Importance: Noninferiority trials test whether a new intervention is not worse than the comparator by a given margin.
Objectives: To study the characteristics of published randomized noninferiority trials in oncology with overall survival as an end point, to assess the association of justification and success in achieving noninferiority with the funding of these trials, and to evaluate the association of such trials with patient survival.
Data sources: A systematic search of PubMed and Google Scholar databases was conducted in March 2018, with no date restrictions.
Study selection: Randomized noninferiority trials of cancer drug therapies with overall survival as an end point were included. Trials of decision support, supportive care, and nondrug treatment in both arms were excluded.
Data extraction and synthesis: This study followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines for meta-epidemiological studies. Studies were screened for eligibility criteria, and data on criteria for noninferiority, funding, success (achieving noninferiority), and hazard ratios with confidence intervals for overall survival were extracted. Hazard ratios for overall survival were pooled across trials using a random-effects model.
Main outcomes and measures: Associations of the justification for using a noninferiority design and success in achieving noninferiority with the source of funding were assessed. Overall pooled hazard ratios and confidence intervals for overall survival were calculated.
Results: Among 74 noninferiority trials of cancer drug therapies, 23 (31%; enrolling 21 437 patients) used overall survival as the primary end point. The noninferiority margins for the hazard ratio of overall survival ranged from 1.08 to 1.33. Noninferiority design was justified in 14 trials (61%) but not in 9 (39%). Overall, 18 trials (78%) concluded with a finding of noninferiority. Industry funding was associated with lack of justification for noninferiority design (P = .02, assessed using the Fisher exact test) but not with success in proving noninferiority (P = .80, assessed using the Fisher exact test). When the hazard ratios across the trials were pooled, there was no beneficial or detrimental association with overall survival, with a pooled hazard ratio of 0.97 (95% CI, 0.92-1.02).
Conclusions and relevance: The findings suggest that a substantial fraction of noninferiority trials in oncology, most of which are industry funded, lack justification for such a design. Greater attention to the use of noninferiority designs in randomized clinical trials of cancer drugs from local and national regulators is warranted.