Strange question from me, but try to follow me. I do not remember or name correctly a type of bias in cohort study which is pretty clear in my mind. I try to explain: Let's assume that I want to test the rate of adverse events for a drug name "ABC" in a cohort of patients, half of which is naive for treatment with "ABC", and other half is being treated with ABC since 10 years. Obviously, the rate of Adverse Events is way higher in the naive cohort, and that's why patients that were being treated for a decade were already selected to be "resistant" to the adverse events of the ABC drug. Obviously, this result is biased.

The question is: how to name this type of bias? My guess is to name it as a type of "immortality bias" (because patients being treated for several years are sort of "immortal" in relation to the adverse events of the drug), but i firmly believe that there is another way to refer to it.

Thanks to anyone that'll try to help me.

  • 1
    $\begingroup$ Immortality bias risks being confused with immortal time bias (bmj.com/content/340/bmj.b5087). Anyway, your scenario seems simply to highlight the risk of selection bias (or a specific subtype of such bias). Eventually, if you are not using randomization with a placebo control, and only comparing naive vs chronic ABC users, your comparative data will be confounded by the fact that the naive group is actually composed (in uncertain %) by chronic ABC user-like subjects, and ABC intolerant subjects. $\endgroup$ Dec 11, 2018 at 16:33
  • $\begingroup$ Still don't know. It may be a case in which this is quite a selection bias, especially if my primary outcome it's the adverse event itself... but what if that's not true and I'm analyzing safety outcomes in a non-naive patients RCT? Maybe the closer name to it it's "Healthy user bias", though this is slightly different from this meaning... $\endgroup$ Dec 11, 2018 at 23:26
  • $\begingroup$ Have a look at bmj.com/content/340/bmj.b5087. $\endgroup$
    – Todd D
    Jan 7, 2019 at 6:15

1 Answer 1


This is attrition bias, which is a form of selection bias - the members of the 10-year group have been selected into that group through their continued use of the drug over a long period of time. For that group, you do not observe the people that started using the drug ten years ago but discontinued treatment (or died).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.