1
$\begingroup$

I’m currently trying to perform a meta-analysis in R via the metafor package using data that does not have a control group. In this case, I am looking to see whether or not a type of surgical intervention improves symptoms in these unique type of patients. Essentially, the only data I have from these studies is whether symptoms of their certain pain has improved.

Would this be a meta-analysis of proportions? Or would I do a standard REML in metafor: For example, one paper gives the data that 20/25 patients had symptoms that improved. Thus, would I have 4 data points, “creating” my own control group?:

  1. Surgery_improvement: 20 patients
  2. Surgery_no_improvement: 5 patients
  3. No_surgery_improvement: 0 patients (obviously, as this doesn’t even exist)
  4. No_surgery_no_improvement: 25 patients

And then perform a standard meta-analysis?

$\endgroup$
1
$\begingroup$

You suggestion that this is a meta-analysis of proportions is correct. If each study gives you the number treated and the number of those who improved then you an analyse those. You second suggestion is flawed because the last group (4) should read no surgery no improvement 0 as that is all you know.

You will probably want to choose one of the options for transforming proportions before doing the meta-analysis especially if they are close to zero or one.

Note that you meta-analysis will be open to the criticism that this is just observational but if that is what the studies available to you have then that would be an unfair criticism.

$\endgroup$

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.