Meta-analysis Help

#1
Dear all,

I have been asked by someone to perform statistical analysis for a meta-analysis.
The aim of the study is to compare between different surgical approaches.
The data I got from this person are only summary data from past medical research results,
for example:

(1) in study No.1, using approach X, there were 400 successes and 50 failures,
using approach Y - there were 300 successes and 60 failures.

(2) in study No.2,, using approach X, there were 300 successes and 40 failures,
using approach Y - there were 200 successes and 60 failures.


My question: Could anyone tell me how do I perform the analysis ? Can I use Chi-square ?

Thank you very much in advance.

Jack
 

hlsmith

Not a robit
#2
Is it time to event data or crossectional results? Do all studies have the same length of followup? How many studies and was treatment assignment randomized? You have to first examine random effects, are you familiar at all with them?
 
#3
Thanks for your reply, hlsmith.

Is it time to event data or crossectional results?
They were all surgical data (outcome = success or failure), so I believe they were not time to event data.

Do all studies have the same length of followup?
I don't think there was any follow-up, as they were surgical data.

How many studies and was treatment assignment randomized?
I don't think they were randomized.

You have to first examine random effects.
What do you mean?
 
Last edited:

hlsmith

Not a robit
#4
Follow-up is important most times, in that I correct a broken leg, success. Well perhaps this is malunion six months down the way, so success was defined too soon. See if this is relevant to your project.

Without randomization you typically have to control for baseline differences between the treatment groups. Each study may control for different variables making pooling an issue.

Random (study specific) effects is what you usually control for in MA. How many studies are there?

You need to read up on MAs. There are more to them than a normal analyses, especially in the observational setting (non-randomized treatments).

Keep question coming if you have them. The go to introductory book for MAs is "Introduction to Meta-Analysis" by Borenstein et al.
 
#5
Hello hlsmith, thank you.
The data I received are entirely summary data, not case data, ie count of successes, count of failures, count of total surgeries (see my first post).
Can I still analysis without individual case data?
 

hlsmith

Not a robit
#6
Depends. You traditionally do use counts in MAs, thus the "Meta" part. I believe you just need to study the method to see the applicability to you potential project.