Clinical Trials

❘❙❚■ Chapter 14 | Noninferiority TrialsNoninferiority trials do not conform to the definition of ‘gold standard’ trials(ie, randomized, double-blind, placebo-controlled trials) [1]. However, it issometimes possible to incorporate a third, placebo, arm into the trial in order todemonstrate internal validity within the trial. Noninferiority trials also suffer fromthe criticism that no naturally conservative analysis exists (in the sense that a truetreatment effect can be diluted by poor investigator conduct and patientcompliance). This is discussed further later in this chapter.Example (continued)A number of antibiotics exist for bacterial infections (such as CAP), offering≥85% clinical efficacy at 4 weeks posttreatment. Demonstrating superior efficacyon the primary endpoint (ie, clinical response at test of cure) in the generalpopulation would be difficult given such an excellent success rate. However, sincedrug resistance is an increasing problem, new antibiotics are required, and so it isconsidered appropriate to show noninferiority to an active comparator and todemonstrate activity against resistant pathogens.How is the noninferiority margin chosen?The choice of δ is a critically important aspect of the study design. The value istypically chosen using clinical judgment, with reference to relevant regulatoryguidance [1,3] and, if appropriate, to guidance for the particular indication.A margin should be chosen such that a difference in treatments of such amagnitude would be considered clinically irrelevant, and anything greater wouldbe unacceptably large. The value of δ is likely to be smaller than the differencelooked for in a placebo-controlled superiority trial, since this would be a valueof undisputed clinical importance.An alternative way to arrive at a δ is to think of the study as attempting to showthat the new compound is superior to a historical control. A δ of half the differencepreviously demonstrated in a superiority trial might, therefore, seem appropriate.Other factors that might influence the choice of δ are the degree of risk associatedwith treatment failure for the indication, relative toxicity, ease of use of the newtreatment compared with the control, and, possibly, feasibility of the study interms of enrolling the required number of subjects.Example (continued)Success rates of 80–90% are typical for currently approved antibiotics in thetreatment of CAP [4]. As such, a δ of –10% is considered clinically irrelevant,so this is the margin recommended in regulatory guidelines [5].134