EMA’s New Draft Guideline: Non-Inferiority and Equivalence Comparisons in Clinical Trials

The European Medicines Agency (EMA) has released a major draft guideline "Non-Inferiority and Equivalence Comparisons in Clinical Trials" that updates how non-inferiority and equivalence comparisons should be designed, justified, and analysed in confirmatory clinical trials. This draft replaces two earlier documents:

• Guideline on the choice of the non-inferiority margin (2005), and

• Points to consider on switching between superiority and non-inferiority (2000).

The new guideline reflects advancements in regulatory science, introduces the estimand framework (ICH E9 R1), and strengthens expectations around assay sensitivity, margin justification, and statistical principles.

Purpose & Scope

According to the draft, the guideline applies to confirmatory randomised controlled trials intended to show:

• Absolute efficacy vs. a putative placebo (via indirect comparison)

• Non-inferior efficacy vs. an active comparator

• Non-inferior safety profiles

• Therapeutic equivalence

• Equivalence in pharmacodynamic properties

• Biosimilarity based on clinical endpoints (comparative clinical efficacy studies)

It does not cover bioequivalence, PK studies, or CMC/quality aspects.

Core Analytical Principles

The guideline clarifies that non-inferiority/equivalence analyses are comparisons, not “trial types”, since one study may pursue multiple objectives across endpoints. The draft clarifies how 95% confidence intervals underpin superiority, non-inferiority, and equivalence testing:

• Superiority → CI must lie entirely above 0

• Non-inferiority → CI must lie above –Δ

• Equivalence → CI must fall fully between –Δ and +Δ

The document stresses that even when statistical criteria are met, regulators still assess total benefit-risk, including trial conduct quality and clinical relevance of outcomes.

Assay Sensitivity: A Critical Requirement

Assay sensitivity—i.e., the trial’s ability to distinguish effective from ineffective treatments—is highlighted as essential. Lack of assay sensitivity can falsely make treatments look “similar”, thus invalidating non-inferiority conclusions. The guideline requires:

• Strong justification of the constancy assumption

• Careful comparison of historic vs. current trial populations, endpoints, variability, intercurrent events, and protocol deviations

• Particular caution in two-arm trials without placebo

If assay sensitivity cannot be guaranteed, EMA advises against attempting a non-inferiority design. Margin justification is one of the most detailed sections of the guideline. EMA integrates ICH E9(R1) by requiring that estimands be aligned to the trial’s regulatory objective. The guideline emphasises that:

• Different objectives (absolute vs. relative efficacy) often require different estimands

• Supplementary estimands may be needed

• Intercurrent events (e.g., rescue medication, discontinuation) must be pre-specified and evaluated for their potential to bias towards similarity

Statistical Analysis Expectations

The guideline reinforces strong statistical discipline:

• Primary analysis must match the primary estimand

• Two-sided 95% CIs required

• Per-protocol analyses are not preferred for primary decisions

• Missing data strategies must avoid bias toward similarity

• Sample size re-estimation can inflate type I error and must be used cautiously

Furthermore, superiority testing after non-inferiority is considered a “margin reduction” and must be included in a controlled testing strategy.

For sponsors designing pivotal trials—whether for novel therapies or biosimilars—this draft represents a critical shift toward methodological rigor, patient-relevant outcomes, and high-confidence decision-making. Please refer to the full EMA Draft Guideline on Non-Inferiority and Equivalence Comparisons in Clinical Trials