The difference between ‘Effect Modification’ & ‘Confounding’
Posted on 4th June 2015 by Deevia Kotecha
Students can often struggle to understand the difference between effect modification and confounding. In order to help with this problem I have put together a simple guide to explain ‘effect modification’ and ‘confounding’ and the difference between the two terms:
Effect Modification
Effect modification is all about stratification and occurs when an exposure has a different effect among different subgroups. Effect modification is associated with the outcome but not the exposure.
For example, imagine you are testing out a new treatment that has come onto the market, Drug X. If Drug X works in females but does not work in males, this is an example of effect modification.
Confounding
Confounding occurs when a factor is associated with both the exposure and the outcome but does not lie on the causative pathway.
For example, if you decide to look for an association between coffee and lung cancer, this association may be distorted by smoking if smokers are unevenly distributed between the two groups. It may appear that there is an association between coffee and lung cancer, however if you were to consider smokers and non-smokers separately for each group this would in fact show no association.
What is the difference?
Confounding factors are a “nuisance” and can account for all or part of an apparent association between an exposure and a disease. Confounding factors simply need to be eliminated to prevent distortion of results.
Effect Modification is not a “nuisance”, it in fact provides important information. The magnitude of the effect of an exposure on an outcome will vary according to the presence of a third factor.
No Comments on The difference between ‘Effect Modification’ & ‘Confounding’
Sorry Cochrane, confounding factors are not a nuisance but the result of observing facts. “Confounding factors simply need to be eliminated to prevent distortion of results” will result in distorting the facts and not eliminating bias but in contrast, introducing bias.
5th August 2023 at 11:02 pmThe pdf very helpful I managed to get the difference between the two
16th September 2022 at 9:53 amThank you for letting the author know, William, much appreciated.
21st September 2022 at 10:47 amLittle more clarification in terms of explaining the example with description of the variable would have been even more easier to understand.
In case of effect modifier: There are 3 variables, Exposure variable is Drug (Given / Not given), Outcome variable is Cure (Cured / Not cured) and the effect modifier is Gender (Male / Female).
Even through it is showing that Drug X helps in curing, By stratification we will find that
Drug (+) Gender (Male) is associated with Cure (Yes)
Drug (+) Gender (Female) is NOT associated with Cure (No)
Drug (-) Gender (Male) is NOT associated with Cure (No)
Drug (-) Gender (Female) is NOT associated with Cure (No)
In case of confounder: The 3 variables are, Coffee (Yes / No) is the Exposure variable, Lung cancer (Yes / No) is the outcome variable and Smoking (Yes / No) is the confounder.
Even though it is apparent that Coffee is associated with lung cancer, By stratification we will find that
Coffee (+) Smoker (+) is associated with Lung cancer
Coffee (+) Smoker (-) is NOT associated
Coffee (-) Smoker (-) is NOT associated
Coffee (-) Smoker (+) is associated with lung cancer
Smoking is associated with Both coffee and Lung cancer.
10th February 2022 at 4:30 amThank you for simplifying these two terms that seemed very difficult to differentiate. I now understand the two terms and their differences and what they do.
29th November 2021 at 5:10 am