Effect modification VS Confounding

[stronghold]

---

Members don't see this ad.

May you give me an example in which we can if we do so and so then it's effect modification, but if we do it in another way then it's a confounding bias.

 

[Myxedema]

I apologize to the epidemiologists and biostatisticians out there for this simple explanation, but here's how I understand it: Let's say you find a variable (coffee drinking) to be a risk factor for a disease (myocardial infarction). It's possible that coffee drinking itself may not be a risk factor, but rather it may be masking the effect of another variable, like smoking: It is possible that people who smoke are more likely to drink coffee. If coffee drinking is not found to be a risk factor for MI after controlling for the effect of smoking, then coffee drinking is a confounding factor.

Let's continue with the same example, but this time, when the effect of smoking is controlled for, coffee drinking is still found to be an independent risk factor. In addition, as it turns out, coffee drinking actually magnifies the effect of smoking: Somebody who both smokes and drinks coffee is found to have a much greater risk than smoking or drinking coffee individually. This would be an example of effect modification.

 

[stronghold]

Thanks man! May you further explain it when we stratify them in groups?

 

[Phloston]

Effect modification is how a medication/treatment plan affects a disease process (e.g. OCP in patients with family Hx of breast cancer). If they suggest in any form that there's a treatment or intervention involved that's related to the disease process, the answer is effect modification.

Confounding is how a lifestyle variable impacts the disease process (e.g. alcoholics are more likely to smoke, which leads to lung cancer, but EtOH itself doesn't cause the cancer). If they suggest in any form that there's a lifestyle choice/decision related to the disease process, completely unrelated to a treatment/medical intervention, the answer is confounding.

(And like Myxedema, that's likely an oversimplified version of it, but that's how I remembered it for Step1)

 

[stronghold]

Thanks BOSS 🙂

 

[stronghold]

Just to return back to my original thread, correct me if I am wrong, after we stratify patients in groups, in case there was no effect, then it's confounding. In case there is still a causative effect, then it's effect modification. Am I right?

 

[magrMPH]

Effect modification is given by an example where an exposure increases an outcome by 10 % or more and the homogeneous null is rejected. For example, a vaccine in a child works for child but not for older child. In the case of independent variable, this will be about confounding such as a characteristic like favoring black coffee with no sugar ,vs with sugar, with milk for HBP outcome. The case when a variable such as indirect smoke exposure is harmful which it is when a direct exposure such as smoking is the vairable and ohter risk factors are present as in HP status for Oropahryngeal Cancer and smoker status, the smoking status is a negative confounder, becuase you may not possibly knwo how bad the risk is ususally worse than expected. Confounding is a bias that is independent and increase the outcome by 10 % or more.
Here are some links below for papers.


http://www.theglobaljournals.com/ijar/file.php?val=MjY2MA==
http://support.sas.com/resources/papers/proceedings12/315-2012.pdf
http://www.lexjansen.com/pharmasug/2011/sp/pharmasug-2011-sp03.pdf
http://analytics.ncsu.edu/sesug/2009/SD018.Agravat.pdf


It seems there was confusion. Confounding can occur possibly from smoking to breast cancer but one must do further investigation s. Effect modification does show that there is a statistically significant relationship to exposure no drinking no smoking or passive exposure to head and neck cancer if Indian race in included on page 31 of this article.

P< 0.02 vs p < 0.10 off Indian race included vs Other race of Hispanic.

http://www.sascommunity.org/mwiki/images/b/b8/GSF_315-2012_-_Amended.pdf

 

[aspiringmd1015]

still confused about this, bc what the Great PHolston is saying is how i understood it, but magrMPH confused me further. Help!

 

[aspiringmd1015]

crystal clear now, thanks!