ISB 2001: Accelerating Patient Benefits through Quantitative Systems Pharmacology (QSP) Guided First-in-Human (FIH) Starting Dose Selection

Introduction: ISB 2001 is a trispecific antibody that binds to CD38 and/or BCMA on tumor cells, and crosslinks to CD3 on T cells to form synapse causing cytotoxicity to tumor cells. ISB 2001 is currently being tested in relapsed/refractory multiple myeloma (r/r MM) patients in Phase 1 study (NCT05862012). The Minimal Anticipated Biological Effect Level (MABEL) approach has been a traditional practice for deriving FIH dose for CD3-T-cell engagers (TCE) but leads to subtherapeutic FIH doses (Saber H et al., Regul Toxicol Pharmacol., 90, 144-152 (2017)). Recently approved BCMA-TCEs, elranatamab and teclistamab, also started at low FIH doses, 0.1 to 0.3 µg/kg, respectively, requiring dose-escalation of approximately 2,150-fold and 100-fold before showing clinical response; further 10,000-fold and 5,000-fold to reach recommended phase 2 dose (RP2D) (Usmani SZ et al., Lancet. 398, 665-674 (2021), ELREXFIO BLA multidisciplinary review). Here, we employed a novel approach, where we used a QSP model to integrate preclinical data, which allowed us to derive the efficacy dose range and minimal pharmacological active dose (MPAD). Optimal FIH dose was determined by correcting the MPAD to minimize potential drug-induced cytokine release syndrome (CRS), but also to avoid underdosing of cancer patients, thereby quickly achieving therapeutic dose in the FIH study. This approach was accepted by heath authorities in US and Australia and is now fully validated by the emerging clinical trial data.

Methods: A fit-for-purpose translational QSP model was developed for ISB 2001 utilizing binding affinities, target expression, cell counts, and pharmacokinetics (PK) to recapitulate the formation of these synapses, calibrated using in vitro and in vivo efficacy data. The model-derived synapse numbers in the tumor microenvironment were normalized to the number of tumor cells, and this was the key parameter linked to efficacy results across preclinical experiments. This parameter was used for translation and prediction of efficacy from preclinical experiments to r/r MM patients. This QSP model was benchmarked with teclistamab preclinical and clinical data to partially validate predictability. The MPAD for ISB 2001 was derived by benchmarking with the lowest teclistamab dose level that showed clinical efficacy signals. This MPAD was then adjusted using the in vitro cytokine release data and included fractionated step-up doses to attenuate any potential CRS incidences. The FIH dose for ISB 2001-101 study (NCT05862012) and the dose-escalation scheme were designed to achieve robust efficacy as predicted by the QSP model in terms of synapse formation. The PK profiles were simulated in a virtual patient population of 100 patients, accounting for inter-subject variability in physiological parameters. The evolving clinical safety, PK, and efficacy data from the ongoing clinical study were compared with these simulations.

Results: A subcutaneous FIH dose of 5.0 µg/kg, preceded by fractionated step-up doses on C1D1 and C1D4, was selected. This dose was approximately 50 to 100-fold higher than the traditional MABEL-based FIH dose.

As of June 25^th, 2024, CRS incidences from 14 patients who received doses ranging from 5.0 µg/kg to 600 µg/kg were available. Out of 14 patients, 7 showed CRS with 6 grade 1 CRS and 1 grade 2 CRS events. Consistent with the predictive modeling, there were no incidences of CRS at the FIH dose, confirming the adequacy of this novel approach that significantly reduced multiple dose escalations vis à vis traditional MABEL based approach, without compromising safety. Objective disease responses were observed starting from the third dose level tested (50µg/kg), demonstrating that this approach minimized the exposure of heavily pretreated patients to non-therapeutic doses. The PK profiles of ISB 2001 were available in 6 patients up to 150 µg/kg. The individual subject serum concentration-time profiles were very close to the simulated serum concentration profiles from the QSP modeling, demonstrating the accuracy of this approach.

Conclusion: Given the technological advancement and tremendous improvement in designing safer TCEs, we believe our novel approach of selecting the FIH dose based on MPAD, as well as dose escalation steps based on QSP modeling is a much more efficient, faster, more patient-centric and safer path forward for developing TCEs, as compared to traditional MABEL-based approaches.

Menon:Ichnos Sciences: Current Employment. Pacaud:Ichnos Sciences: Current Employment, Current equity holder in private company. Holkova:Ichnos Sciences: Current Employment. Gn:Glenmark: Current Employment. Garton:Ichnos Sciences: Current Employment. Pihlgren:Ichnos Sciences: Current Employment. Matsuura:Certara: Current Employment. van der Graft:Certara: Current Employment. Perro:Ichnos Sciences: Current Employment. Konto:Ichnos Sciences: Current Employment.