Thresholds for surfactant use in preterm neonates: a network meta-analysis

Objective To perform a network meta-analysis of randomised controlled trials of different surfactant treatment strategies for respiratory distress syndrome (RDS) to assess if a certain fraction of inspired oxygen (FiO2) is optimal for selective surfactant therapy. Design Systematic review and network meta-analysis using Bayesian analysis of randomised trials of prophylactic versus selective surfactant for RDS. Setting Cochrane Central Register of Controlled Trials, MEDLINE, Embase and Science Citation Index Expanded. Patients Randomised trials including infants under 32 weeks of gestational age. Interventions Intratracheal surfactant, irrespective of type or dose. Main outcome measures Our primary outcome was neonatal mortality, compared between groups treated with selective surfactant therapy at different thresholds of FiO2. Secondary outcomes included respiratory morbidity and major complications of prematurity. Results Of 4643 identified references, 14 studies involving 5298 participants were included. We found no statistically significant differences between 30%, 40% and 50% FiO2 thresholds. A sensitivity analysis of infants treated in the era of high antenatal steroid use and nasal continuous positive airway pressure as initial mode of respiratory support showed no difference in mortality, RDS or intraventricular haemorrhage alone but suggested an increase in the combined outcome of major morbidities in the 60% threshold. Conclusion Our results do not show a clear benefit of surfactant treatment at any threshold of FiO2. The 60% threshold was suggestive of increased morbidity. There was no advantage seen with prophylactic treatment. Randomised trials of different thresholds for surfactant delivery are urgently needed to guide clinicians and provide robust evidence. PROSPERO registration number CRD42020166620.

1. infant newborn.mp. or exp newborn/ 2. extremely low birth weight.mp. or exp low birth weight/ or exp very low birth weight/ or exp extremely low birth weight/ or exp newborn/ or exp prematurity/ 3. extremely-low-birth-weight.mp. 4. (extremely and low and birth and weight).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word] 5. very low birth weight.mp. or exp very low birth weight/ 6. very-low-birth-weight.mp. 7. (very and low and birth and weight).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word] 8. newborn infant.mp. 9. neonate.mp. 10. premature.mp. 11. exp premature labor/ or preterm.mp. or exp gestational age/ 12. elbw.mp. 13. vlbw.mp. 14. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 15 19 21. (random* or factorial* or crossover* or cross over* or cross-over* or placebo* or doubl* blind* or singl* blind* or assign* or allocat* or volunteer*).af. 22. exp crossover-procedure/ or exp double-blind procedure/ or exp randomized controlled trial/ or exp single-blind procedure/ 23. 21 or 22 24. 14 and 20 and 23 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Number of participants with events for binary outcomes, mean and standard deviation for continuous outcomes, number of events and the mean follow-up period for count outcomes and number of participants with events and the mean follow-up period for time-to-event outcomes o Natural logarithm of hazard ratio and its standard error if this was reported rather than the number of participants with events and the mean follow-up period for time-to-event outcomes We collected data at maximum follow-up provided and also at shorter (up to three months) and medium-term follow-up (three months to 1 year) where applicable. We attempted to contact trial authors in the case of unclear or missing information. Any differences in opinion were resolved by discussion.

