aOne article was a follow-up study reporting neurodevelopmental outcomes of a trial (SUPPORT 201062) included in this systematic review.
Each node indicates a ventilation strategy and is sized proportional to the number of infants who received the ventilation strategy. Each line connecting 2 nodes indicates a direct comparison between 2 strategies, and the thickness of each is proportional to the number of trials directly comparing the 2 strategies.
aThe total numbers of trials and infants were not the same as the sum of the numbers of infants and trials in pairwise comparisons because one 3-group trial was included. See Table 1 footnotes for abbreviation expansions.
Network absolute risk difference (RD) was calculated from the network odds ratio (OR) estimates with an assumption that an assumed control risk was the average risk in a control group in the network. The assumed control risk values are presented in eTable 6 in the Supplement. See Table 1 footnotes for abbreviation expansions.
Each line indicates a ventilation strategy. The horizontal x-axis represents the ranking of strategies in which the first through sixth strategies are ranked in numerical order, with the first representing the best strategy. The vertical y-axis represents the probability of each ranking. CI indicates credible interval. The surface under the cumulative ranking curve is a simple summary index with values ranging between 0 (certainly the worst intervention) and 1 (certainly the best intervention).32 See Table 1 footnotes for abbreviation expansions.
eText. Methods of quality-of-evidence assessment for indirect and network estimates
eTable 1. Search strategies and results
eTable 2. Changes of the final review protocol from the original one in PROSPERO and reasons
eTable 3. Information provided by authors of included trials
eTable 4. Risk of bias of included trials
eTable 5. GRADE evidence profile table for direct comparisons
eTable 6. Network meta-analysis for the primary and main secondary outcomes and quality-of-evidence assessment
eTable 7. Network meta-analyses for additional secondary outcomes
eTable 8. Sensitivity analyses excluding trials with high risk of bias
eTable 9. Sensitivity analyses combining LISA and INSURE as LISA/INSURE
eTable 10. Subgroup analyses for the primary outcome
eTable 11. Co-interventions for respiratory management
eFigure 1. Concept of first and second order loops and quality-of-evidence assessment of indirect estimates
eFigure 2. Outcome data for individual trials
Isayama T, Iwami H, McDonald S, Beyene J. Association of Noninvasive Ventilation Strategies With Mortality and Bronchopulmonary Dysplasia Among Preterm InfantsA Systematic Review and Meta-analysis. JAMA. 2016;316(6):611-624. doi:10.1001/jama.2016.10708
Various noninvasive ventilation strategies are used to prevent bronchopulmonary dysplasia (BPD)of preterm infants; however, the best mode is uncertain.
To compare 7 ventilation strategies for preterm infants including nasal continuous positive airway pressure (CPAP) alone, intubation and surfactant administration followed by immediate extubation (INSURE), less invasive surfactant administration (LISA), noninvasive intermittent positive pressure ventilation, nebulized surfactant administration, surfactant administration via laryngeal mask airway, and mechanical ventilation.
MEDLINE, EMBASE, CINAHL, and Cochrane CENTRAL from their inceptions to June 2016.
Randomized clinical trials comparing ventilation strategies for infants younger than 33 weeks’ gestational age within 24 hours of birth who had not been intubated.
Data Extraction and Synthesis
Data were independently extracted by 2 reviewers and synthesized with Bayesian random-effects network meta-analyses.
Main Outcomes and Measures
A composite of death or BPD at 36 weeks’ postmenstrual age was the primary outcome. Death, BPD, severe intraventricular hemorrhage, and air leak by discharge were the main secondary outcomes.
