How to run a pilot study that answers the right questions
A pilot study is not a small trial. Its job is to find out whether the full trial can work: can you recruit, will people stay, do the procedures and questionnaires work? Here is how to design one that gives a clear go or no-go.
- Feasibility objectives
- Progression criteria
- Data that decides
Free sandbox · No credit card · 21 CFR Part 11 aligned
Recruitment
2.1/mo/site
Retention
78%
Adherence
84%
Green
Amber: proceed with changes
Green
The short version
- Pilot and feasibility studies test whether a future trial can be done, not whether the treatment works. Do not power them for efficacy.
- Write feasibility objectives: recruitment rate, eligibility and consent rates, retention, adherence, data completeness, acceptability of procedures.
- Agree progression criteria in advance, often as green, amber and red thresholds that decide go, go with changes, or stop.
- Sample size is justified by the feasibility estimates you need; commonly cited rules of thumb range from around 12 per arm to 30 or more overall.
- Report using the CONSORT extension for pilot and feasibility trials.
Design
Feasibility objectives, not efficacy hypotheses
The most common mistake in pilot studies is treating them as underpowered efficacy trials and reporting p-values that mean nothing. A pilot's value comes from the numbers the full trial needs: how many eligible people there are, how many agree to take part, how many stay, how many complete the questionnaires, how long procedures take and whether participants find them acceptable.
Each of those becomes an objective with a pre-specified measure. "Recruitment rate" becomes "participants randomised per site per month". "Retention" becomes "proportion with the primary outcome at 12 weeks". With clear measures, the pilot ends with an answer instead of an impression.
Progression criteria
Progression criteria turn feasibility results into a decision. A common format is traffic lights: green means proceed as planned, amber means proceed with specific changes, red means do not proceed without major redesign. Agree them with funders and the steering committee before the pilot starts, and publish them in the protocol.
Sample size
Pilots are sized to estimate feasibility parameters (a recruitment rate, a retention proportion, a standard deviation for the future sample size calculation) with reasonable precision, not to detect an effect. Rules of thumb in the literature range from about 12 participants per arm to 30 or more in total, and some approaches size pilots from the planned main trial. Justify the number in the protocol. See sample size calculation.
Site risk
Site 03 needs attention
Action items
2 overdue
What to measure
Common feasibility measures and where the data comes from
| Objective | Measure | Data source |
|---|---|---|
| Recruitment | Participants randomised per site per month | Enrolment dates by site |
| Eligibility | Proportion of screened people eligible | Screening log and screen-failure reasons |
| Consent | Proportion of eligible people who consent | Screening and consent records |
| Retention | Proportion with primary outcome at follow-up | Visit completion |
| Adherence | Proportion receiving the intended dose | Adherence diary or accountability |
| Data completeness | Proportion of questionnaires completed | ePRO compliance |
| Acceptability | Participant and staff views | Short questionnaires, interviews |
In Capture
Feasibility data comes out of the study itself
Screening numbers and screen-failure reasons, consent dates, enrolment by site, visit completion and questionnaire compliance are all recorded as the study runs. The dashboard and status overview show them live, so the steering committee can see progression metrics without a separate tracking spreadsheet.
- Screening wizard with screen-failure reasons.
- Enrolment funnel and compliance on the dashboard.
- Exports by site and date for progression reports.
Set up a pilot with the main trial's forms
Build once in the free sandbox; reuse the configuration for the full trial.
Design choice
Internal or external pilot?
An external pilot is a separate study. Its data is used only to decide whether and how to run the main trial, and it is analysed and reported on its own. An internal pilot is the first phase of the main trial: if progression criteria are met, the trial simply continues, and the pilot participants are included in the final analysis.
Internal pilots save time and participants, but the protocol, forms and procedures must already be final, because changing them after the pilot would mix two versions in one analysis. External pilots allow bigger changes, at the cost of a gap between studies. Choose based on how confident you are in the procedures: if the questionnaires, visit schedule or intervention delivery might change, run an external pilot.
Before the pilot starts
Pilot study checklist
Feasibility objectives
Each with a measure and a data source.
Progression criteria
Green, amber and red thresholds agreed and in the protocol.
Sample size justified
Based on the precision needed for feasibility estimates.
Screening log
Screened, eligible, consented and reasons for non-participation.
Acceptability data
Short questionnaires or interviews planned.
Reporting plan
CONSORT extension for pilot and feasibility trials.
What is the difference between a pilot and a feasibility study?
Terminology varies. Feasibility studies ask whether a trial can be done; pilot studies are a type of feasibility study that runs a smaller version of the future trial.
Should a pilot study test efficacy?
No. Pilots are not powered for efficacy; they estimate the parameters needed to plan the main trial.
How many participants does a pilot need?
Enough to estimate the feasibility measures with useful precision. Rules of thumb range from about 12 per arm to 30 or more overall.
What are progression criteria?
Pre-agreed thresholds, often green, amber and red, that decide whether and how to proceed to the main trial.
Can the pilot configuration be reused?
Yes. Capture study configurations can be exported and imported into a new study.
Keep exploring
Related pages
Pilot with the system you will scale on
Build once, reuse for the main trial. Free sandbox.