Noninferiority margins exceed superiority effect estimates for mortality in cardiovascular trials in high-impact journals

Objectives: To determine, in a sample of noninferiority trials, the noninferiority margins (NIMs) and their relation to effect estimates in superiority trials, the rationale being that in general NIMs should be no larger than effects declared important in superiority trials. Study Design and Setting: To identify cardiovascular trials published in high-impact journals with a statistically significant primary outcome involving mortality, we systematically searched PubMed, Embase, and MEDLINE databases (January 2015-July 2020). We documented the NIMs and determined the proportion of trials with NIMs larger than the median effect estimates across superiority trials. Results: From 1,477 screened titles, 65 (39 noninferiority, 26 superiority) trials proved eligible. The NIMs ranged from risk difference of 0.54-10%. The effect estimate across superiority trials was a median risk difference of 2.1% (interquartile range 1.5-4.9); the NIM was larger than 2.1% in 28 (71.8%) noninferiority trials and larger than 1.5%, which was the lower bound of the interquartile range in 32 (82.1%) noninferiority trials. Conclusion: The very wide range of noninferiority margins and the proportion above a threshold that most would consider an important mortality reduction suggest that clinicians and guideline panels should focus on study results, paying little attention to authors' noninferiority margins. (c) 2023 Elsevier Inc. All rights reserved.

Automated summary

This study aimed to determine the noninferiority margins (NIMs) in noninferiority trials and their relationship to effect estimates in superiority trials. The results showed that the range of NIMs varied widely, ranging from 0.54% to 10%. The majority of noninferiority trials had NIMs larger than the effect estimates in superiority trials, suggesting that clinicians and guideline panels should focus on study results rather than authors' noninferiority margins.