Try referencing the CPIC guidelines for info. I'm not at home, so I can't send the link.
From what I remember, there is really mediocre evidence for support of the use of CYP2C9 and VKORC1 genotyping for dosing. And my understanding is that it is mainly used for initial dosing. You will still use INR monitoring to guide dose adjustments. Warfarin algorithms (warfarindosing.org) I think were recommended over simple Pgx testing. Many of the studies though look at time in therapeutic range and they don't look at clinical outcomes.
Cost is the cost of the test, depending on who's performing the test. Genotyping is relatively inexpensive these days, but insurance certainly isn't going to pay for it unless solid evidence for improved clinical outcomes is shown.
In class we got to do some basic Pgx testing for ourselves. Our school partners up with a diagnostics company to provide genotyping results for a few metabolizing enzymes and some clinically relevant receptors/enzymes -- about 20-25 genes in all.
Our hospital is testing the implementation of CYP2C19 testing to guide antiplatelet selection (clopidogrel v. prasugrel) for patients who present with ACS and undergo PCI. Kinda interesting.