Two poster presentations examine how to assess the accuracy of AI-extracted data from routine cancer care. A featured talk will explore how predictive modeling and digital twins could inform oncology drug development.
Flatiron Health today announced that it will present two research posters and give a featured talk at the ESMO AI & Digital Oncology Congress 2026, taking place November 16–18 in Berlin, Germany. The studies examine how researchers can evaluate the accuracy of clinical variables extracted by large language models (LLMs) from oncology records in Germany and the United Kingdom. The featured talk will explore how predictive models built with real-world data could inform drug development decisions.
As more data from routine cancer care become available for research, AI can help turn information in clinical notes into structured variables for analysis. That speed makes it possible to work across more records, while the usefulness of the resulting data depends on whether those variables accurately reflect what was documented in the chart. The two studies examine complementary approaches to that question, including benchmarking against expert review and using a second model to flag potential inconsistencies for follow-up.
“AI is helping us identify patterns in routine cancer care faster,” said Jessica Davies, Scientific Director, Clinical Development Solutions at Flatiron Health. “To use those insights in drug development, we need to know how reliable the underlying data are, quantify the uncertainty that remains, and identify where expert review is indispensable.”
Research at ESMO AI 2026
Both posters will be presented on Tuesday, November 17, from 16:30 to 17:30 CET in the exhibition area.
-
239P | Development and Evaluation of LLM-based Extraction Workflows to Accelerate Prostate Cancer Evidence Generation in Germany — Daniela Zuluaga Lotero will present research evaluating LLM workflows that extract structured clinical variables from German prostate cancer records. The study used Flatiron’s VALID framework to systematically assess extraction accuracy, with expert review as a benchmark and automated checks as part of the workflow. The work suggests that systematically evaluated LLM extraction could help make prostate cancer data available for research more quickly.
-
172P | Large language model (LLM)-as-a-judge for clinical variable extraction review — Conor Walsh will present a pilot study that used a second LLM to check extracted variables against source documentation in UK and German patient records. The study suggests that automated flagging could complement human review when expert-annotated examples are limited and help reviewers prioritize potential inconsistencies before the data are used in analyses.
Featured session on predictive modeling and digital twins
On Monday, November 16, from 11:45 to 12:05 CET, Jessica Davies will present “Predictive Modeling’s Next Frontier in Oncology Drug Development” on the central stage in the exhibition area. She will discuss how data collected in routine cancer care can be used to build predictive models and digital twins, and where those approaches could help development teams assess trial design choices earlier and make more informed decisions about how to advance a potential therapy.
Flatiron combines real-world oncology data with scientific and clinical expertise to support decisions across drug development. The company applies this approach through predictive modeling, digital twins, and Flatiron Telescope, an AI-powered platform for exploring real-world oncology data and generating insights. Across these applications, rigorous methodology, clinical validation, and expert judgment help determine which questions the data can answer with confidence.
To learn more or request a meeting with the Flatiron team in Berlin, visit the Flatiron Health ESMO AI 2026 event page.
About Flatiron Health
Flatiron Health is a healthtech company working to improve and extend lives by learning from the experience of every person with cancer. Flatiron works with oncology practices, health systems, researchers, and life sciences companies to advance cancer care and accelerate research through technology and real-world evidence. Learn more about Flatiron Health.
Contact
Nina Toor
press@flatiron.com