up vote 3 down vote favorite

I started a meta-analysis which includes studies comparing the complications(yes or no, dichotomous) of different treatment modalities for distal radius fractures. I obtained all the data, but I have troubles doing the statistics. I want to compare all treatment method with the Golden Standard: Plaster Cast Immobilization. I have little to no experience with the pooling of the data per treatment modality. I tried looking on google, youtube, I have read the book of Borenstein, but everybody has their own methods and programs. I just do not know how to pool and compare the data(I tried different methods and programs but I couldn't make it work(revman, R and stata). Can someone help and teach me how to pool the dichotomous data and compare them? I have filled in all data on 2 treatment modalities in Revman, but because all studies do not directly compare both treatment modalities I using revman did not work. enter image description here

share|improve this question
1  
Just looking at your data, these data come from observational studies that don't actually compare the different treatments against each other, but only present complication event rate (n/ N) for different treatments. I'm not going to the go into the issue if it correct or not to compare single arm studies against each other because that's a whole can of worms with a lot of assumptions, but since you are confident that you're not just pooling for the sake of pooling (garbage in, garbage out phenomenon), here's what you need to calculate it in RevMan.... – abousetta Mar 18 at 5:14
1  
First you need to calculate the absolute risk and SE (onlinelibrary.wiley.com/doi/10.1002/9781444311723.oth2/pdf)‌​. Then use the inverse variance function to pool across studies. I would advise you to create subgroups for each intervention and probably not a good idea to pool across subgroups (I'd need more information about your study, but my gut feeling is that you're pooling apples and oranges). Anyways that should get you started. – abousetta Mar 18 at 5:18
1  
In Stata you could have used metaprop, in R you could have used metafor, meta, or one of a number of other packages outlined in the CRAN Task View – mdewey Mar 18 at 12:44
    
Thank you very much for your help! I have calculated the absolute risk, SE, variance and inverse variance for the treatment group of each study separately. Is that the way I should have done it? I am not completely sure on what you meant by creating subgroups per intervention. How can I compare the subgroups in Revman now, because when entering the data I do not know what effect measure to use? I have an absolute risk and can only choose between odds ratio, risk ratio, risk difference, mean difference, std mean difference, hazard ratio or rate ratio? @abousetta – Rotterdam Mar 29 at 13:17

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.