Single-point phenytoin dosage predictions in Singapore Chinese

Phenytoin dosing is often difficult in a clinical situation because of the non-linear nature of phenytoin metabolism at therapeutic plasma concentrations. This study was performed to examine retrospectively five pharmacokinetic methods of adjusting phenytoin dosage based on a single steady-state plasma phenytoin concentration-dose pair in 66 epileptic Chinese children and adults. The methods compared included four fixed-parameter(s) methods (1 = the fixed Km (Michaelis-Menten constant) method; 2 = the fixed Vm (maximum rate of elimination) method; 3 = the fixed Vmax/Km method; 4 = the fixed S (slope of logarithmic growth model) method) and a Bayesian feedback method (method 5). Measures of bias or accuracy (mean error, percentage dose) and precision (root mean squared error, percentage dose) were 20.1/88.0, -1.85/21.5, 2.14/21.9, -1.07/21.1 and -1.98/22.2, respectively. Method 1 was significantly inferior to methods 2-5 with respect to accuracy and precision. The correlation between the predicted and observed doses was higher with methods 2-5 (r = 0.862, 0.855, 0.866 and 0.841, respectively) when compared to method 1 (r = 0.539). All methods had a sizeable number of poor predictions (range: 21.2% for method 4 to 37.9% for method 1), i.e. predictions with an error of greater than 20% of the dose. With respect to the frequency of poor versus good predictions, comparisons of all methods showed no significant differences.