Author: Clemens Grassberger, David (Bo) McClatchy, Harald Paganetti 👨🔬
Affiliation: Department of Radiation Oncology, University of Washington and Fred Hutchinson Cancer Center, Massachusetts General Hospital 🌍
Purpose: While randomized controlled trials (RCTs) are the gold standard for demonstrating efficacy, nearly 50% of late-stage clinical trials fail to meet their endpoint. Tools to study the design of multimodal treatments are needed to increase the chances of successful trials. We aim to show that predictions made by a virtual clinical trial yielded accurate results when compared to clinical trials that were performed after the model was published, demonstrating the utility of this method to aid in RCT design.
Methods: In 2020, we published an in-silico framework to mechanistically model treatment combinations of chemoradiation (CRT) and tyrosine kinase inhibitors (TKI) in virtual populations of EGFR-mutated, locally advanced non-small cell lung cancer (EGFRmut-LA-NSCLC). Since 2020, three publications investigating TKI+CRT regimens in EGFRmut-LA-NSCLC were identified, for which Kaplan-Meier plots were available for statistical comparison. Virtual clinical trials using our existing framework were performed on digital twins representing these trial populations and Kaplan-Meier progression curves were predicted.
Results: The published versus predicted 1&2-year progression-free survival (PFS) rates for 8-week TKI-induction were 58.1%(95CI,33.4%-76.4%) & 36.9%(95CI,16.6%-57.6%) versus 61.7% & 31.5%. For upfront-CRT and TKI maintenance, the published 1&2-year PFS rates were 74%(95CI,64%-80%) & 65%(95CI,56%-73%) versus model-predicted 75.2% & 55.5%, and the published 1&2-year freedom from distant failure (FFDF) rates were 85% and 81% versus model-predicted 84.3% and 67%. Additionally, the published versus modeled hazard ratio of upfront-CRT and TKI maintenance over CRT alone was 0.16(95CI,0.1-0.24) versus 0.2(95CI,0.13-0.31) for PFS and 0.21(95CI,0.11-0.38) versus 0.27(95CI,0.17-0.41) for FFDF.
Conclusion: We demonstrated that in-silico clinical trials based on digital twins and mechanistic modeling can accurately predict outcomes of clinical trials integrating radiotherapy and systemic therapy. Given the limited resources and high cost of running RCTs, in-silico clinical trials could inform the design of multimodal radiation-drug treatments and potentially improve the rate of RCTs meeting endpoints.