We thank Dr. Sethuraman for their interest in our article and the opportunity to address their study design points. First, there was a concern that a non-inferiority design is only suitable for clinical trials in which a treatment or intervention is being compared to another (usually previously existing and well established as effective) treatment. While these types of studies were indeed early examples of noninferiority trials, application of this type of design has expanded rapidly in the last decade.  From a statistical standpoint, it is possible to use a noninferiority design any time two continuous variables are being compared when one is expected to be “not different” (if double-sided) or either not higher or not lower (if single-sided) than the other.

In our study, we hypothesized that the gastric volume of diabetic patients who have followed standard fasting instructions prior to elective surgery would be “no higher” than that of nondiabetic patients. Therefore, we considered a noninferiority design to be more appropriate than a “superiority design” to test our hypothesis.

Dr. Sethuraman is correct to state that to prove non-inferiority, one needs to test the 95% CI of the difference between means and one needs to establish an “a priori” noninferiority margin. In our study, the primary outcome in question was gastric volume per unit of weight (in ml/kg) as estimated using bedside ultrasound. We established the “a priori” noninferiority margin of 0.4 ml/kg as this was 50% of the expected SD of baseline values, which is usually considered to be a small difference, and any difference below that margin could be considered a clinically insignificant difference. Our findings confirmed the hypothesis of noninferiority as the mean difference between the two groups was 0.07 ml/kg and the 95% CI of this difference was –0.24 to 0.1 ml/kg. These results are illustrated and contrasted with the noninferiority margin in figure 3 of our original article.

Finally, we respectfully disagree that our “observational” study would be difficult to reconcile with a noninferiority design. Our study was an observational cross-sectional study without treatment intervention. We merely evaluated the baseline gastric volume of diabetic and nondiabetic patients who have fasted for elective surgery. We did not study the effect of any treatment, but rather compared the effect of an “exposure to a condition” (diabetes mellitus) in an observed biologic outcome (fasting gastric volume). We therefore believe that a noninferiority design is scientifically appropriate for this purpose.