A review of UK publicly funded non-inferiority trials: Is the design more inferior than it should be?

Abstract

Abstract Background: The number of non-inferiority (NI) trials, those aiming to show a new treatment is no worse than a comparator, is increasing. However, their added complexity over superiority trials can create confusion. Most guidance and reviews to date have an industry focus with research suggesting these trials may differ to publicly funded NI trials. This gap in the literature, to understand how well the recommendations are translating to publicly funded NI trials, is the aim of this work by reviewing their design and reporting characteristics. Methods: The International Standard Randomised Controlled Trial Number web registry and the National Institute for Health and Care Research’s Funding and Awards Library and Journals Library were searched using the term non-inferiority and logical synonyms. Inclusion requirements were a UK publicly funded NI randomised controlled trial. Characteristics of the design, analyses and results as available were recorded on a dedicated data extraction spreadsheet. Appropriate summary statistics were used to present the results. Results: Searches completed on the 14th January 2022 identified 477 potential trials which after exclusions resulting in a database of 114 NI trials to be summarised. Non-inferiority margins were defined for most trials with a median of 8% (IQR: 3%-10%) used for risk differences (n=58) and 0.35 (IQR: 0.26-0.43) standardised mean difference for continuous outcomes (n=30). Justifications for the margin chosen (n=62) were more commonly based on the clinical importance (49/62) and less commonly using statistical considerations (13/62). The most prevalent primary analysis population was solely on an intention-to-treat basis (49/114). The superiority of the treatment was well described but not always included as an outcome and only powered for in about a third of cases. Conclusions: Aspects of NI trial design are well described but not always in line with current recommendations. Of particular note, is the absence of statistical considerations when setting the non-inferiority margin, which eliminates the ability to confirm indirect superiority over placebo for the new treatment. Additionally, despite suggestions that it can increase the Type 1 error in NI trials, use of the intention-to-treat alone is most common analysis population. Registration: Research on Research ID: 3171 Funding: Economic and Social Research Council (Reference Number: 205866591)