EPI10-1: Confounding
 

OBJECTIVES
At the end of this session you should:
  • understand the definition of confounding;
  • be able to examine a table of data providing the full data for an exposure of interest and a potential confounder in subjects with and without the disease and to perform stratified analysis of this data to examine for confounding;
  • understand the definition of effect modification;
  • be able to examine a table of data providing the full data for an exposure of interest and a potential effect modifier (EM) in subjects with and without the disease and to perform stratified analysis of this data to examine for effect modification;
  • understand the difference between confounding and effect modification.

CONFOUNDING:

Confounding refers to the effect of an extraneous variable on the association between exposure and the outcome of interest.

Example:

More generally for something to be a confounder three conditions must be satisfied:

  1. The confounder must be associated with the outcome in the unexposed.
  2. The confounder must be associated with the exposure.
  3. The confounder must not be an intermediate variable in a causal chain from exposure to outcome.

Confounding can be controlled either by study design eg. a randomised controlled trial, or in the analysis after the study has been conducted eg by stratified analysis or by standardisation or multivariate analysis.