A simple, novel approach to phenytoin dose adjustment recently proposed by Wagner has been compared with an established Bayesian method. A data set comprising a minimum of two steady-state concentration/dose pairs from each of 43 adult epileptic patients was used. Prediction error analysis demonstrated that the two methods were equally precise but that the Wagner method tended to underpredict concentrations. The slope of the semilogarithmic relationship between concentration and dose was found to be 54% higher than originally reported by Wagner. It may therefore be necessary to adjust Wagner's equation for different patient populations before this method is used in routine clinical practice.