Evaluating Noninferiority Trials for Combination Therapies

• Activity Number: 564
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Heinz Schmidli*+
• Companies: Novartis Pharma AG
• Keywords: meta-analysis ; non-inferiority ; clinical trial ; evidence synthesis ; random effect ; historical information

Abstract

Placebo controlled clinical trials are not always possible for ethical reasons, and hence non-inferiority (NI) trials are used which compare the experimental treatment with an active control. Such NI trials do not provide direct evidence on the superiority of the experimental treatment against placebo, and therefore historical information on placebo and active control treatments is used to estimate the placebo effect and define the NI margin. When the experimental treatment is a combination therapy, then often only limited or no information on the putative placebo is available, which requires more complex meta-analytic methods to estimate the placebo effect. The approach is illustrated by a case study in kidney transplantation, where a pivotal NI trial compared an active control treatment consisting of four immunosuppressant drugs, with an experimental treatment where two of these immunosuppressants were changed. As no historical information on the putative placebo treatment was available, a random-effect logistic meta-regression was used to estimate the placebo effect, and define the NI margin, based on data from 51 randomized controlled trials with a total of 17002 patients.