Phenytoin dosing is critical in cancer patients as to decreased absorption secondary to chemotherapy-induced gastrointestinal toxicity, increased phenytoin metabolism in the liver secondary to chemotherapy, extreme patient profile that falls outside the predicted pharmacokinetic population, frequent hypoalbuminaemia and polydrug treatment. A retrospective study to assess the variability of free phenytoin and the free fraction of phenytoin, as well as the influence of comedication on these parameters was performed in cancer patients by using a population approach. Two hundred fifty-eight data pairs of total phenytoin and free phenytoin were analysed from 155 cancer patients on stable phenytoin using non-linear mixed-effect modeling (NONMEM). Total and free phenytoin were determined using a fluorescence polarization immunoassay. An extensive model building procedure was subsequently used for covariate testing on the free fraction of phenytoin. Mean total phenytoin concentration was 11.7 mg/l, free phenytoin 1.25 mg/l and phenytoin free fraction 0.107. Free phenytoin was <1 mg/l on 132 occasions (51.2%) and >2 mg/l on 37 occasions (14.3%). Total and free phenytoin were significantly correlated (r(S)=0.827, P<0.01). The free fraction of phenytoin was independent of time after drug intake. Serum albumin concentrations and comedication with valproic acid or carbamazepine were identified by NONMEM as significant determinants of phenytoin free fraction. Co-medication with valproic acid and carbamazepine led to a 52.5% and 38.5% increase of the free fraction of phenytoin, respectively, and a 10 g/l decrease of serum albumin to a 15.1% increase of the free fraction of phenytoin. Phenytoin pharmacokinetics could reliably be estimated from oral doses and steady-state concentrations of protein-bound and free phenytoin. The variability in the free fraction of phenytoin could partly be explained by the influence of albumin concentrations and antiepileptic comedication. Significant alterations of the free fraction of phenytoin and free phenytoin by co-administration of valproic acid or carbamazepine suggest therapeutic drug monitoring of free phenytoin to be of potential benefit in cancer patients.