SINGLE-POINT PHENYTOIN DOSAGE PREDICTIONS IN SINGAPORE CHINESE

Abstract

Phenytoin is an antiepileptic drug used widely in the control of seizures. It is cleared from the body mainly by hepatic metabolism which becomes saturated at therapeutic plasma level of the drug. Besides the nonlinear kinetics of phenytoin, its narrow therapeutic index and great interindividual variability in its metabolism are additional contributing factors to much difficulties in its dosage adjustments under clinical situations. Several methods have been proposed to adjust phenytoin dosage based on single-point steady-state drug concentration -dose (Css-dose) pair, using Michaelis-Menten model which is characterised by the two pharmacokinetic parameters, Vmax (the maximum rate at which phenytoin can be metabolised) and Km (the dissociation constant of the enzyme-phenytoin complex). Basically, these methods involve estimation of population Vmax and Km, then use this these estimates and one previous Css-dose data obtained from a subject to predict a dosage that will result in a particular plasma drug level for the subject. This study examined and evaluated the predictibility of five of the proposed methods using the Css-dose data obtained from 66 Singapore Chinese patients. The first method proposed by Richens and Dunlop assumes constant Km for all patients which together with one previous Css-dose pair can give individual estimates of Vmax to be used in subsequent dosage predictions. The second method proposed by Chiba et. al. assumes constant Vmax for patients of a particular age group. This age group-related population Vmax and one previous Css-dose pair can give individual estimates of Km to be used in subsequent dosage predictions. The third method proposed by Martin et. al. makes predictions based on the ratio of a previous observed Css to the expected Css, This method adjusts both Vmax and Km of individuals based on this ratio. The fourth method proposed by Wagner assumes linear relationship between dose and logarithm of Css. The proportionality constant, S, obtained as a population parameter, and a previous Css-dose pair are used in subsequent dosage predictions. The fifth method proposed by Sheiner et. al., known as Bayesian feedback method, makes dosage prediction based on individual's Css and prior information about the population pharmacokinetic parameters of the drug, using Bayesian forecasting technique. Mean prediction error (me) and square root of mean squared prediction error (rmse) were separately calculated for each method to serve as a measure of prediction bias (or accuracy) and precision respectively. It was noted that for all the methods, errors incurred in predictions patients were quite large (>60mg/day). It was thus that intensive drug concentration monitoring in patient is of importance in phenytoin therapy and for some concluded individual whichever method chosen for dosage prediction would only serve as an aid to more accurate and precise dosage adjustment. Among the various methods evaluated, Bayesian feedback method was found to be the most reliable in dosage predictions.