Using Real-World Data to Redefine Evidence for Ultra-Rare Diseases

In ultra-rare diseases, where traditional control arms are often impractical, regulators still require meaningful context to interpret trial outcomes. Here’s how a biotech company leveraged NorstellaLinQ to build a regulatory-grade real-world control dataset that combined deep unstructured clinical notes, AI-enabled data curation, and rigorous propensity score matching. By creating a highly comparable real-world cohort, the company provided critical context for survival outcomes, supporting its FDA submission and strengthening stakeholder confidence to:

Build regulatory-grade real-world evidence with confidence
Create credible real-world control cohorts for rare diseases
Strengthen FDA submissions with unstructured clinical data

Key Results

Create Credible Real-World Control Cohorts

Challenge

For ultra-rare diseases, traditional clinical trial designs often fall short. Small patient populations make randomized control arms impractical, yet regulators still need context to interpret outcomes.

In this case, a biotech company faced that exact challenge: demonstrating whether observed survival outcomes reflected a true treatment effect or simply natural disease variation. The solution wasn’t more patients, it was better context.

The real-world control analysis generated through this work was ultimately submitted to the FDA as part of a regulatory approval package, underscoring the growing acceptance of high-quality real-world data (RWD) as a meaningful component of regulatory decision-making.

This submission reflects a broader shift already underway: regulators increasingly expect real-world evidence to complement clinical trial data, especially in rare and ultra-rare settings. But not all RWD is created equal.

Structured claims and EHR data alone cannot reliably identify relapse status, confirm eligibility criteria, or capture nuanced clinical factors that define whether patients are truly comparable. That level of detail lives in unstructured clinical notes, and requires the right technology and expertise to unlock.

Solution

The company used NorstellaLinQ to build a fit-for-purpose real-world dataset designed specifically to meet regulatory-grade evidentiary needs.

Rather than relying on off-the-shelf datasets, the approach combined:

Deep unstructured clinical notes, enabling identification of relapsed/recurrent patients and granular disease characteristics
AI- and technology-enabled data curation, transforming raw clinical text into analytically usable variables
Rigorous analytic services, including clinically grounded propensity score matching

Each clinical trial patient was matched to five highly similar real-world patients based on tumor location, eligibility criteria, disease timing, and other clinically meaningful factors, ensuring the comparison reflected real-world clinical practice, not just data availability.

Outcome

Even with small sample sizes, the matched real-world comparison delivered what traditional statistics alone could not: interpretability. It provided regulators with essential context for understanding survival outcomes relative to what typically happens outside the trial setting.

That credibility extended beyond regulatory review by:

Supporting FDA discussions and submission strategy
Strengthening investor and board communications
Informing internal development and lifecycle planning
Enhancing confidence among investigators and external stakeholders

Importantly, the dataset itself became a durable strategic asset, reusable, refinable, and adaptable as new questions emerged over time.

This work illustrates where the industry is headed: regulatory-grade evidence increasingly depends on high-quality RWD, unstructured clinical insights, advanced technology, and expert analytics working together.

NorstellaLinQ’s ability to integrate clinical notes, AI-enabled data extraction, and rigorous analytic services allows biopharma teams, especially small and emerging ones, to build credible evidence frameworks earlier, faster, and with greater confidence.

Get Started

Get Closer to the Patient Than Ever Before

NorstellaLinQ links claims, EHR, lab, and electronic prescription data into a single patient view — decision-grade and AI-ready. See what it can do for your team.

Frequently asked questions

Everything You Need to Know

How can real-world data support regulatory submissions for ultra-rare diseases?

High-quality real-world data provides critical context when traditional randomized control arms are not feasible. By creating regulatory-grade real-world control cohorts, biopharma companies can strengthen FDA submissions and better demonstrate treatment effectiveness.

Why are real-world control cohorts important in ultra-rare disease research?

Ultra-rare diseases often have patient populations too small for traditional randomized clinical trials. Carefully matched real-world control cohorts provide regulators with meaningful comparisons that help interpret clinical trial outcomes and survival data.

How do unstructured clinical notes improve real-world evidence?

Unstructured clinical notes capture critical details—such as relapse status, eligibility criteria and disease characteristics—that structured claims and EHR data often miss. AI-enabled data curation transforms these insights into variables suitable for regulatory-grade analyses.

What role does propensity score matching play in real-world evidence studies?

Propensity score matching creates highly comparable real-world control cohorts by matching patients on clinically meaningful characteristics, helping ensure that treatment comparisons reflect real-world clinical practice rather than differences in patient populations.

What outcomes did Norstella deliver in this case study?

Norstella helped the biotech company build a regulatory-grade real-world evidence dataset, create credible real-world control cohorts, strengthen its FDA submission strategy, improve investor and stakeholder confidence, and establish a reusable evidence asset for future development and lifecycle planning.