Among 5598 infants involved in 30 trials, the incidence of the primary outcome was 33% (1665 of 4987; including 505 deaths and 1160 cases of BPD). The secondary outcomes ranged from 6% (314 of 5587) for air leak to 26% (1160 of 4455) for BPD . Compared with mechanical ventilation, LISA had a lower odds of the primary outcome (odds ratio [OR], 0.49; 95% credible interval [CrI], 0.30-0.79; absolute risk difference [RD], 164 fewer per 1000 infants; 57-253 fewer per 1000 infants; moderate quality of evidence), BPD(OR, 0.53; 95% CrI, 0.27-0.96; absolute RD, 133 fewer per 1000 infants; 95% CrI, 9-234 fewer per 1000 infants; moderate-quality), and severe intraventricular hemorrhage (OR, 0.44; 95% CrI, 0.19-0.99; absolute RD, 58 fewer per 1000 births; 95% CrI, 1-86 fewer per 1000 births; moderate-quality). Compared with nasal CPAP alone, LISA had a lower odds of the primary outcome (OR, 0.58; 95% CrI, 0.35-0.93; absolute RD, 112 fewer per 1000 births; 95% CrI, 16-190 fewer per 1000 births; moderate quality), and air leak (OR, 0.24; 95% CrI, 0.05-0.96; absolute RD, 47 fewer per 1000 births; 95% CrI, 2-59 fewer per 1000 births; very low quality). Ranking probabilities indicated that LISA was the best strategy with a surface under the cumulative ranking curve of 0.85 to 0.94, but this finding was not robust for death when limited to higher-quality evidence.
Conclusions and Relevance
Among preterm infants, the use of LISA was associated with the lowest likelihood of the composite outcome of death or BPD at 36 weeks’ postmenstrual age. These findings were limited by the overall low quality of evidence and lack of robustness in higher-quality trials.
Despite the substantial improvement in survival of preterm infants over the last 2 decades, complications of preterm birth remain among the leading contributors to loss of health in the United States.1 Bronchopulmonary dysplasia (BPD) has the highest prevalence of all the major complications of prematurity,2,3 and the incidence is increasing, as shown in 1 study,3 from 32% in 1993 to 45% in the years 2008 through 2012. Bronchopulmonary dysplasia has life-long effects on patients and health care systems due to increased risks of death,4 long-term respiratory problems, and serious neurodevelopmental impairment requiring educational and social support.5,6
Although the pathogenesis of BPD is multifactorial, ventilator-induced barotrauma and volutrauma on the premature lung are major factors.7 Hence, various noninvasive ventilation strategies have been growing in popularity. Although previous randomized clinical trials and systematic reviews reported benefits from these strategies,8- 13 clinicians remain uncertain about which strategy to choose, perhaps because conventional systematic reviews and meta-analyses have focused on head-to-head comparison of 2 interventions without assessing multiple interventions as a whole. Network meta-analyses (or multiple treatment comparison meta-analyses) provide a framework for analyzing and interpreting more than 2 interventions (network of multiple interventions) to understand the evidence of the network of multiple interventions as a whole.14 This systematic review summarizes available evidence from randomized clinical trials using a Bayesian network meta-analysis to compare multiple ventilation strategies simultaneously to identify the best strategy to prevent BPD of preterm infants.
Question What is the best noninvasive ventilation strategy for preventing death or bronchopulmonary dysplasia in the first 24 hours of life in spontaneously breathing preterm infants with or at risk of respiratory distress syndrome?
Findings In this meta-analysis, less invasive surfactant administration was the strategy associated with the lowest odds of the composite outcome of death or bronchopulmonary dysplasia compared with either nasal continuous positive airway pressure or mechanical ventilation.
Meaning Less invasive surfactant administration should be considered as a first-line ventilation strategy for spontaneously breathing preterm infants with respiratory distress syndrome.
Four electronic databases, MEDLINE, EMBASE, CINAHL, and Cochrane CENTRAL, were systematically searched from their inceptions to June 7, 2016, and supplemented by searching the World Health Organization International Clinical Trials Registry Platform15 and reference lists of eligible studies and review articles. (For search strategies, see eTable 1 in the Supplement.) The protocol of this systematic review was registered before the literature search in PROSPERO (Prospero 2015 CRD42015023403).16 Several differences in study methods between the protocol and this article are presented in eTable 2 in the Supplement.
This systematic review included randomized or quasi-randomized clinical trials fulfilling the following 4 criteria: (1) published as a full report in a peer-reviewed journal, (2) enrolled spontaneously breathing preterm infants born at less than 33 weeks’ gestational age (with or at risk of respiratory distress syndrome) who had never been intubated before randomization, which occurred within 24 hours of birth, (3) compared 2 or more of the predetermined 7 ventilation strategies, and (4) reported at least 1 event of the primary or secondary outcomes. Trials in which some of the infants were 33 weeks gestational age or older were included if the mean or median gestational age was less than 33 weeks. Studies enrolling infants within 72 hours of birth were included if a mean or median age was less than 24 hours at study entry. If subgroups of infants fulfilled all the inclusion criteria and their data were available, the studies were included.
