I tend to rely on certain major landmark studies, usually collaborations between many institutions and usually NIH funded, such as the STAR-D study of antidepressants and the CATIE study of antipsychotics. I think those are pretty dependable.
The government has not been entirely passive about all this. Recent regulations require drug companies to "publish" all studies they fund regardless of outcome, but not in journals -- they have to set up a website where all results can be archved. So the data presumably can be found, but one has to go searching for it.