Statistical analysis
Most analyses are descriptive and report proportions (with their 95%
confidence interval [CI] using Wilson score) percentages, medians
(interquartile range, IQR) or means (standard deviation, SD). We
performed three prespecified subgroup analyses. First, we compared the
eight individual usefulness assessments in studies reported before 2000,
between 2000-2009 and between 2010-2019. These time cut-offs were chosen
as we expected a clear improvement on at least the transparency features
due to a wide range of initiatives that started after the millennium.
The trial register clinicaltrials.gov was launched in 2000 and the
International Committee of Medical Journal Editors (ICMJE) started
demanding registration of trials before submission to ICMJE journals in
2005; icmje.org). The comparative analysis compared the most recent time
frame (≧2010) with the time frame before that (<2010) using
risk ratios (RRs) with 95% CI. Second, we assessed the impact factor
(IF) of journals using the Web of Science Journal Citation Report 2019
and pre-specified a ‘high’ IF at a cut-off of ≧4.0 and ‘low’ IF when
<4.(15) This cut-off was chosen to include the top specialty
journals in obstetrics and gynaecology in the ‘high’ group, that
generally have an IF between 4 and 6 (e.g., Obstetrics & Gynecology,
British Journal of Obstetrics and Gynaecology). We expected the most
recent RCTs (≧2010) published in high impact journals to perform better
in some of the usefulness features compared to those in low impact
journals. Third, we stratified usefulness features by whether or not
there was a statistically significant result (P<0.05) in (at
least one) primary outcome. Finally, we counted for each trial for how
many of the 13 usefulness items it scored ‘positive’ to provide an
overview on the number of trials meeting at least half (7/13) of the
items.