Meta-analysis

[Sir Gillies]

Hello,

I was wondering I could ask a few questions to those who have carryed out a systematic review + meta-analysis from scratch.
1) How challenging is it? My stats knowledge is not that great.
2) How long does it usually take (full time)?
3) How well is it seen by attendings and PDs? I appreciate there are many variables, but between a retrospective, cadaveric (max sample size = 20) or meta-analysis what looks better on someone's CV?

Thanks a lot in advance for your help.

---

Members don't see this ad.

 

[surg]

Meta-analysis typically are relatively difficult unless you know that there are very few papers on that topic. It is a relatively stats heavy and evidence based medicine evaluation heavy exercise in trying to decide what constitutes high quality studies. It can take anywhere from a few weeks to forever depending on the number of papers that you are considering and how stringent your criteria area.

The best paper is one that answers a question that people want to know the answer to and uses the appropriate methodology to answer the question. You should not plan your research solely on what kind is going to look best on a CV. Landmark papers come in all sorts, from case reports to large multi-institutional randomized double blinded placebo controlled trials and everything in between.

 

[Sir Gillies]

Thank you very much for your reply and for your input. It's good food for thought : )

 

[PinkIvy08]

I've done one before and am probably gearing up to do another. I think the first meta-analysis is the hardest, but they're really not that bad. As surg said, it depends on the topic you're picking, your inclusion and exclusion criteria and the resulting amount of final papers you include in your analysis. Add in that a properly executed one requires at least one other person to do the same work you're doing...let's just say it can take a while.

I didn't find it too bad, but I'm also getting my MPH in Epidemiology and had learned a lot of the stats and programming. If your institution has a School of Public Health, you might want to trying talking with the epidemiology dept and see if they have a grad student that can assist you - usually they know the stats side well, will take second author, and want it for their own resume.

Gl!

 

[tussy]

I did one as a medical student and got a stats PhD student to help. He got a second author for doing the stats and I did the rest.

It's tedious. I spent a full summer on it (full time) and then over the next 2 years finished it. Got some good milage out of that paper, though this was 1999 and meta-analysis were really popular. They have become less so recently.

 

[Valadi]

My first one took me a month. As they said the first is the hardest.

The most time consuming aspect is sorting through all the papers, then deciding you want to include another data point, and having to go through them again. Carefully determining inclusion/exclusion criteria, as well as data points to be collected will cut the time this takes by 75% (if you're anything like I was).

The stats difficulty depends on what you end up doing. If you're using complex models then you'll probably need someone to help out. Otherwise you can figure out how to do a regression analysis yourself over a weekend.

 

[thinkorswim]

You will absolutely need the help of a statistician if you're looking to publish a decent meta-analysis. A thorough analysis will have much more than just Forest plots and effect measures.

With regards to time, it all depends on the scope of your hypothesis and the number of studies included in your analysis. It can be done in a couple days or it could take a few months.

As for how meta-analyses are perceived, my impression is that they are generally well-received. A good meta-analysis using only randomized, placebo-controlled trials will have a high level of evidence and thus be viewed more favorably.

 

[Sir Gillies]

That's great! Thank you so much for sharing your experience. It's exactly what I needed : )
I'll keep you posted on what I end up doing.