Adaptic Health
Clinical trials cost about $2.6 billion on average and about 90% of them fail.
Core trick
Feeds matched PubMed and ClinicalTrials.gov studies into a step-by-step model reasoning pass, then simulates the failure mechanism with real trial statistics.
Description
Clinical trials cost about $2.6 billion on average and about 90% of them fail. A lot of that failure traces back to one thing: before a single patient signs up, someone picked the wrong research question.
Question Audit checks that question first. You type your trial idea in plain English. Claude Opus 4.8 reads it, pulls in real studies from PubMed and ClinicalTrials.gov, and works through it step by step with its reasoning visible on screen.
The main output is one card: the wrong question, why it fails, the better question, and what changes. An interactive simulator shows the failure mechanism live, using real statistics. Every run shows its work, grades itself against a 7-point rubric, and lets you export a one-page brief or share a link that rebuilds the summary with no backend.
Built around the 2026 NHS Galleri MCED trial: the test worked, but the trial asked the wrong question.
README excerpt
# clinical.dev — Question Audit **Find the flaw before protocol lock.** Live target: https://audit.clinical.dev Question Audit by **clinical.dev** audits the *research question* behind a clinical trial before the first patient is enrolled, using Claude Opus 4.8 to expose hidden assumptions, quantify failure modes, and produce a revised adaptive design — with proof attached. > Product surface is branded **clinical.dev Question Audit**. The model (Claude Opus 4.8) is discussed in this README and the demo/submission docs, but kept out of the product UI on purpose — judges and founders see a product, not a model output page. ## The Problem Clinical trials cost ~$2.6B on average and fail ~90% of the time. Most failures trace to design decisions made before enrollment — wrong endpoint, wrong population, wrong composite, insufficient power for the real effect size. The 2026 NHS Galleri MCED trial is the canonical example: it missed its primary endpoint not because the technology failed, but because it tested a composite all-cancer late-stage endpoint in a general low-prevalence population, where positive predictive value was too low and the per-cancer signal was diluted. A pre-specified subgroup of 12 high-mortality cancers *did* show a stage shift. **A failed endpoint is not a failed technology — the trial asked the wrong question.** ## What Question Audit Does Input a research hypothesis in plain English. Question Audit turns it into a bounded **trial-question audit run**: `