Methodological and Reporting Quality of Non-Inferiority or Equivalence Designs: A Systematic Review of Trial Characteristics, Design Consideration and Interpretation in Breast Cancer Radiotherapy Trials

Abstract

To investigate the methodological and reporting quality of non-inferiority (NI)/equivalence trials of breast cancer radiotherapy and to provide suggestions for future NI/equivalence trials. Prospective phase III randomized controlled trials (RCTs) comparing different radiation modalities in patients with breast cancer and designed or interpreted as NI/equivalence were identified in PubMed, EMBASE and Cochrane library. Two reviewers independently extracted data on trial characteristics, statistical design assumptions and analysis considerations, primary end point results and conclusions. The relationship between the number of published trials and the year of publication was assessed by simple linear regression. Trials with pre-specified NI margins as absolute risk differences were reevaluated using margins as relative risk differences. A total of 1490 records were screened and 41 articles published between January 1, 2001 and May 9, 2022 were selected for full text review. A total of 21 trials were included (18 designed as NI and 3 as equivalence). Publication of these trials increased over time (p = 0.023). Trial interventions included dose fractionation (n = 10), partial/whole breast irradiation (n = 8) and tumor bed boost (n = 3). Eleven (52.4%) trials clearly described the non-efficacy benefits. The primary endpoints included 5-year local recurrence (LR) (n = 11), 5-year locoregional recurrence (n = 3), acute/late toxicities (n = 5), 2-year LR and cosmetic outcome (n = 1), and 10-year LR (n = 1). Only seven (33.3%) trials provided justification of the margins. The absolute and relative risk margins were both mentioned in nine (42.9%) trials' methods and reported in six (28.6%) trials' results. The analyzed populations were intention-to-treat (ITT) in 10, both ITT and per-protocol in 9 trials. Seventeen (81%) trials reported confidence interval (CI), with twelve reporting CI that agreed with the type I error used in sample size calculation, but only eight (38.1%) reported p value for NI/equivalence test. Fifteen (71.4%) trials concluded NI/equivalence. Five (23.8%) trials had misleading conclusions (four for not mentioning small sample size insufficient to confirm NI/equivalence and one for inconsistent with the published results). Thirteen (61.9%) trials reported that the protocol's initial accrual target was not met, with ten (47.6%) owing to overestimation of event rates. For trials that met NI only based on absolute margin, three of eight (37.5%) trials were classified as inconclusive with the assumed relative margins. The use of NI/equivalence trials of breast cancer radiotherapy has dramatically increased recently, but there is substantial room for improvement in the methodological and reporting quality of NI/equivalence trials.