Usefulness features
Figure 2 provides a bird’s eye view of binary usefulness features in proportion to the total body of PTB trials (n=350) and according to publication year and impact factor. In supplementary material (Table S4) more details on the exact numbers and proportions are shown, split in subgroups.
Problem base. The incidence of PTB <37weeks in the control or placebo groups of the trials varied from 0-100% with a median of 13.7% (IQR 7-30%) (Table S4 and Figure S2). This median is comparable with the worldwide incidence of PTB of 11.1% (6), but the range covers the full spectrum, including very low risk, high risk and very high risk populations (i.e. triplet pregnancies).
Context placement. A total of 185/350 trials (53%, 95%CI 48-58%) justified the importance of their study in context of previous systematic reviews (n=101/138 [73%] 95%CI 65-80% published ≧2010) and 18/350 trials (5%, 95%CI 3-8%) performed a systematic review as part of their study or included and updated meta-analyses.
Information gain. Information gain was deemed to be present in 192/350 trials (55%, 95%CI 50-60%). Absence of power calculations decreased from 71/212 trials (34%) among those published before 2010 to 29 /138 trials (21%) among those published ≧2010 (RR 0.63, 95%CI 0.43-0.91) (Table 1). However, from the 250 trials reporting a power calculation, calculations were incomplete and thus un-informative in 107 (43%, 95%CI 40-50%). Examples of non-informative power calculations: ‘we aimed for 80% power with an α of 0.05 and a β of 0.20’ , incoherent information ‘32 infants are necessary to reach a power of 0.05% at the 80% confidence leve l’, or incomplete reporting on the expected (absolute or relative) proportions of the primary outcome with and without the intervention ‘we aimed for a 20% difference’ . The aimed differences for prolongation of gestational age varied from 3 days till 14 days; and birthweight, from 80g till 1500g (Figure S2 and Table S5). Use of composite or surrogate outcomes increased from 28/212 trials (13%) before 2010 to 34/138 trials (25%) ≧2010 (RR 1.87, 95%CI 1.19-2.93)(Table S4).
Pragmatism: A total of 17/350 trials (5%, 95%CI 3-8%) employed a pragmatic design, with no difference over time in their relative frequency.
Patient centeredness and value for money: No trials reported involvement of mothers to reflect patients’ top priorities in research questions or outcomes used. No value of information analysis was reported in any final manuscript.
Feasibility: There were 113/350 trials (32%, 95%CI 28-37%) that did not report their intended sample size, while 34/350 trials (10%, 95%CI 7-13%) were unable to recruit their intended sample size (not counting DSMB interference because of clear benefit or harm n=5).
Transparency: A total of 95/138 trials (69%, 95%CI 61-75%), published ≧2010 were registered and a total of 38/138 (28%, 95%CI 21-36) were preregistered (registration before randomization of first patient). Change over time using ≧ 2010 as a cut-off shows an increase from 0.5% vs 28%, RR 58, 95% CI 8-420%. Protocols were available in 21/138 trials (15%, 95%CI 10-22%). Change over time shows an increase from 0.5% to 15%, RR 32, 95% CI 4-237%. Data sharing statement was reported in 11/138 (8%, 95%CI 5-14%) trials ≧2010 with an increase from 2% to 8%, RR 3, 95% CI 1-10%.
A total of 255/350 trials (73%, 95%CI 68-77%) reported funding sources and 151/350 (43%, 95%CI 38-48%) reported on conflicts of interest.