삐뚤림과 교란 변수를 최소화하기 위한 방법에 대한 고찰

Abstract

Although well-designed randomized controlled trials (RCTs) are the most scientifically rigorous method for testing a hypothesis and have held the uppermost position in the hierarchy of evidence-based medicine as level I evidence, RCTs can be impractical or unethical in clinical settings. Observational studies are the best available alternative when RCTs are unavailable. Well-designed observational studies in the fields of both clinical and epidemiological research have been shown to provide similar results as those of RCTs with level II or III evidence; however, there are major methological limitations of observational studies including biases and confounders. To overcome these limitations, it is very important to consider or avoid possible factors (biases and confounders) which can interfere with adequate interpretation of results. This article provides an overview of selection biases and confounders as well as stratigies to minimize biases and confounders in the design and analytic phase of an observational study. Understanding of selection biases and confounders in observational studies and methods for controlling biases and confounders will assist investigators in conducting well-designed observational studies.