Naoki Kotani (1,2), Nina Wang (2), Preet K. Dhillon (3), Stella Arndorfer (4), Pascal Chanu (5)
(1) Pharmaceutical Science Department, Chugai Pharmaceutical Co., Ltd., Tokyo, Japan; (2) Clinical Pharmacology, Genentech Inc., South San Francisco, CA, USA; (3) Personalized Healthcare, Product Development, Genentech, Inc., South San Francisco, CA, USA; (4) Genesis Research, Hoboken, NJ, USA; (5) Clinical Pharmacology, Genentech/Roche, France
Introduction: Assessment of benefit in clinical trials of systemic therapy for HR+/HER2- mBC is commonly performed by using progression free survival (PFS) as a surrogate efficacy endpoint due to practical limitations for using OS, the gold standard endpoint for clinical benefit in the oncology field [1]. Hence, studies are not powered to detect OS benefit in most cases, whereas the evidence of such clinical benefit is demanded by various stakeholders. The modeling & simulation (M&S) framework linking the relationship between longitudinal tumor size and overall survival has been established and reported in various cancer types [2]. If such M&S framework can be applied for HR+/HER2- mBC and information about potential OS benefit can be supplemented using this approach, it will be of great value. In this framework, the model-based tumor growth inhibition (TGI) metrics can be used as biomarkers to capture treatment effect and to make inference of OS. The aim of this work was to develop a drug-independent link between TGI metrics and OS (TGI-OS model) which enables inference of OS using early tumor size dynamics for HR+/HER2- mBC.
Methods: The model was developed using a pooled dataset including Phase I/II and Phase III clinical studies data of taselisib, a potent, selective inhibitor of class I phosphatidylinositol 3-kinase (PI3K) α-, δ-, and γ-isoforms [3-5]. The OS data from these clinical trials were compared to a real world all-comer HR+/HER2- mBC patient population with similar inclusion/exclusion criteria as the trial, derived from an oncology-based electronic health record (EHR)-derived de-identified database, Flatiron Health. During the study period, the de-identified data originated from approximately 280 US cancer clinics (~800 sites of care). Real world data (RWD) patients were diagnosed with mBC on or after 01/01/2015 and OS was calculated starting at 2L fulvestrant monotherapy index date until 03/31/2020. The individual patient level longitudinal tumor size data from the taselisib’s clinical studies were analyzed to derive TGI metrics using the previously published empirical bi-exponential TGI model [6] modified by including the baseline tumor size estimation. Parameters of the model were estimated with a nonlinear mixed effect modeling (population) approach (NONMEM, version 7.4.3) using the first-order conditional estimation algorithm with interaction. A parametric OS model was developed to describe the OS distribution as a function of TGI metrics and various covariates including potential prognostic factors. The developed TGI-OS model was evaluated in simulating the taselisib’s Phase I/II and Phase III study data which were used for model development (i.e., internal validation).
Results: Both Phase I/II and Phase III studies of taselisib targeted similar HR+/HER2- mBC patient population (2nd or later treatment line), and the observed OS distribution overlapped well with median OS of around 25 months, which supported to pool the data from these studies. The OS distribution was also comparable with that from all-comer RWD patients (median OS: 23.6 months in taselisib’s Phase III study vs 22.3 months in Flatiron Health (95% CI: 14.1, 28.0)). The model based TGI metrics showed wide ranges; median (min, max) of logarithm of tumor growth rate (LogKG) was -2.09 (-3.21, -0.697), and that of time to growth was 22.4 (-50.4, 278) weeks. The final TGI-OS model included LogKG, baseline aspartate transaminase level, baseline Eastern Cooperative Oncology Group (ECOG) performance status, and baseline alkaline phosphatase level as covariates, and the model captured the observed survival data well across various patient groups categorized by each covariate, as well as not significant but slight treatment effect of taselisib.
Conclusions: A TGI-OS model was developed for HR+/HER2- mBC population (in 2L+ settings) using data from taselisib’s Phase I/II and Phase III trials. The TGI-OS model may allow early prediction of clinical efficacy leveraging strong link between TGI metrics and OS, even from short tumor kinetic profiles available at the time of PFS evaluation, and may enable informing clinical benefit in HR+/HER2- mBC for new molecules in clinical development.
References:
[1] Seidman AD et al., Am Soc Clin Oncol Educ Book. (2020).
[2] Bruno R et al., Clin Cancer Res. (2020) 15;26(8):1787-1795.
[3] Juric D et al., Cancer Discov. (2017) 7(7):704-715.
[4] Dickler MN et al., Clin Cancer Res. (2018) 24(18):4380-4387.
[5] Dent S et al., Ann Oncol. (2021) 32(2):197-207.
[6] Stein WD et al., Clin Cancer Res. (2011) 17(4):907-17.
Reference: PAGE 29 (2021) Abstr 9655 [www.page-meeting.org/?abstract=9655]
Poster: Drug/Disease Modelling - Oncology