Direct comparison
Non-Inferiority vs. Superiority Trials
Non-inferiority vs. superiority trial design compared: research question, margin-setting, analysis population, and FDA/ICH E9 guidance for each.
Side-by-side comparison
| Dimension | Non-Inferiority | Superiority |
|---|---|---|
| Research question | Is the new treatment not unacceptably worse than the active control? | Does the new treatment work better than the comparator? |
| Null hypothesis | New treatment is worse than control by at least the margin (M) | No difference between treatments |
| Typical comparator | Active control (established, effective treatment) — placebo generally not usable | Active control or placebo |
| Key design element | Pre-specified non-inferiority margin, justified against the control’s historical effect vs. placebo | Minimum clinically meaningful effect size used to power the trial |
| Can it show the new treatment is better? | Not on its own — only if a pre-specified superiority test is also run and passes | Yes, that is the primary claim being tested |
| Primary analysis population | ITT/Full Analysis Set AND per-protocol as co-primary (both must agree) | ITT/Full Analysis Set (conservative, standard primary analysis) |
| Effect of poor trial conduct/adherence | Dangerous — dilution toward the null can falsely suggest non-inferiority | Conservative — dilution toward the null makes a true effect harder to detect |
| Typical use case | Cheaper, safer, or more convenient alternative to an existing standard of care (e.g., oral vs. infused, shorter course) | Novel mechanism, first-in-class therapy, or any case where the value proposition is efficacy itself |
| Key US guidance | FDA, "Non-Inferiority Clinical Trials to Establish Effectiveness" (finalized November 2016) | General ICH E9/E8(R1) confirmatory-trial framework; no dedicated superiority-specific FDA guidance document |
| Reporting standard | CONSORT extension for non-inferiority/equivalence trials (Piaggio et al., JAMA, 2012) requires the margin and its justification be stated explicitly | Standard CONSORT 2010 reporting |
Common questions
FAQ
Can a trial test for both non-inferiority and superiority?+
Yes, if pre-specified. A common sequential design tests non-inferiority first; if that margin is met, the same data can then be tested for superiority, since superiority is a stronger claim that implies non-inferiority. Testing in reverse order after a failed superiority result is a known source of bias and is not considered a valid non-inferiority claim unless the margin and hierarchy were specified in the statistical analysis plan before unblinding.
Why can’t non-inferiority trials generally use a placebo control?+
Because a non-inferiority claim depends on the active control having a known, reliable effect over placebo to preserve a margin against — without that established effect, showing "non-inferior to the control" provides no evidence the new treatment works at all. Placebo-controlled non-inferiority designs exist in narrow circumstances but are the exception.
Is a "negative" superiority trial the same as a non-inferiority finding?+
No. Failing to show superiority only means no difference was detected — it does not mean similarity was demonstrated within any pre-specified acceptable margin, because no such margin was defined or powered for in a superiority design.
Who sets the non-inferiority margin?+
The sponsor proposes it based on the control’s historical treatment effect and clinical judgment about how much of that effect can acceptably be lost, but it is typically negotiated and agreed with the regulator (e.g., an FDA meeting) before the pivotal trial begins.
Going deeper
