More than 1100 articles now appear annually investigating “independent risk factors” or “independent predictors” for various clinical outcomes. In medical research, independence is generally defined in a statistical sense: a variable is called an independent risk factor if it has a significant contribution to an outcome in a statistical model that includes established risk factors. As such, independence is based on a specific statistical model and depends on the set of established risk factors included in that model. Even when strong statistical evidence indicates that a variable is an independent risk factor for an outcome, this does not necessarily indicate that the risk factor causally contributes to the outcome. The opposite is also true: risk factors that have causal relationships with the outcome will not necessarily prove to be independent risk factors. These are basic statistical principles that are too often given short shrift in medical research. Herein, we discuss the clinical implications conferred by the above definition of independence, primarily using examples from recent cardiovascular literature. A glossary and schema are provided to help clinicians and researchers understand and discuss these matters effectively.
Brotman DJ, Walker E, Lauer MS, O’Brien RG. In Search of Fewer Independent Risk Factors. Arch Intern Med. 2005;165(2):138-145. doi:10.1001/archinte.165.2.138