Skip to content
Scintria
ToolsGrant writerFunding
HomeLiteratureClinical trialsEvidence LabProtocol packData analysisLearnProjectsLibraryReferencesApplications
Start free
ScintriaStart free

Learn / Clinical Research Management / Data management and trial statistics

Sample size, interim analysis and reporting results

13 minute read · 6-question quiz

Download this lesson as a colour Word or PDF study guide with the figures and answers. Included in paid plans; reading and quizzes stay free.Sign in

What you will learn

  • Explain the ingredients of a sample size calculation

  • Calculate a simple sample size for comparing two proportions

  • Describe interim analyses, multiplicity and the clinical study report

A trial that is too small cannot detect a real effect; one that is too large wastes resources and exposes people unnecessarily. The sample size calculation balances these.

Ingredients

  • Primary endpoint and its type (proportion, mean, time to event).
  • Expected effect size: the smallest difference worth detecting, and the expected control group rate.
  • Significance level (alpha): the chance of a false positive, usually 0.05 (two-sided).
  • Power (1 − beta): the chance of detecting a true effect, usually 80% or 90%.
  • Variability for continuous outcomes.
  • Allowance for dropouts and non-adherence.

Less variability, a larger effect or a lower power requirement all reduce the sample size. A smaller effect, higher power or stricter alpha increase it.

Sample size per group for two proportions

n = ((z_(α/2) + z_(β))^2 [p_1(1-p_1) + p_2(1-p_2)])/((p_1 - p_2)^2)

For two-sided alpha 0.05 and 80% power, z_(α/2) = 1.96 and z_(β) = 0.84, so (z_(α/2)+z_(β))^2 ≈ 7.85. For 90% power, z_(β) = 1.28 and the factor is about 10.5.

Worked example: malaria cure rates

The standard treatment cures 80% (p1 = 0.80). A new regimen is expected to cure 90% (p2 = 0.90). Alpha 0.05 two-sided, power 80%.

  • Variance part = 0.80 × 0.20 + 0.90 × 0.10 = 0.16 + 0.09 = 0.25.
  • Numerator = 7.85 × 0.25 = 1.96.
  • Denominator = (0.10)² = 0.01.
  • n per group = 1.96 / 0.01 = 196.

Adding 10% drop-out: 196 / 0.90 = 218 per group, or 436 in total. If the expected difference falls to 5 points (80% vs 85%), the variance part is 0.16 + 0.1275 = 0.2875, the numerator is 2.26, the denominator is 0.0025, and n is 903 per group. A smaller effect needs far more participants.

02505007501000Participants per group (before dropout)19610 point difference9035 point difference
Halving the expected difference increases the sample size by more than four times.

Interim analyses and multiplicity

  • Interim analyses look at accumulating data before the end, usually by an independent DSMB, to stop for harm, overwhelming benefit or futility. Each look increases the chance of a false positive, so protocols use stopping boundaries (such as O'Brien-Fleming) that demand stronger evidence early.
  • Multiplicity: testing many endpoints, subgroups or time points raises the chance of a false positive. Plans specify a primary endpoint, a hierarchy of secondary endpoints, or adjustments for multiple tests.

The clinical study report

After analysis, the sponsor writes a clinical study report (CSR) following ICH E3: methods, populations, results (efficacy and safety), deviations, and conclusions. Regulators rely on it. Results are also registered and published (covered in the closing chapter).

Sample size is a trade-off made visible

It makes explicit what difference the trial can reliably find. Be honest about the effect size: optimistic assumptions lead to underpowered trials.

Underpowered trials waste participants

A trial too small to answer its question exposes people to risk without a reasonable chance of useful results. Ethics committees may reject it.

Check yourself

For a trial you know, find the sample size section. Which effect size and power were used, and how realistic is the effect size?

Summary

  • Sample size depends on effect size, alpha, power, variability and dropout allowance.

  • Smaller expected effects need much larger trials.

  • Interim analyses use stopping boundaries; multiplicity inflates false positives unless controlled.

  • The CSR (ICH E3) reports methods, results and conclusions.

Check your understanding

Question 1
1

Power is the probability of:

Question 2
2

If the expected difference between groups is halved, the sample size:

Question 3
3

196 per group needed; 10% drop-out. The inflated per-group number is about:

Question 4
4

The usual two-sided significance level is:

Question 5
5

Multiplicity refers to:

Question 6
6

The CSR follows which ICH guideline?

Draft study guide. Spotted an error or something unclear? Tell us at hello@scintria.com. This is educational material, not clinical or regulatory advice.

Randomisation, blinding and the statistical analysis planProject planning: scope, timelines and the critical path