Justin J. Wilkins (1), Andre Koenig (2), Lena Klopp-Schulze* (2), Yulia Vugmeyster (3), Samrita De Banerjee (3), Laureen S. Ojalvo (3), Pascal Girard (4), Akash Khandelwal (2) *presenting author
(1) Occams, Amstelveen, The Netherlands; (2) Merck KGaA, Darmstadt, Germany; (3) EMD Serono, R&D Institute, Inc., Billerica, USA and (4) Merck Serono S.A., Institute for Pharmacometrics, Lausanne, Switzerland
Objectives: Bintrafusp alfa* (M7824) is an innovative first-in-class bifunctional fusion protein composed of the two extracellular domains of transforming growth factor-beta (TGF-β) receptor II to function as a TGF-β “trap” fused to a human IgG1 monoclonal antibody against programmed death-ligand 1 (PD-L1). Currently in phase 1, two trials (NCT02517398 and NCT02699515) are ongoing in cancer patients with advanced solid tumors to investigate the clinical potential of bintrafusp alfa as a new immuno-oncology therapeutic agent. The objective of the presented study is to assess benefit-risks of bintrafusp alfa and to support an adequate dose selection for phase II performing exposure-response (E-R) analyses for efficacy and safety.
Methods: For both E-R analyses, individual clearance (CL) parameter estimates and exposure metrices (area under the curve after a single dose [AUC0-336h] and steady-state [AUCss]) were obtained from the population PK model of bintrafusp alfa with time-invariant CL [2]. CL, AUC0-336h and AUCss were investigated separately as predictors of response. For the exposure-safety analyses, derived exposure metrics and adverse event (AE) data from 673 patients with 17 different tumor types treated with 0.3-30 mg/kg, 500 or 1200 mg every two weeks (Q2W) from two phase 1 trials were assessed by applying logistic regression modeling to various AEs: Immune-related AEs (irAE), infusion-related reactions (IRR), treatment-emergent AEs (TEAE), skin AEs related to PD-L1 (sPDAE) and related to TGF-β (sTGAE). For the exposure-efficacy analyses, derived exposure metrics and efficacy data from 80 second-line non-small cell lung cancer patients treated with either 500 mg Q2W or 1200 mg Q2W were evaluated using logistic regression models for best overall response (BOR) and Kaplan-Meier and Cox regression analyses for progression-free survival (PFS). To explore influential or confounding covariates, multivariate logistic and Cox regression analyses were performed using the full-model approach [1] separately for each exposure metric. Reduced logistic models were derived by removing the least informative covariates in a stepwise manner until no further reduction in the Akaike information criterion could be achieved. There was no adjustment for multiplicity for the reported CI corresponding to the different efficacy endpoints, exposure metrics or covariates. The E-R analyses were performed using R (v. 3.2.2).
Results: Both exposure-response analyses for BOR and PFS showed a shallow trend towards a relationship between exposure and efficacy variables. Univariate and multivariate Cox regression analyses detected an association between exposure and PFS (AUC0-336h-PFS multivariate model: HR 0.820 [95%CI 0.692-0.972] per AUC0-336h increase of 10,000 mg·h/L). In multivariate Cox models, metastasis at baseline was identified as influential covariate on PFS (HR 3.307 [95%CI 1.360-8.039]). A definitive relationship between exposure-BOR could not be confirmed, although a weak trend was apparent (AUC0-336h-BOR: OR 1.301 [95% CI: 0.899-1.970] per AUC0-336h increase of 10,000 mg·h/L). Metastasis, smoking status and maximal change of neutrophil/lymphocyte ratio from baseline (not a baseline variable) were detected as potentially influential covariates besides exposure on BOR (these remained in the reduced models). The multivariate exposure-safety analyses found irAE, sPDAE and sTGAE incidences to be positively correlated with exposure (OR of ≤1.354 per AUC increase of 10,000 mg·h/L). Additionally, the number of cycles and concomitant corticosteroids and biologics were positively correlated with AE incidences.
Conclusions: A weak association between exposure and efficacy was indicated by both exposure-efficacy analyses of BOR and PFS. Interestingly, a strong association was observed between time-invariant CL and efficacy (univariate model: OR 0.341 [95% CI: 0.133-0.750] per CL increase of 0.005 L/h), which is likely a manifestation of the disease state on PK [2] and therefore confounding the true relationship between exposure-efficacy. The power to detect influential covariates was limited with most effects being highly uncertain. AE incidences were generally weakly or not correlated with exposure of bintrafusp alfa. Overall, the emerging PK, efficacy and safety data of bintrafusp alfa from preclinical and phase 1 clinical studies [4-5] jointly support the recommended dose selection of 1200 Q2W for future clinical trials.
*Proposed INN.
References:
[1] Gastonguay, M. R. The AAPS Journal. 2004 (6), S1, W4354.
[2] Wilkins J, et al. PAGE 27 (2018) Abstr 8499 [www.page-meeting.org/?abstract=8499]
[3] Wang Y, et al. Clin Pharmacol Therap J doi:10.1002/cpt.628
[4] Strauss J, et al. Clin Cancer Res doi:10.1158/1078-0432.CCR-17-2653
[5] Vugmeyster Y, et al. J Clin Oncol doi:10.1200/JCO.2018.36.15_suppl.2566
Reference: PAGE 28 (2019) Abstr 9086 [www.page-meeting.org/?abstract=9086]
Poster: Drug/Disease Modelling - Oncology