Efficacy of noninvasive respiratory support modes for primary respiratory support in preterm neonates with respiratory distress syndrome: Systematic review and network meta‐analysis

To compare the efficacy of different noninvasive respiratory support (NRS) modes for primary respiratory support of preterm infants with respiratory distress syndrome (RDS).


| INTRODUCTION
The introduction of surfactant had a major impact on improving the outcomes of preterm neonates with respiratory distress syndrome (RDS). 1 There was a major shift in the practice of surfactant therapy in the last decade with studies showing better outcomes with early selective rescue treatment when compared to the previously practiced prophylactic administration. 2 Stabilising neonates with RDS on a noninvasive respiratory support (NRS) such as continuous positive airway pressure (CPAP) and then instituting surfactant therapy in selective neonates who have an increased oxygen requirement has become the standard practice. 3 Newer modalities of NRS strategies that have come into practice in neonatal medicine in the past two decades, include heated and humidified high flow cannula (HFNC), noninvasive positive pressure ventilation (NIPPV), bilevel CPAP (Bi-PAP) as well as nasal high-frequency oscillation ventilation (nHFOV). 4,5 Several systematic reviews compared different NRS strategies in pair-wise meta-analysis, however, only one network meta-analysis (NMA) evaluated different NRS strategies in preterm neonates with RDS. [6][7][8][9][10] The NMA by Isamaya et al 10  and mechanical ventilation [MV] following surfactant) along with CPAP and NIPPV. 10 In this systematic review, we critically review the different modes of NRS and compare their effects in an NMA.

| METHODS
The efficacy and safety of four NRS modalities used as primary respiratory support in preterm neonates with RDS were compared: HFNC, CPAP, BiPAP, and NIPPV. The systematic review protocol was registered with PROSPERO (CRD42020177474). 11 The reporting of this review is consistent with the PRISMA for network metaanalyses guidelines. 12

| Outcomes
The primary outcomes were: (a) requirement of invasive MV within the first 7 days of randomization; (b) treatment failure, defined as requirement of an additional form of respiratory support for various reasons such as respiratory acidosis, hypoxemia, or severe apnea within the first 7 days of randomization. The secondary outcomes included incidence of mortality (neonatal and before discharge), incidence of bronchopulmonary dysplasia (BPD) defined as oxygen requirement at 36 weeks of postmenstrual age, incidence of mortality or BPD, incidence of air leak, incidence of severe intraventricular haemorrhage defined as grade more than 2, 13 incidence of necrotising enterocolitis (NEC) stage ≥2, 14 incidence of patent ductus arteriosus requiring medical therapy or surgical intervention, incidence of severe retinopathy of prematurity defined as those requiring laser therapy and or intra-vitreal antivascular endothelial growth factor and/or stage ≥3 as per International committee for the classification of retinopathy of prematurity 15 and incidence of nasal injury. | 2941 the two review authors for further evaluation. In case of any conflicts, a third author's (KM) opinion was sought.

| Assessment of risk of bias in included studies
The Cochrane risk of bias tool version 1 was used to assess the risk of bias of the included studies by two review authors independently (VVR and KM). 16 The risk of bias was evaluated based on the following five domains-selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias. Any disagreement between the reviewers was resolved by discussion or consultation with the third author (PBH).

| Data synthesis
The characteristics of the included studies were tabulated and reviewed to exclude those studies that might result in intransitivity. NMA was done by the Bayesian approach using a random-effects model with Markov chain Monte Carlo simulation with vague priors (GEMTC, BUGSnet) using the R-software (Version-R 3.6.2). 17,18 Generalized linear models with four chains, burn-in of 50 000 iterations followed by 100 000 iterations with 10 000 adaptations were used. 18 The geometry of the networks was assessed using network plots with the size of the nodes being proportional to the number of subjects included in the intervention and the thickness of the arms connecting the different intervention nodes corresponding to the number of studies included in the comparison. Model convergence was assessed using Gelman-Rubin plots as well as by analyzing the trace and density plots. 19 Inconsistency was assessed by node-splitting. 20 Pair-wise meta-analysis evaluating the direct evidence for the different NIV modalities was also done and heterogeneity was assessed using I 2 statistic and Cochran Q test. The results of the NMA were expressed as risk ratios (RR) with 95% credible intervals (CrIs) in league matrix tables and forest plots. The league matrix tables display the RR of the outcome parameter for the intervention in the row vs that in the column in the lower triangle and vice versa in the upper triangle. The comparison of direct and indirect evidence using node-splitting are expressed as odds ratios with 95% CrIs. Surface under the cumulative ranking curve (SUCRA) was used to rank interventions for all the outcomes. SUCRA is an index with values from 0 (least effective intervention) to 1 (best intervention). 21 SUCRA should always be interpreted with 95% CrIs as well as the quality of the evidence. The confidence in the final estimates for all the outcomes was assessed using the GRADE approach as recommended by the GRADE working group. 22

| Meta-regression and sensitivity analysis
The following sensitivity analyses were done-i. Excluding trials with high risk of bias ii. Excluding trials that had enrolled neonates who had already received surfactant prior to randomization iii. Assessing the interventions NIPPV and BiPAP based on synchronization-synchronized NIPPV (S-NIPPV), nonsynchronized NIPPV (NS-NIPPV), synchronized BiPAP (S-BiPAP) and nonsynchronized BiPAP (NS-BiPAP) Meta-regression was done using age as the covariate. The results of the meta-regression were illustrated with regression plots. The network estimates (expressed as RR [95% CrI]) for different gestational ages were depicted using forest plots.

