For many people, the concept of noninferiority studies may not be intuitive. It is therefore of concern that the article cited by Nicole Le Saux and associates1 as methodologic support for their study design did not explicitly mention the term “noninferiority” or “margin of noninferiority,”2 although it did include the operational definition. Equivalence and noninferiority studies are usually considered related, yet distinct epidemiologic entities. Indeed, “equivalence trials … aim to match the action of an established therapy and prove this to statistical significance within a predetermined range ([delta]), in both positive and negative directions (a 2-sided test) … [whereas] noninferiority trials are statistically based on a 1-sided comparison to an active control in the positive direction (a 1-sided test).”3
It is unclear to me whether, on theoretical grounds, this was an appropriate methodologic choice for the amoxicillin study. “[E]quivalence and noninferiority trials rely on the premises that the superior efficacy of the active control over placebo has been previously proven for a given indication, and that this efficacy will be preserved under the conditions of the trial. If either of these assumptions is not valid, the results become suspect unless the primary analysis demonstrates differences between the experimental and active-control treatments.”3 Since the conclusion of Le Saux and associates may also be a necessary assumption for the internal validity of the trial, perhaps this trial was a “self-fulfilling prophecy.”
Furthermore, “a fundamental assumption in the design and analysis of an active-control noninferiority trial is that the active control is truly effective.”4 This assertion suggests the following corollary for this study: amoxicillin must be superior to placebo (i.e., has proven efficacy). Since the investigators set out to determine whether placebo was noninferior to amoxicillin, one has to assume that it was still unclear whether the drug was superior. Therefore, the apparently crucial premise of proven efficacy of the active control (amoxicillin) was not met.
Unless the investigators can provide a satisfactory explanation for these apparent illogicalities, one additional question remains: To what extent does such a discrepancy affect their conclusions?
Footnotes
-
Competing interests: None declared.