Correspondence

    loading  Checking for direct PDF access through Ovid

Excerpt

To the Editor:
I read with interest the article by Alberti and la Cour1 titled Non-supine Posturing in Macular Hole Surgery. Although there are now registry-based data suggesting acceptable noninferiority of nonsupine posturing in smaller holes, it is important that surgeons continue to consider this question critically, particularly for larger holes.2
There is considerable responsibility on the authors of a noninferiority study to predefine a clinically appropriate noninferiority definition. The authors' choice of a noninferiority approach was the correct one; however, I feel their choice of a noninferiority margin of 15% is unsupportable.
The wide noninferiority margin enabled the authors to use a small sample size. This is only appropriate if clinicians and patients agree that nonsupine positioning would be preferable if the success rate was no more than 15% worse—in other words, it is OK for facedown positioning to be withheld even if as many as 1 in 7 holes fail to close as a direct result of nonsupine positioning. If readers do not accept this assertion, then they cannot accept the noninferiority conclusion.
There are now 2 published studies that have independently either used or proposed a 5% noninferiority margin when evaluating this question.2,3 Using this noninferiority cutoff would have required a much larger sample size and possibly made the study by Alberti and la Cour nonviable. A difficult sample size, however, should not be a justification to relax the noninferiority margin.
A prospective randomized study investigating this issue is underway, and these authors have pragmatically chosen a 15% risk difference as the basis for their sample size calculation.4 These investigators, however, are using a superiority design, and a hypothetical negative result will not result in a conclusion of noninferiority.
Again, I commend the authors in their ongoing efforts to address this as-yet unanswered question.
    loading  Loading Related Articles