Rare Disease Trial Design: Controls, Endpoints, Power

The Problem

You’re staring at a rare disease trial readout. The company enrolled 20 patients. There’s no placebo group. The “control” is a dataset from a patient registry you’ve never heard of. The primary endpoint is a biomarker you can’t pronounce. The stock is up 40% and you need to decide whether to buy or run.

Rare disease trials don’t look like standard Phase 1-2-3 programs. Evaluate them with the same framework you use for a 500-patient randomized study and you’ll get the risk wrong in both directions: panic at n=20 when the trial is well-designed, or dismiss the risks because “orphan” doesn’t mean “easy approval.”

The Solution

Learn what makes rare disease trials different: natural history studies as external controls, accelerated approval on surrogate endpoints, adaptive designs that combine phases, and sample sizes that look tiny but may be statistically defensible.

Step-by-Step

Step 1: Find the Trial on ClinicalTrials.gov and Read the Protocol

Go to ClinicalTrials.gov and search the NCT number from the press release. Read the “Study Design” section. Is it randomized? Single-arm? Open-label? These fields tell you the evidence level before you read any data.

A single-arm trial has no control group: every patient gets the drug. The company compares results to something external, which is where the trouble starts. Single-arm trials are common in rare disease because randomizing half the patients to placebo when the disease is fatal or progressive is ethically difficult.

The FDA requires trial sponsors to register applicable clinical trials on ClinicalTrials.gov within 21 days of enrolling the first patient under FDAAA 801. The primary endpoint is locked at registration. Check the “History of Changes” tab: if the company changed the primary endpoint mid-trial, that’s a red flag. For a deeper guide, see the how to read ClinicalTrials.gov walkthrough.

Step 2: Identify the Control Arm (or Lack of One)

The control-arm question determines everything else about how you read a rare disease trial. Three possibilities:

1. Randomized controlled trial: Patients are randomly assigned to drug or placebo. The gold standard, but rare in ultra-rare diseases because the patient population may be too small to split.
2. External control arm (ECA): The trial is single-arm, but the company compares treated patients to data from a natural history study, patient registry, or real-world data. The FDA’s February 2023 draft guidance on externally controlled trials is cautious: “in many situations, the likelihood of credibly demonstrating the effectiveness of a drug of interest with an external control is low.”
3. No control at all: Single-arm, no external comparison. The company argues the drug’s effect is so large that no control is needed. The weakest evidence and the hardest to get past the FDA.

Natural history studies track how a disease progresses without treatment. A good one uses the same endpoint assessments, patient population, and time frame as the treatment trial. If any don’t match, the comparison is biased.

Real example: uniQure’s AMT-130 gene therapy for Huntington’s disease uses the Enroll-HD natural history dataset as an external control. Enroll-HD is a prospective observational study tracking Huntington’s in thousands of patients worldwide. The comparison is credible because the dataset is large, well-characterized, and uses similar outcome measures. Prior analysis of the AMT-130 BLA filing.

Step 3: Check the Primary Endpoint: Surrogate or Clinical?

A surrogate endpoint is a biomarker reasonably likely to predict clinical benefit but isn’t itself a clinical outcome. Examples in rare disease: enzyme levels for lysosomal storage disorders, bile acid levels for cholestatic diseases, dystrophin for DMD.

The FDA’s Accelerated Approval Program lets drugs for serious conditions with unmet need get approved based on surrogate endpoints, with the requirement that the company run a confirmatory trial to verify clinical benefit. If the confirmatory trial fails, the FDA can withdraw approval. Under FDORA (2022), the FDA gained expedited withdrawal authority, no longer needing years of negotiation with the sponsor.

This matters for investors: accelerated approval carries withdrawal risk. About 15% of accelerated approvals in oncology have been withdrawn (Mehta et al., JCO 2024). In rare disease, the confirmatory trial may be hard to run because the patient population is small, meaning delays and lingering withdrawal risk.

Real example: Ipsen’s Bylvay (odevixibat) received accelerated approval for progressive familial intrahepatic cholestasis. The confirmatory BOLD trial in biliary atresia missed its primary endpoint and the stock dropped. Coverage of the Bylvay BOLD Phase 3 miss. For primers, see surrogate endpoints and accelerated approval.

