Study On Prediction Model Of Drug Concentration Of Aripiprazole And Its Metabolites

Object:(1)A method for the determination of aripiprazole(APZ)and its active metabolite dehydroaripiprazole(DAPZ)in human body was established by high performance liquid chromatography-tandem mass spectrometry.Application in the individualized drug administration of patients with antipsychotic disorders.(2)Discuss the influence of multiple interpolation method on the construction of GA-BP neural network prediction model.(3)Construct GA-BP neural network prediction model to predict the concentration of APZ+DAPZ.Methods:(1)HPLC-MS/MS was established.The chromatographic conditions were as follows: Xbrige BEH(C18 3.0 × 50 mm,2.5μm)column.Mobile phase: formic acid water/formic acid acetonitrile;Mass spectrum conditions: electrospray ion source;MRM positive ion scanning mode,and verify the other law;(2)174 patients who received regular APZ treatment in the People’s Hospital of Autonomous Region from July 2021 to August 2022 were retrospectively collected,The original data(including missing values)and missing data were treated with multiple interpolation respectively using stepwise regression method to screen variables that were correlated with the blood concentration of APZ+DAPZ(3)174 samples were randomly divided into a modeling group(139 cases)and a verification group(35 cases).Three GA-BP neural network APZ+DAPZ concentration prediction models were constructed using the modeling group.(1)Samples were modeled without multiple interpolation and screening variables;(2)The samples were selected by stepwise regression method for variable modeling;(3)Samples were modeled by multiple interpolation and screening variables.The precision and accuracy of the model were evaluated when the mean prediction error(MPE),mean absolute error(MAE)and percentage of prediction error(PE%)of the three model verification groups were less than ± 5%.Draw a scatter plot of correlation between predicted values and measured values.Results:(1)The standard curve regression equation y = 0.00626x-0.00507(r = 0.9997)was determined by HPLC-MS/MS.The linear range was 20-1200 ng/m L.The accuracy in/between batches was 96.28～99.22%,and the precision was 3.31～7.33%.DAPZ standard curve regression equation y = 0.00958 x-0.0137(r =0.9998),linear range of 20～1200 ng/m L,accuracy 95.89～99.22%,precision 1.44～5.15%.Matrix effect did not affect the quantitative analysis of analytes.(2)The factors that were correlated with the blood concentration of APZ+DAPZ by stepwise regression were height and the number of days of medication(all P < 0.05).The sample data were screened by stepped-regression method after multiple placement using MCMC method,and the factors related to the blood concentration of APZ+DAPZ were height,nationality,body weight,total bilirubin,serum glutamate pyruvate transaminase,red blood cells,neutrophils,platelets,days of medication,drug combination,gamma-glutamyltranspeptide,and drug dose(all P < 0.05).(3)MAE = 0.6895,MPE =-0.0926 and PE% less than ±5% were 100.0%,and the correlation coefficient between predicted and measured blood concentration was r = 0.997.The variables selected by stepwise regression method were established to verify that MAE =28.2560,MPE = 6.9352 and PE% less than ±5%were 34.30% and r = 0.751.MAE = 0.4853,MPE = 0.0957 and PE% less than ±5% were 100.0% in the model established by screening variables after 5 times of multiple interpolation,and the correlation coefficients between predicted and measured blood concentration were all 0.997.Conclusions:(1)In this study,HPLC-MS/MS technology was used to establish a method for the determination of APZ and DAPZ therapeutic drug concentrations in human body.(2)Multiple interpolation of missing data can be used as an effective method to filter variables.(3)GA-BP neural network prediction model can better predict the blood concentration of APZ+DAPZ.