The State of Being Noninferior

Abstract

This issue presents the results of a study by Diestelhorst et al that was designed to demonstrate that intraocular pressure (IOP) reduction from a fixed combination of latanoprost and timolol given once daily is “noninferior” to the 2 drugs administered separately. A noninferiority design also is being used in a large, ongoing phase II clinical trial comparing anecortave acetate with photodynamic therapy in 530 patients with age-related macular degeneration (AMD). If you are somewhat confused about what is meant by noninferior, rest assured that you are in the large majority. The purpose of this editorial is to bring some clarity to this increasingly prevalent trial design. In the absence of a treatment that has been proven effective, clinical trials of proposed treatments have made use of the placebo-controlled randomized clinical trial, the hallmark design of drug development for decades. Such studies are designed to show that the treatment being tested is superior to the placebo control. Today’s world, thankfully, often presents a different scenario, with effective treatments available for many diseases. Drug development has now shifted to more frequent use of active-controlled randomized clinical trials, wherein the goal is either to improve upon an already effective treatment (which invokes the “superior to” hypothesis) or to demonstrate that, compared with the current treatment, the new treatment is equivalent (i.e., about the same) or not inferior (i.e., possibly better, but definitely no worse). Equivalence and noninferiority trials are attractive because the new treatment can be declared a success even if it does not beat the standard treatment. Such trials are used to test new treatments that are expected to have an efficacy similar to that of the standard treatment, but may have an advantage such as fewer side effects, lower cost, or easier administration. In a noninferiority trial, a margin of noninferiority ( ) must be specified such that a new treatment that is either better than the standard treatment or worse by no more than is considered acceptable. For example, Diestelhorst et al specified a noninferiority margin of 1.5 mmHg in mean daily IOP reduction—that is, they wished to test if the once-a-day combined treatment would result in a mean reduction in IOP between baseline and week 12 no more than 1.5 mmHg less, on average, than the twice-a-day (standard) treatment. They concluded that the once-a-day treatment was noninferior to the standard treatment because, although the IOP reduction with the once-a-day treatment was 0.3 mmHg less than the reduction with the standard treatment, the confidence interval (CI) for the difference in IOP reduction ( 0.1 to 0.7) did not include 1.5. In general, a conclusion of noninferiority is made when the CI for the treatment effect lies entirely in the noninferiority region.