Clinical factors, demographic variables and variations in two genes, CYP2C9 and VKORC1, have been shown to contribute to the variability in warfarin dose requirements among adult patients. Less is known about their relative importance for dose variability in children. A few small studies have been reported, but the results have been conflicting, especially regarding the impact of genotypes. In this article, we critically review published pharmacogenetic-based prediction models for warfarin dosing in children, and present results from a head-to-head comparison of predictive performance in a distinct cohort of warfarin-treated children. Finally we discuss what properties a prediction model should have, and what knowledge gaps need to be filled, to improve warfarin therapy in children of all ages.