Clinical Trials

❘❙❚■ Chapter 12 | Equivalence Trialswhere each subject receives both treatments in a random order – the betweensubjectvariability is minimized because the same subject is given both treatments.Assuming that the subject remains in similar health at the beginning of each armof the study, he/she acts as his/her own reference or control group. This makes thecrossover design more efficient in terms of sample size. If a parallel design wereused, more volunteers would be needed to reach equivalence with the same power.In some cases, if the within-subject variability is high (eg, nifedipine and acyclovirhave highly variable effects whereby the same dose and drug given to the samesubject on different occasions will have differing absorption, metabolism, andclinical effects), the advantage of using a crossover design is minimal.NoninferiorityThere are instances in which the efficacy of a drug needs to be shown to be notinferior to that of another drug, and equivalence trials designed to detect this aresaid to be noninferiority trials. TARGET (Do Tirofiban and ReoPro Give SimilarEfficacy Outcomes Trial) is an example of a noninferiority trial. This studycompared two glycoprotein IIb/IIIa receptor blockers with similar mechanismsof action but different chemical structures, platelet-cell adherence profiles,and durations of action [7]. Although the trial was designed to show thenoninferiority of tirofiban as compared with abciximab, the trial resultsdemonstrated that tirofiban had higher ischemic event rates than did abciximab,failing to show noninferiority.Interpreting resultsEquivalence trialsThe definition of an endpoint can also vary depending on the drug/disease inquestion. In an equivalence trial in a disease that has a major impact on thepatient, an absolute reduction of events of within 1% of the established/referencetreatment might be acceptable for equivalence. So, if the established/referencetreatment achieved a 10% reduction in events, the new treatment must achieve anabsolute reduction of events of between 9% and 11%. However, for a disease witha relatively small impact on the patient, such as the common cold, a result within5% of the reference treatment might be considered equivalent.Noninferiority trialsWhen two treatments are compared, a single trial gives one estimate of the truedifference between them. A range can be determined from that single estimateto capture the true effect (see Chapter 18). Consider an example trial that showsthat a new treatment is 10% less effective than an established treatment, with116