Calculating Odds Ratio from Relative Risk

[Reza Kahlaee]

Hello

I have a set of Relative risks (RR) and Confident intervals and I want to convert them into Odds ratio and Confidence interval, I wonder how is it possible to calculate ORs from RRs?
I would be grateful if you have any answers!

Kind regards.

Reza

[Sven Trelle]

All conversions I know of require an assumption about the baseline/control group risk. See this reference as an example: http://www.bmj.com/content/348/bmj.f7450.

Best
Sven

[Clyde Schechter]

What Sven Trelle says is absolutely correct. However, if the baseline risk is small, which generally means < 10%, then OR = RR is often a reasonable approximation for practical purposes.

[Rich Goldstein]

I'm a little confused - getting RR's from OR's (which is what the cited paper in #2 discusses) requires the mentioned assumption; however, the OP is asking about getting OR's from RR's; in certain circumstances, this is easy - but it depends on what other information the OP has (e.g., total N) - or, have I misread something?

[Clyde Schechter]

Rich is right. I misinterpreted the original request.

[Sven Trelle]

Me too. Sorry.
If there is no more information available I think it might be difficult/impossible. If information on the risks in the two groups is available it is easy as already mentioned. If not, it might be possible to deduce some information from the formula for the standard error of the log relative risk (that you get from the confidence interval). Just a thought.

[Andrew Lover]

The place this usually pops up is in the context of meta-analysis (perhaps that's your study?), so you might find some ideas in that literature, eg:

http://handbook.cochrane.org/chapter...e_variance.htm

[Reza Kahlaee]

Thanks all for your answers and comments, yes it pops in Meta-analysis when we need to convert RR into OR, and we have RR and CI, so we need to extract OR and Ci from it! I wonder is it possible ?

[Rich Goldstein]

as said above, it is possible - if you have some other information it is even easy - however, you haven't told us whether you have any other information (e.g., the total N's or the marginal N's or ...)

[Reza Kahlaee]

Yes I have total numbers in the article and the relative risk and confidence interval for different adjustment models.