Chen Wang 1, Han Zheng 2, Zhi Zhang 1, Lin Hu 1, Qi Zhang 1, Jing Nie 1
1 Abbisko Therapeutics (Shanghai, China), 2 Shanghai BioGuider Medical Technology Co., Ltd (Shanghai, China)
Objectives:
FGF19 overexpression (+) occurs in ~30% of hepatocellular carcinoma (HCC) cases and is associated with poor prognosis. The FGF19/FGFR4 signaling axis is a promising therapeutic target. ABSK-011, an oral, highly selective, and potent FGFR4 inhibitor, exhibited a promising efficacy and safety profile for FGF19+ HCC in a phase I trial (ABSK-011-101) as monotherapy and a phase II trial (ABSK-011-201), in combination with Atezolizumab.
Progression-free survival (PFS) has been identified to be moderately correlated with OS in HCC[1], and it remains a critical endpoint due to its earlier evaluability. Although TGI-OS models have been reported in HCC for predicting overall survival from early tumor dynamics [2], the joint modeling of PFS remains exceptionally rare. A recent integrated framework incorporating tumor growth dynamics, non-target disease progression, and dropout has demonstrated strong predictive performance for PFS in other solid tumors [3]. Adopting the semi-mechanistic Tumor growth dynamic PFS (TGD-PFS) framework in HCC patients could meaningfully improve the predictive performance of PFS as a surrogate endpoint and allow for earlier efficacy evaluations before survival data mature and potentially accelerate early clinical development.
This work aims to develop a unified joint modeling framework that predicts PFS by integrating TGD, target progression hazard, non-target progression hazard, and dropout, enabling quantitative prediction of PFS for ABSK-011 in advanced FGF19+ HCC patients.
Methods:
A joint TGD-PFS modeling framework was established to characterize longitudinal tumor dynamics and predict progression-free survival in FGF19+ HCC patients treated with ABSK011. The model comprised three interconnected components:
First, tumor growth dynamics were described using an adapted TGD model, with drug effect incorporated into the tumor shrinkage rate constant (ks). Dose-response relationships were explored across multiple dosing regimens (60–400 mg QD, 160–300 mg BID) using data from 71 patients in the ABSK-011-101 study.
Second, to account for the PFS due to nontarget progression, the additional hazard was described using a parametric time-to-event (TTE) model, with the hazard parameter linked to the tumor growth rate of the target lesion.
Third, a parametric TTE dropout sub-model was included to account for patient discontinuation. PFS was right-censored if a patient withdrew from the study before experiencing either target lesion progression or non-target progression.
Progression events were defined as the earliest occurrence of RECIST 1.1-defined target lesion progression (≥20% increase in sum of longest diameters from the nadir, with an absolute increase of ≥5 mm) or unequivocal non-target lesion progression or new lesion. Model performance was evaluated using visual predictive checks (VPCs) that compared observed versus simulated Kaplan-Meier curves for PFS, while accounting for censoring and preserving the protocol-specified visit schedule in the simulations. The analyses were conducted using Maspectra v2.10.3.
Results:
The model adequately characterized the dose-response relationship for ABSK-011 across multiple dosing regimens in FGF19+ HCC patients, effectively describing the longitudinal tumor growth dynamics. Progression beyond fixed threshold-based target lesion progression, together with significant growth in non-target lesions or appearance of new lesion was characterized by the proposed model. In addition, an innovative VPC approach was applied—accounting for the impact of the timing of the tumor assessment and discontinuations due to both disease progression and dropout —allowing an informative evaluation of model predictability. The diagnostic plots, including VPC, have demonstrated that the established joint TGD-PFS model could capture the observed progression-free survival.
Conclusions:
A joint Dose-TGD-PFS model was successfully developed for ABSK-011 in FGF19+ HCC patients, directly linking tumor growth dynamics to both components of RECIST-defined progression while accounting for early drop-out. This approach provides a more coherent and robust framework for simulating PFS outcomes and may inform the design of future pivotal trials in the target population.
References:
1. Llovet JM et al,. J Hepatol(2019) Jun;70(6):1262-1277.
2. Antonio G et al PAGE 32 (2024) Abstract 10849 [www.page-meeting.org/?abstract=10849]
3. Yu J et al,. CPT Pharmacometrics Syst Pharmacol.(2020) Mar;9(3):177-184.
Reference: PAGE 34 (2026) Abstr 12177 [www.page-meeting.org/?abstract=12177]
Poster: Drug/Disease Modelling - Oncology