Fenja Klima (1,2), Thomas Helland (3,4,5), Robin Michelet (1), Wilhelm Huisinga (6), Daniel Hertz (3), Charlotte Kloft (1) for the CYP2D6 Endoxifen Percentage Activity Model in Breast Cancer (CEPAM) consortium
(1) Department of Clinical Pharmacy and Biochemistry, Institute of Pharmacy, Freie Universitaet Berlin, Germany, (2) Graduate Research Training Program PharMetrX, Berlin/Potsdam, Germany, (3) Department of Clinical Pharmacy, University of Michigan College of Pharmacy, Ann Arbor, United States of America, (4) Hormone Laboratory, Department of Medical Biochemistry and Pharmacology, Haukeland University Hospital, Bergen, Norway, (5) Department of Clinical Science, University of Bergen, Norway, (6) Institute of Mathematics, University of Potsdam, Germany
Introduction: Tamoxifen (TAM) is a prodrug established in the treatment of oestrogen receptor-positive breast cancer, yet 25% of patients experience recurrence [1]. Polymorphic CYP2D6 plays a crucial role in the formation of the most active metabolite Z-endoxifen (Z-EDX), and genotype frequency varies across biogeographical groups [2]. For improving TAM therapy, the quantitative impact of CYP2D6 genotypes on Z-EDX formation warrants better understanding. In addition, not all studies bioanalytically separated Z-EDX from inactive metabolites (mainly Z4’-EDX), thus quantifying total EDX. To meet these challenges and integrate all available pharmacokinetic (PK) EDX data, we leveraged a large clinical database comprising studies conducted around the world with either measured Z-EDX or total EDX. By extending a published parent-metabolite nonlinear mixed-effects (NLME) PK model [3], this work aimed to quantify allele activities for relevant CYP2D6 genotypes on a continuous scale from 0 (no activity) to 1 (wildtype) [4] to further understand variability in TAM treatment.
Methods: The CEPAM database comprised 36 clinical studies, 8451 patients and 10574 blood samples. Patient exclusion criteria were male, <1 month on TAM, missing CYP2D6 genotype or missing or considerably higher/lower TAM or EDX concentrations compared to other studies (16% of patients). Missing time after last dose (TALD) (59.5% of samples) were imputed as 23 h in line with study protocols. After estimating parameters of a published NLME PK model, patients with conditional weighted residuals>±5 and imputed TALD were excluded. To include measured total EDX, the published NLME PK model was expanded by one compartment for Z4’-EDX. Total EDX was predicted as sum of Z-EDX and Z4’-EDX. Model performance was assessed via goodness-of-fit (GOF) plots, median prediction error (MPE) and median absolute PE (MAPE). To allow for estimation of CYP2D6 allele activities, Z-EDX formation was split into a CYP2D6-dependent and CYP2D6-independent process and the relative contribution of each process estimated. Previously included covariates were allocated to each process by physiological plausibility. Non-wildtype CYP2D6 alleles with a frequency of ≥1% in the dataset were included as covariates on CYP2D6-dependent Z-EDX formation as relative change relation. Gene duplications were assumed to multiply the allele activity.
Results: The final dataset comprised 31 studies, 7081 patients and 9131 samples with measured TAM and EDX. In 74.5% of patients, only Z-EDX was measured, whereas Z-EDX+Z4’-EDX and total EDX were measured in in 15.0% and 10.5% of patients. The extended NLME PK model comprised one compartment for TAM, Z-EDX and Z4’-EDX each, with first-order absorption, elimination, and metabolite formation. All molecular species were predicted adequately as indicated by GOF plots. Individual predictions (IPREDs) of TAM and Z-EDX were with low bias (0.14% and 0.47% MPE) and precise (5.97% and 4.67% MAPE). IPREDs of total EDX and Z4’-EDX showed low bias (5.85% and 3.36% MPE) and acceptable precision (28.8% and 25.3% MAPE). Covariates on Z-EDX formation were allocated as follows: CYP2D6 allele activities and inhibitors impacted CYP2D6-dependent Z-EDX formation and CYP2C9 phenotype impacted CYP2D6-independent Z-EDX formation; body weight impacted both pathways jointly. Interindividual variability was 51% CV on both Z-EDX formation pathways jointly. The typical fraction of CYP2D6-dependent Z-EDX formation was 70.1%. The activity of non-functional alleles *3-*5 was fixed to 0 as estimation led to high RSEs and values close to 0. Activity of reduced-function alleles (*9, *10, *17, *41) ranged from 0.119 to 0.641 (≤20% RSE except for *10 with 45% RSE) and the activity of normal-function alleles *2 and *35 was 0.737 and 0.818 (≤7% RSE).
Conclusions: An NLME PK model of TAM and Z-EDX was extended to include total EDX and leveraged to estimate CYP2D6 allele activities, using a large clinical database representing the distribution of CYP2D6 genotypes across biogeographically diverse populations. The predominant role of CYP2D6 in Z-EDX formation was confirmed and a wide range of allele activities identified, highlighting the potential of CYP2D6 polymorphism to contribute to suboptimal TAM treatment. The developed framework will be applied to investigate the impact of CYP2D6 genotypes on TAM treatment outcome and has the potential to inform genotype-based treatment individualisation.
References:
[1] Davies et al. Lancet 2013.
[2] Mürdter et al. Clin. Pharmacol. Ther. 2011.
[3] Mc Laughlin et al. Clin. Pharmacol. Ther. 2024.
[4] Agema et al. Biomed. Pharmacother. 2023.
Reference: PAGE 32 (2024) Abstr 11069 [www.page-meeting.org/?abstract=11069]
Poster: Drug/Disease Modelling - Oncology