NMA results: dexamethasone vs. betamethasone
The geometry of the treatment networkis presented in Figure 2 , and the direct, indirect and NMA (mixed) effect estimations for the main eight outcomes of the comparison dexamethasone vs. betamethasone are presented in Figure 3 and Table 1 of Summary of finding. We also present in the Appendix S5 the direct comparison forest-plots, in Appendix S6 the summary of finding tables of each corticosteroid vs. control, in Appendix S7 the pairwise meta-analysis forest-plots by type of corticosteroid against placebo and in Appendix S8 the NMA outputs and SUCRA values.
-Chorioamnionitis (6,698 patients, 15 studies)
Compared with placebo/no treatment (control), dexamethasone probably increases chorioamnionitis but in one side of the confidence interval could also reduce it: OR 1.46 (95%CI 0.81-2.66). On the contrary betamethasone reduces chorioamnionitis: OR 0.63 (95%CI 0.41-0.95). The test for subgroup differences by corticosteroid type showed this disparity (P= 0.010, I² 84.2%).
Compared with betamethasone, dexamethasone probably reduces chorioamnionitis but in one side of the CrI is slightly detrimental:OR 0.70 (95%CrI 0.45-1.06), moderate-CE . We found serious incoherence between direct and indirect evidence (Ratio of OR [ROR] 3.18 [95%CI 1.26-8.02]).
-Endometritis/puerperal sepsis (4,030 patients, 10 studies)
Dexamethasone may increase endometritis/puerperal sepsis and betamethasone probably have little or no effect against control: OR 1.93 (95%CI 0.8534.41) and 0.94 (95%CI 0.47-1.87), respectively (Test for subgroup differences suggest disparities; P: 0.16, I² 49.8%).
There was no report of direct evidence regarding this outcome. Indirect evidence suggest that compared with betamethasone, dexamethasone may increase endometritis/puerperal sepsis, but in one side of the CrI is protective: OR 2.04 (95%CrI 0.72-6.06), low-CE .
-Neonatal Death (8697 patients, 23 studies)
Both dexamethasone and betamethasone reduce neonatal death against control: OR 0.60 (95%CI 0.37-0.94) and OR 0.57 (95%CI 0.39-0.80), respectively (Test for subgroup differences shows no disparity; P: 0.81, I² 0%). We found no incoherence; ROR 1.15 (95%CI 0.44-2.96), therefore we considered the NMA evidence the most reliable estimation. Compared with betamethasone, dexamethasone probably has no effect on neonatal death, but the CrI is compatible with beneficial or detrimental effect:OR 1.05 (95%CrI 0.62-1.84), moderate-CE .
-Foetal death (3857 patients, 13 studies)
Dexamethasone may reduce foetal death and betamethasone probably have little or no effect against control: OR 0.86 (95%CI 0.32-2.16) and 1.05 (95%CI 0.58-2.15), respectively (Test for subgroup differences shows no disparity; P: 0.70, I² 0%). There was no report of direct evidence regarding this outcome. Indirect evidence suggest that compared with betamethasone, dexamethasone may reduce foetal death, but the CrI limits is compatible with large beneficial or detrimental effect: OR 0.81 (95%CrI 0.24-2.41), low-CE .
-Respiratory distress Syndrome (9784 patients, 30 studies)
Both dexamethasone and betamethasone may reduce neonatal death against control: OR 0.64 (95%CI 0.47-0.90) and 0.47 (95%CI 0.35-0.60), respectively (Test for subgroup differences suggest disparities; P: 0.11, I² 54.7%).
We found no serious incoherence; ROR 1.14 (95%CI 0.71-2.75), therefore we considered the NMA evidence the most reliable estimation. Compared with betamethasone, dexamethasone probably increases RDS but the CrI is compatible with a small protective effect: OR 1.38 (95%CrI 0.96-2.11), moderate-CE .
- Neurodevelopmental disability (2628 patients, 3 studies)
We did not find direct evidence for betamethasone vs. placebo. Dexamethasone may reduce neurodevelopmental disability against control: OR 0.39 (95%CI 0.01-8.08).
Compared with betamethasone, dexamethasone probably has no effect on neurodevelopmental disability, but the CrI is compatible with large beneficial or detrimental effect: OR 1.14 (95%CrI 0.24-13.86). Two of the included studies had rare events. The frequentist analysis suggested more precise and reliable estimation an OR 1.03 (95%CI 0.80-1.33), moderate-CE .
-Intraventricular haemorrhage (IVH) (7449 patients, 17 studies)
Both dexamethasone and betamethasone reduce IVH: OR 0.473 (95%CI 0.281-0.738) and 0.381 (95%CI 0.191-0.668), respectively (Test for subgroup differences shows no disparity; P: 0.88, I² 0%).
We found no serious incoherence; ROR 1.54 (95%CI 0.57-4.16). The NMA evidence suggested that compared with betamethasone, dexamethasone may reduce IVH but the CrI is compatible with beneficial or detrimental effect: OR 0.812 (95%CrI 0.420-1.427), low-CE. However, we found mild heterogeneity (I2 31%) and important subgroup differences by corticosteroid type (I² 63.5%).
Considering the very high risk of attrition bias (43% of non-analysed infants), the unique marked effect favouring dexamethasone of the study Elimian 200746 and the incoherence that this study had generated, we performed a post-hoc sensitivity analysis excluding it. Both the heterogeneity and subgroup differences changed to an I² of 0%. The new estimation, stilllow-CE , OR 1.04 (95%CrI 0.56-1.78) was more consistent with the indirect evidence (ROR 1.14 (95%CI 0.51-2.57) and therefore we considered it as the most reliable estimation (forest-plots in Appendix S5 ).
-Mean birthweight (8645 patients, 23 studies)
Both dexamethasone and betamethasone have no effect on birthweight against control: MD -17.04gr (95%CI -75.48; 41.41) and -9.74 gr. (95%CI -43.11; 23.63), respectively (Test for subgroup differences shows no disparity; P: 0.80, I² 0%). We found no serious incoherence; ROR 1.15 (95%CI 0.44-2.96) and both direct and indirect evidence were considered as high certainty evidence, therefore we considered the NMA evidence the most reliable estimation. Dexamethasone has no effect mean birthweight:mean difference + 5.29gr (95%CrI -49.79, 58.97) high-CE .