M-PROTEIN DYNAMIC METRICS TO SUPPORT EARLY DRUG DEVELOPMENT DECISIONS IN MULTIPLE MYELOMA

Pascal Chanu ¹, Mathilde Marchand ², Federico Mattiello ³, Mark Yan ⁴, Monica Susilo ⁵, Divya Samineni ⁵, Shweta Vadhavkar ⁵, Dale Miles ⁵, Chunze Li ⁵, Jin Jin ⁵

1 Clinical Pharmacology Genentech/Roche (Lyon, France), 2 Certara Drug Development Solutions (Paris, France), 3 Biostatistics, Roche (Basel, Switzerland), 4 Biostatistics, Roche (Mississauga, Canada), 5 Clinical Pharmacology, Genentech (South San Francisco, USA)

Objectives
Model-derived tumor dynamic metrics have shown promise in predicting overall survival (OS) [1,2,3] to support late-stage drug development (DD). However, there is a need to support early decision-making, which generally relies on the objective response rate. A modeling framework was developed to assess the operating characteristics (OCs) of tumor growth inhibition (TGI) metrics and support early decision-making in non-small cell lung cancer [4]. This approach was applied to multiple myeloma (MM) using model-derived M-protein dynamic metrics [5]. Progression-free survival (PFS), the primary endpoint in MM, was also simulated using TGI-PFS modeling [2,3,5]. The work supported the decision, though some inconsistent results were observed: the growth constant (KG), being the best predictor of PFS, demonstrated poor operating characteristics. We revisited this analysis by improving the execution of OCs and extending the TGI-PFS simulations to further support the next readout.

Methods
Building on previous works [5], we used APOLLO trial data [6] accessed through the YODA project [7] to evaluate M-protein dynamic metrics for quantifying the Daratumumab, Polamidomide, dexamethasone (D-Pd) treatment benefit over Pd. A realistic DD scenario was designed with 1,000 subsampled datasets with D-Pd (N=30/arm, 24-week follow-up) versus Pd (full sample size and follow-up) generated via bootstrapping. The effect size between the two arms was assessed by:
-Null Hypothesis (H0): Resampling twice from the Pd arm to mimic a negative study,
-Alternative Hypothesis (H1): Resampling from D-Pd and Pd.
A TGI model [8] was fitted to each dataset to estimate parameters (growth/shrinkage) and derive metrics (e.g., M-protein ratio, time-to-growth) using nlmixr2 [9]. Metric performance was assessed by the probability of achieving geometric mean ratio (GMR) thresholds e.g. GMR < 0.70. Compared to the previous analysis [5], a more parsimonious model (removing one random effect) was used to better fit the limited data in the D-Pd subsamples. In parallel, a TGI-PFS model [2,3] was developed to perform clinical trial simulations (CTS) and assess the probability of technical success (PTS) of a virtual APOLLO Phase 3 trial. Trial success was defined by PFS hazard ratio (HR) <0.75. We conducted 500 CTS, each with 200 APOLLO design replicates, by resampling from the respective subsamples. A different D-Pd subsample was used for each of the 500 CTS to capture the PTS variability relevant to Phase 3 Go/No-Go decisions. The CASTOR trial comparing daratumumab, bortezomib, dexamethasone (D-Vd) vs Vd served as a validation dataset [10]. Results The M-protein ratio to baseline at weeks 8, and 12 demonstrated strongest OCs, a GMR < 0.70 yielded a correct Go rate > 80% and an incorrect Go rate < 20%. KG provided comparable performance, in contrast to earlier findings [5]. A validation was performed on the CASTOR trial using the same resampling methodology to assess whether a Go decision would have been supported based on limited Phase 1b/2 data. KG and the M-protein ratio to baseline at week 8 and 12 met the correct Go criteria (GMR<0.7) in 94.1%, 74.3% and 87.5% of subsamples respectively while the corresponding incorrect Go rates were: 11.8%, 3.8% and 7.6%. These results demonstrated the predictive ability of the selected metrics and gating criteria for decision making. The TGI-PFS model, developed and validated using APOLLO data, identified log(KG) as the best predictor of PFS, with baseline hemoglobin as an additional positive prognostic factor. This model was then applied to quantify variability in PTS predictions. Across 500 APOLLO CTS, the median [95% prediction internal] of medians of HR was 0.69 [0.25;1.09] (observed HR=0.63), yielding a PTS of 71% [6,100]. Comparable results were obtained for CASTOR (observed HR=0.39), with a median PTS of 100% [19,100]. Conclusion This analysis suggests that model-derived M-protein dynamic metrics can support early decision-making in MM by assessing treatment benefit using limited Phase 1b/2 data. A robust application of OCs helped to reconcile the totality of the results, confirming KG as a reliable predictor. CTS demonstrated the high variability in PTS illustrating the risk of making Go/No Go to Phase 3 decisions based on single small/short follow-up Phase 1b/2 data. References: [1] Claret L. et al. Model-based prediction of phase III overall survival in colorectal cancer based on phase II tumor dynamics. J Clin Oncol 27:4103-4108 (2009). [2] Claret L. et al. A model of overall survival predicts treatment outcomes with atezolizumab versus chemotherapy in non-small cell lung cancer based on early tumor kinetics. Clin Cancer Res 24:3292-3298 (2018). [3] Chanu P. et al. A disease model for Multiple Myeloma developed using Real World Data and validated on Phase 3 clinical trials. PAGE 32 (2024) Abstr 11094 [www.page-meeting.org/?abstract=11094]. [4] Bruno R. et al. Tumor dynamic model-based decision support for Phase Ib/II combination studies: A retrospective assessment based on resampling of the Phase III study IMpower150. Clin Cancer Res 29:1047-1055 (2022). [5] Marchand M. et al. Operating Characteristics of M-protein Dynamic Metrics to Support Early Decisions in Multiple Myeloma. PAGE 33 (2025) Abstr 11562 [www.page-meeting.org/?abstract=11562]. [6] Dimopoulos M.A. et al. Daratumumab plus pomalidomide and dexamethasone versus pomalidomide and dexamethasone alone in previously treated multiple myeloma (APOLLO): an open-label, randomised, phase 3 trial. Lancet 22:801–812 (2021). [7] Krumholz H.M. et al. A historic moment for open science: the Yale University Open Data Access Project and Medtronic. Ann. Intern. Med. 158:910–911 (2013). [8] Stein W. D. et al. Tumor Regression and Growth Rates Determined in Five Intramural NCI Prostate Cancer Trials: The Growth Rate Constant as an Indicator of Therapeutic Efficacy. Clin Cancer Res 17:907–917 (2011). [9] Fidler M. et al. Nonlinear Mixed-Effects Model Development and Simulation Using nlmixr and Related R Open-Source Packages. CPT: Pharmacometrics & Systems Pharmacology, 8(9):621–633 (2019). [10] Palumbo A. et al. Daratumumab, Bortezomib, and Dexamethasone for Multiple Myeloma. N Engl J Med. 375(8):754-66 (2016).

Reference: PAGE 34 (2026) Abstr 12308 [www.page-meeting.org/?abstract=12308]

Poster: Drug/Disease Modelling - Oncology