In the evolving landscape of clinical research, the integration of advanced technologies has become crucial. Synapsis AI, a large language model-based system trained in medical applications, has revolutionized the screening process for patients eligible to participate in a phase III clinical trial for polycythemia vera (PV). By significantly reducing the time and effort required for patient screening, Synapsis AI has enhanced both the volume of patients screened and those successfully enrolled in the trial, surpassing traditional methods.
“Once we integrated the [clinical trial] protocol into the LLM and specified the study, we received a list of approximately 30 eligible patients from our medical system within seconds,” explained Aaron T. Gerds, MD, MS, a physician in the Department of Hematology/Oncology at Cleveland Clinic.
The ASCO Post interviewed Dr. Gerds regarding the capabilities of Synapsis AI and the research he and his colleagues shared at the 2025 American Society of Hematology (ASH) Annual Meeting and Exposition.1
Understanding Clinical Trial Screenings
Finding suitable candidates for clinical trials is often a resource-intensive endeavor. This challenge is particularly pronounced in studies involving rare diseases, where identifying a sufficient number of patients can be difficult.
Dr. Gerds, who also holds a position as Assistant Professor of Medicine at Case Western Reserve University and serves as Deputy Associate Director for Clinical Research at the Case Comprehensive Cancer Center’s Developmental Therapeutics Program, noted that standard methods of patient chart reviews can take over 30 minutes, depending on the trial’s inclusion/exclusion criteria and the length of the patient’s medical history. Candidates are usually identified only during active clinical care, which further constricts the pool available for upfront treatment trials.
Aaron T. Gerds, MD, MS
Researchers at the Cleveland Clinic explored utilizing Synapsis AI for the automation of patient eligibility identification. The team assessed the system’s accuracy and calculated its positive predictive value. They then compared the number of patients identified through Synapsis AI with those found via traditional methods.
“The significant difference in patient identification underscores the efficiency and time-saving benefits of Synapsis AI’s automated prescreening, compared to conventional approaches,” Dr. Gerds emphasized. “By streamlining recruitment, tools like Synapsis AI can enhance trial efficiency and accelerate the development of life-saving therapies, even in the realm of rare diseases.”
The ongoing phase III clinical trial is designed to compare treatment outcomes between the class I and class II histone deacetylase inhibitor givinostat and hydroxyurea in patients with JAK2 V617F-positive high-risk polycythemia vera (GIV-IN-PV; NCT06093672).
Dr. Gerds explained that Synapsis AI systematically filtered through the Cleveland Clinic Health System’s records to pinpoint eligible trial candidates. Initially, the system looked for cancer-related ICD-10 codes to identify polycythemia vera patients treated within the last three years. It then analyzed the patients’ electronic health records against the trial’s criteria, utilizing both structured and unstructured data.
Results
Within the Cleveland Clinic’s health system, there are 4.7 million active electronic health records. Among these, 28,200 belong to patients diagnosed with cancer in the past three years, with 904 specifically diagnosed with polycythemia vera.
Synapsis AI completed a full assessment of all eligible patients within one week, successfully identifying 22 individuals who met the inclusion criteria. In total, the system flagged 50 candidates before the trial enrollment closed.
Research staff then validated the eligibility of the candidates identified by the system, confirming that all patients were indeed qualified, yielding a positive predictive value of 100%.
In contrast, traditional methods resulted in clinical staff prescreening only nine patients, with four enrolled and three treated over a span of 12 months. This amounted to an average enrollment of about one patient every four months.
Synapsis AI achieved a sevenfold increase in candidate identification, all within a fraction of the time required by the medical team, thus significantly lightening the workload for the clinical trial team and accelerating the trial timeline.
“If we can enroll more patients in a shorter timeframe for studies involving rare diseases, we can dramatically expedite drug development. Instead of taking 5, 10, or even 15 years, the timeline could be halved,” Dr. Gerds remarked. “As these systems become further developed, clinical research efficiency will significantly improve.”
DISCLOSURE: Dr. Gerds disclosed research funding from Kratos Pharma, Ionis Pharmaceuticals, Imago BioSciences, Constellation Pharmaceuticals, Accurate Pharmaceuticals, and CTI Biopharma, as well as consultancy roles for various organizations.
REFERENCE
1. Gerds A, et al: 2025 ASH Annual Meeting. Abstract 4340. Presented December 7, 2025.
EXPERT POINT OF VIEW: Kenneth L. Kehl, MD, MPH
The role of artificial intelligence in analyzing clinical data is anticipated to grow significantly in the coming years, particularly in identifying candidates for cancer clinical trials. Various AI tools are currently in development.
Kenneth L. Kehl, MD, MPH
It will be crucial to assess how these tools can be integrated into clinical research workflows, especially across different cancer types and institutions. A primary challenge in this area is the absence of a central source detailing which trials currently have openings for various patient types, given the dynamic nature of trial availability. The development of AI infrastructure to suggest suitable trial options is often the easier task. The real challenge lies in effective implementation, connecting these systems to electronic health records and clinical trial databases, and proving their tangible impact on trial enrollment outcomes.
DISCLOSURE: Dr. Kehl reported no conflicts of interest.
Kenneth L. Kehl, MD, MPH, serves on the faculty at the Lowe Center for Thoracic Oncology and the Division of Population Sciences at Dana-Farber Cancer Institute in Boston.
