According to this paper[1], prior work has already established that 42% of the d...

According to this paper[1], prior work has already established that 42% of the dose variance is controlled by 3 SNPs. It isn't 100%, but 42% is not insignificant.

[1]: http://www.ncbi.nlm.nih.gov/sites/entrez?cmd=Search&db=p...

edit: Also on 23andMe's page regarding Wafarin:

  In January 2010 the FDA updated warfarin's label to say
  that information on these variants can assist physicians in 
  selecting a starting dose of the drug. The agency also 
  provided initial dosage recommendations for patients with 
  different variant combinations. The FDA does not, however, 
  require that genetic testing be done before prescribing
  warfarin.