A method of predicting the risk of a patient for developing phenytoin-induced adverse drug reactions (ADRs), including Stevens-Johnson syndrome (SJS), toxic epidermal necrolysis (TEN), or drug reactions with eosinophilia and systemic symptoms (DRESS) is disclosed. Genetic polymorphisms of CYP2C genes (including rs1057910 (CYP2C9*3) and rs3758581 on CYP2C19), and HLA alleles (including HLA-B*1502, HLA-B*1301, and HLA-B*5101) can predict adverse reactions caused by phenytoin or fosphenytoin. Accordingly, the present invention provides a kit to assess the risk of a patient for developing adverse reactions in response to phenytoin-related drugs, which comprises the determination of the presence of a specific allele selected from the group consisting of rs1057910 (CYP2C9*3), rs3758581 on CYP2C19, HLA-B*1502, HLA-B*1301, and HLA-B*5101, wherein the presence of at least one allele is indicative of a risk for the adverse drug reactions.