🔬 The Persistent Challenges in Rare Disease Clinical Trials
Rare diseases affect millions worldwide, yet developing effective treatments remains a monumental task. These conditions, defined by the National Institutes of Health (NIH) as impacting fewer than 200,000 people in the United States per disorder, number over 7,000 known types. Despite this prevalence—estimated to burden 30 million Americans—more than 90% lack FDA-approved therapies. A primary bottleneck lies in clinical trials, where patient recruitment often drags on for months or years due to small patient pools, stringent eligibility criteria, and the labor-intensive process of manually reviewing electronic medical records (EMRs).
Traditional chart review requires clinicians and research coordinators to sift through vast amounts of unstructured data, including clinical notes, lab results, and imaging reports. This process can take hundreds of minutes per patient, limiting scalability across large health systems. For rare diseases like transthyretin amyloid cardiomyopathy (ATTR-CM), a progressive heart failure condition primarily affecting older adults through amyloid protein buildup in the heart, identifying suitable trial candidates becomes even more daunting. ATTR-CM exemplifies how diagnostic delays and geographic barriers exacerbate underrepresentation, particularly among diverse populations not routinely connected to specialists.
Health systems like Cleveland Clinic, with its expansive network of 25 hospitals and 250 outpatient centers spanning Ohio, Florida, and Nevada, face these hurdles daily. Recruitment inefficiencies not only delay trial timelines but also hinder equitable access, as traditional methods favor patients already engaged with specialized care. Statistics reveal that clinical trials for rare diseases often recruit fewer than 10 patients per site, with global starts comprising nearly half of all studies yet facing high failure rates due to enrollment shortfalls.
🌐 How AI Transforms Patient Identification for Trials
Artificial intelligence (AI), particularly large language models (LLMs) trained on medical data, offers a paradigm shift in addressing these challenges. Medically trained LLMs excel at natural language processing (NLP), enabling them to parse complex, unstructured EMRs with human-like comprehension. Unlike rule-based systems, these models learn contextual nuances, such as subtle symptom descriptions or evolving lab trends, to answer trial-specific eligibility questions accurately.
In the context of rare disease trial identification, AI-driven chart review automates the pre-screening phase. Systems like Synapsis AI from Dyania Health integrate seamlessly into EMR platforms, querying structured data alongside free-text notes. This end-to-end approach provides not just yes/no decisions but detailed, auditable justifications, allowing clinicians to validate outputs efficiently. Early pilots, including those for melanoma and polycythemia vera at Cleveland Clinic, demonstrated AI completing assessments in minutes versus hours, paving the way for broader rare disease applications.
By scaling across health systems, AI uncovers hidden eligible patients—those outside specialist networks or from underserved communities—accelerating enrollment while maintaining rigorous safety standards through human oversight.
📊 Spotlight on the Cleveland Clinic Study
A groundbreaking study published in The Journal of Cardiac Failure, the official journal of the Heart Failure Society of America, showcases this technology in action. Collaborating with Dyania Health, Cleveland Clinic deployed Synapsis AI for the DepleTTR-CM Phase 3 trial targeting ATTR-CM. The trial sought patients with confirmed ATTR-CM diagnosis, specific cardiac metrics, and exclusion criteria like comorbidities that could confound results.
Over one week, the AI reviewed EMRs of 1,476 patients, flagging 46 potential matches. Clinicians then confirmed eligibility, revealing 30 true positives—29 of whom were previously unidentified through 90 days of traditional recruitment. This efficiency led to seven enrollments in just six days, surpassing the site's goal before halting further screening. Cleveland Clinic's newsroom details the full methodology.
The study's rigorous validation involved 7,700 trial-specific questions across nine domains, from diagnosis confirmation to organ function assessments.
📈 Impressive Accuracy and Performance Metrics
The AI system's precision stood out: it achieved 96.2% accuracy overall, with a 99% negative predictive value (NPV) by correctly excluding 198 of 200 non-eligible patients. Every justification provided was 100% accurate and physician-interpretable, fostering trust in the technology.
| Metric | AI-Driven Screening | Traditional Screening |
|---|---|---|
| Patients Screened | 1,476 (1 week) | Not specified (90 days) |
| Potential Matches | 46 | Lower yield |
| Enrollments | 7 (6 days) | 10 (90 days) |
| Accuracy | 96.2% | Baseline human |
| NPV | 99% | N/A |
Lead investigator Trejeeve Martyn, M.D., Director of Heart Failure Population Health at Cleveland Clinic, noted, “This study shows how medically trained AI can support chart review at scale, transforming what has traditionally been a labor-intensive process.” Eirini Schlosser, CEO of Dyania Health, added that AI addresses bottlenecks by surfacing overlooked patients, especially from underrepresented groups. For more insights, see the Medical Xpress coverage.
🎯 Boosting Diversity and Equity in Trials
One of the study's most compelling outcomes was enhanced diversity. Among 30 AI-confirmed matches, 36.6% identified as Black, compared to just 7.1% in traditional screening. Additionally, only 60% had prior heart failure specialist connections (versus 92.8% traditionally), democratizing access for primary care patients across geographies.
- Racial diversity: AI captured 5x more Black patients.
- Geographic reach: Screened patients from multiple states without specialist bias.
- Equity impact: Reduced barriers for underrepresented minorities in rare disease research.
This aligns with NIH priorities for inclusive trials, where diverse data improves generalizability and outcomes for all populations.
🚀 Broader Implications and Future Directions
Beyond ATTR-CM, Synapsis AI holds promise for observational research, disease registries, and deploying underutilized therapies. Cleveland Clinic's investment in Dyania Health signals scalability, with pilots expanding to oncology and hematology. As AI refines rare disease trial identification, it could slash timelines, cut costs, and fill therapeutic gaps.
For academics and researchers, this underscores opportunities in research jobs blending AI and medicine. Explore clinical research jobs or postdoc positions advancing such innovations. Professionals can leverage tools like those on how to write a winning academic CV to enter this field.
In summary, AI-driven chart review is reshaping rare disease clinical trials, offering speed, accuracy, and inclusivity. Share your thoughts in the comments—what role do you see AI playing in future research? Check Rate My Professor for insights on leading AI educators, browse higher ed jobs, or visit higher ed career advice and university jobs for opportunities. Post a job at our recruitment page.