Circulation. 2026 May 20. doi: 10.1161/CIR.0000000000001440. Online ahead of print.
ABSTRACT
Resources for observational comparative research have expanded enormously in recent years to include very large sources of granular, routinely collected health care data and modern statistical, epidemiologic, and econometric techniques. This scientific statement provides an overview of best practices and analytic considerations in observational comparative studies from the perspective of investigators, sponsors, publishers, and consumers of observational research. Observational comparative research is a component of the research landscape that fulfills a role distinct from that of interventional studies in the evaluation of drugs, surgical procedures, medical devices, and health policies. Sources of systematic error (ie, bias) in observational comparative studies include selection bias, information bias, and confounding. Principles from statistical science and econometrics can potentially be used to make causal conclusions from observational data. Target trial emulation is a useful framework to guide the rational design and illuminate the limitations of observational studies. As with interventional research, a formal study protocol should be prepared before every observational study to enhance rigor, reduce data manipulation, and promote transparency of study reporting. Selection of the study data source is a key decision early in the design stage of a study, and should be chosen on the basis of concordance between the needs of the specific study question and the properties of the data set. We recommend the use of causal directed acyclic graphs to clearly specify the study exposure, end points, confounders, colliders, moderators, and mediators. Taken together, these recommendations promote rational design choices and cautious interpretation of the results of observational comparative studies.
PMID:42158988 | DOI:10.1161/CIR.0000000000001440