Another thing one can do is what I did quite some time back. Except now, you'd have to (for the process to be practical) introduce a date limitation into the search. I'd suggest searching only up till say 1996.
This can be done by picking the Advanced search in Pubmed.
Namely, use the names of each common anabolic steroid, e.g.
Testosterone OR nandrolone OR stanozolol OR boldenone OR....
and pull up every abstract. At least skim through them all, and read thoroughly the ones that aren't obviously useless. Those that on reading seem especially important, save the reference and go read the actual article later, if having access to a suitable library.
There are thousands of hits, but this is not undoable.
Secondly, if having access to a suitable library, you will find it not at all impractical to check the contents, and read all interesting articles, of earlier years such as in the 50s, 60s, and 70s for journals such as the Journal of Medicinal Chemistry; Steroids; and the Journal of Steroid Biochemistry and Molecular Biology.
For both the Pubmed search and directly checking the journals for every article, the reason for the date limitation is that these days there is too much stuff published to wade through everything. With the vast majority of it being useless.
That is not to say that nothing new is worth reading -- some of it certainly is -- but such a broad approach will turn up too much. It would take you a great deal of time to thumb through just the last year of the above three journals: in the same length of time you could thumb through an entire decade's worth of the journals in those past decades. With a far higher hit rate of useful stuff.