Can meta-analysis be trusted?

Until around 25 years ago the only way to assimilate and evaluate research evidence was through discursive literature reviews, in which someone with an interest in a given research topic would accumulate and subjectively evaluate the importance of research findings in that area. These reviews, although informative, are highly reliant on the discretion of the author who, with the best will in the world, could be unaware of important findings or could give particular importance to studies that others might believe to be relatively less important (see Wolf, 1986). The failure of literature reviews to provide objective ways to assimilate scientific evidence led scientists to look a statistical solution. The groundbreaking work of Glass (1976) and Rosenthal and Rubin (1978) paved the way for what we now know as meta-analysis: a statistical technique by which findings from independent studies can be assimilated.