Clinical research & EBM · Reference

What is a non-inferiority trial?

A non-inferiority trial is designed to show that a new treatment is not unacceptably worse than an established comparator, by a pre-specified margin. Unlike a superiority trial, it does not aim to prove the new treatment is better — only that it is not meaningfully worse.

Non-inferiority versus superiority

A conventional randomized controlled trial usually tests for superiority — whether a new intervention produces a better outcome than its comparator. A non-inferiority trial reframes the question: it sets out to show the new intervention is not meaningfully worse. This is the right design when a new option is unlikely to be more effective but is attractive for other reasons — fewer side effects, lower cost, simpler administration — so the aim is to confirm it does not sacrifice too much efficacy in exchange for those advantages.

The non-inferiority margin

The crux of the design is the non-inferiority margin (often written Δ): the largest reduction in effect, fixed in advance, that would still be regarded as acceptable. The trial concludes non-inferiority if the confidence interval for the difference between treatments lies entirely on the acceptable side of this margin. Choosing the margin is a careful judgement, informed by how much of the comparator’s established benefit must be preserved; setting it too wide can declare an inferior treatment "non-inferior". Because the margin is defined before the data are seen, it must be justified in the protocol.

Why analysis and rigour differ

Non-inferiority trials demand particular care because sloppiness biases the result towards the conclusion sought. In a superiority trial, anything that blurs the difference between groups makes a true effect harder to detect; in a non-inferiority trial, the same blurring can make two genuinely different treatments look similar, falsely supporting non-inferiority. For this reason both an intention-to-treat and a per-protocol analysis are usually reported and expected to agree, and reporting follows an extension of the CONSORT statement specific to non-inferiority and equivalence trials.

Key facts

At a glance

Goal: Show new treatment is not unacceptably worse
Contrast: Differs from superiority (proving better)
Key value: Pre-specified non-inferiority margin (Δ)
Conclusion: Confidence interval stays on acceptable side
Analysis: ITT and per-protocol both reported
Reported via: CONSORT non-inferiority extension

Common questions

FAQ

How is a non-inferiority trial different from a superiority trial?+

A superiority trial tries to show a new treatment is better than its comparator. A non-inferiority trial only tries to show it is not worse by more than a pre-specified margin, which suits a new option valued for advantages such as safety, cost or convenience rather than greater efficacy.

What is the non-inferiority margin?+

The non-inferiority margin is the largest acceptable reduction in effect, set in advance, below which the new treatment would be judged unacceptably worse. Non-inferiority is concluded when the confidence interval for the treatment difference lies entirely on the acceptable side of this margin.

Why do non-inferiority trials report both ITT and per-protocol analyses?+

Because anything that blurs the difference between groups can falsely support non-inferiority, these trials are held to extra rigour. Reporting both intention-to-treat and per-protocol analyses, and expecting them to agree, guards against a conclusion of non-inferiority that arises from loose conduct rather than genuine similarity.

Going deeper

Related on CASRAI

Sources

The step most authors miss

Doing CRediT right? Don’t stop at the statement.

A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.

Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.