Equivalence Trial

An equivalence trial is a clinical trial designed to demonstrate that two treatments produce outcomes that are sufficiently similar, within a predefined margin of clinical acceptability, that neither is meaningfully inferior or superior to the other. Unlike superiority trials, which test whether a new treatment is better than a comparator, equivalence trials test whether two treatments are therapeutically interchangeable. The predefined equivalence margin (delta) defines the maximum difference in outcomes that is considered clinically acceptable.

Not every new treatment needs to be superior to existing options. A new treatment that is therapeutically equivalent but offers advantages in other dimensions, such as lower cost, fewer side effects, a more convenient route of administration, or reduced pill burden, may be preferred by patients and healthcare systems. Generic drug approvals, biosimilar evaluations, and comparisons of established treatments all involve equivalence questions. Without equivalence trials, clinicians would lack the evidence needed to confidently substitute one treatment for another. The design and interpretation of equivalence trials differ substantially from superiority trials, and applying the wrong framework can lead to incorrect conclusions.

The equivalence margin is the critical design parameter. It represents the maximum difference between treatments that would be considered clinically unimportant, below which the treatments are considered interchangeable. The margin must be specified before the trial begins, justified on clinical grounds, and smaller than the smallest clinically important difference. For example, if two blood pressure medications differ by no more than 3 mmHg in their effect on systolic blood pressure, they might be considered equivalent.

The statistical framework differs from superiority trials. In a superiority trial, the null hypothesis is that there is no difference, and the trial seeks evidence to reject this hypothesis. In an equivalence trial, the null hypothesis is that the true difference lies at or beyond the equivalence margin in either direction, and the trial seeks evidence that it falls within the margin. Equivalence is established when the confidence interval for the treatment difference falls entirely within the range defined by negative delta to positive delta.

Sample size calculations for equivalence trials often require larger samples than superiority trials because the goal is to demonstrate with precision that the difference is small, which requires narrow confidence intervals. Equivalence is usually tested with two one sided tests (TOST), one against each margin, and is concluded only if both reject; ICH E9 notes that this is operationally the same as checking that the confidence interval lies within the margins. Each one sided test at 2.5% corresponds to a two sided 95% confidence interval, and each at 5% to a 90% interval, the convention in pharmacokinetic bioequivalence studies. Power is calculated for both tests succeeding together.

In superiority trials, intention to treat analysis is standard and conservative. In equivalence trials it is not necessarily conservative: factors that dilute the treatment difference (poor adherence, crossover, protocol deviations) make treatments appear more similar than they actually are, biasing toward a false conclusion of equivalence. Per protocol analysis, which includes only participants who adhered to the protocol, is not a guaranteed fix, because excluding people after randomization can introduce its own bias. ICH E9 therefore recommends planning both analyses in confirmatory trials, and the CONSORT extension for noninferiority and equivalence trials asks authors to state whether the conclusion rests on intention to treat, per protocol, or both analyses, and whether it holds across them. Equivalence is most convincing when both analyses support it.

Equivalence trials are appropriate when the objective is to demonstrate that two treatments are interchangeable, when a new treatment is expected to be similar in efficacy but may offer other advantages, or when comparing established treatments to determine whether they can be used interchangeably. Common applications include generic drug versus branded drug comparisons, comparisons of different formulations or routes of administration, and evaluations of simplified versus standard treatment regimens. Equivalence trials are not appropriate when the goal is to determine whether a new treatment is better (superiority trial) or merely not unacceptably worse (noninferiority trial).

A common mistake is setting an equivalence margin that is too wide, which makes it easy to demonstrate "equivalence" between treatments that actually differ meaningfully. Another error is interpreting a superiority trial that fails to show a significant difference as evidence of equivalence. Absence of evidence of a difference is not the same as evidence of no difference; a trial that was not designed or powered for equivalence cannot demonstrate it. Failing to report both intention to treat and per protocol analyses deprives readers of the information needed to evaluate the robustness of the equivalence conclusion. Using an inappropriately small sample size produces wide confidence intervals that cannot exclude clinically meaningful differences, resulting in an inconclusive trial. Finally, confusing equivalence with noninferiority is common; noninferiority tests only one direction (whether the new treatment is no worse), while equivalence tests both directions (whether it is neither worse nor better beyond the margin).

A randomized equivalence trial compares a new once daily oral antibiotic formulation to the standard twice daily formulation for uncomplicated urinary tract infections. The equivalence margin, set in the protocol and justified on clinical grounds, is 10 percentage points for clinical cure rate at day 14. The trial enrolls 800 women, 400 per formulation. In the intention to treat analysis, the cure rate is 86% (344 of 400) in the once daily group and 88% (352 of 400) in the twice daily group, a difference of negative 2 percentage points (95% CI: negative 6.7 to positive 2.7). In the per protocol analysis of 300 women per group, it is 88% (264 of 300) versus 90% (270 of 300), again negative 2 percentage points (95% CI: negative 7.0 to positive 3.0). Because both confidence intervals fall entirely within the predefined margin of negative 10 to positive 10 percentage points, equivalence is established, supporting the use of the more convenient once daily formulation.