The 7 eligible ventilation strategies were (1) nasal continuous positive airway pressure (CPAP) alone, in which infants continued nasal CPAP with surfactant selectively given only when infants met a certain criteria of nasal CPAP failure; (2) intubation and surfactant administration followed by immediate extubation (INSURE), in which infants were intubated, given surfactant, and extubated within 1 hour received nasal CPAP13; (3) less invasive surfactant administration (LISA), in which infants continued nasal CPAP without intubation and surfactant was given via thin diameter tubes or catheters (eg, feeding tubes, vascular catheters) directly placed into infants’ tracheas using laryngoscopes with or without Magill forceps17; (4) noninvasive intermittent positive pressure ventilation (NPPV), defined as any noninvasive strategy that provided intermittent increased airway pressure in addition to nasal CPAP including biphasic nasal CPAP18; (5) nebulized surfactant administration while receiving nasal CPAP; (6) surfactant administration via laryngeal mask airway, in which surfactant was given via a laryngeal mask airway as a conduit without intubation after which infants received nasal CPAP; and (7) mechanical ventilation via endotracheal tube. Because the use of positive end-expiratory pressure or nasal CPAP was considered standard respiratory management, ventilation strategies without it were excluded.19,20 No language restrictions were applied.
Nine outcomes were selected a priori and rated on a 1-to-9 scale (7-9, critical; 4-6, important; and 1-3, of limited importance) based on their importance for patients and clinicians according to a method proposed by the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) group.21 A composite outcome of death or BPD at 36 weeks’ postmenstrual age (a critical outcome rated 8) was selected as the primary outcome because BPD is the most important respiratory morbidity of preterm infants and death is a competing risk. Because several similar but slightly different definitions exist for BPD and because some studies used more than 1 definition, the order of priority for selecting a definition of BPD in each trial was decided a priori as follows, in descending order: (1) oxygen use, positive pressure support, or both at 36 weeks’ postmenstrual age; (2) oxygen use at 36 weeks’ postmenstrual age; (3) oxygen use, pressure support, or both at 36 weeks’ postmenstrual age along with oxygen use at 28 days of age; and (4) oxygen use at 36 weeks’ postmenstrual age along with oxygen use at 28 days of age.
Bronchopulmonary dysplasia at 36 weeks’ postmenstrual age (a critical outcome rated 7), death at 36 weeks’ postmenstrual age or before discharge (a critical outcome rated 9), severe intraventricular hemorrhage (grade 3 or 4 based on the Papile criteria,22 a critical outcome rated 7), and air leak including pneumothorax or pulmonary interstitial emphysema before discharge (an important outcome rated 5) were selected as the 4 main secondary outcomes because previous studies indicated that early ventilation strategies might affect these outcomes.23,24 Other secondary outcomes included necrotizing enterocolitis (stage 2 or higher based on the Bell criteria,25 a critical outcome rated 7) and severe retinopathy of prematurity (stage 3 or higher based on the international classification, or treated disease,26 a critical outcome rated 7) before discharge; neurodevelopmental impairment at 18 months or later (a critical outcome rated 8); and a composite outcome of death or neurodevelopmental impairment at 18 months or later (a critical outcome rated 8).
Two reviewers (T.I. and H.I.) independently screened all titles and abstracts identified in the literature search, reviewed full texts of eligible articles, and extracted data from the selected articles using a pretested data extraction form. The 2 reviewers independently assessed the risk of bias of each trial for each outcome including selection bias (inadequate random sequence generation, failure to conceal allocation), performance bias (inadequate blinding of patients and personnel), detection bias (failure to blind outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective outcome reporting), and other bias (publication bias, stopping early for apparent benefit, etc) according to the Cochran Handbook.27 If 1 or more components of the risk of bias assessment were judged high risk, the trial was judged high risk of bias. Any disagreement between the reviewers was resolved by discussion or consultation with the third author (S.M.).
