Clinical Trial Designs · Cluster-Randomized Trials

Cluster-Randomized Non-Inferiority Trial — Continuous Outcome

Plan a two-arm parallel cluster-randomized trial in which a continuous outcome must rule out a prespecified unacceptable loss relative to the reference treatment.

How this calculator works

The treatment contrast is always experimental minus reference. You enter a positive maximum acceptable loss and identify whether higher or lower outcome values are favorable. The calculator derives the signed non-inferiority boundary and evaluates candidate cluster allocations using a cluster-level one-sided noncentral-t approximation.

The calculator does not choose or justify the margin. Margin selection, assay sensitivity, constancy, analysis population, and sensitivity analyses require protocol-specific clinical and statistical justification.

1. Outcome and non-inferiority hypothesis

Primary outcome label

Optional reporting label.

Outcome unit

Optional. The expected difference, margin, and SD must use this same scale.

Favorable outcome direction

Higher Lower

This orients the hypothesis without changing the experimental-minus-reference subtraction order.

Expected mean difference: experimental − reference

Enter a signed value. Do not reverse the subtraction order when lower values are favorable.

Maximum acceptable loss relative to reference

Enter a positive magnitude in the original outcome unit—not a signed boundary.

2. Outcome and clustering assumptions

Expected common individual-level outcome SD

Use evidence from a comparable outcome, population, and measurement schedule.

Intracluster correlation coefficient

Use a credible outcome- and setting-specific estimate.

Analyzable mean participants per retained cluster

This is the expected analyzable size, before recruitment inflation is reported.

Cluster-size assumption

Equal Variable

Variable mode applies a common coefficient-of-variation design-effect approximation.

3. Operating characteristics and recruitment

One-sided Type I error

A 2.5% one-sided alpha corresponds to a 95% two-sided confidence interval.

Target power

Experimental:reference cluster ratio

The realized ratio may differ slightly after upward integer rounding.

Participant attrition

Inflates recruited participants per cluster.

Whole-cluster attrition

Inflates randomized clusters separately by arm.

4. Results

Review the signed hypotheses, enter the assumptions, and select Calculate.

Analyzable design and recruitment target

Arm	Analyzable clusters	Analyzable participants	Recruit clusters	Recruit participants

ICC sensitivity analysis

ICC	Design effect	Experimental clusters	Reference clusters	Total clusters

Choose the planning ICC from credible evidence, not because it produces a convenient design.

Suggested sample-size wording

Formula, assumptions, limitations, and methodological note

Example clinical scenario

Primary-care clinics are randomized to an experimental care programme or reference care. The trial is designed to rule out a clinically unacceptable loss in mean patient-reported health score at six months.

Formula and methodological note

Let δ be experimental minus reference, let d be +1 when higher values are favorable and −1 when lower values are favorable, and let M be the positive maximum acceptable loss. The signed null boundary is b = −dM and the favorable gap is g = d(δ₀ − b). The design requires g > 0.

For equal cluster sizes, DE = 1 + (m − 1)ρ. Variable-size mode uses DE = 1 + ([(1 + CV²)m] − 1)ρ. For candidate cluster counts kE and kR, SE = σ√[(DE/m)(1/kE + 1/kR)], df = kE + kR − 2, and noncentrality = g/SE. Power is the upper tail beyond t_1−α,df.

This is a cluster-level t planning approximation. It does not represent every mixed-model, GEE, matched, stratified, stepped-wedge, multilevel, repeated-outcome, or informative-cluster-size design.

Technical results

Method references

Piaggio G, et al. Reporting of noninferiority and equivalence randomized trials: extension of the CONSORT 2010 statement. JAMA. 2012;308:2594–2604.
Campbell MK, et al. CONSORT 2010 statement: extension to cluster randomised trials. BMJ. 2012;345:e5661.
Eldridge SM, Ashby D, Kerry S. Sample size for cluster randomized trials: effect of coefficient of variation of cluster size and analysis method. Int J Epidemiol. 2006;35:1292–1300.

Privacy: Calculations run locally in your browser. Values entered are not uploaded by this calculator.

Planning disclaimer: This tool supports research planning and does not replace protocol-specific statistical consultation.

Suggested citation

Yaacob, Najib Majdi. “Cluster-Randomized Non-Inferiority Trial — Continuous Outcome.” Web calculator. NajibMY, 2026. Accessed August 19, 2026. http://127.0.0.1:4321/tools/sample-size/clinical-complex-study-designs/cluster-randomized-non-inferiority-continuous-outcome/.