We appreciate the thoughtful comments by Sethuraman and Kurhekar [1] on our propensity score-matched, non-inferiority study [2].

The letter argues that our study cannot be classified as a “non-inferiority study.” Importantly, we did not describe our study as a randomized controlled trial (RCT). Instead, we explicitly stated that this was a retrospective non-inferior “study” using propensity score matching to mimic an RCT [2], not a non-inferiority “trial.” Most of the points raised in the letter are relevant; however, they pertain primarily to RCT standards and requirements [3]. Because our study was not a prospective RCT, we do not have to comply with the same criteria as those for non-inferiority trials.

The fact that propensity score-matched studies typically do not employ a non-inferiority design may have contributed to this misunderstanding. However, we adopted a non-inferiority design in this study because we needed to consider both the potential advantages and disadvantages of remimazolam compared with dexmedetomidine. Although remimazolam was expected to offer benefits such as hemodynamic stability and minimal effects on cardiac conduction, delirium was also a concern. Consequently, a comprehensive evaluation was necessary, which led to the development of the non-inferiority hypothesis.

This letter highlights the general principle of the non-inferiority margin. Non-inferiority margins are recommended to be based on previous superiority trials comparing the standard treatment with a placebo [4]. However, in the context of perioperative outcomes, comparing the standard sedative against a placebo is not feasible owing to ethical and practical considerations. Therefore, no historical data were available to compare transcatheter aortic valve replacement outcomes between dexmedetomidine and placebo. Consequently, we established a non-inferiority margin following clinical judgment [5]. Specifically, we set the margin to −10%, which we believed to be appropriate based on both clinical context and sample size considerations.

Although further criticism of this −10% margin may be justified, introducing a stricter non-inferiority margin would have required an extension of the study period, potentially increasing the risk of temporal bias owing to the distinct periods during which remimazolam and dexmedetomidine were administered. Thus, the tradeoff between increasing statistical power and reducing temporal biases was carefully considered, and we determined that a −10% margin would provide the best compromise.

We hope that this explanation clarifies our rationale for designing and conducting our study in this manner. We appreciate the opportunity to provide you with this response and look forward to further discussion.