Bayesian hierarchical random-effects network meta-analysis was conducted to compare all interventions simultaneously using Markov chain Monte Carlo simulation with noninformative prior distribution. Network meta-analyses generate direct pair-wise effects estimates (eg, A vs B) and also indirect effects estimates (eg, A vs C via B using 2 comparisons of A vs B and B vs C) to estimate network effects (or mixed effects) combining direct and indirect effects and rank the interventions, enabling selection of the best intervention.14,28,29 The method estimates relative effects of multiple interventions simultaneously whether or not they have been directly compared with each other in previous trials.14 The analyses used generalized linear models with a logit link function with 4 chains and 100 000 iterated simulations discarding the initial 5000 iterations as burn-in. Convergence was assessed using the Brooks-Gelman-Rubin statistic.30 Model fit was assessed by comparing the mean sum of residual deviance to the number of independent trial groups. Odds ratios (ORs) and 95% credible intervals (95% CrIs) were estimated from the medians and 2.5th and 97.5th percentile of the posterior distributions in the simulations, respectively. A network absolute risk difference (RD) was calculated from the network OR estimates using an assumed control risk27 that was the average risk in the control group in the network derived by dividing the total event number by the total infant number in the control groups in the network. The I2 statistic and Cochran Q test were used to assess heterogeneity of trials within each direct comparison. Node-splitting was used to assess incoherence between direct and indirect comparisons.31
Rank probabilities that interventions were the best, second best, third best, etc were calculated from proportions of Markov chain cycles in which the interventions had the lowest, second lowest, third lowest odds ratios (ORs), respectively. Surface under the cumulative ranking curve (SUCRA) for each intervention was calculated from a cumulative ranking probability that an intervention is above a certain ranking.32 SUCRA is a simple summary index indicating the degree to which an intervention is better or worse than others, taking a value between 0 (certainly the worst intervention) and 1 (certainly the best intervention).32 All the analyses were conducted using R version 3.1.2 (R Project for Statistical Computing) with R packages (gemtc, metafor, and rjags), and JAGS version 3.4.0.
The quality of evidence of each direct, indirect, and network effects estimate was evaluated for the primary and main secondary outcomes according to the GRADE method.33,34 The quality of evidence of direct-effects estimates started as high and was decreased to moderate, low, or very low based on risk of bias, imprecision, heterogeneity, indirectness, and publication bias.33 For the assessment of precision, a sample size required to detect a 25% relative risk reduction, called optimal information size, was calculated for each comparison for each outcome based on a total event rate in the control group35 using the PS Power and Sample Size Calculation software version 3.0.36 Publication bias was assessed by inspecting asymmetry of funnel plots visually. The quality of evidence of indirect and network effects estimates were derived from those of direct- effects estimates by evaluating network geometry, intransitivity, and incoherence (for details, see the eText and eFigure 1 in the Supplement).
For the primary and main secondary outcomes, 2 preplanned sensitivity analyses were conducted: excluding trials with high risk of bias and combining INSURE and LISA as 1 strategy.
Four preplanned subgroup analyses were conducted for the primary outcome, stratifying by potential effects modifiers including mean (or median) gestational age at birth (≤28 or >28 weeks), timing of interventions (≤1 or >1 hour after birth), thresholds of fraction of inspired oxygen (≤40% or >40%), and backup measures (mechanical ventilation or INSURE and LISA) for treatment failure of noninvasive ventilation strategies. Between-subgroup differences in effects estimates were assessed by the Z test.
