Meta analysis is a useful tool in the scientist's arsenal, since it provides a quantitative, and somewhat objective, way to combine the results of various studies on the same topic. This allows scientists to try to analyze how the results of the studies, which may be very different, combine to form a more complete picture. Combining studies can also allow scientists to find out whether a small effect was significant, something that may be difficult to do in a single study that has fewer enrollees.
The biggest weakness of meta-analysis is that it relies only on pre-existing studies rather than any original data collection. This means that any flaws that consistently run through the early studies may be magnified in the meta-analysis. Meta analysis may also be hampered by publication bias -- if there is a tendency for journals to only publish positive results (and not studies that show little to no effect), then the magnitude of an association seen in a pooled analysis may seem to be larger than it actually is.