An integrated pharmacometric dialysis model to evaluate the pharmacokinetic impact of renal replacement therapy
A. Broeker (1), S.G. Wicha (1)
(1) Dept. of Clinical Pharmacy, Institute of Pharmacy, University of Hamburg, Germany
Objectives: Renal replacement therapy (RRT) may alter the pharmacokinetic (PK) profile. Information on the impact of RRT on the PK is commonly obtained in small patient collectives. The aim of this project was to (i) develop an integrative, pharmacometric analysis method for PK studies in RRT in “small-n” studies, and (ii) evaluate the approach in simulation-estimation studies as well as using data from clinical trials with tigecycline and doripenem receiving continuous veno-venous haemodialysis (CVVHD) or -diafiltration (CVVHDF).
Methods: An integrated pharmacometric dialysis model was developed and implemented in NONMEM® 7.4.1, which simultaneously included all measureable dialysis specimens (plasma pre-/post filter, effluent, collected effluent). For a hypothetical drug (V: 30 L, CL: 2‑3 L/h, CLdial: 0.3‑3 L/h), study scenarios were created and either the integrated pharmacometric dialysis model or conventional models, evaluating only one dialysis specimen at a time, were utilised in a stochastic simulation and estimation study facilitated by PsN (Version 4.7.0). Based on published information, adsorption to the dialysate membrane and drug degradation were included as potential influential factors on the measured dialysis specimens. The performance of all models was compared considering the power to detect a dialysis clearance in a small-n study (10 dialysis patients and 10 non‑dialysis patients, rich sampling schedule), as well as attained accuracy, expressed as relative bias (rBias) and precision, expressed as relative root mean squared error (rRMSE) of all estimated parameters. The integrated model was applied to clinical datasets of tigecycline and doripenem.
Results: The integrated model was superior over the conventional approaches at the same total patient number: While in the plasma-pre-filter setting, a dialysis clearance had to be 60 % of the intrinsic clearance to be detectable with an 80 % power, the integrated approach was >1000-fold more sensitive and estimated dialysis clearance was unbiased (rBias: <0.2 %) and highly precise (rRMSE <2.5 %). In addition, the integrated model allowed quantifying binding to the dialysis cartridge, as well as degradation of the drug in the collected effluent without affecting accuracy and precision. In contrast, for conventional models, adsorption to the dialysis cartridge lead to significant bias in estimated dialysis clearance (-65 %) and even affected the structural PK model parameters (e.g. up to 13.5 % rBias for V). For typical dialysate collection intervals of 12 hours, degradation in the effluent caused bias of ‑20.4 % using the conventional approach. Instead, using the integrated model the degradation process was quantifiable (rBias 1.3 %, rRMSE 27.1 %) and dialysis clearance was estimated accurately and precisely (rBias 0.15 %, rRMSE 16.9 %). Results were confirmed in the application study with tigecycline, where plasma pre-/post filter and effluent were estimated simultaneously, and doripenem, where collected effluent was modelled as well.
Conclusions: The integrated dialysis model better exploits PK information in RRT studies than conventional approaches and was successfully applied to clinical studies with tigecycline and doripenem. Estimation of biased dialysis clearance due to adsorptive loss or drug degradation can be avoided while power, accuracy and precision of all PK parameters can be increased using the integrated pharmacometric dialysis model. Application of the approach to further studies is highly warranted to increase the information obtained from “small-n” studies in RRT.