There are lots of studies on studies, and in general they are a good idea. Here's my take on that (from SoylentNews), slightly paraphrased to hopefully demonstrate why meta-studies can be good:
Keeping track of information is difficult, and journals generally don't like people to pepper their articles with too many citations. If the same information gets spread around, then the chance of citation drops for any particular article that contains that information. This is a problem, even with Watson-level recall, and even the very best papers will suffer from this issue.
Let's say there's a wonderful paper published in a journal that reviews a whole bunch of things. It survives for about 6 months with citations ramping up, but then someone discovers something new and interesting about one of those things. Then, people who would previously cite the big paper and therefore let others know about it, might decide that in their particular area, the new paper is a more appropriate citation.
About 6 months after that, the paper has hit its "peak citation rate", as the popularity of the paper is eroded in many different areas by the smaller, newer papers. Pick any one of those new papers, and you could easily say the earlier paper is better. However, pick any one of those many things, and you can probably find a better paper for the that particular area of study. Funding sources encourage this behaviour — being better than some previous paper, and fragmenting the research knowledge as much as possible.
People could read the single big paper and get a great overview, but over time they become more likely to know about the smaller papers which give excellent detail, but are very specific. Over time, the general knowledge of readers is reduced, and they lose track of related work outside their area of expertise.