Target-response relationship of bevacizumab may be more relevant than exposure-response - a target-mediated drug disposition (TMDD) model

Sarah Lobet (1), Morgane Caulet (2), Nicolas Azzopardi (3), Gilles Paintaud (4, 5), Thierry Lecomte (1, 2), David Ternant (4,5)

(1) Tours University, Inserm UMR 1069, Nutrition Croissance et Cancer (N2C), Tours, France ; (2) Department of Gastroenterology and Digestive Oncology, CHRU de Tours, Tours, France; (3) Tours University, EA7501 GICC, Team PATCH, Tours, France, Tours, France; (4) Pharmacology Department, Tours University Hospital, Tours, France; (5) Tours University, EA4245 Transplantation, Immunologie, Inflammation, Tours, France.

Introduction: Bevacizumab (BV), an anti-angiogenic monoclonal antibody targeting vascular endothelial growth factor (VEGF), has been approved for the treatment of metastatic colorectal cancer (mCRC).  BV pharmacokinetics was shown to be mediated by its target: its elimination is influenced by target (VEGF) levels as well as by tumour burden (1) and is also variable according to circulating VEGF levels (2). In addition, the relationship between serum BV concentrations and clinical response (1) was questioned, because of confounding factors, including tumor size (3) or sarcopenia (4). Indeed, sarcopenia may trigger an increase in BV clearance independently from target, and tumor size may be associated with increased VEGF production (5). Therefore, it may be hypothesized that target amount is more relevant to explain clinical efficacy than antibody serum concentrations. This amount may be assessed using target-mediated drug disposition (TMDD).

Objectives: This study aimed at describing latent target kinetics of BV using TMDD modeling and investigating the relationship between latent target kinetics and clinical response.

Methods: Patients with mCRC and liver metastasis received four cycles of 5 mg/kg every 2 weeks (Q2W) BV (1). Blood samples were collected before the first injection, at day 1 and at weeks 2, 4 and 8 after first injection and were mesured using a validated enzyme-linked immunosorbent assay (ELISA) (6). BV and latent target kinetics were described using a two-compartment TMDD model with quasi-steady-state (QSS) approximation (7). Model parameters were estimated using nonlinear mixed-effects modeling and Monolix software (8). The relationship between latent target kinetics and tumor burden biomarkers was assessed using the following baseline covariates: carcinoembryonic antigen (CEA) levels, circulating VEGF (cVEGF) and presence of extra-hepatic (EH) metastases. Model parameter estimates were used to compute for each patient the following metrics, i.e. the trough BV concentrations and unbound target levels two weeks (C_2W and R_2W) and 8 weeks (C_8W and R_8W) after the first injection. The association of these metrics with overall (OS) and progression-free (PFS) survival were investigated using Cox models. The influence of covariates on structural kinetic and Cox model parameters was assessed using likelihood ratio or Wald’s test (α<0.05). Finally, parameter estimates of the final TMDD model were used to simulate median and 90% prediction intervals of BV concentrations and unbound target levels for various values of baseline target levels (R₀).  Simulations of four cycles of 5 mg/kg Q2W BV were computed for 1000 virtual patients and by using Simulx software (9).

Results: A total of 486 concentration measurements were available in 130 patients. BV kinetic parameters (estimate, interindividual standard deviation) were: central (V₁=4.1 L, ω_V1=0.33) and peripheral (V₂=4.7 L) volumes of distribution, and systemic (CL = 0.16 L/d, ω_CL=0.28) and intercompartment (Q=1.0 L) clearances. Target-mediated kinetic estimates were: initial target levels (R₀=8.4 nM, ω_R0=0.30), first-order elimination rate (k_deg=0.93 day^-1), steady-state dissociation rate (K_SS=10 nM) and first-order elimination rate of complexes (k_int=0.52 day^-1). Covariate analysis revealed an increase in R₀ with baseline CEA (p=0.022), cVEGF (p=0.016) and presence of EH metastases (p=0.029). R_2W is the metric most strongly associated with both OS and PFS which were respectively doubled (12.0 vs. 34.8 months, HR [90% CI] = 2.3 [1.5-3.5]; P=6.10^-5) and tripled (8.9 vs. 18.4 months, HR [90% CI] =1.8 [1.2-2.6]; P=0.0019) in patients with R_2W superior to median value. Simulations showed decreased unbound target levels for increased R₀ values (median R2W ranging from 0.26 nM to 1.3 nM for R₀ values ranging from 3.5 nM to 8.7 nM, respectively).

Conclusions: This is the first study that allowed a quantification of the latent target levels of BV over time.  This estimation is relevant since associated with circulating VEGF concentrations and tumor burden biomarkers (CEA, presence of EH  metastases) on one hand, and survival (OS, PFS) on the other hand. The estimation value of initial target levels might help optimizing BV dosing regimen for each patient individually.

References:

1. Caulet M, Lecomte T, Bouché O, Rollin J, Gouilleux-Gruart V, Azzopardi N, et al. Bevacizumab Pharmacokinetics Influence Overall and Progression-Free Survival in Metastatic Colorectal Cancer Patients. Clinical Pharmacokinetics. 2016;55(11):1381‑94.
2. Panoilia E, Schindler E, Samantas E, Aravantinos G, Kalofonos HP, Christodoulou C, et al. A pharmacokinetic binding model for bevacizumab and VEGF165 in colorectal cancer patients. Cancer Chemother Pharmacol. 2015;75(4):791‑803.
3. Bensalem A, Ternant D. Pharmacokinetic Variability of Therapeutic Antibodies in Humans: A Comprehensive Review of Population Pharmacokinetic Modeling Publications. Clin Pharmacokinet. 13 mars 2020;
4. Dolly A, Dumas J-F, Servais S. Cancer cachexia and skeletal muscle atrophy in clinical studies: what do we really know? J Cachexia Sarcopenia Muscle. déc 2020;11(6):1413‑28.
5. Kut C, Mac Gabhann F, Popel AS. Where is VEGF in the body? A meta-analysis of VEGF distribution in cancer. Br J Cancer. 8 oct 2007;97(7):978‑85.
6. Ternant D, Cézé N, Lecomte T, Degenne D, Duveau A-C, Watier H, et al. An enzyme-linked immunosorbent assay to study bevacizumab pharmacokinetics. Ther Drug Monit. oct 2010;32(5):647‑52.
7. Gibiansky L, Gibiansky E, Kakkar T, Ma P. Approximations of the target-mediated drug disposition model and identifiability of model parameters. J Pharmacokinet Pharmacodyn. 1 oct 2008;35(5):573‑91.
8. Monolix version 2020R1. Antony, France: Lixoft SAS, 2020. http://lixoft.com/products/monolix/.
9. Simulx version 2020R1. Antony, France: Lixoft SAS, 2020. http://lixoft.com/products/simulx/.

Reference: PAGE 30 (2022) Abstr 9989 [www.page-meeting.org/?abstract=9989]

Oral: Drug/Disease Modelling - Oncology

PDF poster/presentation (click to open)