Strengths and limitations
Among the strengths of our work we can mention that we followed the Cochrane guidelines Cochrane19, the PRISMA-NMA extension15for reporting and we registered the study protocol in advance. Our work is the most updated and complete systematic review assessing clinical effectiveness and safety of corticosteroids. Our exhaustive search strategy, included clinical trials registries and contacted experts for additional relevant evidence. Although we did not hand-search conference proceedings it is unlikely that our search strategy missed RCTs not included in biomedical databases nor the trials registers.
This NMA added two small trials13,14and one of large good quality trial that compared directly dexamethasone with betamethasone12 to the body of evidence. It provided new indirect estimations and increased the precision of the estimations, still low for most outcomes, by combining direct and indirect evidence. The prespecified meta-regression, subgroup and sensitivity analyses reinforced the robustness of our results.
We assessed the certainty of the evidence by the GRADE-NMA approach33,34, the validity of the transitivity assumption by comparing the distribution of potential effect modifiers across comparisons and the coherence assumption by the design-by-treatment interaction model and loop-specific approaches.22,24
The results of the NMA were mostly coherent, except for chorioamnionitis, may be due to differences between populations included in indirect and direct evidence, and differences in RoB. The indirect evidence came mostly from mothers with ruptured membranes36,37,41,47,49,51,52,55-57,60,65,67,69,72while the direct evidence from a mix of mothers with intact and ruptured membranes12. However, meta-regression, subgroup and sensitivity analyses did not explain this incoherence. Therefore, for chorioamnionitis, following the GRADE approach33, we considered the direct evidence the most reliable estimation of 23 fewer cases (43 fewer or 5 more) per 1000 women treated with dexamethasone.
We included studies conducted in a range of 50 years and healthcare advances, specifically in neonatology, could be an extra-source of heterogeneity that could partially explain the contradictory direction of effect for some outcomes, but the effect modifiers or RoB did not provide a solid explanation of the effects. The contradictory beneficial or detrimental effect of different outcomes warranted our decision to explore the patients’ perspectives about our findings comparing corticosteroids trough a focus group (Appendix 3). Briefly, women failed to make a decision about which corticosteroids they would choose because the trade-off between risk and benefits were very complex for them. They agreed that it would be a decision that they would share or delegate to a professional with whom they established a bond of trust.
The evidence shows limitations, regarding its generalizability to lower-resource countries, since only three14,42,47(7%) out of the included RCTs were from lower-MICs and none from LICs. Trials have been largely conducted in tertiary hospitals and recruited highly selected populations.83 Concerns about safety and efficacy in low-resource settings were supported by the adverse findings in neonatal deaths and maternal infection of ACT, a community-based, cluster-RCT conducted in six LMICs.84 However we did not find important differences by income country classification and by GNI per capita. Hopefully, the ongoing ACTION study could answer this question.6