I congratulate CMAJ and Boutis and colleagues for a brilliant research paper.1 Intention-to-treat analysis is a comparison of the treatment groups that includes all patients as originally allocated after randomization. This is the recommended method in superiority trials to avoid any bias. For missing observations, “last value carried forward” is the recommended method.
Per-protocol analysis is a comparison of treatment groups that includes only those patients who completed the treatment originally allocated. If done alone, this analysis leads to bias.
In noninferiority trials, both intention to treat and per-protocol analysis are recommended; both approaches should support noninferiority. In the article by Boutin and colleagues, intention to treat should have included 50 patients in either group as per randomization or at least 45 in the group with splints (in 4 patients, the diagnosis was wrong) and 50 in the group with casts; this may change the results to indicate a borderline effect. In that article, the analysis was done with 43 patients in the splint group and 49 in the cast group, which appears to be a per-protocol analysis, though it was called an intention-to-treat analysis. Hence, noninferiority can be concluded only after analysis by both approaches.
Reference
- 1.Boutis K, Willan A, Babyn P, et al. Cast versus splint in children with minimally angulated fractures of the distal radius: a randomized controlled trial. CMAJ 2010;182:1507–12 [DOI] [PMC free article] [PubMed] [Google Scholar]