1
$\begingroup$

I am conducting a meta-analysis. One of the studies have both OR and its upper confidence limit equal infinity (control group have zero outcome events). How can I combine such study with other studies in a meta-analysis?

Thanks in advance.

$\endgroup$
  • $\begingroup$ What other studies do you mean? $\endgroup$ – Karel Macek Jan 18 '17 at 12:00
  • $\begingroup$ Do you have the raw frequencies? $\endgroup$ – mdewey Jan 18 '17 at 12:11
  • $\begingroup$ I mean the other studies that I included in the meta-analysis. $\endgroup$ – user-anna Jan 18 '17 at 12:44
  • $\begingroup$ Yes, the study shows the raw frequencies. $\endgroup$ – user-anna Jan 18 '17 at 12:44
1
$\begingroup$

If you have the raw frequencies for all the studies then you have a number of options.

You can use the Mantel-Haenszel method which will give you the overall summary but will not give you an estimate for that study so it will not be included in your forest plot.

You could use an alternative to the odds ratio like the arc sin difference as suggested by Rücker and colleagues in a paper entitled "Why add anything to nothing? The arcsine difference as a measure of treatment effect in meta--analysis with zero cells" available here and also I think elsewhere on the web. As her title implies this is an alternative to the most common method which is to add a small constant to the cells (0.5 is usually chosen). Bradburn and colleagues have compared several methods in a paper entitled "Much ado about nothing: a comparison of the performance of meta--analytical methods with rare events" available here. There have been numerous other papers on this topic.

You have tagged this Stata, I am not sure how Stata handles this but I feel confident that some of these methods are available. If not you can swap to R where they are all included in one or more of the available packages.

$\endgroup$
  • $\begingroup$ Thank you so much. Very informative with great suggested papers. $\endgroup$ – user-anna Jan 18 '17 at 17:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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