AI-powered social media listening of oncologist conversations at ASCO 2025.

Abstract

9045 Background: Major oncology congresses generate extensive expert-driven discussion on social media around clinical trial data. However, standardized methods to systematically capture, contextualize, and evaluate these conversations remain limited. LARVOL CLIN is an AI-powered platform that analyzes oncology-focused discussions on X (formerly Twitter), enabling systematic evaluation of clinical trial discourse through assessment of trial-level activity, oncologist sentiment, and engagement. Methods: This observational, descriptive analysis evaluated oncology-related posts on X associated with ASCO 2025 (from Apr 23, 2024, to Jan 14, 2025). Posts were identified using predefined conference- and trial-specific keywords and underwent manual validation to confirm relevance to oncology clinical trials. Text-based sentiment analysis was conducted exclusively on posts from oncologists, while trial-related images and polls from both oncologist and non-oncologist X accounts were reviewed for relevance. Validated content was analyzed using oncology-trained large language models, based on ChatGPT 5.1, to interpret clinical context, including efficacy, safety, endpoint status, and potential practice relevance. Sentiment was categorized as negative, neutral, positive, or strongly positive. Results: A total of 4,621 oncology clinical trial–related X posts from 651 digitally active oncologists were analyzed for ASCO 2025. Sentiment profiles varied across trials (Table 1). Some studies demonstrated a higher proportion of positive or strongly positive sentiment while others were characterized primarily by neutral assessments reflecting cautious clinical interpretation. Neutral sentiment constituted a substantial share of oncologist posts, often reflecting data interpretation, contextual discussion, or pending clinical relevance. Negative sentiment was typically associated with limited efficacy signals, safety considerations, or unmet expectations. Conclusions: This ASCO 2025 observational study demonstrates that AI-driven social listening can systematically capture, contextualize, and quantify expert oncology discourse surrounding clinical trial presentations. LARVOL CLIN enables real-time assessment of oncologist sentiment supporting medical affairs, outcomes research, and strategic communication. Digital oncologist activity for selected ASCO 2025 trials. Trial (Abstract ID) Views (K) Not Rated Negative Neutral Positive Strongly Positive Total X Posts PACIFIC15 (8516) 798 47 10 14 25 14 110 DESTINY-Breast09 (LBA1008) 477 32 4 14 24 34 108 SERENA-6 (LBA4) 358 36 9 15 39 26 125 ASCENT-04 (LBA109) 355 22 2 5 25 34 88 ATOMIC (LBA1) 290 25 4 8 27 40 104 LARVOL CLIN represents cumulative views across all X posts for each trial. Sentiment categories reflect AI-generated classification of oncologist X posts.