I've come across a series of association studies that report odds and asymmetric confidence intervals. The standard meta-analytic techniques I've used assume symmetrical confidence intervals to calculate the standard error. I'm not sure if there are standard practices for this and here my naive thoughts:

  1. Use the smaller CI and calculate the SE from it. This would assume it was calculated using a consistency standard error and that the other CI was calculated using an agreement standard error.

  2. Calculate the log(OR) and its SE assuming that both boundaries were calculated using a consistency standard error (which we know it's true).

  3. Use some other approach (haven't found any in the literature) and/or software functions.

Data looks like this:

1.03 1.00 1.06
1.012 1.011 1.014
1.35 1.16 1.59
1.89 1.1 3.26


1 Answer 1


Option 2 is the standard way to go especially since you will be doing your meta-analysis on the scale of log odds. You may end up presenting it as OR of course but the underlying analysis would be on the log scale.


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.