You can't take a dozen lousy studies and put them together to get better statistical significance than a single well-designed study. If the methodology of a study is bad, the results have to be scrapped entirely, and peer review is supposed to be the process through which we filter out bad methodology.

Without peer review we don't know which studies to eliminate from our informal meta-analysis, which means we can't come to any conclusions about the data.