Please explain what you want to do. I have never heard of a proportional meta-analysis. Do you mean one like in which a ratio is used as an effect size (e.g., odds ratio, risk ratio, response ratio)?
I´ve came across this while reading the article (attached). Apparently, it is a mean of quantitatively cumulating results from treatment arms/cohorts regardless of the comparator? But I am not sure if i understand it correctly.
I want to gather results on efficacy of different treatments in HCV. I have results on treatment A+B+C from two different trials (one single-arm, second A+B+C vs A+B+C but with different duration of therapy) and was wondering if cumulating quantitatively would be appropriate. Another thing is that there are therapies like A+D and A+E studied in different trials, randomized or single-arm cohort (A is the common element). What are my options to compare A+B+C, A+D and A+E and draw a conclusion? Is network meta-analysis the answer?
yes you can perform proportion meta-analysis. MA without comparators are also called one group or one arm meta-analyses. I Agree with Wisit about using R. I suggest the "metafor" package that allow you to perform also moderator analysis to explain heterogeneity. In this type of meta-analysis the real aim often is exactly that.
They are very common to answer epidemiology questions. You can use stats Direct Software to make it easy or Stata but you transformations and back transformatiosn before and after the transformation.
Also, just to briefly add, if you're doing a meta-analysis of proportions with very rare or very common events (e.g., 3% mortality rate or 92% survival rate) then would recommend a logit transformation. Although you can do a meta-analysis of proportions in Stata (you just need to have the lower and upper confidence intervals in addition to the estimate), logit transformations then back transformations are not done automatically, so you would need to re-do the Forest Plot in a different program once you have all of the estimates. This is why I think it's best to use the R package 'metaprop' as you can specify the type of transformation. The Freeman-Tukey transformation is an alternative to the logit transformation, and is better for small sample sizes as it helps with variance stabilization... however, you should look into whether or not this would be the right option for you (http://www.metafor-project.org/doku.php/analyses:miller1978). Good luck!
Please read the following paper ": Jaiswal N, Singh M, Thumburu KK, Bharti B, Agarwal A, et al. (2014) Burden of Invasive Pneumococcal Disease in Children Aged 1 Month to 12 Years Living in South Asia: A Systematic Review. PLoS ONE 9(5): e96282. doi:10.1371/journal.pone.0096282"