The aim of the present study was to develop a population pharmacokinetic model of carbamazepine from routine therapeutic drug monitoring data. Steady-state carbamazepine plasma concentrations determined by homogenous enzyme immunoassay technique, dosing history including cotherapy, schedule of blood sampling, and patients' demographic characteristics were collected retrospectively from patients' chart histories. A one-compartment model was fitted to the data using nonlinear mixed effects modeling. The influence of weight, age, gender, smoking, allergy, carbamazepine daily dose, and cotherapy on clearance (CL/F) was evaluated. Additionally, bioavailability of controlled-release relative to immediate-release tablets was assessed. Two hundred sixty-five patients (423 concentrations) were used to develop a population pharmacokinetic model. The population estimate of CL/F from the base model was 5.14 L/h with interindividual variability of 50.20%. Patients' gender, age, smoking, allergy, cotherapy with lamotrigine and benzodiazepines had no effect on CL/F. Patient weight (WT), daily carbamazepine dose (DCBZ), daily dose of phenobarbitone (DPB) and valproic acid (VPA), when its daily dose exceeded 750 mg, significantly influenced CL/F and were included in the final model:[equation: see text] where VPA is 1 if dose is greater than 750 mg or 0 otherwise. No difference in bioavailability of carbamazepine between controlled- and immediate-release tablets was detected. The model predictions in the validation set had no bias and satisfactory precision. The model can be used for estimation of carbamazepine CL/F in individual patients in the postautoinduction phase and for selection of optimum dosing regimen in routine patient care.