The Case for Meta-analysis Posted on January 20, 2009 Antonio Cepeda-Benito (bio) outlines a successful paradigm for writing a meta-analysis paper.

I think that's the most important question for a meta-analysis or for anything else. So imagine the best possible scenario and say, "Now that I know this, so what?"

Then you have to sort of make sure that it hasn't been done already. So do a very thorough search of the literature to find out that it has not been done. And then when you determine (a) it's a worthwhile question to find the answer to and (b) the answer is not already there, then I think that you educate yourself very well on the topic for meta-analysis.

You really need to read lots of articles on this subject first because once you really know the literature well and you know what people do and what are the signs. Then you can start thinking about, how am I going to code the results? How am I going to characterize this body of literature in a systematic way?

So that I can go, in each article, I can say, its using women and men, its used in, and then coding for that. And then is the therapy involved long-term therapy versus short-term therapy? And do you know what you need to code for?

And then you just do a very good search and get to work. I've done two meta-analyses, and both are highly cited. And it's a lot of work, but so are other things. It can be done with very little funding. And it's fun to do, I mean, because it's kind of solving the puzzle. You have to do sort of finding for the structure of what is the skeleton of this literature and then so that you can code for it, you can then look for it and find out.

Not all studies are measuring the same constructs, and that might be something interesting to find out. If you use this measure or if you validate, for example, in studies of treatment outcome for smoking, do these studies validate whether the person has smoked or not? Or are they just going by the person's word?

That could make a difference. If it doesn't, you can say, "Well, it doesn't make a difference. People tend to tell the truth." That's another thing you know. But because you notice that some do and some don't, you code for that, because that's a potential variable that is going to have an impact on the outcome of the report.

And you learn the field. I mean, then you are the expert. You know, if somebody wants to know something about what's on, they call you because they know that you have read 300 articles on the subject and that basically you read them very carefully.