Background and Objectives
Selective reporting can bias estimates of effect, yet methods to detect such biases are limited. Statistical methods for detecting publication bias (e.g., funnel plots, Beggs rank correlation) are underpowered. Comparing outcomes listed under Methods versus those reported under Results in published manuscripts is an expedient but crude method for detecting reporting bias. Another method is to search ClinicalTrials.gov (CT.gov) and (a) compare studies identified there to published studies (to detect publication bias) and (b) compare planned analyses and outcomes reported in CT.gov to those reported in the final publication (to detect reporting bias). The EPC guidance recommends this approach. While conceptually sound, this approach may be labor-intensive, and its utility is uncertain.
The overall goal of this project was to evaluate the utility of CT.gov for detecting selective reporting, and to determine the impact of selective reporting on the estimates of treatment effect. A secondary goal was to estimate the person-hours required to complete these analyses.
Williams JW Jr., Eaton JL, Gierisch JM, Masilamani V, von Isenburg M, Chobot M. Supplemental Project To Assess the Transparency of Reporting for Trials Evaluating Treatment for Infertility. Methods Research Report. (Prepared by the Duke University Evidence-based Practice Center under Contract No. 290-2015-00004-I) AHRQ Publication No. 17-EHC022-EF. Rockville, MD: Agency for Healthcare Research and Quality. June 2017. DOI: https://doi.org/10.23970/AHRQEPCMETH4.