Evaluation of Binding Equation Method for Unbound Serum Concentration Prediction of Valproic Acid in Polytherapy...

In a previous study, we determined the in vivo population binding parameters of valproic acid (VPA) to serum proteins at single dose in nine healthy young and defined a binding equation that was derived from the Scatchard equation. Schever et al. and Yu also determined the in vivo population binding parameters of VPA in patients with epilepsy receiving VPA monotherapy or polytherapy. In this study, we retrospectively evaluated the ability of a binding equation using each of population mean binding parameters of Kodama et al. (Method 1), Scheyer et al. (Method 2), or Yu (Method 3) to predict the unbound serum VPA concentration in 23 pediatric patients with epilepsy receiving polytherapy. Mean prediction error, mean absolute prediction error (MAE), and root mean squared error (RMSE) were separately calculated for each method and served as a measure of prediction bias and precision. All three methods were significantly biased, showing a bias to overpredict unbound serum VPA. The MAEs and RMSEs for Methods 1 and 2 were similar in magnitude (Method 1: MAE = 16.5, RMSE = 23.1; Method 2: MAE = 14.0, RMSE = 17.5). On the other hand, each value for Method 3 was extremely high (MAE = 32.8 and RMSE = 36.3). Method 3 is apparently inferior to other two methods in accuracy and precision. For Methods 1 and 2, similar tendency to overprediction was observed in total concentration above 400 &mgr;mol L(minus sign1). These two methods may be useful to patients with total VPA lower than 400 &mgr;mol L(minus sign1).