In rare disease, these challenges are amplified. With only 5% of rare diseases having an FDA-approved treatment, the need for effective trials is urgent, but the populations involved are small, geographically dispersed, and often difficult to identify through traditional clinical pathways. Standard site models and recruitment strategies are frequently insufficient in these settings. However, recent advances in AI, genetic screening, and flexible site activation models are enabling more reliable identification of eligible patients and reducing the structural barriers that delay trial access.
At a roundtable hosted by Sano at the World Orphan Drug Congress in April 2025, clinical operations, translational medicine, and advocacy stakeholders examined the structural failure points in rare disease recruitment and what operational alternatives are beginning to address them. The following draws on that discussion to examine each in turn.
Key Takeaways
- AI-Driven Matching: Large Language Models (LLMs) and hotspot mapping are significantly increasing the accuracy of identifying eligible rare disease patients.
- Proactive Screening: Genotype-first approaches allow for the detection of pathogenic variants in patients who remain undiagnosed through traditional clinical pathways.
- Agile Site Models: Just-in-Time (JIT) and virtual site activations reduce setup delays and allow patients to participate regardless of geographic location.
- Global Scale: Leveraging international registries and regions with high disease prevalence is essential to overcome the limitations of small, dispersed patient populations.
Patient identification using AI and genomic tools
In rare disease, the most fundamental recruitment challenge is that eligible patients often remain unidentified. Many are undiagnosed, misdiagnosed, or simply unaware that relevant trials exist. Traditional referral pathways are insufficient for populations this small and dispersed. New strategies are emerging to close this gap, particularly with advancements in AI models. For instance, a large language model (LLM) demonstrated high accuracy in identifying and matching patients to clinical trials for which they were eligible. Experts also mentioned the relevance of hotspot mapping techniques, which layer geographic and epidemiological data to prioritise regions with the highest probable patient density.
AI can also be used to model trial feasibility before protocol design, helping identify eligible patient populations and optimal sites. In rare disease, the small size of available datasets currently limits what these models can achieve. But as natural history studies and real-world data collection expand, the training data available to these models will grow, making feasibility modeling increasingly practical for even the smallest patient populations. In addition, the genotype-first screening approach is enabling the proactive detection of individuals carrying rare pathogenic variants, many of whom remain undiagnosed through traditional clinical pathways. Together, these methods improve the likelihood of identifying eligible patients before protocol finalization and reaching patients who would otherwise be missed by conventional site-based approaches.
Just-in-time and virtual approaches to site activation
Trial setup in rare disease is often associated with delays and inefficient start-and-stop phases. But operational friction is only part of the problem. For patients, the physical location of a study site is one of the most influential factors in whether they participate at all. When trials require travel to distant centers, many otherwise eligible patients never enroll. The just-in-time (JIT) approach was created to reduce unnecessary delays and make transitions between phases smooth. This entails engaging with study sites early in the process, prior to formal site approval. This approach prioritizes patient identification and streamlines the rest of the process thereafter. Virtual site models extend this flexibility further, enabling participation even when patients are geographically distant or unable to travel. Mobile sites offer an alternative way to deliver trial-related services closer to patients' homes. However, these models can introduce complexities around clinical oversight, sample handling, and consistency of the patient experience. Because patients weigh the nature of study procedures heavily when deciding to participate, maintaining trust and quality in decentralized delivery is essential.
Adapted from Lynam et al., 2012.
Global recruitment strategies
Since many rare diseases are inherited, families in regions with higher disease prevalence represent a unique opportunity for recruitment. This is particularly true in countries where consanguinity is present and inherited diseases may be more easily identified within extended family networks. Targeting these populations can increase the density of eligible candidates per recruitment effort, reducing the time and cost associated with identifying qualified participants.
To overcome the persistent challenge of patient recruitment in rare disease, stakeholders also highlighted the promise of cross-border strategies. Beyond expanding the eligible population, global approaches help ensure that trial cohorts reflect the diversity of patients who will ultimately use the therapy. Without participants from a wide range of backgrounds, treatment efficacy cannot be confidently generalized. A notable example of a global strategy is Sanofi’s Rare Disease Registry that includes data on patients from over 65 countries, representing a diverse population. However, effective and widespread cross-border coordination remains a limiting factor.
Establishing global patient registries and partnerships will require:
- Long-term stakeholder commitment
- International regulatory alignment
- Data interoperability standards
- Significant infrastructure investment