Step 4: Assess the Sample Size and Statistical Power

Rare disease trials are small. That’s not automatically a problem. The question is whether the sample size is adequate to detect a meaningful effect given the expected effect size and variability.

A trial with n=20 can have adequate power if the expected effect is large: if every untreated patient declines and every treated patient improves, n=20 is enough. But if the effect is modest, n=20 may be underpowered, and a p=0.03 result on 12 patients may be noise.

Check: What’s the expected effect size stated in the protocol? What’s the variability (rare disease endpoints can be noisy because the population is heterogeneous)? Is the analysis Bayesian (borrowing strength from prior data) or frequentist? Bayesian designs can make small trials more efficient, but prior assumptions matter enormously. For hazard ratios, confidence intervals, and p-values, see the guide to clinical trial readouts.

Step 5: Determine If the Trial Is Registrational

A registrational trial is designed to support an NDA or BLA submission. Not every trial is. Companies sometimes run exploratory Phase 2 trials that generate interesting data without being sufficient for approval. Check the ClinicalTrials.gov entry for “registrational” or investor materials for “designed to support regulatory submission.” For more on what makes a trial registrational, see the explainer on what is a pivotal trial.

Step 6: Check for Priority Review Voucher Eligibility

Rare disease drugs, especially for rare pediatric diseases, often qualify for a priority review voucher (PRV): a transferable FDA coupon that converts a standard 10-month review into 6-month priority review. PRVs sell for $67M to $180M. For a small-cap biotech, PRV cash can exceed the company’s market cap. The Rare Pediatric Disease PRV program sunsets September 30, 2029, creating a scarcity premium. See the priority review vouchers guide for the full economics.

Rare disease trials often combine phases (Phase 2/3 designs, adaptive designs that expand cohorts mid-trial). Don’t assume standard Phase 1 to Phase 2 to Phase 3 progression. Check what the trial calls itself on ClinicalTrials.gov. For standard phase definitions, see clinical trial phases.

Common Mistakes

Assuming small means underpowered. A trial with 20 patients can have adequate power if the expected effect is large and the endpoint is objective. Read the sample size calculation before discounting n=20.

Assuming orphan means safe. Orphan designation means the disease affects fewer than 200,000 people in the US. It says nothing about drug safety. The safety database in a 20-patient trial is thin.

Ignoring the external control’s quality. An ECA is only as good as the data behind it. If the natural history study used different endpoint assessments, enrolled patients at a different disease stage, or was collected in a different era of standard of care, the comparison is biased.

Treating surrogate endpoint success as clinical proof. A drug that lowers a biomarker has not been proven to help patients feel better, live longer, or function better. The confirmatory trial is the real test.

Forgetting that accelerated approvals can be withdrawn. FDORA gave the FDA expedited withdrawal authority. If the confirmatory trial fails or is delayed, the drug can be pulled.

Overlooking PRV economics. A PRV can be worth $100M or more. For a company with a $200M market cap, that is dilution avoidance and cash runway extension. For sector pricing context, see the guide to orphan drug pricing.

Final Checklist

Before you buy a rare disease biotech stock on trial data, confirm:

• Trial is registered on ClinicalTrials.gov with a locked primary endpoint
• You know whether the trial is randomized, single-arm with ECA, or single-arm with no control
• Primary endpoint is identified as surrogate or clinical outcome
• If surrogate: you know what the confirmatory trial needs to show
• Sample size and expected effect size are stated in the protocol
• Trial is registrational (designed to support NDA/BLA)
• Drug has orphan designation and possible rare pediatric PRV eligibility
• You’ve checked the “History of Changes” tab for endpoint modifications
• You understand which clinical trial phases are being combined
• You’ve read the rare disease PDUFA catalysts overview for sector context

Rare disease trials trade one kind of risk (small sample, unconventional design) for another (ethical necessity, large unmet need). The investors who make money here can tell the difference between a well-designed 20-patient trial with a matched external control and a sloppy single-arm study with no credible comparator. Read the protocol first.