When you can’t randomize, the control arm is the submission. We design, build, and defend external control arms, hybrid designs, and small-sample trials for programs that cannot randomize — rare disease, oncology, gene therapy, pediatrics — with synthetic controls taken through NDA, Bayesian methods in rare disease, and natural-history comparators built to agency standard. Statisticians with a minimum of 15 years of regulated-industry experience.
Too few patients, no ethical placebo, an accelerated pathway, a therapy that changes the disease course. The agency still needs a comparator it believes — in rare disease, in oncology, in gene therapy, in pediatrics. That comparator is what we build.
Natural-history comparators for exon-skipping, gene therapy, and small-molecule programs where placebo is unethical and every patient counts.
CAR-T, AAV, and ex-vivo programs where the external control decides the label — from first-in-human dose escalation to BLA.
Prior-trial and real-world comparators for single-arm oncology programs, built to the precedent the agency has already accepted.
Pediatric extrapolation, platform trials, and hybrid designs that augment a small randomized control with external data — Bayesian borrowing that learns fast from few patients.
The sequence the FDA’s externally controlled trials guidance and EMA reviewers expect to see — run end to end, reproducible end to end.
Natural history, registries, EHR, prior trials — selected for fit to your eligibility, era, and endpoints.
Treatment effect, population, index date, and intercurrent-event strategy under ICH E9(R1), pre-specified.
Eligibility alignment, propensity and doubly robust adjustment, Bayesian borrowing where the prior is defensible.
Tipping point, E-values, alternative comparators — every question the reviewer will ask, answered before it is asked.
Briefing document, meeting package, and the statistician in the room.
Fixed-fee engagements with defined scope. Each begins with a signed analysis plan and ends with documentation an agency reviewer can follow line by line.
Comparator source selection, eligibility alignment, propensity and doubly robust adjustment, Bayesian borrowing, and the full sensitivity suite — tipping point, unmeasured confounding, index-date alignment.
Delivered as a submission-ready package: protocol section, SAP, analysis, and the briefing-document argument.
Rare-disease, first-in-human, and platform designs: hierarchical and borrowing models, adaptive and group-sequential plans, estimand framing, and power by simulation — not by table.
Statistical sections of the submission, ISS/ISE, Type B and C meeting packages, responses to information requests, and the room-ready statistician at the agency meeting.
Comparative-effectiveness and natural-history studies from EHR, claims, and registry data — target-trial emulation, causal methods, and uncertainty quantification. Try the Conformal Engine →
What changes when the comparator is external — and what the agency asks instead.
| Randomized control arm | External control arm, built by InterClin AI | |
|---|---|---|
| Comparator | Enrolled, randomized, concurrent | Natural history, registry, EHR, or prior trials — matched to your protocol |
| Bias control | By design | By method: eligibility alignment, propensity/doubly robust adjustment, index-date rules |
| Regulatory basis | Standard | FDA externally controlled trials guidance; precedent in rare disease, gene therapy, oncology |
| Sample size | Full control arm enrolled | Control patients largely external; every enrolled patient can be treated |
| What the agency asks | Was randomization intact? | Is the comparator credible? — answered in the sensitivity suite we pre-specify |
| Who has done it | Everyone | Statisticians who have taken synthetic controls through an NDA |
The precedent record — every FDA approval built on an external control — and where our statisticians have been in it: Our experience →
Describe your study in plain English; validated formulas do the math; get an IRB- and protocol-ready justification, group-sequential plan, and DSMB outline.
Upload calibration data, get statistically proven coverage intervals with empirical validation — the uncertainty quantification we use in RWE models. Runs in your browser.
InterClin AI is an independent statistical practice built for programs that cannot randomize. Every statistician on our engagements brings a minimum of 15 years in regulated industries — rare disease, neuromuscular and DMD, cell and gene therapy, oncology, and CNS — with synthetic control arms taken through NDA, Bayesian methods in rare-disease submissions, natural-history comparators, and FDA/EMA interactions from first-in-human to approval.
Senior statisticians with a minimum of 15 years in regulated industries, and the people who help sponsors find us. Remote, engagement-based.
A written read on comparator sources, matching strategy, regulatory precedent, and a fixed-fee scope for the full build — free of jargon, ready for your CMO.