Nonpublication and Selective Outcome Reporting of Oral Health Trials

Abstract

Less than one-quarter of oral health trials are registered in a public registry. However, no study has assessed the extent of study publication and selective outcome reporting bias in the field of oral health. We identified oral health trials registered between 2006 and 2016 in ClinicalTrials.gov. We assessed whether results of early discontinued trials, trials having an unknown status, and completed trials had been published and, among published trials, whether outcomes differed between the registered record and the corresponding publication. We included 1,399 trials, of which 81 (5.8%) were discontinued, 247 (17.7%) had an unknown status, and 1,071 (76.6%) were completed. The registration was prospective for 719 (51.9%) trials. Over half the registered trials were unpublished ( n = 793, 56.7%). To explore the association between trials publication and characteristics of trials, we performed a multivariate logistic regression analysis. Trials conducted in the United States ( P = 0.003) or Brazil ( P < 0.001) were associated with increased odds of publication, whereas trials registered prospectively ( P = 0.001) and industry-sponsored trials ( P = 0.02) were associated with decreased odds. Among the 479 published trials with completed status, the primary outcomes of 215 (44.9%) articles differed from that registered. Major discrepancies consisted of the introduction of a new primary outcome in the published article (196 [91.2%]) and the transformation of a registered secondary outcome into a primary outcome (112 [52.1%]). In the remaining 264 (55.1%) trials, primary outcomes did not differ from that registered, but 141 (53.4%) had been registered retrospectively. Our study highlights the high rate of nonpublication and selective outcome reporting in the field of oral health. These results could alert sponsors, funders, authors of systematic reviews, and the oral health research community at large to combat the nondisclosure of trial results.