Efficient Parameter Estimation in Preclinical Animal Pharmacokinetic Studies

Ette, Ene I (1991) Efficient Parameter Estimation in Preclinical Animal Pharmacokinetic Studies. PhD thesis, University of Glasgow.

Full text available as:
[img]
Preview
PDF
Download (9MB) | Preview

Abstract

An estimation of the average value of pharmacokinetic parameters in a group of animals provides limited information if there is no good measure of the variability of each of the parameters. The traditional approach used in the analysis of animal pharmacokinetic data obtained from studies involving the use of small laboratory animals (rats or mice) in which each animal supplies only one concentration - time point does not provide this, nor can it assess the influence of physiology (or pathology) on pharmacokinetics. The consideration of variability within the same species during interspecies scaling has been advocated (Vocci & Farber, 1988). Thus, provision should be made for the estimation of variability inherent in an animal population in analysing data obtained by "destructive sampling". The NONMEM approach does, however, provide estimates of both average values of pharmacokinetic parameters and their statistical distribution within the population. In this thesis data were generated by simulation (assuming no covariance), and analysed using the NONMEM program. The efficiency of this approach is the focus of this thesis. Experimental error, number of samples taken, and the arrangement of samples in time are factors which must be taken into account in designing experiments for efficient parameter estimation. In addition, appropriate methods of data analysis must be used to extract the required information from the data. Simulated data sets were used to investigate the effect of various design features on the efficiency of parameter estimation using the one observation per animal design. In addition, the efficiency with which parameters could be estimated given a range of parameter values and variability was investigated. Several methods were used to determine the efficiency of parameter estimation. Prediction error (bias and precision) was useful in assessing the efficiency with which individual parameters were estimated. In addition, the 99% individual and joint confidence intervals containing the true parameter 95% of the time for all parameters were introduced as aids to judging the efficiency of estimation of individual and all parameters of a model, considered as a set. Confidence interval tables were constructed to reveal the influence of bias and standard error on parameter estimation. Also, the design number, a new statistic which combines the contributions of bias and precision in judging the efficiency of parameter estimation, was introduced to complement bias and precision, and confidence intervals methods of analysis. The design number also allowed the efficiency with which all parameters of a model were estimated as a set to be judged. The incidence of high pairwise correlations of parameter estimates was also taken into account in assessing the acceptability of estimates and the adequacy of model parameterization. Assuming IV bolus injection with the monoexponential pharmacokinetic model, simulation studies were carried out to investigate the influence of interanimal variability on the estimation of population pharmacokinetic parameters and their variances. The range of variability investigated was similar to that expected in real studies, and sampling was done at set times. The efficiency of estimation of the structural model parameters (Cl and V) was good, on average, irrespective of the variability in Cl and V. However, the estimation of these parameters was associated with negative bias which was attributed to the nature of the NONMEM program (i. e. estimation error since negative bias was also observed in subsequent studies in which dE was set to 0%). The variance parameters were mostly inefficiently estimated in this study and all other studies using the one observation per animal design. This was attributable to the lack of information in the data set about dE. When the effect of the arrangement of concentrations in time on parameter estimation was studied with the two sample point design, efficient parameter estimates were obtained when the first sample was obtained as early as possible (5 min. ) and the second sample was located at > 1.4 times the simulated t1/2 (84min. ) of the drug. When three or four sample points were used the exact location of the third or fourth sample was not critical to efficient parameter estimation. The efficiency of parameter estimation was investigated given a range of parameter values, concentration measurement error, and sampling schedules with the two compartment model parameterized as A, a, B, B and assuming IV bolus injection with animals sampled at set times. The parameters, considered as a set, were efficiently estimated when a was in the range of 2.0 to 4.0 h-1 and the A:B ratio in the range of 2.5 to 30.0. These results were attributed to the distribution of data points between the distribution and elimination phases of the plasma concentration - time profile. Concentration measurement error greater than 10% yielded variance parameter estimates with a greater degree of bias and imprecision.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: Medicine, Pharmacology
Date of Award: 1991
Depositing User: Enlighten Team
Unique ID: glathesis:1991-78383
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 28 Feb 2020 12:09
Last Modified: 28 Feb 2020 12:09
URI: http://theses.gla.ac.uk/id/eprint/78383

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year