Population Pharmacokinetic Estimation Of Prograf Apparent Clearance Through Multiple-Linear Regression

Objective: The goal was to estimate the value of Prograf apparent clearance (CL/F) in adult liver transplant recipients and found out the factors affecting CL/F as well as develop a population pharmacokinetic model for CL/F by statistical methods.Method: Prograf blood drug concentration, individual condition, sampling time, and biochemical inspection were obtained from 21 liver transplant recipients (16 male and 5 female) retrospectively. Then established Concentration database using Microsoft Access, in which two new tables, Information (including Name, ID, Gender, Operation time, diagnosis, diabetes mellitus, hypertension) and Data (including ID, ID + Sampling point, blood drug concentration, sampling time and biochemical indices), were built. All data would be analyzed in Concentration database. After that, develop a population pharmacokinetic model for CL/F through three different methods. Method one, introduce data into multiple-linear regression model directly; Method two, transform biochemical indices according to normal range in test sheet, if the value is normal, set 1, otherwise set 0; Method three, compute natural logarithm (ln) of CL/F and biochemical indices, such as GPT. Hereafter, develop statistical model through multiple-linear regression provided by SPSS. Assess the correlation among all factors and CL/F in terms of F test, and develop model through stepwise method provided by SPSS. Finally, introduce 5 external datasets (male 4 and female 1, including 24 simple points) into developed model to predict the value of CL/F, check errors between predicted values and measured values, and calculate prediction error percentage (PE%) andabsolute prediction error percentage (|PE%|).Result: model from method one is Cl/F=-0.047+0.002xALB +0.000xTime (the coefficients of AST and ALB are too little to produce validation model); model from method two is CL/F=0.11+0.21xAST; model from method three is Ln(CL/F)=-6.167-0.016xTime+1.438xLnALB+0.288 xLnGGT-0.515xLnCrea-0.368xLnAST. It is proved that FK506 apparent clearance is related to AST, ALB, GGT and post-transplant time, while is independent to other factors, such as age and gender. For method two, mean values of errors in two groups are 12.25% and 23.69%, respectively. The errors are so big that method two is invalidation. For method three, mean values are 2.33% and 8.41% respectively, therefore, the model is robust enough to guide the rational use of drug and customize individual dosage regimen of FK506.Conclusion: if fixed effects and random effects have been taken into account adequately, optimized population pharmacokinetic model could be developed by multiple-linear regression, which would provide theoretical foundation for regulating dosage regimen in clinic.