Background: The majority of laboratories measure total phenytoin concentration for therapeutic drug monitoring. However, there are substantial interindividual variations in free phenytoin concentrations, the pharmacologically active component. Methods: We describe the process and data used to implement monitoring of free phenytoin only in an urban medical center. Over a 6-week period, total and free phenytoin concentrations were measured, clinical charts reviewed, and indications for alterations in the percentage of free phenytoin fraction were determined. Results: Of the 189 phenytoin requests from 139 patients, 136 data points were analyzed. Free phenytoin concentrations were 6.8-35.3%, with 50% outside the expected range of 8-12%. Clinical indications likely responsible for variations were hypoalbuminemia, drug interactions, uremia, pregnancy, and age. Overall, 30% of patients demonstrated a discrepancy between therapeutic, subtherapeutic, or supratherapeutic concentrations between free and total phenytoin concentrations. The largest discordance (53%) occurred in the patient group with free phenytoin <8% or >12%. Coclusions: This study supports previous clinical findings that monitoring total phenytoin is not as reliable as free phenytoin as a clinical indicator for therapeutic and nontherapeutic concentrations. Thus, we recommend that therapeutic monitoring should use free phenytoin concentrations only. (C) 2000 American Association for Clinical Chemistry.