Development of a methodology to make individual estimates of the precision of liquid chromatography-tandem mass spectrometry drug assay results for use in population pharmacokinetic modeling and the optimization of dosage regimens.

Gellért Balázs Karvaly,István Vincze,Krisztián Kovács,Roger W Jelliffe,Michael N Neely,Barna Vásárhelyi,Jed N Lampe

doi:10.1371/journal.pone.0229873

Abstract

The clinical value of therapeutic drug monitoring can be increased most significantly by integrating assay results into clinical pharmacokinetic models for optimal dosing. The correct weighting in the modeling process is 1/variance, therefore, knowledge of the standard deviations (SD) of each measured concentration is important. Because bioanalytical methods are heteroscedastic, the concentration-SD relationship must be modeled using assay error equations (AEE). We describe a methodology of establishing AEE's for liquid chromatography-tandem mass spectrometry (LC-MS/MS) drug assays using carbamazepine, fluconazole, lamotrigine and levetiracetam as model analytes. Following method validation, three independent experiments were conducted to develop AEE's using various least squares linear or nonlinear, and median-based linear regression techniques. SD's were determined from zero concentration to the high end of the assayed range. In each experiment, precision profiles of 6 ("small" sample sets) or 20 ("large" sample sets) out of 24 independent, spiked specimens were evaluated. Combinatorial calculations were performed to attain the most suitable regression approach. The final AEE's were developed by combining the SD's of the assay results, established in 24 specimens/spiking level and using all spiking levels, into a single precision profile. The effects of gross hyperbilirubinemia, hemolysis and lipemia as laboratory interferences were investigated. Precision profiles were best characterized by linear regression when 20 spiking levels, each having 24 specimens and obtained by performing 3 independent experiments, were combined. Theil's regression with the Siegel estimator was the most consistent and robust in providing acceptable agreement between measured and predicted SD's, including SD's below the lower limit of quantification. In the framework of precision pharmacotherapy, establishing the AEE of assayed drugs is the responsibility of the therapeutic drug monitoring service. This permits optimal dosages by providing the correct weighting factor of assay results in the development of population and individual pharmacokinetic models.

Highlights

It is increasingly accepted that the utility of therapeutic drug monitoring (TDM) can be improved significantly by integrating the assayed drug concentrations into population and individualized clinical pharmacokinetic models
In the framework of precision pharmacotherapy, establishing the assay error equations (AEE) of assayed drugs is the responsibility of the therapeutic drug monitoring service
Our aim is to present an approach to develop assay error equations (AEE’s) which allows the most precise calculation of the standard deviations (SD)’s for each result obtained in routine clinical LC-MS/MS drug assays. For this purpose we studied precision profiles using an in-house LC-MS/MS method developed for the analysis of carbamazepine (CBZ), fluconazole (FLU), lamotrigine (LAM) and levetiracetam (LEV) [2,16,17]

Summary

Introduction

It is increasingly accepted that the utility of therapeutic drug monitoring (TDM) can be improved significantly by integrating the assayed drug concentrations into population and individualized clinical pharmacokinetic models This can be optimized by using correct weighting of the data in making such models [1,2,3,4], enabling the maximally precise calculation of individual patient dosages using precision pharmacotherapy software. The relationship between SD and analyte concentration can be estimated quantitatively for each level measured by the analytical method This approach was first proposed by Ekins et al for an assay of aldosterone, revisited by Sadler, and proposed as the experimentally quantified error term in both population pharmacokinetic (PK) modeling and Bayesian posterior models in individual patients by Jelliffe and Tahani [9,10,11].

Objectives

Methods

Results

Conclusion