Non-inferiority describes a clinical trial strategy in which the objective is not to demonstrate superiority over a comparator treatment, but to show that a new treatment is not meaningfully worse in terms of efficacy. The key difference from a superiority trial lies in the research question: instead of asking whether the new therapy works better, the question is whether it falls within a predefined acceptable tolerance limit. This limit is referred to as the non-inferiority margin and is the most methodologically critical component of a non-inferiority trial. The margin must be defined before the study starts and is based on clinical considerations as well as historical data on the efficacy of the comparator therapy. Without a clearly defined and well-justified margin, the evidentiary value of a non-inferiority trial is not scientifically defensible.
When is non-inferiority used?
Non-inferiority trials are used when a new treatment is not expected to be more effective than the established standard therapy, but may offer advantages in other attributes. These include a more favorable safety profile, an easier route of administration (for example, oral instead of intravenous), lower costs, or improved patient compliance. In these situations, a classic superiority trial would be neither ethically nor scientifically meaningful, because the new therapy is never intended to be more effective.
Typical areas of application include antibiotic trials, anticoagulants, immunotherapies in oncology, and vaccine trials. For vaccines, for example, the immunogenicity of the new vaccine is compared with that of an approved reference vaccine, and demonstrating non-inferiority is considered sufficient evidence for approval. Non-inferiority trials are also well established in dermatology, cardiology, and the treatment of infectious diseases. In every case, it is essential that the new product offers clear added value that justifies not demonstrating higher efficacy.
The non-inferiority margin
The non-inferiority margin (also referred to as delta, Δ) is the maximum tolerable difference between the new treatment and the standard therapy that is still considered clinically acceptable. Defining this margin is both regulatorily and methodologically demanding: on the one hand, it must be based on the historically demonstrated effect of the comparator versus placebo (the so-called M1 step), and on the other hand, it must be chosen so that a clinically meaningful proportion of this effect is preserved (M2 step).
Both the EMA and the FDA have published guidelines on defining the non-inferiority margin. The EMA guideline CPMP/EWP/2158/99 and the FDA guidance “Non-Inferiority Clinical Trials to Establish Effectiveness” (2016) emphasize that the choice of margin must be scientifically justified and documented in the study protocol. A margin that is set too wide can result in an ineffective therapy being classified as non-inferior—an issue referred to as “bio-creep.”
Statistical analysis
The statistical analysis of a non-inferiority trial is generally based on confidence intervals. A one-sided test or a two-sided 95% confidence interval is calculated, and non-inferiority is considered demonstrated if the lower bound of the confidence interval lies above the negative non-inferiority margin. In contrast to superiority trials, the ITT (intent-to-treat) analysis is not necessarily the most conservative analysis in non-inferiority trials: per-protocol analyses are also of central importance, and both results must be consistent.
A common interpretive error is confusing statistical non-inferiority with clinical equivalence. Non-inferiority merely means that the lower bound of the confidence interval does not cross the margin—not that both treatments are identically effective. Regulators and reviewers therefore scrutinize the biological plausibility of the results carefully. In addition, study sensitivity must be demonstrated, meaning it must be shown that the study would have been able to detect a relevant difference if one had existed. This requirement is referred to as assay sensitivity and is a key criterion in regulatory review.
Regulatory Requirements
Regulatory authorities such as the EMA and FDA require non-inferiority trials to be planned prospectively. The margin, the statistical method, and the primary endpoints must be defined in the protocol and the statistical analysis plan before the study begins. Retrospective adjustments to the margin are considered a serious protocol deviation and may lead to rejection of the marketing authorization application.
As part of a Marketing Authorisation Application (MAA) procedure, the methodological quality of the non-inferiority trial is reviewed in depth by the authority’s scientific advisors. Among other aspects, the authority assesses whether the margin was chosen conservatively enough, whether the study population is representative, and whether the results are internally consistent. Full-service CROs such as mediconomics support sponsors in study design, defining the margin as part of the scientific advice process, and the statistical planning and analysis of non-inferiority trials across all therapeutic areas.
A practically relevant aspect is also the choice of comparator in non-inferiority trials. Ideally, the active comparator therapy should have demonstrated and reproducible efficacy versus placebo in the population studied and under the trial conditions. If this is not the case—for example, because the comparator’s efficacy has only been shown in historical studies with different designs—it becomes difficult to derive a valid margin. Regulators review this aspect particularly critically, and it should be clarified early in the scientific advice process.
Frequently Asked Questions (FAQ)
Can superiority be derived retrospectively from a non-inferiority trial?
Yes, if a hierarchical testing procedure is specified in the protocol from the outset: non-inferiority is tested first, and if it is demonstrated, superiority can be tested in a second step. However, this approach must be planned prospectively and documented in the statistical analysis plan.
What is the difference between non-inferiority and equivalence?
An equivalence trial assesses whether two treatments fall within a tolerance margin in both directions (neither superior nor inferior). A non-inferiority trial assesses only whether the new therapy is not meaningfully worse. Equivalence trials are used primarily in biosimilar development, whereas non-inferiority trials are used for efficacy-based approvals.