Two multisource phenytoin products (Antepil^® and Comitoína^®) available for oral administration in Uruguay were authorized to continue in the market provided they could demonstrate similar pattern of efficacy and safety. To accomplish this, fifty-seven epileptic patients under chronic treatment with one of the two brands were enrolled in a bioequivalence parallel design study in which saliva concentration-time profiles of phenytoin were evaluated twice in each subject. Maximum and mean steady state concentrations (C_max and C_ss) and peak trough fluctuation (PTF) were obtained from the saliva concentration-time profiles. Two 90% confidence intervals (90%CI) for the ratio of the geometric brand means were calculated for each parameter using the total and the residual variance. The results show a narrower 90%CI when the residual instead of the total variance is used, making it possible to include the 90%CI obtained from the residual variance within the bioequivalence interval [0.80-1.25], in opposition to the wider 90%CI obtained from the total variance for the parameters C_ss and C_max. Regarding PTF, as the residual variability was similar to the total one, none of the 90%CI could be included within the bioequivalence interval. However, for parameters with high intrinsic coefficient of variation, a wider bioequivalence interval has been accepted by the WHO. The results obtained in this study allow us to conclude that Antepil^® and Comitoina^® are bioequivalent. Moreover, the procedure of parallel assay, with replicate evaluation of drug exposure, should be considered as a valuable solution to demonstrate bioequivalence of multisource drug products.