| RESULTS
The electronic database search revealed a total of 9032 studies.
After screening for suitability, 35 studies were included in the final synthesis.  The PRISMA flow diagram is depicted in Figure 1.
Thirty-three studies (3994 neonates) and thirty-two studies (3867 neonates) were analyzed for the primary outcomes of treatment failure and requirement of MV, respectively. The mean gestational age of the neonates was 31 weeks (E- Figure S1). Five studies had enrolled neonates who had already received surfactant before randomization. The time cuts-offs for treatment failure and MV were within the first 72 hours after randomization for most of the included studies.
The NRS settings (initial and maximum) used in the included studies is given in E- Table S2. The characteristics of the included studies are given in Table 1. Some of the studies that were excluded for varying reasons are given in Table 2. 58-89
Meta-regression showed a trend similar to the outcome of MV (E- Figure S6).

| Secondary outcomes
The geometry and other characteristics of the networks for the different secondary outcomes are displayed in Figure 2 and E- Figure S7 and Table 3, respectively. The network assessing the outcomes mortality and NEC were inconsistent as assessed by node-splitting. The SUCRA plots for the secondary outcomes are given in the Figure 4 and E- Figure S8

| Air leak
NIPPV was associated with lesser incidence of air leak when com-   (Figure 3 and Table 4).

| Mortality or BPD
NIPPV was associated with a decreased risk of the combined outcome of BPD or mortality when compared to CPAP (0.74 [0.52, 0.98]) ( Figure 3 and Table 4).    Figure S9 and Table 4).

| Excluding studies with high risk of bias
When studies with high risk of bias were excluded, there was no difference in efficacy between any of the NRS modalities for the outcome MV. The results were unchanged for other outcomes (E- Figures S10 and S11).

| Excluding studies which had enrolled neonates who had already received surfactant
The results were unchanged after excluding trials that had enrolled neonate who had already received surfactant before randomization (E- Figures S10 and S11).     table in E-Table S3.
NS-BiPAP, S-NIPPV, and NS-NIPPV had decreased incidence of treatment failure when compared to CPAP as well as HFNC.
There were no statistically significant differences between NS-  table in E-Table S3 ( Figure 5).

| Quality of evidence
The overall confidence in the NMA effect estimate for the primary outcomes of treatment failure and requirement of MV was moderate for HFNC vs CPAP; CPAP vs NIPPV; HFNC vs NIPPV comparisons and low to very low for other comparisons. The quality of evidence was very low to moderate for all other secondary outcomes for the different comparisons. The quality of evidence for all the comparisons across outcomes is given in Table 5. NIPPV being superior to CPAP in preventing treatment failure as well as MV. 9 The relative risk reduction for both the primary outcomes was much larger than that reported by Lemyre et al 9 with

| DISCUSSION
narrower CrIs. Reasons for this could be that this NMA had included more recently published studies and also that the modalities BiPAP and NIPPV were evaluated as separate interventions. Also, this was an NMA where apart from the direct synthesis, the indirect evidence also contributed toward the overall effect estimate. It is evident from the included studies that the peak inspiratory pressure and hence the mean airway pressure (MAP) that was delivered with NIPPV was much higher than the positive end-expiratory pressure generated with CPAP. [41][42][43][44][45][46][47][48][49][50][51][52][53][54] This might be one of the reasons for NIPPV being more effective than CPAP. The fact that the incidence of air leak, as well as that of the combined outcome of BPD or mortality, was much lesser with NIPPV when compared to CPAP might suggest that the use of a relatively higher MAP with NIPPV is not deleterious.  The analysis of secondary outcomes reveals that both BiPAP and CPAP were associated with an increased risk of air leak and mortality when compared to NIPPV. Also, the risk of combined outcome of mortality or BPD was higher in CPAP compared with NIPPV. Isayama et al 10 in their NMA of different invasive and noninvasive modalities along with different methods of surfactant administration in preterm neonates with RDS had found no differences in the incidence of air leak, mortality, or BPD between CPAP and NIPPV. 10 The differences in the findings between this NMA and Isayama et al's 10  The increased risk of air leak with BiPAP when compared to NIPPV could be explained by the different mechanisms of flows used by these two interventions. 91 While NIPPV uses a fixed flow using a ventilator, BiPAP is a variable flow device. Some of the BiPAP studies have used very high pressure high of upto 15 cm H2O which might require a very high gas flow rate. 92 Also, the inspiratory times are typically higher in BiPAP compared to NIPPV which might result in the alveoli being exposed to higher pressures for a longer period of time as well as increasing the risk gas trapping, especially when higher respiratory rates are used. The risk of mortality or BPD was higher in CPAP compared with NIPPV in this NMA. This was not seen in the Isayama et al's 10 NMA. This might be due to the differences in the inclusion criteria between the two meta-analyses as specified above. Also, the quality of evidence for most of the secondary outcomes of this NMA was low to very low and hence should be interpreted with caution.

| STRENGTHS AND LIMITATIONS
This is one of the largest NMA evaluating the different NRS modalities used as primary support for preterm neonates with RDS. It is PRSIMA NMA extension compliant. The quality of evidence for all the outcomes was done in a very robust manner as per the GRADE working group recommendations.
Limitations in this NMA include that this did not include two of the recently introduced NRS modalities in neonatal respiratory care namely, nasal high-frequency oscillation ventilation (nHFOV) and neurally adjusted Ventilatory assist. Also, two of the secondary outcomes (NEC and mortality) had inconsistent networks. Finally, the event rates and the optimal information size for most of the secondary outcomes were low with the quality of the evidence being downgraded to low to very low for these outcomes.

| CONCLUSIONS
The overall quality of evidence for the primary outcomes was moderate to very low for the different comparisons. NIPPV appears to be the most effective primary NRS modality in preterm neonates with RDS to prevent MV and respiratory failure in the first few days of life.