Among 6082 records identified in the literature search, 31 articles of 30 trials met inclusion criteria and involved 5598 infants8- 10,18,37- 63 (Figure 1). One article62 was a follow-up study of an included trial.8 Sample sizes ranged from 24 to 1316 infants; mean or median gestational age at birth, 25 to 32 weeks; mean or median timing of enrollment, less than 24 hours; antenatal corticosteroid exposure, 15% to 99%; and threshold for backup measures in the nonventilation group of fraction of inspired oxygen (Fio2; 0.3 to 1.0 (Table). Except for one 3-group trial,37 all others were 2-group trials. Noninvasive positive-pressure ventilation was only compared with nasal CPAP alone. There was only 1 eligible trial for nebulized surfactant administration58 and surfactant administration via laryngeal mask airway.63 The authors of 17 trials provided additional information or data for this systematic review (eTable 3 in the Supplement).9,18,38,42,50- 53,56,57,59- 61 The authors of 5 trials provided data on a subset of infants who were eligible for the review.18,51,53,57,63
Seven trials did not report or conduct allocation concealment and were considered at high risk of bias.39,40,54,55,57,60,61 Three trials were stopped early due to significant findings in interim analyses. Because early stopping may overestimate intervention effects, the trials were considered high risk of bias (eTable 4 in the Supplement).38,42,43 High risk of attrition bias (missing data >10%) was found in 1 trial for both the primary outcome and for BPD,48 3 trials for severe intraventricular hemorrhage,43,48,53 and 2 trials for retinopathy of prematurity.8,53 Among 30 trials included, 16 trials were at low risk of bias for all outcomes assessed,9,10,18,37,41,44- 47,50- 52,56,58,59,63 and 3 trials were at low risk of bias for some outcomes (eTable 4 in the Supplement).8,48,53
A total of 21 trials including 4987 infants reported the primary composite outcome of death or BPD (Figure 2). The incidence of the primary outcome was 33% (1665 of 4987 infants) with 505 infant deaths and 1160 infants with BPD. LISA was associated with a lower likelihood of the primary outcome than was mechanical ventilation (OR, 0.49; 95% CrI; 0.30-0.79; absolute RD, 164 fewer per 1000 infants; 95% CrI, 57-253 fewer per 1000 infants; moderate quality of evidence) and nasal CPAP alone (OR, 0.58; 95% CrI, 0.35-0.93; absolute RD, 112 fewer per 1000 infants; 95% CrI, 16-190 fewer per 1000 infants; moderate quality of evidence) (Figure 3). INSURE was associated with a lower likelihood of the primary outcome than was mechanical ventilation (OR, 0.71; 95% CrI, 0.50-0.98; absolute RD, 83 fewer per 1000 infants; 95% CrI, 5-160 fewer per 1000 infants]; moderate quality-of-evidence). The individual trial-level outcome data are in eFigure 2 in the Supplement.
The network meta-analyses for the 4 main secondary outcomes included 19 to 30 trials involving 4455 to 5587 infants (Figure 2). The incidence of BPD was 26% (1160 of 4455); death, 10% (542 of 5368 eFigure 2 in the Supplement); severe intraventricular hemorrhage, 8% (389 of 4796); and air leak, 6% (314 of 5587). LISA was associated with a lower likelihood of BPD and severe intraventricular hemorrhage than was mechanical ventilation (moderate quality-of-evidence) (Figure 4). LISA and INSURE were associated with a lower likelihood of air leak than was nasal CPAP alone (very low-quality of evidence). There were no other significant differences between interventions in the likelihood of the main secondary outcomes (Figure 3 and Figure 4) or other secondary outcomes (eTable 7 in the Supplement).
For the main outcomes, LISA had the highest probability of being the best strategy for supporting respiration in preterm infants with or at risk of respiratory distress syndrome (Figure 5). Additionally, LISA was the best strategy for all outcomes based on SUCRA (SUCRA, 0.85-0.94). INSURE was the second best strategy for the primary outcome (tied with NPPV), BPD, air leak, and severe intraventricular hemorrhage (SUCRA, 0.63-0.81).
Among a total of 34 direct comparisons for the primary and main secondary outcomes, the quality of evidence was down rated for serious risk of bias in 10 comparisons, for serious heterogeneity in 10 comparisons, and for serious or very serious imprecision in all 34 comparisons (eTable 5 in the Supplement). The sample size of the meta-analyses did not reach the optimal information size in most direct comparisons (31 of 34) resulting in down rating due to imprecision (eTable 5 and eTable 6 in the Supplement). Node splitting found no significant incoherence in comparisons for any outcomes (eTable 6 in the Supplement). The inspection of effects modifiers found potential intransitivity in 3 comparisons (NPPV vs mechanical ventilation for death, LISA vs INSURE for severe intraventricular hemorrhage, and INSURE vs mechanical ventilation for air leak), and the quality of evidence for their indirect effects estimates was down rated (eTable 6 in the Supplement). Based on these results, the quality of evidence for network effects estimates was judged as moderate in 7, low in 26, and very low in 38 comparisons (Figure 3, eTable 6 in the Supplement).
