Quality of reporting noninferiority/similarity in research studies of diagnostic imaging.

PURPOSE: To evaluate the quality of reporting noninferiority/similarity in research studies of diagnostic imaging.
MATERIALS AND METHODS: All 661 original research articles published in Radiology and European Radiology in 2012 were hand searched to find original studies of diagnostic imaging that included a claim of noninferiority/similarity as a study conclusion. A separate MEDLINE search was performed to identify original studies of diagnostic imaging designed as noninferiority studies, defined as studies at least embodying noninferiority margin, published in English between April 2007 and December 2012. Thirty-eight articles in the first category and 30 in the second category reporting a total of 31 noninferiority studies were finally collected after screening for eligibility. The quality of the reported analytic methodology was evaluated largely according to the extended Consolidated Standards of Reporting Trials, or CONSORT, statement to cover noninferiority studies. In the 31 designated noninferiority studies, the authors evaluated the following key areas of compliance required for proper reporting: noninferiority margin defined and justification provided, sample size calculated according to noninferiority methodology, and use of proper statistical analysis.
RESULTS: Of the 38 studies concluding noninferiority/similarity, one study (3%) used proper analytic method. Twenty-nine of the 38 studies (76%), 28 of which also omitted sample size estimation, claimed noninferiority/similarity for P ≥ .05 from conventional two-sided superiority testing. Of the 31 designated noninferiority studies, seven (23%) fulfilled the requirements for proper reporting. Justification for noninferiority margin was provided in 15 of the 31 studies (48%). Proper sample size estimation was done in 14 of the 31 studies (45%).
CONCLUSION: The quality of reporting noninferiority/similarity in recently published research studies of diagnostic imaging was deficient. © RSNA, 2013.