Tom C Zwart(1), Aline GJ Engbers (2), Ruth E Dam (3), Paul JM van der Boog (3), Johannes W de Fijter (3), Henk-Jan Guchelaar (1,4), Aiko PJ de Vries (3), Dirk Jan AR Moes (1,4)
(1) Leiden University Medical Center, Department of Clinical Pharmacy and Toxicology, Leiden, The Netherlands; (2) Leiden University, Leiden Academic Centre for Drug Research, Division of Systems Biomedicine and Pharmacology, Leiden, The Netherlands; (3) Leiden University Medical Center, Department of Nephrology and Leiden Transplant Center, Leiden, The Netherlands; (4) Leiden Network for Personalised Therapeutics, Leiden, The Netherlands
Objectives: Iohexol plasma clearance (IPC)-based glomerular filtration rate (GFR) estimation is a promising strategy for renal function evaluation. This method, which encompasses a single intravenous injection of iohexol, is particularly viable for populations without chronic kidney disease in which estimated GFR is not accurate enough or in which 24 h creatinine clearance is not feasible [1]. Current IPC-based GFR estimation methods typically rely on extrapolation of the iohexol area under the concentration-time curve (AUC) of only the linear terminal elimination phase, mostly using linear regression and the Bröchner-Mortensen correction to estimate the full AUC [2, 3]. The iohexol AUC divided by the dose then yields its plasma clearance, which reflects the patients’ GFR. The need to obtain one or multiple blood samples up to 8 h postdose, however, poses a drawback and has hampered implementation in routine clinical care. Here, a population pharmacokinetic (popPK) model and limited sampling strategy (LSS) were developed to provide a pragmatic method for IPC-based GFR estimation.
Methods: Blood samples (n=328) drawn at 5 min to 4 h postdose were available from 49 renal transplant donor candidates and renal transplant recipients routinely screened at Leiden University Medical Center. Nonlinear mixed-effects modelling was applied to develop a popPK model using NONMEM, and the PsN Toolkit[4] and Piraña[5] as modelling environment. Graphics and LSS statistics were performed in R. The FOCE-I method was applied throughout the model development. Model selection was based on a statistically significant change in the objective function value (dOFV) between a modified model and its precursor, with dOFV>6.63 (p<0.01, df=1, χ2 distribution) resulting in selection of the modified model. Model evaluation was performed using standard diagnostic plots and prediction corrected visual predictive checks (VPC). Internal validation of the final model was performed using a bootstrap analysis (n=1000). The final model was used to develop LSS based on the individual predicted GFR. The individual predicted GFR as calculated with the final model and the full dataset (GFRfull) was compared to the GFR as calculated with 1-4 samples drawn during the first 3 h postdose (GFRlss). LSS predictive performance was assessed using the Pearson correlation coefficient (r2), mean percentage prediction error (MPE), root mean squared percentage prediction error (RMSE) and the percentage of GFRlss exceeding the 5% margins around the GFRfull.
Results: Iohexol pharmacokinetics were best described by a 2-compartmental first order elimination model with proportional residual error and a full variance-covariance matrix of random effects. Total body clearance (CL), intercompartmental clearance (Q) and the volumes of distribution of the central (Vc) and peripheral (Vp) compartments were 4.89 L h-1 (6% RSE), 7.26 L h-1 (25%), 9.20 L (6%) and 5.48 L (14%), respectively. Interindividual variability for CL, Q, Vc and Vp was 34.4% (14% RSE; 0% shrinkage), 86.2% (18%; 11%), 35.2% (12%; 6%) and 41.7% (44%; 9%), respectively. Random residual variability was 5.4% (42%; 25%). The VPC of the final model showed a complete overlap between predicted and observed intervals. The bootstrap analysis showed adequate parameter reliability. The best GFRlss for one, two, three and four sample-based strategies were iohexol C2h (r2: 0.968; MPE: 3.04%; RMSE: 2.52%; <5% discordance: 73.47%) , C30min,3h (0.997; 0.97%; 0.77%; 95.92%), C5min,30min,3h (0.999; 0.67%; 0.41%; 95.92%) and C30min,1h,2.5h,3h (0.996; 0.62%; 0.60%; 97.96%), respectively. Two LSS were of particular clinical interest; C5min,2h,3h and C5min,1h,2h,3h, as these could provide options for blood draw alignment with abbreviated AUC-based TDM of tacrolimus (Ctrough,2h,3h)[6] and mycophenolic acid (Ctrough,1h,2h,3h)[7, 8] in transplant recipients. Both C5min,2h,3h (0.993; 0.16%; 0.81%; 95.92%) and C5min,1h,2h,3h (0.997; 0.36%; 0.79%; 95.92%) showed excellent predictive performance.
Conclusions: The iohexol popPK model and LSS provide accurate and pragmatic options for IPC-based GFR estimation and pave the way for implementation of this method in routine clinical care. This approach fills a clear gap for clinical situations in which current GFR estimation methods are not accurate enough or not feasible.
References:
[1] Delanaye P, Melsom T, Ebert N, Back SE, Mariat C, Cavalier E, et al. Iohexol plasma clearance for measuring glomerular filtration rate in clinical practice and research: a review. Part 2: Why to measure glomerular filtration rate with iohexol? Clin Kidney J 2016; 9(5): 700-4
[2] Delanaye P, Ebert N, Melsom T, Gaspari F, Mariat C, Cavalier E, et al. Iohexol plasma clearance for measuring glomerular filtration rate in clinical practice and research: a review. Part 1: How to measure glomerular filtration rate with iohexol? Clin Kidney J 2016; 9(5): 682-99.
[3] Brochner-Mortensen J. A simple method for the determination of glomerular filtration rate. Scand J Clin Lab Invest 1972; 30(3): 271-4.
[4] Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed 2005; 79(3): 241-57.
[5] Keizer RJ, van Benten M, Beijnen JH, Schellens JH, Huitema AD. Pirana and PCluster: a modeling environment and cluster infrastructure for NONMEM. Comput Methods Programs Biomed 2011; 101(1): 72-9.
[6] Woillard JB, de Winter BC, Kamar N, Marquet P, Rostaing L, Rousseau A. Population pharmacokinetic model and Bayesian estimator for two tacrolimus formulations – twice daily Prograf and once daily Advagraf. Br J Clin Pharmacol 2011; 71(3): 391-402.
[7] Langers P, Press RR, Inderson A, Cremers SC, den Hartigh J, Baranski AG, et al. Limited sampling model for advanced mycophenolic acid therapeutic drug monitoring after liver transplantation. Ther Drug Monit 2014; 36(2): 141-7.
[8] Staatz CE, Tett SE. Maximum a posteriori Bayesian estimation of mycophenolic acid area under the concentration-time curve: is this clinically useful for dosage prediction yet? Clin Pharmacokinet 2011; 50(12): 759-72.
Reference: PAGE 28 (2019) Abstr 8928 [www.page-meeting.org/?abstract=8928]
Poster: Drug/Disease Modelling - Other Topics