0
$\begingroup$

To examine the association between an exposure and outcome, for example body mass index as exposure and all-cause mortality as outcome, some of studies reported relative risk as a continuous form. For example, reported relative risk only for 1 unit increment in body mass index. Now, how we can rescale relative risk and estimate it for 2 unit increment? I am trying to conduct a systematic review and meta-analysis but I dont know how we can do this. And what is such solution for case-control studies?

another example,

association between Mediterranean dietary pattern score and all-cause mortality: score range from 0 to 9 / number of participants: 1849 / number of death: 553 / relative risk of all-cause mortality for one unit increment in MED score : 0.95 (95%CI: 0.91, 1) /

know, relative risk for 2 unit increment in MED score?

Improve this question
$\endgroup$
2
  • $\begingroup$ You should fix your keyboard---the caps key obviously is not working. For now, I edited in caps for you. $\endgroup$
    – kjetil b halvorsen ♦
    Oct 26, 2016 at 14:52
  • $\begingroup$ Body mass index has units that are not physical. It is heuristic. If you want to predict anything, I would suggest you employ a physically relevant measurement. $\endgroup$
    – Carl
    Oct 26, 2016 at 15:34

1 Answer 1

Reset to default
0
$\begingroup$

Honestly, it all depends on how the relative risk is calculated. This answer describes how it's done using linear models. If the relative risk is an average over multiple stratification groups, computed using e.g. The Mantel Haenszel Method, then you cannot infer anything about a unit increase in any of the stratification variables, since they are typically discretized.

Improve this answer
$\endgroup$
1
  • $\begingroup$ thank you for your answers. but my question is how to convert relative risk per 1 unit or 1.5 unit increment in exposure to relative risk per two unit increment, if the data for multiple categories were not reported in articles. $\endgroup$
    – ahmad
    Oct 26, 2016 at 21:04

Your Answer

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

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