Excluding studies with high risk of bias, the lower odds of the primary outcome and severe intraventricular hemorrhage in LISA compared with mechanical ventilation remained significant (eTable 8 in the Supplement). The other significant findings in the primary analyses became nonsignificant (eTable 8 in the Supplement). As in the primary analysis, LISA had the highest probability of being the best strategy and had the highest SUCRA among all strategies for all the main outcomes except for death (eTable 8F in the Supplement), for which NPPV had the highest SUCRA.
LISA and INSURE together were associated with lower odds of the primary outcome and BPD than was mechanical ventilation and lower odds of air leak than nasal CPAP alone (eTable 9 in the Supplement). LISA and INSURE had the highest probability of being the best strategy and had the highest SUCRA among all strategies for all the main outcomes except for death (eTable 9F in the Supplement).
The 4 preplanned subgroup analyses did not find any significant differences between subgroups for the primary outcome (eTable 10 in the Supplement).
This network meta-analysis including 30 trials with 5598 nonventilated spontaneously breathing preterm infants with or at high risk of respiratory distress syndrome simultaneously estimated relative effects of 7 currently used noninvasive or invasive ventilation strategies. The use of LISA was associated with a lower likelihood of the primary outcome of death or BPD and secondary outcomes of BPD and severe intraventricular hemorrhage than mechanical ventilation and lower likelihood of the primary outcome and air leak than nasal CPAP alone. INSURE was associated with a lower likelihood of the primary outcome than mechanical ventilation and lower likelihood of air leak than nasal CPAP alone. Ranking probabilities supported that LISA was the best strategy among all strategies for all outcomes assessed. INSURE was the second best strategy to prevent the primary outcome (tied with NPPV), BPD, air leak, and severe intraventricular hemorrhage. Although significant findings for the primary outcome had moderate quality of evidence, the evidence for the secondary outcomes was, overall, of low quality. When limited to high-quality trials, the best strategy remained LISA for all main outcomes except for death, for which NPPV was the best strategy.
Several systematic reviews have evaluated various noninvasive ventilation strategies using conventional pair-wise comparisons. Three systematic reviews reported that early nasal CPAP use, avoiding intubation, reduced the composite outcome of death or BPD compared with early intubation with or without early surfactant administration.11,12,23 Because these systematic reviews did not differentiate between INSURE (or LISA) and mechanical ventilation in their intubation groups,11,12,23 they did not address the important question of whether early surfactant administration with INSURE or LISA, avoiding prolonged mechanical ventilation, is more effective than nasal CPAP alone.
Among noninvasive strategies with early surfactant administration, INSURE has been most intensively investigated. Since Verder and colleagues42,43 originally reported 2 randomized clinical trials during the 1990s, several other trials evaluated the efficacy of INSURE. A recent systematic review found no significant differences in efficacy or safety between early INSURE and nasal CPAP alone in preterm infants.13 LISA has been developed as an alternative to INSURE, with potential benefits including maintenance of spontaneous breathing of infants while receiving nasal CPAP during the procedure,17,64 complete avoidance of intermittent positive-pressure ventilation via endotracheal tubes,17,64 and reduction of traumatic airway injuries caused by intubation with semi-rigid endotracheal tubes.65 A previous systematic review64 evaluating LISA that included 4 observational studies and 2 randomized clinical trials reported that all 6 studies demonstrated that LISA reduced the need for mechanical ventilation, and 1 study reported a reduction in BPD incidence.59 The study however did not conduct a meta-analysis of the data.
To our knowledge, our study is the first systematic review conducting meta-analyses to evaluate the outcomes and adverse effects associated with LISA. Other modes of surfactant administration, via either laryngeal mask airway or nebulizer, have suggested promise.64 However, there was only 1 eligible small trial for each of these strategies, and their effectiveness has yet to be adequately assessed.
