Preliminary data have suggested that phenytoin systemic clearance may increase during initial therapy in critically ill patients. The objectives for this study were to model the time-variant phenytoin clearance and evaluate concomitant changes in protein binding and urinary metabolite elimination. Phenytoin was given as an intravenous loading dose of 15 mg/kg followed by an initial maintenance dose of 6 mg/kg/day in 10 adult critically ill trauma patients. Phenytoin bound and unbound plasma concentrations were determined in 10 patients and urinary excretion of the metabolite p-hydroxyphenyl phenylhydantoin (p-HPPH) was measured in seven patients for 7 to 14 days. A Michaelis-Menten one-compartment model incorporating a time-variant maximal velocity (Vmax) was sufficient to describe the data and superior to a conventional time-invariant Michaelis-Menten model. Vmax for the time-variant model was defined as V'max + Vmax delta (1 - e(-kindt)). Vmax infinity is the value for Vmax when t is large. The median values (ranges) for the parameters were Km = 4.8 (2.6 to 20) mg/L, Vmax infinity = 1348 (372 to 4741) mg/day, and kind = 0.0115 (0.0045 to 0.132) hr-1. Phenytoin free fraction increased in a majority of patients during the study period, with a binding ratio inversely related to albumin. Measured urinary p-HPPH data were consistent with the proposed model. A loading and constant maintenance dose of phenytoin frequently yielded a substantial, clinically significant fall in plasma concentrations with a pattern of apparently increasing clearance that may be a consequence of changes in protein binding, induction of metabolism, or the influence of stress on hepatic metabolic capacity.