Renwei Zhang1, Tingjie Guo1, Li Qin2, Rik Greef2, Matthew Zierhut2, Laura Zwep1, J. G. Coen Hasselt1
1Systems Pharmacology and Pharmacy, Leiden Academic Centre for Drug Research, Leiden University, 2Certara Inc.
Introduction/Objectives Combination therapies are commonly for treatment of various cancers. Methods to predict expected therapeutic effects of combination therapies are therefore of high clinical relevance. Recently, it has been shown that the therapeutic benefits for selected combination anticancer treatments as observed by progression-free survival (PFS) may be predicted by an individual patients’ best response to the most effective drug in a combination treatment [1], a principle which is also known as the independent drug action (IDA) hypothesis [2]. At a population level, the IDA hypothesis can explain combination treatment effects suggesting additive or synergistic drug effects, due to the increased response rates at an individual level. So far, the IDA hypothesis has only been explored for a limited number of drugs using PFS data from randomized controlled trials (RCTs)[3, 4]. Limited availability of RCTs reporting both monotherapy and combination treatment arms limit a more comprehensive evaluation of the IDA hypothesis for prediction of combination treatment effects. In the current study we demonstrate how a novel model-based meta-analysis (MBMA) modelling approach can be used to address this challenge, as it allows exploitation of abundantly available single-arm trial data from literature, in addition to RCTs, to evaluate the independency or interaction effects of drugs for combination therapies in oncology. Specifically, we aimed (1) to implement an MBMA workflow to predict combination treatment effects assuming the IDA hypothesis using aggregate PFS monotherapy data and (2) to evaluate the predictive performance of the developed workflow through application for a broad range of combination therapies Methods Data: Literature-reported aggregate PFS data were sourced from published clinical trials in non-small cell lung cancer as available in the CODEx database [5]. Clinical trials reporting commonly-used anticancer agents as monotherapy and combination therapy (carboplatin, docetaxel, paclitaxel, gemcitabine, pemetrexed, erlotinib, bevacizumab, pembrolizumab) were included, involving 256 trials, 315 arms and 49427 patients. Model implementation: Mixed-effects survival models were developed for all available monotherapies, accounting for between-study variability. PFS for individual patients receiving either monotherapy or combination therapy were predicted using corresponding survival models. Following the IDA hypothesis, the predicted PFS for individual patients on combination therapy were set to equal the best expected PFS for monotherapy [1]. Evaluation metrics: Predicted PFS curves for all combination therapies were compared against the observed data. The hazard ratio (HR) between the combination therapy and the best monotherapy was estimated to assess the effect of a combination treatment. We estimated the HRs using both predicted and observed PFS curves for combination therapy and compared them accordingly. Results We successfully applied the IDA hypothesis within the MBMA framework. We show that MBMA-based evaluation of the IDA hypothesis for drug combination treatment data is possible and allows prediction of combination treatment effects based on monotherapy data. Specifically, the model predictions for PFS aligned closely with observed data in more than half of evaluated combination therapies (9/16). The observed HR of combination therapies over best monotherapies fell between 1 and the predicted benefit HR (median 0.60), suggesting that assuming IDA may be sufficient for predicting combination treatment effect using monotherapy trial data. For several combination therapies (4/16), especially concerning erlotinib and chemotherapeutic agents (3/16), models assuming IDA were not able to adequately capture the PFS. Models underpredicted PFS with an observed HR lower than predicted HR for three combinations of pembrolizumab and chemotherapy (?HR: -0.20, -0.17, -0.14), which suggest potential synergistic or additive effect between these treatment classes, highlighting how the methodology can allow differentiation between drug interaction mechanisms. Conclusions The current study demonstrates the utility of a novel IDA hypothesis-driven MBMA strategy to enable systematic evaluation and prediction drug interaction effects in anticancer combination therapies using aggregate monotherapy clinical trial results.
[1] Palmer & Sorger. Cell. (2017); 171:1678–1691. [2] Gaddum J.H. Pharmacology, First Edition (Oxford University Press). (1940). [3] Palmer C et al. J Immunother Cancer. (2020); Oct;8(2):e000807. [4] Palmer C et al. Clin Cancer Res. (2022); 28:368–77. [5] https://codex.certara.com/Â
Reference: PAGE 33 (2025) Abstr 11766 [www.page-meeting.org/?abstract=11766]
Poster: Methodology - New Modelling Approaches