Aim: To make programs for population pharmacokinetic analysis and to assess the ability of this method in pharmacokinetic parameter estimation and in the prediction of serum concentrations.
Methods: Data of amikacin as a model drug were collected from 42 neonates with 142 serum samples. A one-compartment open model was used to describe the kinetics of amikacin after the intravenous infusion. Following Sheiner's idea of population pharmacokinetics, we made the programs to evaluate population parameter and individual parameter. The target function minimality was obtained from Monte Carlo algorithm. The validation of the population analysis was performed using classic pharmacokinetic program 3p87 for antithesis. The predictability of the developed method was evaluated by computing precision and accuracy of serum concentration predicted using the parameter estimates.
Results: The stability of our self-made program was good. The population parameters obtained from this approach were in conformity with those from 3p87, and the interindividual variability was relatively small. For the learning sample and the validation sample, predicted and observed concentrations were all close with correlation coefficient 0.995 and 0.990, respectively. Most of predicted errors were found < +/- 1 mg/L, and RMSD and BIAS were 0.58 and -0.07 for the validation sample, respectively. The choice of blood sampling time was an important factor for the predictive performance. An early sampling time after the infusion was observed to be the best sampling time.
Conclusion: The estimation program of population parameter and individual parameter made by us ran stably, and allowed us to use sparse data to estimate population pharmacokinetic parameters. It provided accurate estimates of these parameters and satisfactory ability of serum concentration prediction. Therefore, it can be used for the population pharmacokinetic analysis and individualization of dosage regimen.