Rare Diseases by the Numbers
It is a sad, but unfortunate reality of human life in 2025: despite years of fighting to find cures for rare diseases, challenges remain. Defined as conditions that affect fewer than 200,000 individuals in the United States, rare diseases individually are small, but collectively they afflict over 30 million Americans. The US Food and Drug Administration (FDA) says there are 7,000 identified rare diseases1, while patient advocacy groups peg that number even higher: 10,000 rare disorders affecting one out of every ten people in the US.2
These sobering numbers represent a profound challenge for payers, from insurance providers to healthcare systems to the patients themselves who struggle with out-of-pocket expenses. For each rare disease, their inherent low prevalence makes systemic inefficiencies worse. Institutional payers grapple with escalating costs, diagnostic uncertainties, and a lack of therapeutic alternatives. As the industry moves further towards artificial intelligence (AI) for help, the newest offshoot—agentic AI—offers even greater hope for faster development and healthier outcomes.
Undertreatment of Rare Diseases
Central to payer struggles is the sheer scale and undertreatment of rare diseases. Of the rare disorders identified as of 2025, only about 5% possess FDA-approved treatments.
“The US Food and Drug Administration (FDA) says there are 7,000 identified rare diseases1, while patient advocacy groups peg that number even higher: 10,000 rare disorders affecting one out of every ten people in the US.2”
Why? The reason is simple: high failure rates in drug development. According to one analysis of Biomedtracker data, approximately 52% of candidates advance from Phase I to Phase II clinical trials, but that number falls quickly, dropping to 29% from Phase II to Phase III.3 Rare disease trials fare worse, with 30% more planned patient visits, 23% longer start-up timelines, and 19% extended treatment durations compared to non-rare trials.4 These brutal and expensive numbers often result in underpowered studies and outright trial terminations.
Rare Disease Impact on the Payer Ecosystem
Payers face heightened scrutiny over reimbursing high-cost orphan drugs, whose average return on investment has plummeted to 2.5% for launches since 2022.5 Furthermore, insurance providers face ballooning premiums, while healthcare systems contend with budget strains from gene therapies costing millions per patient. Patients, meanwhile, endure out-of-pocket expenses averaging thousands annually during prolonged diagnostic journeys, often spanning years.
A pivotal challenge lies in patient identification and recruitment. Rare diseases, by their very nature, have genetic variability. In many cases, these variables are clustered by ancestry and geography. For instance, Gaucher disease type 1 occurs in one out of every 450 Ashkenazi Jewish births versus one out of every 40,000–60,000 globally.6 Even when armed with that knowledge, trial administrators still struggle with filling trial slots, due to low awareness in the populace, misdiagnosis, or simply the fact that patients are dispersed worldwide.7 These trials need global outreach and culturally sensitive protocols. This scarcity of patients delays evidence generation. That leaves payers with not enough real-world data to assess a treatment’s value.
“Coverage decisions become more strict, exacerbating access barriers. Payers worry about absorbing the upfront costs for gene therapies, which can eventually cause policies that shift out of alignment with FDA indications, or administrative hurdles that deny or delay reimbursements.”
Consequently, coverage decisions become more strict, exacerbating access barriers. Payers worry about absorbing the upfront costs for gene therapies, which can eventually cause policies that shift out of alignment with FDA indications, or administrative hurdles that deny or delay reimbursements. Those issues, in turn, affect patients fortunate enough to find those therapies: they can face portability issues across insurers, forcing out-of-pocket payments or abandoning the therapy altogether.
“In 2025, a new hope has emerged; one that takes that AI foundation and applies it with surgeon-like precision: agentic AI. These are autonomous systems that act independently, refine hypotheses, and integrate data streams without the need for constant human prompting or intervention.”
The Role of Agentic AI in Solving Rare Disease Challenges
These issues underscore the need for innovative thinking. For the past few years, the industry has turned to AI as a transformative force. AI excels in processing vast, unstructured datasets, from genomics to wearables, accelerating diagnoses by identifying patterns in symptoms, lab results, and histories that human clinicians may overlook. In drug development, AI-led pattern recognition can slash failure rates by pinpointing viable candidates early. For patient recruitment, AI algorithms match individuals to trials via electronic health records, reducing geographic and awareness barriers.
But in 2025, a new hope has emerged; one that takes that AI foundation and applies it with surgeon-like precision: agentic AI. These are autonomous systems that act independently, refine hypotheses, and integrate data streams without the need for constant human prompting or intervention. Agentic AI’s benefits are particularly notable in the fight against rare diseases. Unlike passive tools, agentic AI orchestrates entire workflows. For diagnosis, agents can work in a hierarchy, with one that extracts phenotypes, another to prioritize variants, and yet another to synthesize evidence. A new hope, yes. But like everything in life, nothing comes easy, and agentic AI is no exception. It faces challenges in areas like data bias and hallucinations, issues that have historically plagued AI systems in healthcare. Data bias arises from underrepresented populations in training datasets, leading to skewed diagnostic outcomes for diverse patient groups affected by rare diseases. Hallucinations, where AI generates plausible but inaccurate information, pose risks in high-stakes medical decisions.
“In treatment, agentic AI personalizes plans by autonomously analyzing genetic profiles and histories for rare conditions. For payers, this means robust evidence for value-based reimbursements.”
To mitigate these, the industry is adopting innovative approaches. Retrieval-Augmented Generation (RAG) introduces external, verified knowledge bases into AI workflows, grounding outputs in reliable sources and reducing AI’s misinterpretations by cross-referencing real-time data. Small language models (SLMs), which are more efficient and specialized than larger counterparts, are being finetuned on domain-specific datasets to minimize bias while maintaining high performance. Additionally, medicalspecific large language models (LLMs), such as those trained exclusively on peer-reviewed literature, electronic health records, and clinical guidelines, enhance accuracy and trust.
diagnosis, these interventions enable hierarchical agents to incorporate bias-detection mechanisms, audit trails for hallucinations, and continuous learning loops that refine models based on clinician feedback, ensuring safer and more equitable applications in rare disease management.
An agentic framework, supercharged with these hallucination mitigation efforts, can shorten a diagnostic investigation from years to days, significantly cutting costs for the payers in this part of the journey.
Similarly, drug discovery agents autonomously design compounds for rare targets, reining in development costs and failures. In treatment, agentic AI personalizes plans by autonomously analyzing genetic profiles and histories for rare conditions. For payers, this means robust evidence for valuebased reimbursements. AI-driven recruitment expands trial diversity, producing genetically inclusive data that justifies coverage decisions. Beyond these complex applications, agentic AI can be applied right now in simpler, more administrative tasks, such as claims processing and patient data management, realizing significant cost savings for payers. Integrating agentic AI could further reduce payer burdens through predictive analytics forecasting disease progression, minimizing hospitalizations, and optimizing resource allocation.
Ultimately, while rare diseases continue to impose formidable economic and access challenges on payers, AI’s evolution, led by cutting-edge agentic systems, offers a beacon in the dark.
Enhanced diagnosis, recruitment, and evidence generation result in more strategic capital spending and more equitable outcomes. Embracing agentic AI’s role ensures rare disease management becomes more sustainable and, most critically, patient-centered.
