Subgroup analyses
Publications in high impact journals (impact factor ≧4) after 2010 vs low impact journals after 2010 showed potential higher rates of context placement (52/67 [78%] vs 49/71 [69%]; RR 1.1, 95%CI 0.9-1.4), information gain (47/67 [70%] vs 32/71 [45%]; RR 1.56, 95%CI 1.15-2.10) and transparency features: trial registration (62/67 [93%] vs 33/71 [47%]; RR 1.99, 95%CI 1.54-2.58), published protocols (17/67 [25%] vs 4/71 [6%]; RR 4.50, 95%CI 1.60-12.7), and data availability (9/67 [13%] vs 2/71 [3%]; RR 4.77, 95%CI 1.07-21.27). However, only 36 of these 67 high impact journal publications (54%, 95%CI 42-65%) reported a complete power calculation and 17/67 (25%, 95%CI 16-37%) used a surrogate and/or composite outcomes as their primary outcome (Table S4). Some examples of surrogate outcomes: “cervical dilation” in women receiving home uterine monitoring for early PTB detection; “number of hospital antenatal visits” in women receiving routine doppler ultrasound; biochemical test results like HbA1c values in diabetic pregnant women or CRP values in pregnant women treated for a (vaginal) infection (Table S6).
Out of the 350 trials, 259 (74%) reported a primary outcome of which 97/259 (38%) reported a statistically significant result (P<0.05) in the primary outcome(s). Trials with a non-significant finding in the primary outcome(s) most frequently satisfied several usefulness criteria. This is most prominently for information gain (RR 1.27, 95%CI 1.03-1.56) and transparency features like trial registration (RR 1.31, 95%CI 0.97-1.78), preregistration (RR 1.34, 95%CI 0.73-2.4) and protocol availability (RR 2.0, 95%CI 0.76-5.34) (Table S4).