
GE Healthcare Study Shows that AI and Imaging May Play a Role in Checkpoint Inhibitor Selection for Cancer Patients
11-3-23 (by: Scott Gleason) GE HealthCare has unveiled new data validating the potential of Artificial Intelligence (AI) in forecasting patients’ responses to immunotherapies, with the software offering a 70 to 80 percent accuracy. The AI models, developed in collaboration with Vanderbilt University Medical Center (VUMC) and the University Medicine Essen (UME) in Germany, were originally crafted based on a cohort of over 3,000 immunotherapy patients from VUMC and subsequently tested on a cohort of 4,000 patients from UME. The AI models not only predict the efficacy outcomes but also assess the probability of an individual patient experiencing an adverse reaction to immunotherapies. This significant development could impact precision cancer care by enabling clinicians to identify the most suitable personalized treatment pathways, potentially sparing patients from unnecessary side effects and expense.
Immunotherapies have the potential to use the body’s immune system to identify and target cancer cells, often providing more effective treatment than traditional methods. However, low response rates and severe side effects have been common challenges. With approximately 5,000 immunotherapies currently in development, the AI models have the potential to streamline patient selection, aiding drug developers in accelerating clinical trials and increasing the likelihood of success. GE HealthCare plans to commercialize these models for use in pharmaceutical drug development and clinical decision support once regional regulatory approvals have been obtained. The landscape of cancer immunotherapy is marked by significant milestones and growing potential. In 2011, the first immunotherapy for cancer received FDA approval and by 2017, the FDA had already approved four immunotherapy checkpoint inhibitors, showcasing their rapid integration into clinical practice. However, while immunotherapy holds promise, only 15-20% of patients achieve durable results with this approach, indicating the need for further refinement. Additionally, there is high cost with most immunotherapy drugs costing $150,000 to $250,000 per year.
The development of these AI models involved the retrospective analysis of immunotherapy treatment responses in thousands of VUMC cancer patients by GE HealthCare and VUMC. They correlated the responses with a range of patient data, including demographic, genomic, tumor, cellular, proteomic, and imaging information, all of which were anonymized. Notably, these AI models employ routinely acquired data from patients’ electronic health records.
Dr. Travis Osterman, Associate Vice President for Research Informatics and Associate Chief Medical Information Officer for Vanderbilt University Medical Center, lauded the results, highlighting that the unpredictability of reactions to immunotherapy has led to increased morbidity and costs. The AI models offer a path toward better patient selection for the therapy, potentially improving outcomes and reducing side effects.
The AI models are an integral part of GE HealthCare’s immuno-oncology development portfolio, which includes the creation of novel PET tracers. Recently, the company announced the initiation of a Phase I clinical trial for a fluorine-18 PET radiopharmaceutical specific for CD8, a subpopulation of white blood cells that play a role in the action of almost all immunotherapies. The trial aims to enhance the understanding of patient responses to immune checkpoint inhibitors, the main class of immunotherapies. This study aims to identify if the presence of CD8+ T cells within tumors is indicative of a higher likelihood of positive response to immunotherapies. Furthermore, the study will facilitate the early assessment of responses to immunotherapies by utilizing sequential whole-body imaging to monitor CD8 changes over time. This capability enables physicians to promptly switch patients who do not respond to alternative treatment options.