Tom C Zwart (1), Dirk Jan AR Moes (1,2), Paul JM van der Boog (3), Nielka P van Erp (4), Johan W de Fijter (3), Henk-Jan Guchelaar (1,2), Ron J Keizer (5), Rob ter Heine (4)
(1) Leiden University Medical Center, Department of Clinical Pharmacy and Toxicology, Leiden, The Netherlands; (2) Leiden Network for Personalised Therapeutics, Leiden, The Netherlands; (3) Leiden University Medical Center, Department of Internal Medicine (Nephrology), Leiden, The Netherlands; (4) Radboud University Medical Center, Radboud Institute for Health Sciences, Department of Pharmacy, Nijmegen, The Netherlands; (5) InsightRX, San Francisco, CA, USA
Background: The mammalian target of rapamycin (mTOR) inhibitor everolimus is increasingly prescribed in renal transplantation. Its extensive interindividual pharmacokinetic (PK) variability and narrow therapeutic index necessitate therapeutic drug monitoring (TDM)-guided dose adaptation. Whereas most centers rely on conventional TDM, more advanced pharmacometric model-based approaches are available [1]. Here, we aimed to externally validate a previously published mechanistic population PK model intended for model-informed precision dosing (MIPD) of everolimus in renal transplantation and oncology [2].
Methods: The validation dataset included trough concentrations (C0) and short PK curves (0-6h), collected retrospectively from routine clinical TDM data. The evaluated population PK model accounted for saturable erythrocyte binding of everolimus in whole blood. Plasma PK parameters were estimated using paired everolimus and haematocrit observations, enabling haematocrit-normalization of whole blood concentrations. Overall model appropriateness was evaluated with a prediction-corrected visual predictive check (VPC) [3]. A fit-for-purpose analysis was conducted to evaluate model performance for its intended clinical use, namely to predict a future C0 or area under the concentration-time curve (AUC) based on the prior C0 or AUC and patient characteristics available at that point. Model performance for predicting C0 and AUC was evaluated separately and stratified on the early (≤6 months) and late (>6 months) posttransplant periods. The fit-for-purpose prediction bias and precision were expressed as the mean normal (MPE), mean absolute (MAPE) and root mean squared prediction error (RMSE) [4]. The percentage of predictions within ±10-30% of observations (P10-P30) were assessed to aid clinical interpretation. Lastly, TDM target attainment of observed and haematocrit-normalized C0 and AUC were assessed, and the added value of allometric scaling and absorption interoccasion variability was evaluated using reduced versions of the model. Pharmacokinetic modelling was performed with nonlinear mixed-effects modelling software package NONMEM V7.3, using Pirana and Perl speaks NONMEM as modelling interface [5].
Results: A total of 4132 PK observations were available from 173 renal transplant recipients. The prediction-corrected VPC showed adequate overall model appropriateness. The fit-for-purpose performance was good for C0 in the early (MPE: 8.1 ± 2.5%, MAPE: 26.8 ± 2.1%, P10: 30.9%, P30: 71.1%) and late (MPE: 4.7±2.0%, MAPE: 25.4 ± 1.4%, P10: 28.2%, P30: 70.7%) period, whereas AUC predictive performance was excellent in the early (MPE: -9.7 ± 5.1%, MAPE: 13.3 ± 3.9%, P10: 50.0%, P30: 90.6%) and late (MPE: -0.13 ± 4.8%, MAPE: 13.3 ± 2.8%, P10: 37.5%, P30: 95.8%) period. Haematocrit-normalization yielded a different TDM target attainment decision on 16.4% and 15.5% of occasions as compared to TDM based on the observed whole blood C0 and AUC. Inclusion of allometric scaling or interoccasion variability in the model did not affect its performance, as the RMSE varied only 0.1-2.8% between the full and reduced model versions. This validation study was conducted in renal transplant recipients exclusively. Our results can, however, likely be extrapolated to the oncology setting as everolimus PK in these populations is largely similar [2].
Conclusion: Our model shows potential for MIPD of everolimus, as we demonstrated that it can accurately and precisely predict future everolimus exposure. A prospective evaluation is, however, needed to determine the added value of this approach over conventional everolimus TDM. Lastly, TDM based on haematocrit-normalized whole blood concentrations poses a promising strategy to further optimize everolimus therapy and warrants further investigation.
References: [1] Keizer RJ, Ter Heine R, Frymoyer A, Lesko LJ, Mangat R, Goswami S. Model-Informed Precision Dosing at the Bedside: Scientific Challenges and Opportunities. CPT Pharmacometrics Syst Pharmacol 2018; 7(12): 785-7.
[2] Ter Heine R, van Erp NP, Guchelaar HJ, de Fijter JW, Reinders MEJ, van Herpen CM, et al. A pharmacological rationale for improved everolimus dosing in oncology and transplant patients. Br J Clin Pharmacol 2018; 84(7): 1575-86.
[3] Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Prediction-corrected visual predictive checks for diagnosing nonlinear mixed-effects models. AAPS J 2011; 13(2): 143-51.
[4] Sheiner LB, Beal SL. Some suggestions for measuring predictive performance. J Pharmacokinet Biopharm 1981; 9(4): 503-12.
[5] Keizer RJ, Karlsson MO, Hooker A. Modeling and Simulation Workbench for NONMEM: Tutorial on Pirana, PsN, and Xpose. CPT Pharmacometrics Syst Pharmacol 2013; 2: e50.
Reference: PAGE () Abstr 9350 [www.page-meeting.org/?abstract=9350]
Poster: Drug/Disease Modelling - Other Topics