Guidelines for measuring free drug concentrations in serum have become necessary due to the easy availability of these assays as a result of the introduction of commercial kits. The present study was performed to identify patients or groups of patients in whom the serum free phenytoin fraction varied from normal, such that they might benefit from measurement of serum free phenytoin. Three hundred fourteen samples submitted for routine phenytoin analysis were studied by enzyme-modified immunoassay technique (EMIT). Thirty-eight patients on phenytoin monotherapy and without other factors thought to affect protein binding of this drug had a mean (+/- SD) free phenytoin fraction of 9.8 +/- 1.8% of total concentration (mean serum albumin concentration 43.4 +/- 3.9 g/L). The free fraction was elevated by administration of comedications which are themselves highly protein bound, and in those patients who were hypoalbuminaemic (serum albumin less than 30 g/L). The groups studied were not mutually exclusive, but stepwise regression analysis showed that other factors known to affect serum albumin (e.g., age greater than 65 years, liver or renal disease, or pregnancy) did not, in themselves, produce a significant effect on free phenytoin fraction. Similarly, an elevated total serum phenytoin concentration was not a significant factor in producing an elevation in free phenytoin fraction.(ABSTRACT TRUNCATED AT 250 WORDS)