USMLE Forums banner

Confounding and Effect Modification‏

4K views 4 replies 5 participants last post by  praru 
#1 ·
1- Confounding bias :(Middle factor that is related to both Exposure and Outcome)

Alcohol ( exposure ) Oral cancer ( outcome )

Smoking is a confounding bias here.
Because smoking is related to BOTH the exposure and the outcome. People who smoke will more often drink, and people who have oral cancer, were probably smokers ( smoking is a known risk factor for oral cancer).

2- Effect modification :(Factor that is ONLY related to Outcome; has got nothing to do with exposure).

OCPs ( exposure ) Breast cancer (outcome )

Family history of breast cancer is an effect modification here. Because family history will definitely modify the effect and will fasten the process of Breast cancer. OCP(Exposure) in this case, has got nothing to do with Family history.

Asbestos ( exposure ) Lung cancer(outcome)

Smoking is an effect modification; Smoking is not related to asbestos; Smoking is a risk factor for Lung Cancer(Outcome)

Estrogens ( exposure ) DVT (outcome )

Smoking is an effect modification; Smoking is not related to Estrogen; Smoking is a risk factor for DVT(Outcome)

So do you notice the difference ?? The effect modification is ONLY related to the outcome, but NOT to the exposure. Think with me here : smoking does not effect neither asbestos exposure nor estrogen levels or intake,BUT definitely is a risk for lung cancer and DVT !!

Essential Exam Points for fast answering:
-Confounding is a bias(error), Effect modification is NOT.
-Think of Confounding bias for the ENTIRE population in the study, whereas Effect Modification is stratified. Stratified means that it is for a particular age group, or a particular SUBGROUP within the study population.
-Sometimes in exam they give an example in which both effect modification and confounding bias are present. These questions are tricky(250+), but you can answer them with the following example:

A good example is: Drug X worked on children but did not work on older adults.If they ask you what is this? then the answer is effect modification as the drug x worked on a particular subgroup(stratified) and not on the entire population. If they ask you, why is the result not statistically significant when we study the whole population(children+adults) for Drug X? the answer is because AGE is the confounding bias. Once you confound for age, you will get a statistically significant result!
Hope this helps!
 
See less See more
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top