Lind Partners with Investigators at NCI on AACR Abstract: Structuring the unstructured: How protocol quality and precision drive clinical trial completion
At the AACR Annual Meeting 2025, Lind presented Abstract LB373, applying large language models and machine learning to 12,096 oncology trials. The framework predicted trial completion with ~81% accuracy and showed that simpler, guideline-aligned protocols with fewer restrictive eligibility criteria enroll and complete more successfully.
Lind co-founders Oggie Nikolic and Pranav Singh, with collaborators including Richard L. Schilsky, presented Abstract LB373 at the AACR Annual Meeting 2025. The study transformed unstructured protocols from 12,096 interventional oncology trials into machine-readable data, then used a Random Forest model to predict trial completion (81.4% accuracy). The takeaway: larger cohorts and longer recruitment windows help, while excessive eligibility criteria and deviations from FDA/NCTN guidelines hurt completion — a scalable framework for designing trials that actually finish.
Citation: Nikolic O, Singh P, Meyer A-M, Silkensen S, Rodriguez-Watson C, Schilsky RL. Cancer Res (2025) 85 (8_Suppl_2): LB373. doi.org/10.1158/1538-7445.AM2025-LB373