Noninvasive positive-pressure ventilation is another noninvasive alternative to nasal CPAP alone. A previous systematic review comparing NPPV and nasal CPAP in ventilated and nonventilated preterm infants66 found a reduction of the need for mechanical ventilation in the NPPV group but no differences in the rates of BPD and other major outcomes. This review focused on infants who had never been intubated before study entry, which is essential to evaluate the effect of avoiding ventilator-induced lung injury. Although this systematic review found no significant differences in outcomes between NPPV and other strategies, sample sizes did not reach the optimal information size and NPPV was only compared with nasal CPAP alone.
The overall inferiority of mechanical ventilation to other strategies found in this systematic review suggests that routine use of this strategy should not be recommended. However, because previous studies reported that substantial proportions of infants initially managed with LISA,67 INSURE,68 or nasal CPAP alone69 required mechanical ventilation later, it may be reasonable to use mechanical ventilation strategies if noninvasive strategies are expected to fail.68,69
The American Academy of Pediatrics recommends early nasal CPAP as an alternative to routine intubation and surfactant administration.19 The European consensus guidelines recommend early routine use of nasal CPAP along with early rescue surfactant administration for infants with respiratory distress syndrome.20 Both the American and European guidelines recommend the use of INSURE for rescue surfactant administration if infants seem to tolerate immediate extubation; however, neither include a recommendation for the use of LISA.19,20 The lowest likelihood of adverse outcomes associated with LISA found in this systematic review could inform future updates of these clinical guidelines.
This study has several strengths, especially the use of network meta-analysis to enable comparisons among currently used respiratory strategies, while increasing statistical power by taking advantage of indirect network pathways. This systematic review used robust methods, guided by the Cochrane Handbook27 and the GRADE approach for network meta-analyses.34 The Bayesian statistical methods provided ranking probabilities and allowed comparison of all the strategies simultaneously using SUCRA. Two sensitivity analyses assessed the robustness of the study findings. The authors of 17 original articles provided data to help assess the study designs, reduce missing data, clarify outcome definitions, and enable inclusion of clinically important subgroups.
This study has some limitations. Although this systematic review is the largest yet performed, most of the direct comparisons had smaller sample sizes than the optimal information sizes. The wide 95% CrIs of the network ORs in many comparisons, especially those including NPPV and surfactant administration via nebulizers or laryngeal airway mask, indicate that further trials are needed to obtain more precise effects estimates. There were some differences in baseline characteristics of included trials that could lead to biased results.34 This issue was incorporated in the quality-of-evidence assessment by evaluating the I2 statistic and Cochrane Q test for heterogeneity within each direct comparison, inspecting differences in effects modifiers for transitivity between comparisons, and conducting node splitting to assess incoherence between direct and indirect comparisons.34 Furthermore, differences in the infants’ baseline characteristics indicate variations in severity of respiratory distress syndrome in the included trials. Because the severity of respiratory distress was reported to predict nasal CPAP failure,69 the strategies with early surfactant administration (LISA, INSURE, and surfactant administration via nebulizer or laryngeal mask airway) may be more effective for those with severe respiratory distress. The preplanned subgroup analyses assessed this possibility; however, the small sample sizes in subgroups limited the assessment. Therefore, which infants need early surfactant administration via LISA or INSURE is yet to be addressed. Because the lack or delay of surfactant administration is a drawback of the nasal CPAP-alone and NPPV strategies, it is possible that the nasal CPAP alone and NPPV with early, appropriate but selective surfactant administration may be as or more effective than LISA and INSURE. Also, there are potential cointerventions that may affect primary ventilation strategies, such as premedications before LISA, INSURE, or surfactant administration via laryngeal mask airway (eg, atropine, sedatives)17 and prophylaxis with methylxanthines (eg, caffeine) to prevent apnea.70 Because many of the included trials did not report cointerventions (eTable 11 in the Supplement), future trials should describe them. When limited to high-quality trials, some of the study findings changed. This study incorporated the fragility of these results by rating down their quality of evidence.
