Falling in the margin

In a randomised controlled trial (RCT) we are usually trying to answer the question of which treatment is best. Sometimes, though, may be considering that do not believe will perform better than an existing option but has other benefits: it could cheaper, easier deliver or more acceptable patients. For example, need for urodynamic testing compared with simple office evaluation before stress incontinence surgery was evaluated in RCT (Nager et al. N Engl J Med 2012;366:1987–97). The authors believed office-based would less invasive and also reduce risk urinary tract infections. It expected unlikely success. key consideration how much worse terms outcome going acceptable. There no consensus about this should evaluated, intuitively difference small take into account nature outcome, balanced potential benefits. One ask clinicians patients what level cause them switch treatments. This termed non-inferiority margin determines size trial. Confidence intervals used determine whether new ‘not unacceptably worse’ standard treatment. If upper end confidence interval (CI) effect smaller margin, then can conclude (scenario B Figure 1). A number design issues considered setting why some consider trials have weaknesses superiority trials. These include: (i) ensuring efficacy comparator been proven on similar population; (ii) selecting appropriate analysis population light reverse hypothesis (i.e. demonstrate similarity, per-protocol populations alongside intention-to-treat populations); (iii) taking care conduct, as carelessness (such poorly executed randomisation) lend itself conclusion similarity. Further reading these recommended (see below). primary study described above success symptoms substantially improved) at 12 months. set 11 percentage points. found tested urodynamics rate 76.9% (203/264), undergoing 77.2% (200/259). between two groups –0.3% (95% CI –7.5% 6.9%). tells us might 6.9% (upper CI) 7.5% (lower urodynamics. results therefore indicated certainly 11% worse. declared non-inferiority, recommending performed. None declared. Completed disclosure interests form available view online supporting information. Please note: publisher responsible content functionality any information supplied by authors. Any queries (other missing content) directed corresponding author article.