Pillar 2 of 6 · Design
Is the trial designed to win?
Your design against the indication norm — and against every comparable trial that failed or succeeded before you, with the design features that separated them.
Is the trial designed to win?
What it answers
Is the trial designed to win?
A protocol does not fail on a spreadsheet. It fails because the endpoint was wrong for the phase, the enrolment could not detect the effect, the eligibility drifted mid-study, or twenty other sponsors were already competing for the same patients. This pillar puts your design next to the ones that already ran.
Who else is running this?
The interventional trials competing with yours for the same patients, in the same condition — where they are, what stage they are at, and who is sponsoring them.
What killed the comparable trials?
Not a narrative. The design features that actually separate the failed cohort from the succeeded one in the posted results — and where your design sits against that split.
Is the design coherent for the indication?
Where your endpoints, arms and enrolment sit inside the empirical distribution of sibling trials — as a percentile, not a hand-wave against a median — and whether the planned enrolment can detect a meaningful effect at all.
Inside the pillar
6 analyses, each one named and cited
These are the analyses that actually run — not a category list. Each states what it computes and what it needs, because an analysis that quietly degrades is worse than one that reports it could not run.
Competitive landscape
Fans out across ClinicalTrials.gov and the WHO ICTRP registry network for interventional trials in the same condition and reduces them to who is running what, where, at what stage, and at what recruiting status — the trials competing with yours for the same patients.
Needs
An indication, and a phase when you have one. Your own sponsor pipeline can be overlaid on the public set, but the public number stays the primary figure.
Why comparable trials failed
Reads the posted results of those competing trials, classifies each as success, failure or excluded, then compares the failed cohort against the succeeded one across masking, allocation, phase, sample size, endpoint significance and serious-adverse-event burden — and puts your own design against the pattern as quantified lessons.
Needs
Competitor trials that have posted a results section. When competitors exist but none carry results yet, that state is reported distinctly from a genuine finding of no failures — the two are not allowed to look alike.
Design vs. indication norm
Pulls your primary and secondary endpoints, arm structure and stratification factors from the registry record and fits the empirical distribution of sibling trials in the same indication and phase, placing this trial inside it per axis — a percentile position for enrolment, a categorical share for arm count and primary endpoint.
Needs
A registered design or a protocol synopsis. Multiplicity hierarchies and alpha-spending functions never appear in a public registry record, so those are read from your own uploaded statistical plan when you provide one.
Statistical power
A deterministic two-sample adequacy screen: from planned enrolment and arm count it computes the minimum detectable effect and the power achieved against an indication-typical standardized effect-size prior, at a two-sided alpha.
Needs
Planned enrolment and the number of arms. The effect size is a prior drawn from the indication, not your declared effect — this is a directional adequacy screen, never a substitute for the statistical analysis plan's own power calculation.
Regulatory-precedent timeline
Lays out prior approvals and assessments for the molecule and its class on a per-regulator swim-lane — FDA approvals and review documents, EMA European public assessment reports, NICE health-technology appraisals and Health Canada records — so precedent is a timeline rather than an anecdote.
Needs
A drug, or a therapeutic class when the asset has no public name. With no molecule named, the lane runs at class level and labels itself that way.
Protocol-amendment risk
Reads the registry's own version history to time-stamp every change to eligibility, endpoints, sample size and arms, and scores the resulting hazard. Optionally mines a bounded sample of sibling trials' version histories to derive a per-field empirical amendment base rate and a predicted amendment probability for this trial.
Needs
A registered trial with a version history, or siblings in the same indication and phase to derive the base rate from.
Where the evidence comes from
Public sources, named
Registry coverage is the constraint here, not our reach. Trials registered outside the major networks, and trials that never posted results, are invisible to any analysis built on public records — including this one. The assessment reports how much of the comparable set it could actually read.
ClinicalTrials.gov
U.S. National Library of Medicine · Daily
Registry of public and private clinical studies worldwide — enrollment, status, sponsors, sites, and outcomes.
clinicaltrials.gov (opens in a new tab)AACT
Clinical Trials Transformation Initiative (CTTI) · Daily
Relational, queryable database of all ClinicalTrials.gov content — the analytical backbone for site and enrollment statistics.
aact.ctti-clinicaltrials.org (opens in a new tab)PubMed
NCBI / U.S. National Library of Medicine · Daily
Biomedical literature index of 36M+ citations — used to ground indication baselines and standard-of-care evidence.
pubmed.ncbi.nlm.nih.gov (opens in a new tab)FDA
U.S. Food & Drug Administration · Weekly
Approvals, labels, guidance, and adverse-event signals (openFDA) used to anchor regulatory-uplift and safety risk.
open.fda.gov (opens in a new tab)EMA
European Medicines Agency · Weekly
European approvals, EPARs, and scientific guidance used to model EU regulatory pathways and country benchmarks.
ema.europa.eu (opens in a new tab)Also read by this pillar
What you get
What lands in the assessment
Every competing trial links to its registry record; every failure classification links to the posted results it was read from; every amendment links to the registry version that made it.
- A Design tab with the competing-trial set, the failure-pattern comparison, your position in the sibling distribution, and the amendment ribbon
- A power screen stating the minimum detectable effect and the prior it was computed against
- A per-regulator precedent timeline for the molecule or its class
Read by these expert roles
A role drafts from this pillar; a human reviews and signs off. The sign-off is a 21 CFR Part 11 e-signature on a hash-chained record.
How the roles workHonest limits
What this pillar cannot tell you
Every assessment method has a boundary. Publishing ours is the point — a number you cannot check is worth less than a gap you can see.
It reads the design, not the science of your endpoint.
Norms and percentile positions come from what comparable trials actually did. A genuinely novel endpoint has no distribution to sit inside, and the tile says so rather than manufacturing one.
Power here is a screen, not a statistical analysis plan.
The effect size is an indication-typical prior. Dropout, multiplicity, adaptive rules, interim analyses and the real variance of your endpoint are yours to model — this only tells you whether the planned enrolment is in the right order of magnitude.
Stopped is not the same as failed.
Registries record what a sponsor chose to post. A trial halted for business reasons and one halted for futility can look identical in the record. The classifier abstains when the stated reason is ambiguous rather than scoring a censored trial as a failure.
Results coverage is uneven.
A large share of registered trials never post results at all. The failure-pattern analysis can only compare the trials that did, and it reports the size of the set it worked from.
The rest of the assessment
No pillar decides alone
Each pillar answers one question. The Verdict composes all six into a single go/no-go, with the risks that could still change it.
See a verdict you can actually check.
Send us a protocol — or just a molecule and an indication. We'll return a fully cited feasibility assessment you can trace, line by line, back to public data — yours to defend in a bid, take to your board or investment committee, or hand to a regulator.