A research protocol is the written plan for a study: why it is needed, what it will measure, how, in whom, and how the data will be analysed. For a clinical study, write it before any participant is approached, cover every section an ethics committee checks, and treat it as the document your final paper must match.
The protocol does three jobs. The ethics committee reads it to decide whether the study is justified and safe. The trial registry record is built from it. Months later, a journal reviewer compares your paper with it. A vague protocol causes trouble at all three points, and the last one is the hardest to repair.
This guide is for the hospital-based clinician writing a first protocol, and covers the sections, SPIRIT 2025 and amendments.
A research protocol fixes the key decisions before any data exist
The value of a research protocol is its timing. It records the primary outcome, the eligibility criteria and the analysis before anyone knows the results. A decision made after the data have been seen can look like choosing whatever worked.
For a clinical trial, that plan must also be public. Under the policy of the International Committee of Medical Journal Editors (ICMJE), a trial is registered at or before the time the first participant consents to enrolment, and the registry record repeats the protocol's key fields. Our guide to clinical trial registration and its deadline covers the registry side. For observational and confirmatory studies, preregistration and Registered Reports offer a similar public timestamp.
Use SPIRIT 2025 for trial protocols and its structure for other designs
SPIRIT is the reporting guideline for trial protocols. The current version is SPIRIT 2025, which replaced the 2013 edition and is written for randomised trials. It sets out what a protocol should contain, not how to design the trial. The checklist is linked from the EQUATOR Network.
Any study that prospectively assigns people to an intervention to evaluate a health outcome is a trial, whether or not it is randomised. A single-arm study of a new technique therefore needs registration, and the SPIRIT 2025 items still give its protocol a sound structure.
For a cohort, case-control or cross-sectional study, SPIRIT is not required, but its headings make a sound scaffold. STROBE, or RECORD for routinely collected data, will shape the final paper. A systematic review protocol follows PRISMA-P 2015 instead.
A section-by-section research protocol checklist
Use this table while drafting. If your committee has its own template, map these sections onto its headings rather than leaving any out.
| Protocol section | What to include | What the ethics committee and reviewers will look for |
|---|---|---|
| Title and identifiers | A title naming the design, population and setting; version number and date; the registry number once issued | A title that matches the design actually proposed; a version and date on every page |
| Background and rationale | What is known, the specific gap, and why your setting can answer it | Whether the question is already answered, and why participants should be asked to take part |
| Objectives | One primary objective, the secondary objectives, and a hypothesis if the study is confirmatory | Objectives that map one-to-one onto the outcomes |
| Study design | The named design (randomised trial, prospective cohort, retrospective chart review) and any allocation or blinding | Whether the design can support the claim the objectives make |
| Setting | Each site, the data collection period, and each site lead | Feasibility, and whether every site holds its own approval |
| Eligibility criteria | Inclusion and exclusion criteria defined operationally, and how eligibility will be confirmed | Exclusions without a reason, such as by age, sex or language, and any vulnerable groups |
| Interventions or exposures | What is done, by whom and when, and the comparator; for an observational study, how the exposure is defined and measured | What differs from routine care, and who delivers it |
| Outcomes | The primary outcome with its definition, measurement method, unit and time point; each secondary outcome defined the same way | A single named primary outcome with its timing stated |
| Sample size | The calculation with every assumption and its source, or a justification for a fixed sample | Sourced assumptions and an allowance for loss to follow-up |
| Recruitment | How participants are identified, who approaches them, and the plan if recruitment is slow | Pressure on patients who are approached by their own treating clinician |
| Data collection and management | The data collection form, who enters data, storage, access, retention and de-identification | Who can see identifiable data, and where it is kept |
| Statistical analysis plan | The primary and secondary analyses, missing data handling, pre-specified subgroup and sensitivity analyses, and software | Analyses fixed in advance, with effect sizes and confidence intervals planned |
| Ethics and consent | Who takes consent and how, the information sheet version, and the reasons for any request to waive consent | Consent documents that match the protocol, and a justified waiver request |
| Safety reporting | Definitions of adverse and serious adverse events, the grading system and version, and who reports what to whom, and when | Clear reporting lines, timelines and any stopping rules |
| Funding and interests | The funder, the sponsor's role, and declared interests | Whether the funder can influence the analysis or publication |
| Timeline | Start and end of recruitment, follow-up, analysis and reporting | A duration that fits the recruitment estimate |
| Dissemination | A commitment to report any result, authorship criteria and data sharing | A plan to publish negative or inconclusive results |
| Amendment history | A log of each version: number, date, what changed and why | That the version under review is the latest one |
Define each outcome so two people would measure it the same way
The outcome definitions are what a reviewer later tests your paper against, so vagueness here is costly. An outcome definition needs four parts: what is measured, how, by whom, and at what time point. If a complication is graded, name the system, such as Clavien–Dindo for surgical complications, and for the Common Terminology Criteria for Adverse Events (CTCAE) state the version you will use.
The outcome below is invented to illustrate the structure; it does not describe a real study.
Weak: Infection rates after surgery.
Specific: Primary outcome: surgical site infection within 30 days of the index operation, defined using [named definition, with reference], assessed by [role] at the [day] clinic review or by structured telephone call, and recorded as present or absent.
A reviewer can check your results table against the specific version.
Plan the sample size and statistical analysis in the protocol
The sample size is calculated prospectively from the primary outcome and the smallest clinically meaningful difference. Write every assumption into the protocol with its source. Our guide to sample size calculation for small clinical studies gives the inputs and a reporting template.
The statistical analysis plan names the test for the primary outcome, how missing data will be handled, and which subgroup or sensitivity analyses you intend. Plan to report effect sizes with confidence intervals. An analysis that appears in the paper but not in the protocol should be labelled exploratory.
Settle ethics, consent and site agreements before you submit
Ethics approval, consent to take part and consent for publication are three different things, and the protocol should say which apply. A retrospective study still needs approval or a documented waiver of consent. Cite the World Medical Association's Declaration of Helsinki as the ethical framework; its latest revision is October 2024. Our post on reporting ethics approval and informed consent shows the wording a manuscript needs.
If more than one hospital is involved, agree data ownership, data transfer and authorship before recruitment starts. The multi-centre study agreement guide lists what to settle, and the protocol should name each site.
Keep the protocol and the final paper consistent
Reviewers compare the paper with the protocol and the registry record. Check these points before you submit:
- The primary outcome in the paper has the same wording and time point as in the protocol.
- The target sample size and the number actually recruited are both reported.
- The analyses match the statistical analysis plan, and any extra analyses are labelled exploratory.
- Every deviation from the protocol is reported with its date and reason.
- The paper states the protocol version it followed and where a reader can see it: the registry, a supplementary file or a published protocol.
A difference you explain is a transparent change. A difference a reviewer discovers looks like outcome switching.
Manage amendments with version control
Protocols change during a study, and that is acceptable when the change is controlled. Version control means every copy traces to a numbered, dated version.
- Put the version number and date in the footer of every page.
- Record each change in the amendment log, with the reason.
- Submit substantial amendments to the ethics committee and wait for approval before applying them. Check your committee's own definition of substantial.
- Update the registry record for a trial, including any change to outcomes or sample size.
- Keep every superseded version, so you can show what was approved at each point.
- In the paper, report changes made after the study began, with dates.
If you spot an error in this guide, please tell us how to correct it.