Finding Eligible Patients in a Rare Metabolic Disorder Trial

Background

A rare disease sponsor developing an enzyme replacement therapy for a rare metabolic disorder — estimated US prevalence approximately 1 in 80,000 — needed to recruit 42 eligible patients across 7 academic medical centers for a pivotal study. The primary challenge was not site quality or trial design: it was the extreme rarity of the target patient population combined with the complexity of diagnosis confirmation criteria.

The Challenge

Patients with this condition are often misdiagnosed or underdiagnosed. Their EHR records frequently show diagnoses coded under broader metabolic categories rather than the specific ICD-10 code for the target condition. Standard protocol-matching approaches based on diagnosis codes alone were surfacing only 20–30% of the actual eligible population.

Additionally, the eligibility criteria included specific enzyme activity thresholds documented in lab results — values that require pattern-matching across lab test naming conventions that vary significantly between EHR systems.

What TrialVyx Did

TrialVyx developed a phenotype model that combined multiple signals to identify patients who were likely eligible even when the specific target diagnosis code was absent:

• Pattern detection across broader diagnostic code families associated with the metabolic pathway
• Lab result matching normalized across 7 different lab test name conventions used across the 7 sites
• Clinical notes NLP to identify symptom clusters and specialist referral patterns consistent with the target condition
• Exclusion flagging for comorbidities listed in the protocol's exclusion criteria

Results

The TrialVyx phenotype engine identified 4.1x more candidate patients than the baseline approach (diagnosis-code-only chart review). The trial reached full enrollment in 6 months — well ahead of the sponsor's 18-month projection based on historical rare disease enrollment rates.

Screen failure rate was 38% — high by general standards, but significantly below the 65–70% rates typical for rare disease trials with complex diagnostic criteria. The majority of screen failures were due to lab threshold confirmation, not patient willingness or undisclosed contraindications.