To the Editor In their August 2014 editorial, “Warfarin, Genes, and the (Health Care) Environment,” Kazi and Hlatky1 point out a number of plausible reasons why pharmacogenetic data in this setting did not make a significant difference in the clinical outcomes measured. They then argue that effects on “surrogate markers, such as drug levels” are insufficient to recommend widespread pharmacogenetic testing, yet such testing, like any diagnostic testing, should not be performed in a “widespread” way but only when clinically indicated.