eMethods -Data Synthesis
A network meta-analysis was conducted to compare thresholds of FiO2 simultaneously for each of the primary and secondary outcomes. Network meta-analysis combines direct evidence within trials and indirect evidence across trials [1]. Our analysis was based on guidance by the National Institute for Clinical Excellence (NICE) Decision Support Unit (DSU). [1][2][3][4] We obtained a network plot to ensure that the trials were connected by interventions [3]. We excluded any trials unconnected to the network from the meta-analysis and reported only the direct pair-wise meta-analysis for such comparisons. We conducted a Bayesian network meta-analysis using the Markov chain Monte Carlo method. We used a fixed-effect model and random-effects model for the network meta-analysis. For each pair-wise comparison in a table, we reported the fixed-effect model if the two models reported similar results; otherwise, we reported the more conservative model. We used a hierarchical Bayesian model using three different initial values, employing codes provided by NICE DSU [5]. We used a normal distribution with large variance (10,000) for treatment effect priors (vague or flat priors). For the random-effects model, we used a prior distributed uniformly (limits: 0 to 5) for between-trial standard deviation but assumed the same between-trial standard deviation across treatment comparisons [5]. We used a 'burn-in' of 10,000 simulations, checked for convergence (of effect estimates and between-study heterogeneity) visually (i.e. whether the values in different chains mix very well by visualisation) and ran the models for another 10,000 simulations to obtain effect estimates. If we did not obtain convergence, we increased the number of simulations for the 'burn-in'.
We estimated the probability that each intervention ranks at one of the possible positions using the NICE DSU codes [5]. Analysis was carried out using OpenBUGS, version 3.2. 3 We assessed inconsistency (statistical evidence of the violation of transitivity assumption) by fitting both an inconsistency model and a consistency model. We used inconsistency models employed in the NICE DSU manual, as we used a common between-study standard deviation [2]. In the presence of inconsistency, we assessed whether the inconsistency was due to clinical or methodological heterogeneity. We performed the direct comparisons using the same codes and the same technical details Subgroup/sensitivity analysis: Subgroup analysis was planned based on 1) trials at low risk of bias compared to trials at high risk of bias, 2) gestational age, 3) Current best practiceuse of antenatal steroids and NCPAP. Due to a paucity of data these could not be carried out. A sensitivity analysis of current best practice was performed. No trials reported only per-protocol analysis results, therefore no best-worst case scenario/worst-best case scenario analyses as sensitivity analyses were required. No imputations were required for mean or standard deviation, therefore sensitivity analysis excluding same was not required.

eResults -Excluded Studies
None of the excluded studies met the inclusion criteria. 5 of the studies were identified as review articles or systematic reviews [6][7][8][9][10]. 1 study is an ongoing trial assessing surfactant thresholds for treatment [11]. We were unable to translate 2 studies and the abstracts did not provide sufficient information for inclusion [12,13]. 23 were not randomised control trials . 6 trials met the inclusion criteria but did not list an fio2 for treatment with selective surfactant [37][38][39][40][41][42]. 55 did not meet the inclusion criteria of a trial assessing prophylactic treatment with surfactant vs selective treatment with surfactant . 10 of the references are trial register or published abstracts of an included trial: [97-106]. 3 references were abstracts without a published trial found despite attempts to contact the author [107][108][109].
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

eResults -Primary Outcome Mortality
A random-effect model was used for the network meta-analysis because it was more conservative. Deviance Information Criteria (DIC) for fixed model was 171.1, random 172.3. Median between-study standard deviation for the random-effect model 0.23 (95% CrI 0.011, 0.742), variance 0.055. Model used for direct comparisons are included in Table 1

eResults -Sensitivity Analysis of Current Best Practice
Six studies met the criteria. This included 2554 patients. 1268 were in the combined prophylaxis arm and were compared with 138 (one study) in the 30% threshold arm, 183 (2 studies) in the 40% arm, 727 (two studies) in the 50% arm and 216 (one study) in the 60% arm. eTable 4 shows the odds ratio for each comparison within the analysis, along with the model of comparison used. Most conservative model was used in each case. Fixed-effects model was used for all outcomes, except pneumothorax, where random-effects model was used. DIC, between-study variance with 95% CrI and variance where applicable are shown in eTable 5. There was no statistically significant difference seen in mortality, BPD, pneumothorax, or grade 3/4 IVH. There was an increased rate of major morbidity in the 60% threshold group-31 more per 1000 (95% CrI intervals 136 more to 572 more). Each comparison was deemed to be at very-low quality of evidence.