Objective: To assess the factors affecting the free fraction of phenytoin in serum and the per formance of currently available equations used to predict normalized total phenytoin concentrations in Korean patients and to identify the most appropriate equation for Korean patients. Design : Patients with both total and free phenytoin concentrations measured were enrolled by a retrospective chart review. Setting : Samsung Medical Center in Seoul, Korea; a 1400-bed university teaching hospital. Participants : Total of 227 inpatients or outpatients were enrolled: 196 patients with norma albuminemia and 31 with hypoalbuminemia. Patients receiving drugs known to displace phenytoin or those with renal failure or abnormal liver enzymes were excluded in evaluating the performance of the equations (n = 115). Main outcome measures: Correlation between free fraction of phenytoin and factors such as age, sex, serum albumin, BUN, Scr, AST, ALT and concurrent medications was assessed by multiple regression analysis. Predictive performance of the equations was evaluated using bias (MPE ; mean prediction error) and precision (RMSE ; root mean squared error), the methods proposed by Sheiner and Beal. Results : Free fraction of phenytoin correlated well with all factors ($r^{2}$=0.51). However, only albumin and concurrent medication showed statistically significant correlation(p<0.05). Normoal buminemia patients exhibited a weak correlation ($r^{2}$=0.16) and hypoalbuminemia patients exhibited a strong correlation ($r^{2}$=0.69) between free fraction of phenytoin and the factors. Albumin and concur rent medication were shown to be the significant factors in these subgroups as well. The age especial ly had shown a significant correlation with free fraction of phenytoin in hypoalbuminemia patients. Bias (MPE) and precision (RMSE) of the equations of predicting normalized total phenytoin concentrations were as follows: the Sheiner-Tozer equation (MPE=2.1, RMSE=5.82), the revised Winter Tozer equation (MPE=-0.74, RMSE=5.09), the new equation using the affinity constant of 0.24 (MPE=-0.26, RMSE=5.07). Conclusions: There was a high correlation between free fraction of phenytoin and the factors affecting protein binding such as albumin and concurrent medication. In addition, the age was found to have a significant impact on free fraction of phenytoin in hypoalbuminemia patients. The affinity constant of 0.24 had smaller bias than 0.20 or 0.25 in Korean patients. However, it needs to be reassessed in a larger study with Korean patients.