Enhancing Oncology Clinical Trial Prescreening at UPenn with Mendel AI
At Mendel AI, we’re proud to partner with leading academic institutions to bring transformative AI solutions to real-world clinical challenges. Recently, researchers from the Perelman School of Medicine at the University of Pennsylvania leveraged our state-of-the-art AI product to enhance the prescreening process for oncology clinical trials—demonstrating both increased efficiency and maintained accuracy in patient eligibility assessments.
The Challenge: Manual Prescreening in Oncology Trials
Identifying eligible patients for oncology clinical trials has traditionally relied on manual chart reviews performed by clinical research coordinators (CRCs). This labor-intensive process is not only time-consuming but also susceptible to human error, with studies suggesting that up to 70% of eligible cancer patients are inadvertently overlooked. Such delays and inaccuracies can impede trial enrollment and ultimately slow the pace of innovative cancer research.
The Mendel AI-Powered Solution
To address these challenges, the research team incorporated Mendel AI’s advanced natural language processing (NLP) platform into their clinical trial prescreening workflow. Our solution automates critical steps of data extraction and analysis by:
- Optical Character Recognition (OCR): Converting handwritten or printed text in electronic health records (EHRs) into machine-readable data.
- Entity & Relationship Extraction: Identifying and structuring key clinical details such as cancer type, stage, treatment history, and actionable biomarkers.
- Natural Language Inference: Utilizing deep learning and symbolic AI techniques to contextualize and interpret complex clinical narratives.
By integrating these capabilities, Mendel AI’s product allowed the team to automatically flag 13 common eligibility criteria in real-world patient records for non-small cell lung cancer (NSCLC) and colorectal cancer (CrCa).A Closer Look: The Study DesignIn a preplanned interim analysis of a paired-design noninferiority trial, the research compared three methods of chart review:
- Human-alone: CRCs manually reviewed charts without AI support.
- Human+AI: CRCs reviewed charts supplemented with Mendel AI’s NLP-generated annotations.
- AI-alone: Automated chart reviews performed solely by the AI system.
The primary outcome focused on chart-level accuracy—comparing 13 eligibility items determined by the CRCs against a gold standard established by blinded clinicians. The secondary outcome measured timeliness, specifically the time required to review each chart.Key Findings
- Maintained Accuracy:
The Human+AI approach achieved an overall accuracy of 78.7%, which was statistically noninferior to the Human-alone method at 76.7%. Notably, AI-alone recorded a lower accuracy of 63.5%, underscoring the value of combining human expertise with AI insights. - Enhanced Efficiency:
The median review time per chart was significantly reduced—from 43.9 minutes for Human-alone to 34.1 minutes for the Human+AI approach. This improvement in efficiency means that more patients can be screened in less time, accelerating the clinical trial process. - Targeted Performance Gains:
The combined Human+AI team demonstrated superior performance in evaluating RECIST response criteria, a critical measure for assessing tumor response to treatment.
Impact and Future Directions
This study is a compelling example of how AI can streamline clinical operations without compromising on accuracy. By reducing manual workload and speeding up the prescreening process, Mendel AI’s solution not only helps identify eligible patients faster but also supports the broader goal of accelerating cancer research and clinical trial enrollment.The platform is already in use for eligibility assessment in ongoing clinical trials, and we’re excited to see the continued impact of our technology in real-world applications. Our collaboration with the Perelman School of Medicine highlights the potential of human+AI teams to redefine traditional workflows in the healthcare sector.Join Us in Transforming Clinical TrialsAt Mendel AI, our mission is to empower healthcare professionals with cutting-edge AI tools that enhance precision, efficiency, and patient care. We are honored to contribute to advancements in clinical trial prescreening and look forward to partnering with more institutions to drive innovation in medical research.
