Prediction Study of Serum Concentration of Levetiracetam in Children with Epilepsy of Uygur Nationality in Xinjiang Based on Artificial Neural Network Model

OBJECTIVE To establish artificial neural network model for predicting the steady-state serum concentration of levetiracetam(LEV) in Uygur children with epilepsy in Xinjiang, so as to provide reference for clinical individualized drug administration. METHODS The steady-state serum drug concentration of levetiracetam in 330 cases Uyghur children with epilepsy in Xinjiang was determined, the clinical data was collected, and the artificial neural network was used to construct the prediction model of LEV serum concentration. RESULTS The results of model verification showed that the mean prediction error was (-2.15±6.97)%(<5%), the ratio of prediction error <±20% was 96.00%(47/50), mean absolute prediction error was (1.11±2.23)%(<5%), mean square prediction error was (52.16±106.81)%(<100%), and root mean square prediction error was (5.27±4.99)%(<10%) in the serum concentration of LEV in 50 children with epilepsy of Uygur nationality. There was a high correlation between the predicted and measured values of serum concentration after oral administration of LEV in children with epilepsy of Uygur nationality(r=0.986 1). These results showed that the model had good prediction performance and could be used to predict the serum concentration of LEV. CONCLUSION It is feasible to predict the serum concentration of LEV by using artificial neural network, and it can be used in the study of individual drug administration of levetiracetam to promote the rational use of drugs in clinic.