Sreenath Krishnan1, Koorosh Korfi2, Sylvia Herter2, Marie-Helene Wasmer3, Isabel Prieto4, Heather Hinton5, Wouter Driessen1, Georgios Kazantzidis5, Akiko Tagawa5, Natalie Dimier6, Katharina Lechner3, Candice Jamois1
1Pharma Research and Early Development, Roche Innovation Center Basel, 2Pharma Research and Early Development, Roche Innovation Center Zurich, 3Pharma Research and Early Development, Roche Innovation Center Munich, 4Pharma Development, Roche Innovation Center Welwyn, 5Pharma Development, Roche Innovation Center Basel, 6Pharma Research and Early Development, Roche Innovation Center Welwyn
Background: Englumafusp alfa (ENGLU) is a CD19-4-1BBL antibody-like fusion protein targeting CD19 on B cells and 4-1BB on immune cells that elicits a costimulatory signal (signal 2) that augments and prolongs T-cell activity in presence of a signal 1 provider, such as glofitamab (GLOFIT), a CD20xCD3 T-cell engaging bispecific antibody. In a Phase I dose-escalation study (NCT04077723), where ENGLU was evaluated in combination with GLOFIT, promising anti-tumour activity across all ENGLU dose levels in patients with relapsed/refractory aggressive Non Hodgkin Lymphoma was reported [1]. To optimize dose selection and maximize patient benefit over risk, a totality of evidence mindset with comprehensive data integration and quantitative modeling approaches is essential [1,2,3] . Objective: The authors demonstrate how preclinical and clinical data were leveraged using a population PK matching exposure strategy and exposure-response analyses to enable the selection of ENGLU doses. Methods: Preclinical and clinical data (efficacy, safety, PK, PK/PD, biomarkers) were integrated to characterize dose-exposure-response (DER) relationships. In vitro assays and in vivo dose-finding efficacy studies using WSU-DLCL2 and SUDHL8 humanized mouse models were used to define an efficacious dose range that was assumed to be translatable to humans. The immune system activation observed in these models was assumed to reflect similar pharmacodynamic responses in patients, especially with respect to co-stimulation via 4-1BB. Species differences in PK and immune cell sensitivity were implicitly accounted for by matching clinical exposures to the preclinical efficacious exposure range (obtained using a translational PK model), rather than direct dose translation. This population PK-matching strategy assumes a similar concentration-effect relationship in humans as observed in humanized mouse models. Pharmacokinetic (PK) data from 122 patients and pharmacodynamic (PD) biomarker data from 99 patients treated with ENGLU (0.36–75 mg) were analyzed. Population PK modeling using NONMEM 7.5 [4], logistic regression, and graphical exploration assessed PK properties and exposure-response correlations. Virtual patient simulations in R (mrgsolve) predicted steady-state average concentrations (Cavg,ss) and exposure overlap between doses. Statistical and graphical analyses in R (version 4.4) informed dose selection based on defined criteria: Distinct PK profiles with minimal exposure overlap; Matching Cavg,ss with preclinical efficacious range; PD biomarker DER consistent with co-stimulator mechanism of action (MoA). Results: Preclinical experiments showed a bell-shaped dose-response curve. Complete tumor regression and increased CD8+ T-cell infiltration were observed for doses X-W, referred to as in vivo efficacious dose range with greater add-on effect compared to GLOFIT alone. Clinically, the combination was well tolerated across doses, with no dose-dependent or synergistic safety concerns. Efficacy was observed among all doses tested during dose escalation with no relationship with exposure. A 2-compartment PK model with linear and Michaelis-Menten elimination adequately captured ENGLU target-mediated drug disposition and dose-dependent elimination across dose levels and demonstrated excellent predictive performance through visual predictive checks. Fixed effect parameters were 0.186 L/day, 3.4 L, 2.82 L, 0.781 L/day , 0.167 mg/day and 183 ng/day respectively for clearance; central and peripheral volumes of distribution (V1, V2), maximum elimination capacity (Vm) and Michaelis Menten constant (Km). Between subject variabilities were 42, 33, 58 and 37% for CL, V1, V2 and Vm respectively with a residual unexplained variability of 62%. Relative Standard Errors were below 30%, ensuring model robustness and reliability. Body weight (allometric scaling) was found to statistically influence clearance and volume parameters (effect not considered clinically relevant in absence of DER relationship). Simulations identified Dose Gx as achieving the highest proportion of patients within the preclinical efficacious range (88%, CI: 80–95%), followed by Dose Hx (60%, CI: 50–72%) and Dose Fx (53%, CI: 40–63%). The analyses revealed significant exposure overlap between Dose Gx and Dose Hx (61%), that was reduced between Dose Fx and Dose Hx (35%). Changes in peripheral PD1+ CD8+ effector memory RA (Temra) cells and soluble CD25 were observed, consistent with MoA. An inverse bell-shaped dose response indicated optimal immune activation for intermediate doses tested (Dose Ex to Dose Hx). Conclusion: In the absence of DER, a quantitative clinical pharmacology approach integrating preclinical data along with clinical PK, and PD biomarkers was employed to select two doses, Dose Fx and Dose Hx, for further clinical evaluation. These doses demonstrated distinct PK profiles, minimized exposure overlap, and showed optimal PD biomarker responses, paving the way for a randomized study to refine the DER analysis and optimize patient benefit-risk. References: [1] 66th American Society of Hematology Annual Meeting (2024) Abstract 990. [https://ash.confex.com/ash/2024/webprogram/Paper200096.html] [2] Samineni D et al., Clin Pharmacol Ther. (2024). [3] Cucurull-Sanchez et al., CPT: Pharmacometrics & Systems Pharmacology (2019). [4] NONMEM Users Guides, (1989-2021). Icon Development Solutions, USA.
[1] 66th American Society of Hematology Annual Meeting (2024) Abstract 990. [https://ash.confex.com/ash/2024/webprogram/Paper200096.html] [2] Samineni D et al., Clin Pharmacol Ther. (2024). [3] Cucurull-Sanchez et al., CPT: Pharmacometrics & Systems Pharmacology (2019). [4] NONMEM Users Guides, (1989-2021). Icon Development Solutions, USA.
Reference: PAGE 33 (2025) Abstr 11481 [www.page-meeting.org/?abstract=11481]
Poster: Drug/Disease Modelling - Oncology