Ask Mallard

Noninferiority and equivalence trials

A noninferiority trial asks whether a new treatment is not worse than an existing one by more than a prespecified margin, rather than asking whether it is better.
NHMRC EVIDENCE LEVEL · 2009Level II · InterventionProvisional design classification. Study quality and certainty of findings require separate appraisal.
Why this level?Participants or clusters are randomly allocated to concurrent comparison groups. NHMRC 2009 source

Allocation is randomised, so the design sits where any trial sits. The margin, not the level, is what this design lives or dies on.

What question does this help answer?

It suits a new treatment expected to be about as effective as the standard while being cheaper, safer, simpler or more acceptable. The clinical claim is that giving up a defined and tolerable amount of efficacy buys something worth having.

When to use it

Use it when an effective standard treatment exists, so a placebo comparison would be unethical, and when the new option has a non-efficacy advantage that justifies accepting slightly less benefit. An equivalence design is the two-sided version, used when being better would also matter.

When not to use it

Do not use it when a superiority question is the honest one, and never adopt a noninferiority framing after a superiority trial failed: switching afterwards converts a disappointing result into an apparently positive one without new evidence.

How it works in plain language

A margin is fixed in advance as the largest loss of effect that would still be acceptable. The trial then asks whether the confidence interval for the difference excludes that margin. The margin does all the work, and it is a clinical and regulatory judgement rather than a statistical one, so it must be justified before the data exist.

Intervals read against the margin

Five confidence intervals for the difference between a new and a standard treatment, with the favourable direction to the right, placed against a noninferiority margin, no difference, and an upper equivalence margin: superior, noninferior, equivalent, inconclusive, and inferior.Noninferiority and equivalence are read against the margin, not against no differenceDifference: new minus standard, favouring the new treatment to the rightNoninferiority marginNo differenceUpper equivalencemarginSuperiorNoninferiorEquivalentInconclusiveInferiorNoninferior: the whole interval lies on the favourable side of the margin. Equivalent: it lies inside bothmargins. The margin is a clinical judgement fixed in the protocol.
Conceptual diagram. It shows the structure of the idea, not data from any study.

Why Mallard may recommend it

Mallard may recommend a noninferiority framing when the stated goal is to show a new option is acceptable rather than superior, and when an active comparator with established benefit is available.

Key assumptions and design requirements

  • The margin is prespecified and clinically justified, not chosen to make the sample size affordable.
  • The active comparator has demonstrated benefit in a setting similar to this trial (the constancy assumption).
  • Both intention-to-treat and per-protocol analyses are planned, because neither alone is conservative here.
  • Adherence and follow-up are protected: sloppy conduct biases a noninferiority trial TOWARD its conclusion.
  • The conclusion is stated as noninferiority within the margin, never as equivalence or as evidence of no difference.

What to watch out for

The margin does the work and is a clinical judgement, not a statistical one, so a margin chosen to make the trial feasible has assumed the answer. Sloppy conduct biases a noninferiority trial TOWARD its conclusion, which is the opposite of a superiority trial, so poor adherence and loss to follow-up make inferiority easier to miss.

Clinical example

Question: Is a once-daily oral regimen no worse than the standard twice-daily one for treatment success? Design: Randomize to the two regimens, prespecify a margin of five percentage points in success as the largest acceptable loss, and conclude noninferiority only if the confidence interval for the difference lies entirely above that margin.

Code for this method
Loading…
Recent studies that used this
Searching PubMed…
Related Learn pages
Sources & further reading
  1. Reporting of noninferiority and equivalence randomized trials: extension of the CONSORT 2010 statement (opens in a new tab) · JAMA · doi 10.1001/jama.2012.87802
  2. Overview of clinical study designs (opens in a new tab) · Clin Exp Emerg Med · doi 10.15441/ceem.23.036
The CONSORT extension covers reporting and explains why the margin must be justified. It does not choose the margin for you, and no statistical method can: that is a clinical judgement.
Last source review: 2026-09-06
AI-generated educational content
Mallard uses reputable methodological and reporting sources, but this material may contain errors or omit important context. Verify important methodological decisions with the cited sources and, when appropriate, a qualified research methodologist or statistician.