Identification and quantification of variability in the microdialysis technique using mixed-effects modelling within an integrated in vitro and ex vivo approach: a linezolid case study.
Mueller, F. (1, 2), Ilia, L. (1, 2), Bindellini, D. (1,2), Mikus, G. (3), Aulin, L.B.S. (1), Michelet, R. (1), Kloft, C. (1)
(1) Dept. of Clinical Pharmacy and Biochemistry, Institute of Pharmacy, Freie Universitaet Berlin, Germany, (2) and Graduate Research Training program PharMetrX, Germany, (3) Department of Clinical Pharmacology and Pharmacoepidemiology University Hospital Heidelberg, Heidelberg, Germany
Objectives:
In the treatment of infections, the unbound drug concentration at the target site, i.e. the interstitial space fluid (ISF), is driving the antimicrobial effect. The quantification of this concentration is essential to evaluate and improve dosing [1] and can be determined using the microdialysis (µD) sampling technique, where molecules diffuse across a semipermeable membrane into the µD catheter. A constant flow through the catheter does not allow an equilibrium between perfusate and ISF concentration and these catheters need to be calibrated. Catheter-specific calibration is commonly conducted by retrodialysis, using the delivered drug fraction of a highly concentrated perfusate into the tissue to calculate the relative recovery (RR), assuming recovery is equal to delivery. However, high variability in RR has been observed in several clinical studies, there among a study investigating the pharmacokinetics (PK) of linezolid (LIN) in obese and non-obese patients [3]. As the RR is used to calculate ISF concentrations from µD concentrations, high RR-related variability is making the quantitative use and interpretation of µD data challenging [2].
To support the optimisation of therapy based on µD measurements by reducing variability in µD trials, the influence of (i) different catheter-surrounding matrices in vitro and (ii) the positioning of µD catheters ex vivo on RR-related variability was investigated and analysed utilising (non-)linear mixed-effects ((N)LME) modelling.
Methods:
To assess RR-related inter- and intracatheter variability in Ringer’s solution (RS) and artificial ISF (diluted human plasma, aISF) for LIN a dynamic in vitro µD system (dIVMS) was used. Three µD catheters (60 mDialysis) were used (n=4) to sample in vitro from an in vivo mimicked PK profile of LIN followed by retrodialysis. The data was analysed by NLME modelling in NONMEM (version7.4.3) and inter- and intracatheter variability was quantified. The variability was implemented as fixed values in a previously developed clinical NLME model of LIN [3] and the remaining parameters were re-estimated. Model performance was assessed by goodness of fit plots (GOF), visual predictive checks (VPC) and objective function value (OFV).
The impact of catheter positioning was investigated by performing retrodialysis ex vivo in human s.c. adipose tissue close to the dermis and in deep s.c. adipose tissue with 6 catheters. The data were analysed using LME modelling in R utilising the “lmer4” package to quantify intercatheter and residual unexplained variability (RUV).
Results:
The in vitro mimicked PK profile was best described by a one-compartment model with linear elimination and linked to three µD measurement compartments accounting for RR per catheter. Estimated ISF compartment volume of distribution and clearance were 91.1 mL (RSE=3.5%) and 155 µL/min (5%), in line with experimental in vitro settings (flask volume of 100 mL, pump rate of 145 µL/min). Inter- and intracatheter variability were estimated to be low (aISF: 4.10% CV (RSE=17%) and 2.55% CV (47%), RS: 2.80% CV (23.7%) and 2.60% CV (32.1%), respectively), compared to 26.1% CV (19.5%) and 27.2% CV (10.4%) for the clinical NLME model. Informing the clinical NLME model with in vitro variability decreased the predictive performance of individual µD concentrations (VPCs, GOF plots). Additionally, fixing the variability allowed the estimation of RR-related interindividual variability (12.6% CV, RSE=19%) and RUV (56.2% CV, 13%), decreasing the OFV by 3329, however not improving predictive performance of individual µD concentrations compared to the clinical NLME model.
In the ex vivo setting quantified variability using LME modelling, split into inter-catheter and RUV (including intracatheter variability), of 9.20% CV (RUV=7.35%) close to the dermis and 21.4% CV (RUV=10.4%) in the deep s.c. adipose tissue was determined compared to 16.0% CV (RUV=18.1%) in vivo.
Conclusions:
The results indicated that target site characteristics contribute more to the overall variability than catheter-related parameters. This finding was corroborated by the decrease in accuracy of individual predicted µD concentrations upon fixing estimated to in vitro derived RR-related variability in the NLME model. Furthermore, the retrodialysis variability in the deep s.c. adipose tissue was comparable with those obtained from in vivo studies, which illustrated the impact and importance of catheter position in clinical trials.
References:
[1] Committee for Human Medicinal Products (CHMP). Guideline on the use of pharmacokinetics and pharmacodynamics in the development of antibacterial medicinal products. EMA (2016) (last access 10.03.2023)
[2] D. Busse et al., Eur. J. Pharm. Sci. 1;157:105607 (2021)
[3] L. Ehmann et al., Clin. Microbiol. Infect. 26;1222-1228 (2020)