How to justify your sample size in a protocol or ethics application
Last reviewed 16 October 2026
Ethics committees and funders ask the same question: is the study large enough to answer its question, and no larger than necessary? A short, specific paragraph answers it.
Open the sample size calculatorWhat the paragraph must contain
The primary outcome and the comparison it supports, the smallest difference that matters and why, the variability or baseline rate and its source, the significance level and power, the formula or software, the adjustments (dropout, clustering) and the final number.
- Primary outcome and comparison
- Smallest important difference, with a reason
- Baseline rate or standard deviation, with a source
- Significance level and power
- Formula or software used
- Adjustments for dropout and clustering
- Final number per group and in total
Model wording
"The primary outcome is infection at 12 months. Assuming 30% infection with standard care (source), the trial is designed to detect an absolute reduction to 20%, which clinicians in this setting consider the smallest worthwhile effect. With a two-sided significance level of 0.05 and 80% power, 294 participants per arm are needed. Allowing for 10% loss to follow-up, 327 per arm (654 in total) will be enrolled."
Mistakes that get applications sent back
Using the difference you hope for instead of the smallest that matters. Giving no source for the baseline rate or the standard deviation. Adding a percentage for dropout instead of dividing. Forgetting clustering in a cluster design. Calculating power for a secondary outcome and presenting it as the primary.
Show your sensitivity
State how the number changes if your assumptions are wrong, for example a smaller effect or a larger standard deviation. It shows the committee that the study is robust to reasonable error.
Common questions
- Do I need a statistician to sign off?
- Many committees expect the calculation to be reviewed by someone with statistical training, especially for trials. A transparent calculation with sources makes that review quick.
- What if my sample is limited by feasibility?
- Say so, and report what the study can detect with the sample you can recruit. Honest limits are better than a calculation reverse-engineered to a convenient number.
- Which software should I cite?
- Name the formula and the tool, and state the inputs. Scintria's calculators show every step and are checked in public against SciPy, statsmodels and R.
Keep reading
General methods guidance, not a substitute for advice from a statistician on your specific study.