For investigators, sponsors and CROs before the protocol is written
Trial Protocol DeskThe sample size worked out, the gaps found, then a critique or a draft.
Paste a study synopsis. Free in your browser: the sample size - t-test or two-proportion, superiority or non-inferiority, allocation, dropout, sensitivity - and the power your planned enrolment really buys; the synopsis checked for an endpoint whose timepoint is a visit, a comparator, blinding, a margin, measurable criteria, safety oversight and an analysis plan; the Schedule of Activities drawn. A run critiques the design or drafts the core protocol sections, held to the page's numbers.
1 Your synopsis free
2 The numbers and the check free
Nothing leaves your browser for this part. The sample size is preliminary - a biostatistician signs it off.
The sample size
Where the gaps are
Schedule of Activities
Findings
3 Critique the design, or draft the protocol metered
The reply, as it came
A drafting aid for qualified trial professionals, not medical, statistical or regulatory advice. Do not use it for treatment or enrolment decisions; a biostatistician, the sponsor, the ethics committee and the regulator review the real protocol.
Design review
Sample size
Questions for the team
The draft
Open questions
The draft as text
Findings
The page's calls
Caveats
Checked against the page
Saved studies
How it works
The sample size follows the calculator that ships with the clinical-trial-protocol skill: a two-sample t-test (normal approximation) for a continuous endpoint, a two-proportion z-test for a binary one, two-sided alpha for superiority and one-sided for non-inferiority (where the margin replaces the effect), enrolment inflated for dropout arm by arm (so the total is always the sum of the arms - the upstream script inflates the total separately, which can come out one lower), and sensitivity at 90% power and at a 10% smaller effect. For unequal allocation the page uses the standard arm-size formula (the upstream script's unequal-allocation branch does not reduce to its own 1:1 formula). The check reads the synopsis against the protocol template the skill follows: one primary endpoint with a timepoint that is a scheduled visit, a comparator, randomisation and blinding, measurable criteria, a visit schedule, safety monitoring and a data monitoring committee where the risk calls for one, an analysis method and population.
The critique and the draft are written by a language model from your synopsis and the page's reading. The page holds the reply to it: every error and warning answered by rule id; every sample-size number one of the page's; no invented trial registrations, guidance titles or publications; the primary timepoint named in every section that needs it. A drafting aid for qualified people - the sample size needs a biostatistician, the protocol needs the sponsor, the ethics committee and the regulator.