Interpreting interval estimates in superiority and non-inferiority trials: a unified framework implemented in the Conclusion Generator.

BACKGROUND: Confidence intervals are increasingly recommended to be interpreted as compatibility intervals rather than dichotomised according to statistical significance. However, no uniform framework exists for interpreting interval estimates across superiority and non-inferiority trials, particularly when different combinations of superiority and non-inferiority margins are specified. METHODS: We developed a unified framework for interpreting interval estimates in superiority and non-inferiority trials based on the position of a user-provided interval estimate relative to prespecified superiority and non-inferiority margins. The framework was implemented in an updated version of the freely available online Conclusion Generator (https://apps.biostat.au.dk/erikparner/Conclusion_app/) and in standalone R functions available through GitHub (https://github.com/erikparner/conclusion_generator). Both implementations automatically generate standardised concise and elaborated conclusion text for superiority, non-inferiority, and combined trial designs. RESULTS: The framework classifies all possible interval estimates into predefined interpretation scenarios, providing consistent interpretations of treatment effects while avoiding dichotomous terminology based on statistical significance. The updated Conclusion Generator supports trials specifying superiority margins, non-inferiority margins, or both, and extends its application by adding support for mean differences and refined standardised wording. CONCLUSIONS: This unified framework provides a consistent approach to interpreting interval estimates across superiority and non-inferiority trials and is freely available as a web application and standalone R functions.