Rare Disease Natural History
Real natural-history data for a rare disease is scarce almost by definition — small registries, short follow-up, and analysis pipelines that often go untested simply because there isn't enough real data to test them on. This category exists to fix the second problem without pretending to fix the first: every registry here is a synthetic, multi-state model of how a disease actually progresses, built to be a safe, unlimited substrate for methodology — not a stand-in for a real cohort.
One registry follows the same disease-progression engine across three genuinely different conditions — Duchenne muscular dystrophy, Huntington's disease, Gaucher disease — with the biologically correct detail preserved in each: real recovery under enzyme replacement therapy in Gaucher, no recovery from a skeletal event once it happens, no recovery at all in Duchenne's ambulatory decline. An external check against a published hazard ratio for Huntington's closed roughly half the gap to the literature value through a mechanism predicted in advance; the other half is reported as unexplained, not quietly closed by further tuning.
What you'll typically find here
- Multi-state progression registries — patients moving through real clinical stages over years, with the enrolment design (incident cohort vs. prevalent cross-section) stated explicitly, because it changes the numbers by double digits.
- Known-truth benchmarks — the generating process is exactly known, so an estimator's error can be measured against ground truth, not against another estimator.
- Honest regulatory framing — every registry states plainly that it is not a natural history study in the FDA guidance sense, and is not fit for use as a standalone external control arm.
Where a check against the literature does not fully pass, that's stated as a numbered fact, not folded into the marketing copy.