Study Designs in Multi-arm Trials for Breast Cancer: A Systematic Literature Review of Major Journals

Background: Several articles showed that statistical efficiency of multi-arm randomized clinical trials (RCTs) is much better than conventional two-arm RCTs. Multi-arm RCTs attract interest mainly when the experimental treatment regimen is not optimized or several pipelines under development exist. Breast cancer is a possible candidate disease. Our aim was to elucidate the current study designs and multiplicity adjustment methods in multi-arm RCTs. Methods: A search of the PubMed database revealed 468 articles on breast cancer RCTs published from 2010 to 2016. Information on study designs and analysis methods was collected from 4 major journals. Results: A total of 202 RCTs were selected, 48 were multi-arm and 29 were three-arm RCTs. In two of the target journals, multi-arm RCTs have been increasingly reported since 2013. Compared with two-arm RCTs, three-arm RCTs were frequently conducted in neoadjuvant settings (7.7% vs 33.3%). The number of trials performed in perioperative settings was 46 in two-arm and 15 in three-arm RCTs. Of these, the proportion of industry-sponsored trials in two-arm and three-arm RCTs was 26.1% and 53.3%, respectively. Shared control designs (SCDs) which randomized to a common control arm and multiple experimental arms comprised 54.2% of 48 multi-arm RCTs. For SCDs, detailed information on multiplicity adjustment methods was seldom reported. The Bonferroni adjustment method together with alpha-spending functions was commonly used. Conclusion: Breast cancer multi-arm RCTs have been increasingly reported. The majority of multi-arm RCTs are industry-sponsored trials using SCDs in neoadjuvant settings. Detailed description about multiplicity adjustment methods is required for multi-arm RCTs.