Tutorials : confounders

What is a confounder?

​A confounder is a variable which is not involves in the causal chain from the exposition to the outcome, but wich is the cause of wrong estimation of the impact of the exposition on the outcome. Formally : a confounder is associed to both the exposure and the outcome but is not on the pathway between the exposure and the outcome.

Example: a false confounder

In kidney transplantation, consider the study of the association between the cold ischemia time and the graft survival. The delayed graft function should not be consider as a confounder because it is a consequence of the cold ischemia time (Debout et al. 2015).

Example: a true confounder

Consider a study of the impact a diet on the incidence of cardiovascular disease. The obesity can be considered as a confounder because it is a possible cause of the diet and it constitutes a risk factor of cardiovascular disease.

Which variables has to be consider for adjustement

A strategy is to retain all the determinants of the outcome, not only the true confounder as defined previously. This strategy was validated by simulations-based study, more particularly in the marginal context (Ali et al. 2015 ; Austin 2011).