AI-Enabled Extraction of eGFR and BRAF Testing Patterns in Advanced NSCLC: IMPACT-NSCLC, a Multicenter Real-World Study (2019-2024)

Author(s)

Asmaa Zkik, MSc1, ISABELLE DELAROZIERE, PhD2, Lorraine HOUVET, MSc2, Silvia AVILES, PharmD1, Arnaud Panes, PharmD, PhD3, Abir Tadmouri, MPH, PhD1.
1Pierre Fabre, Boulogne Billancourt, France, 2OSPI, Paris, France, 3Artificial intelligence and cancers association, Paris, France.
OBJECTIVES: Cancer treatment is increasingly complex, driven by advances in personalized medicine and the emergence of mutation-targeted therapies. While these innovations offer improved outcomes, they also introduce significant challenges for testing patients at the optimal time point to provide efficient care, both in terms of health outcomes and resource use. Leveraging artificial intelligence (AI) to gather and process large-scale, real-world data from rare-mutation populations could accelerate evidence generation and refine treatment and testing strategies.IMPACT-NSCLC aims to describe testing of EGFR- and BRAF- mutated advanced or metastatic non-small cell lung cancer (NSCLC). The underlying technical objective is to demonstrate the feasibility of an AI-based data collection process on large scale patient-pathway data.
METHODS: This multicentric retrospective cohort study employs AI and NLP technologies (NER, LLM) algorithms, to extract and structure variables from unstructured data sources from approximately 14 hospitals in France, between 2019 and 2024. AI models are designed to capture into a harmonised dataset the complexity of NSCLC patient care pathways, including diagnosis information, timing of key events such as treatments, biopsies and molecular testing, and follow-up outcomes. After data pseudonymization by study sites, the data preparation using AI includes iterative refinement of detection algorithms and metrics generation from manual annotations. Monocentric databases are aggregated into a multicentric dataset, undergoing data cleaning, derived variable creation, and consistency checks.
RESULTS: Although the study is in its initial phase, the AI-based approach promises to significantly accelerate data collection and processing, enabling reconstructing detailed treatment trajectories for patients with rare mutations across 14 study sites, despite variability in documentation formats and clinical workflows
CONCLUSIONS: AI has demonstrated strong potential to enhance data collection and structuration in oncology research, even for low-prevalence genomic subgroups. The approach paves the way for large-scale analyses of rare-mutation NSCLC and supports broader deployment of AI-enabled evidence generation in precision oncology.

Conference/Value in Health Info

2025-11, ISPOR Europe 2025, Glasgow, Scotland

Value in Health, Volume 28, Issue S2

Code

HSD5

Topic

Health Service Delivery & Process of Care, Methodological & Statistical Research, Study Approaches

Disease

Oncology

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×