Confounding by Indication in Clinical Research
This JAMA Guide to Statistics and Methods reviews the use of methods to control for confounding by indication in clinical studies that assess the potential effect of a treatment of risk factor on a patient outcome.
In the assessment of the effect of a treatment or potential risk factor—termed an exposure—on a patient outcome, the possibility of confounding by other factors must be considered.1 For example, if researchers studied the effect of coffee drinking on the development of lung cancer, they might observe an apparent association between these 2 variables. However, because drinking coffee is also related to smoking, the observed association between coffee drinking and lung cancer does not represent a true causal relationship but is rather the result of the association of coffee drinking with smoking—the confounder—which is the true cause of lung cancer.
This illustration is a simple example of the very complicated and multifaceted phenomenon of confounding. Distortion from a confounder can appear to strengthen, weaken, or completely reverse the true effect of an exposure. In addition, multiple factors can interact to cause confounding in both epidemiologic and clinical research. Notwithstanding these complexities, a confounding variable can be readily identified if it meets 3 important criteria.1 First, a confounder must be an independent risk factor for the outcome, either a causal factor or a surrogate for a casual factor (eg, smoking for lung cancer). Second, a confounder must be associated with the exposure (eg, smoking and coffee drinking). Third, a confounder cannot be an intermediate variable between the exposure and the outcome (eg, smoking is not caused by drinking coffee).
A particularly important type of confounding in clinical research is “confounding by indication,” which occurs when the clinical indication for selecting a particular treatment (eg, severity of the illness) also affects the outcome. For example, patients with more severe illness are likely to receive more intensive treatments and, when comparing the interventions, the more intensive intervention will appear to result in poorer outcomes. This is called “confounding by severity” to emphasize that the degree of illness is the confounder. Because the degree of severity affects both treatment selection and patient outcome and is not an intermediate between the treatment and outcome, it fulfills the criteria for confounding.
As an example, the nonrandomized assessment of tracheal intubation vs bag-valve-mask ventilation for pediatric cardiopulmonary arrest reported by Andersen et al2 is likely to be complicated by confounding by indication. Clinical conditions (eg, asthma, cystic fibrosis, and upper airway obstruction) existing before and during a patient's cardiopulmonary resuscitation will both affect the patient's outcome and influence the type of airway management.2 In other words, it is likely that children with more severe disease and worse overall prognosis for survival had a greater probability to be intubated.2 This possibility is especially great because severity ...