Vincent F. S. Dubois(1), Paul G. Baverel(1), Paolo Vicini(1), ChaoYu Jin(2), Yanan Zheng(2), Rajesh Narwal(3), Lorin Roskos(3)
(1)MedImmune, Cambridge, UK; (2)MedImmune, Mountain View, CA, USA; (3)MedImmune, Gaithersburg, MD, USA
Background: All anti-PD1/PD-L1 monoclonal antibodies approved in immuno-oncology (IO) utilized an empirical time-varying CL population PK model to inform their labelling [1]. Limitations surrounding the use of such model were recently highlighted [2] and a more mechanistic model was proposed to delineate the gradual increase of drug exposure observed over time in cancer patients benefiting from IO therapy. However, a quantitative evaluation of the relative performance of these two alternative models was not formally conducted.
Objectives: The objective of this analysis and subsequent simulation framework was to qualitatively and quantitatively compare a recently proposed semimechanistic PK model and an empirical counterpart model.
Methods: Two candidate population PK models [2,3] of durvalumab, a fully human anti-PD-L1 monoclonal antibody recently approved in urothelial carcinoma and stage III non-small cell lung cancer, were developed in NONMEM, version 7.3.0, based on clinical data from 2 trials (NCT01693562, NCT02087423) that comprised a pool of 1409 patients with solid tumors and 7407 PK observations. Dose levels ranged from IV infusion of 0.1 to 20 mg/kg administered either bi-weekly, every 3 weeks, or monthly. Duration of therapy was limited to 12-month or contingent on clinical benefit or until unacceptable toxicity. Both models shared common aspects (a two-compartment structural form with both linear and non-linear clearances, banded-matrix stochastic model, and identical pool of statistically significant covariates, identified by stepwise covariate modelling (SCM)[4]), but differed in the implementation of the time-course of durvalumab non-specific clearance mechanisms. While the empirical model relied on a sigmoid Tmax-type model to mimic the decrease in clearance over time, the semimechanistic model associated durvalumab CL with longitudinal (not only baseline) biomarker patients’ individual data. A comparison of model statistical fit, parameters and precision estimates, predictive performance (VPCs), and simulation performance was undertaken through Monte-Carlo simulations based on bootstrap replicates from both models. Since simulations of longitudinal biomarkers were necessary to derive the semimechanistic model CL time-course, a simplistic nonlinear fixed-effects modeling of longitudinal biomarker data was performed in R, version 3.3.1.
Results: Durvalumab exhibited non-linear PK with saturable target-mediated clearance at doses 50 with a typical estimate of 173 days [95%CI: 74.2; 395]. Data supported individualization of Tmax (ω2=0.0548), but this parameter suffered from high η-shrinkage (67%). Population PK analysis identified several statistically significant covariates that were however not clinically relevant, including body weight, sex, serum albumin, tumor size, post-baseline antidrug antibodies, creatinine clearance, ECOG performance status, and soluble PD-L1 levels [2]. Overall, the statistical fit in NONMEM (ΔOFV=-368), parsimony principle (4 less degrees of freedom), and predictive performance favoured the semimechanistic model over the empirical model. Simulations depicting the time-course of CL across replicated trials confirmed that the semimechanistic model, in addition to not being trial or trial duration dependent, provided more certain CL predictions.
Conclusion: A semimechanistic population PK model of durvalumab incorporating longitudinal biomarker in cancer patients provides superior predictive capabilities than a well-established empirical time-varying CL model, and supports the hypothesis that patients benefiting from therapy have reduced proteolytic catabolism.
References:
[1] Liu C, Yu J, Li H, Liu J, Xu Y, Song P, et al. Association of Time-Varying Clearance of Nivolumab With Disease Dynamics and Its Implications on Exposure Response Analysis. Clin Pharmacol Ther. 2017;101(5):657–66.
[2] Baverel PG, Dubois VFS, Jin CY, Zheng Y, Song X, Jin X et al. Population pharmacokinetics of durvalumab in cancer patients and association with longitudinal biomarkers of disease status Clin Pharmacol Ther. 2017 Dec [epub ahead of print]
[3] Bajaj G, Wang X, Agrawal S, Gupta M, Roy A, Feng Y. Model-Based Population Pharmacokinetic Analysis of Nivolumab in Patients With Solid Tumors. CPT Pharmacometrics Syst Pharmacol. 2017;6:58–66.
[4] Jonsson EN, Karlsson MO. Automated Covariate Model Building Within NONMEM. Pharm Res. 1998;15(9):1463–8.
Reference: PAGE 27 (2018) Abstr 8709 [www.page-meeting.org/?abstract=8709]
Poster: Drug/Disease Modelling - Oncology