In brief

• Real-world claims data can serve as a reliable evidence foundation for Pompe disease, with literature-based symptoms confirmed as both identifiable and measurable within routine healthcare data.
• Clinician-defined clinical trial endpoints can be mapped to and monitored within routine claims data, strengthening the real-world evidence base for regulatory and HTA purposes.
• Machine learning can discover clinically meaningful disease features in Pompe disease that no pre-specified framework had defined, informing the design of future studies and trials.

New research presented at ISPOR Global 2026 in Philadelphia demonstrates that machine learning can map clinician-defined endpoints to real-world claims data in Pompe disease and surface disease manifestations beyond pre-specified frameworks. The study, conducted by Volv Global in collaboration with Sanofi, was conducted in a US administrative claims database.

Pompe disease is a rare, chronically debilitating metabolic disorder in which enzyme replacement therapy has now extended patient survival, bringing new long-term manifestations not captured by endpoints established earlier in its treatment history. Many clinically meaningful endpoints do not map to routine healthcare codes, leaving a gap between what patients experience and what the evidence base reflects – with consequences for disease monitoring, HTA submissions, and trial design.

The research addresses that gap through three sequenced methodological contributions:

• The prevalence of literature-based disease symptoms was compared between the Pompe patient cohort and a matched control population without Pompe disease, confirming that the correct patients are represented in the claims data and that these symptoms are reliably measurable within it – a foundational validation step underpinning all subsequent analyses.
• Machine learning models mapped 46 of 67 pre-specified clinical endpoints to diagnosis, procedure, and treatment codes in claims data, demonstrating that endpoints designed for clinical trials can be reliably tracked in routine healthcare data.
• An unsupervised discovery analysis identified novel cardiovascular, respiratory, and systemic features highly prevalent in the Pompe cohort but absent from any pre-specified framework, confirmed against the same control population and offering candidates for endpoint design in future natural history studies and trials.

Volv Global's proprietary machine learning methodology was applied across all three components, providing a reproducible framework applicable across rare diseases where treatment advances have outpaced existing evidence frameworks.

"Rare disease evidence frameworks are often frozen at the moment of first approval," said Christopher Rudolf, CEO and Founder of Volv Global. "This research demonstrates that machine learning can systematically close that gap – confirming what clinicians know, independently discovering what the data reveals, and making this a tractable problem for any team building an evidence strategy in a disease where treatment has changed the clinical picture. This is precisely the work Volv Global exists to do."

On the study
Patients were identified in a US administrative claims database using confirmed diagnosis and/or treatment records. The control population comprised patients with mimic disease codes and no Pompe disease history in the preceding seven years. Of 67 pre-specified clinical endpoints, 46 were successfully mapped to claims codes; the 21 unmapped endpoints reflect the limits of administrative claims data coding, and are themselves informative for teams assessing the feasibility of claims-based real-world evidence strategies. All findings are based on retrospective analysis; prospective validation has not been conducted and is not claimed. Research conducted in collaboration with Sanofi. To the abstract