Aurélie Lereclus (2,3), Andréa Boniffay (3), Olivier Blin (1,2), Sylvain Benito (3), Romain Guilhaumou (1,2).
Service de Pharmacologie clinique et Pharmacovigilance, Hôpital de la Timone, France (1), Aix Marseille Université, Institut de neurosciences des systèmes, Inserm UMR 1106, France (2), EXACTCURE, France (3).
Introduction: Lithium is recommended in the treatment of bipolar disorder and displays a narrow therapeutic range. Thus, treatment optimization through therapeutic drug monitoring and population pharmacokinetics (PK) model’s application is needed. For such a purpose, several lithium population pharmacokinetic models have been introduced in the last decades[1]. However, few external evaluation studies have been published whereas it is required prior to apply the model in a clinical setting. Indeed, building a robust pharmacokinetic dataset in clinical psychiatric population remains challenging. Furthermore, few of these PK models include renal status as a covariate whereas lithium clearance has been described to be renal status dependent[2].
Objectives: The aims of this study were: 1) to evaluate the predictive performance of previously developed lithium population pharmacokinetics models using individual pharmacokinetics data from literature; 2) to assess the impact of models’ updating by integration of glomerular filtration rate (GFR) effect on lithium clearance (Cl_li)[3].
Methods: A systematic review of literature for lithium population pharmacokinetics models was performed, models for which we did not have enough information to perform simulations were excluded. Predictive performances of the models were evaluated with external evaluation based on data retrieved in non-compartmental pharmacokinetics literature. Prediction error was calculated to determine models bias and inaccuracy (MDPE% and MDAPE%). GFR was added as Cl_li covariate according to three different literature equations: (1) Cl_li=0.235*GFR[4], (2) Cl_li=4.03+0.19*GFR[5] and (3) Cl_li=6.47+0.161*GFR[5]. Updated MDPE% and MDAPE% were then calculated.
Results: A total of 8 models were included in this study. An evaluation dataset of 89 individual pharmacokinetics data was constituted from non-compartmental pharmacokinetics literature and used for external evaluation, including 39 data with concomitant GFR. For literature models, MDPE% ranged from -5.5 to 99.8 and a MADPE% ranged from 24.4 to 99.8. After GFR addition (n=57), predictive performance was: MDPE% ranged from 0.5 to -41.7 and MADPE% ranged from 17.9 to 41.7 with equation (1); MDPE% ranged from 0.6 to -41.6 and MADPE% ranged from 17.8 to 41.6 with equation (2); MDPE% ranged from 1.3 to -41.0 and MADPE% ranged from 18.3 to 41.3 with equation (3).
Discussion/Conclusion: Our results of external evaluation, performed with non-compartmental pharmacokinetics literature individual data, illustrate that previously developed lithium population PK models present limited predictive performance. PK models performances can be improved by adding new covariates such as GFR into clearance estimation thanks to models’ updating methods. Although the usual way of rebuilding models from own data is efficient, our results highlight that using models’ updating would be a powerful way to improve model’s performance and propose to clinicians adapted tools for therapeutic drug monitoring. In addition, this procedure is much easier to set up compared to more traditional pharmacometrics approaches since neither additional data nor further model training are required.
References:
[1] J. Wen, D. Sawmiller, B. Wheeldon, et J. Tan, « A Review for Lithium: Pharmacokinetics, Drug Design, and Toxicity », CNS Neurol. Disord. Drug Targets, vol. 18, no 10, p. 769-778, 2019, doi: 10.2174/1871527318666191114095249.
[2] K. Yoshida et al., « Prediction Model of Serum Lithium Concentrations », Pharmacopsychiatry, vol. 51, no 03, p. 82-88, mai 2018, doi: 10.1055/s-0043-116855.
[3] K. G. M. Moons et al., « Risk prediction models: II. External validation, model updating, and impact assessment », Heart, vol. 98, no 9, p. 691-698, mai 2012, doi: 10.1136/heartjnl-2011-301247.
[4] A.-L. Kamper, N.-H. Holstein-Rathlou, P. P. Leyssac, et S. Strandgaard, « Lithium Clearance in Chronic Nephropathy », Clin. Sci., vol. 77, no 3, p. 311-318, sept. 1989, doi: 10.1042/cs0770311.
[5] T. Motoki et al., « Calculation of Lithium Clearance for Clinical use Based on Renal Excretion in Japanese Patients », Int. J. Clin. Pharmacol. Pharmacother., vol. 2016, mai 2016, doi: 10.15344/2456-3501/2016/107.
Reference: PAGE 29 (2021) Abstr 9662 [www.page-meeting.org/?abstract=9662]
Poster: Methodology - New Modelling Approaches