If we consider this article here: http://www.sciencemag.org/content/342/6154/60.full
Then the system of peer review seems to be suspect, not only for open access journals, but also for the "closed" journals. This study shows that the peer review
accepted an article with built in errors for publication for almost of half of the journals under investigation. Explanations could be that there are too many articles to be reviewed and to little number of reviewers. So the errors were not detected.
On the other side the high impact factors of natural science journals are a discrimination of human science articles, which also appear in peer reviewed journals.