1
$\begingroup$

I want to perform a pooled analysis of studies including only a treatment arm, so no proper control groups are available.

Studies present data about subjects before the treatment and after treatment, at different follow up time.

Outcome measures are both categorical and numerical (with mean and SD expressed).

What I extracted from studies and want to analyze are:

  • vitality status at follow up. this is 100% at baseline of course. here basically i want to know the proportion of subjects alive at follow up, as all subjects were alive at the beginning of the studies.

  • one categorical variable (status of interest) which varies between baseline and follow up. I want to know if the proportion of subjects with this status before and after treatment differ significantly

  • several continuous variables, for which I have n, mean and sd. I have these both before and after treatment. what i wish to know is if there is a difference between pre and post treatment for these variables

I am pretty new to meta-analysis. I am familiar with r, but willing to learn how to use RevMan if necessary.

What kind of analysis should I perform ? Do you have any materials to share where I can learn and study ?

Thanks, C

$\endgroup$
  • $\begingroup$ There is no particular problem doing this. See stats.stackexchange.com/questions/241960/… or stats.stackexchange.com/questions/207426/… or look at the documentation for the metafor package available from CRAN. $\endgroup$ – mdewey Aug 16 at 15:08
  • $\begingroup$ Thanks! The materials and discussions are very useful and I m reading them carefully. Nonetheless, i was wandering: in my case, i have pre-treatment and post-treatment variables, and my true goals would be to pool these variables (both continuous and categorical) from different studies, and see if there is a difference between pre-treatment and post-treatment. From what I read so far, proportion meta analysis mainly pool prevalence, but does no comparison between 2 time points. is it correct ? or am I missing something ? $\endgroup$ – cloud91 Aug 17 at 0:20
  • $\begingroup$ I think you need to edit to clarify exactly what you have extracted from the primary studies. $\endgroup$ – mdewey Aug 17 at 12:29
  • $\begingroup$ done! hope it’s more clear now $\endgroup$ – cloud91 Aug 17 at 21:40
  • $\begingroup$ dear all, has anyone else any clues and can help me ? $\endgroup$ – cloud91 Sep 3 at 8:55
0
$\begingroup$

For your first question about meta-analysing the proportion of survivors at follow-up this is a duplicate of Meta-analysis of proportions which gives details.

For your second question you need to have the data from the matched pairs (the same person before and after). You then need to calculate the appropriate odds ratio from the table. If you are doing this in R using metafor you need to look at the documentation for "Outcome measures for change or matched pairs" and then look for "Measures for dichotomous variables" which gives you options and several references to look at.

For your third question just go a bit further down the document for "Measures for quantitative variables" where again there are several options and references for them.

$\endgroup$
  • $\begingroup$ Thx! it was very clear . Since I have mainly group data (mean and SD before and after procedure) do you think it would be acceptable from a statistical point of view to approximate individual’s patient values to the mean of the group ? $\endgroup$ – cloud91 Oct 3 at 13:00

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.