Based on the best available evidence, this systematic review found that early surfactant administration via LISA was the best management strategy, and INSURE likely the second best, for nonventilated spontaneously breathing preterm infants with or at high risk of respiratory distress syndrome, along with early nasal CPAP application. However, when limited to high-quality evidence, some significant findings for LISA compared with other strategies became nonsignificant, and the lower likelihood of death associated with LISA was not robust. Therefore, to confirm the overall lower likelihood of the primary and secondary outcomes associated with LISA found in this systematic review, further well-designed trials with large sample sizes comparing LISA with nasal CPAP alone are warranted, and some are currently under way.71 Because LISA and INSURE are similar procedures, INSURE can be an alternative to LISA, especially for clinicians not familiar with LISA.
Among preterm infants, the use of LISA was associated with the lowest likelihood of the composite outcome of death or BPD at 36 weeks’ postmenstrual age. These findings were limited by the overall low quality of evidence and lack of robustness in higher quality trials.
Corresponding Author: Tetsuya Isayama, MD, MSc, Clinical Epidemiology & Biostatistics, McMaster University, Room HSC 2C, Hamilton, ON, Canada L8S 4K1 ON (email@example.com).
Correction: This article was corrected September 13, 2016, to change the corresponding author’s address.
Author Contributions: Dr Isayama had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Isayama, McDonald, Beyene.
Acquisition, analysis, or interpretation of data: Isayama, Iwami, Beyene.
Drafting of the manuscript: Isayama.
Critical revision of the manuscript for important intellectual content: All Authors.
Statistical analysis: Isayama, Beyene.
Study supervision: McDonald, Beyene.
No additional contributions: Iwami.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.
Previous Presentation: Presented as an abstract at the Pediatric Academic Societies 2016 Annual Meeting, April 30, 2016, Baltimore, Maryland..
Additional Contributions: We thank the authors of 17 included trials for providing additional study information for this systematic review. None of the authors received financial compensation for providing the information for this study: (Haresh Kirpalani, MSc [Children's Hospital of Philadelphia, University Pennsylvania, Philadelphia] for the NIPPV study group,18 Jiajun Zhu, PhD [Women's Hospital, School of Medicine, Zhejiang University, Zhejiang, China],62 Sourabh Dutta, PhD [The Postgraduate Institute of Medical Education and Research, Chandigarh, India],52 Gianluca Lista, MD [V. Buzzi Children's Hospital, Milan, Italy],53 Jucille Meneses, PhD [Instituto de Medicina Integral Professor Fernando Figueira, Recife, Brazil],54 Gozde Kanmaz, MD [Zekai Tahir Burak Maternity Teaching Hospital, Ankara, Turkey],60 Kayavan Mirnia, MD [Tabriz Medical University, Alzahra Teaching Hospital, Tabriz, Iran],61 Mohammad Heidarzadeh, MD [Tabriz Medical University, Alzahra Teaching Hospital, Tabriz, Iran],61 Ghassan Salama, MD [Jordanian Royal Medical Services, Amman, Jordan],58 Carlo Dani, MD [Careggi University Hospital of Florence, Florence, Italy],39 Amir Kugelman, MD [Bnai Zion Medical Center, Haifa, Israel],51 Michael Dunn, MD [Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada],38 Hemasree Kandraju, MD [Fernandez Hospital, Hyderabad, Telangana, India],48 Srinivas Murki, DM [Fernandez Hospital, Hyderabad, Telangana, India],48 Carl D'Angio, MD [Golisano Children's Hospital, University of Rochester School of Medicine and Dentistry, Rochester, New York],45 Henrik Axel Verder, DMSc [Holbaek University Hospital, Region Zealand, Denmark],43 Mehmet Yekta Oncel, MD [Zekai Tahir Burak Maternity Teaching Hospital, Ankara, Turkey],9 Teresa Aguiar, MD [Hospital Prof Doutor Fernando Fonseca, Amadora, Portugal],57 Joaquim M. B. Pinheiro, MPH [Albany Medical College, Albany, New York]).64 We also thank Russell de Souza, ScD (McMaster University, Hamilton, Ontario, Canada) for reviewing the study protocol and manuscript and providing comments; Yue Yao, MD (Hospital for Sick Children, Toronto, Ontario, Canada) for translating the Chinese articles. Neither Dr de Souza nor Yao received financial compensation for their contribution to this manuscript.