Results of the search

The initial literature search returned 1160 potential articles or abstracts, of which 37 were of particular interest (full‐text review). See Figure 1.

Included studies

We identified 10 studies that met our inclusion criteria: Armanian 2014, Bisceglia 2007, Kirpalani 2013, Sai Sunil Kishore 2009, Kugelman 2007, Lista 2009, Meneses 2011, Ramanathan 2012, Salama 2015, and Wood 2013.

Armanian 2014 performed a quasi‐randomized trial that enrolled 98 infants (< 1501 grams and/or < 35 weeks) with respiratory distress syndrome (RDS) and a compatible chest x‐ray requiring noninvasive respiratory support after birth. Investigators randomized 44 infants to NIPPV and 54 to NCPAP, according to their medical chart number (odd or even number). All infants received aminophylline. A ventilator (unspecified) in the nonsynchronized mode provided NIPPV. The control group received bubble CPAP. Both groups used short binasal prongs. The primary outcome measure was respiratory failure with need for intubation within 48 hours. Secondary outcomes included need for surfactant, duration of oxygen, CLD, time to full feeds, length of hospital stay, pneumothorax, IVH, and PDA.

Bisceglia 2007 enrolled 88 preterm infants (28 to 34 weeks' gestational age) with mild to moderate RDS in the first four hours of birth. Investigators randomized 46 infants to NCPAP alone and 42 to NIPPV. They did not treat infants with aminophylline or caffeine, and they performed ventilation of both groups via nasal cannulae with the Bear Infant Ventilator CUB 750 (Ackrad Laboratories, Cranford, NJ, USA). Study authors did not synchronize NIPPV with spontaneous breathing. The primary outcome measured was number of infants needing intubation, and secondary outcomes included total duration of respiratory support, number of apneic episodes, and variation in blood gas levels. Upon receiving correspondence, study authors provided data for mortality at 28 days; presence of BPD, IVH, and necrotizing enterocolitis; and duration of hospital stay.

Kirpalani 2013 enrolled 1009 preterm infants (< 1000 grams and < 30 weeks). Researchers provided NIPPV by ventilator or bilevel device. Synchronization was allowed but was not mandatory. The primary outcome was death or moderate to severe BPD at 36 weeks. Among studied infants, 185 received noninvasive respiratory support from birth and were randomized in their first 24 hours (95 infants received NIPPV; 90 received NCPAP). We obtained data from study authors about these 185 infants for inclusion in this review.

Sai Sunil Kishore 2009 enrolled 76 infants (28 to 34 weeks' gestational age) with RDS in an RCT. Investigators randomized 37 infants to NIPPV and 39 NCPAP alone. They provided nasal respiratory support for infants by using VI P Bird‐R Sterling (Viasys Health Care, Conshohocken, PA, USA) and Drager Babylog 8000 (Drager Medical Inc, Lubeck, Germany) ventilators. NIPPV was nonsynchronized. Primary outcomes measured were failure of noninvasive respiratory support and need for intubation within the first 48 hours of randomization. Secondary outcomes included failure of nasal respiratory support within the first seven days, duration of respiratory support, duration of oxygen, duration of hospitalization, time to full feeds, air leaks, septicemia, upper airway injury, feed intolerance, abdominal distention, and bowel perforation.

Kugelman 2007 performed an RCT of 84 infants (24 to 34 weeks' gestational age) with RDS. Study authors randomized 41 infants to NCPAP alone and 43 to NIPPV. They administered both modes of respiratory support using the SLE 2000 (SLE Ltd, Surrey, UK) via nasal prongs and synchronized NIPPV with spontaneous breathing and with the infants' breathing. The primary outcome measured was need for mechanical ventilation, and secondary outcomes included blood pressure, heart rate, respiratory rate, blood gases, time needed to stop respiratory support, incidence of IVH, incidence of BPD, time to full feeds, and length of hospital stay.

Lista 2009 conducted an RCT of 40 infants (28 to 34 weeks' gestational age) with RDS diagnosed in the first hour of life. Researchers randomized 20 infants to NIPPV and 20 to NCPAP. They administered both treatments using Infant Flow (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) devices, and they synchronized NIPPV. The primary outcome measured was cytokine levels, and secondary outcomes included heart rate, blood pressure, oxygen saturation, blood gas levels, length of ventilation, incidence of PDA, treatment with ibuprofen, treatment with surfactant, pneumothorax, IVH, and oxygen dependency at 28 days (or at 36 weeks' gestational age).

Meneses 2011 enrolled 200 infants (26 to 33 and 6/7 weeks' gestational age) with RDS in an RCT. Investigators randomized 100 infants to NIPPV and an equal number to NCPAP alone. They performed nasal respiratory support of infants using the InterMed continuous‐flow and bubble CPAP systems (InterMed Inc, São Paulo, Brazil) via binasal prongs. NIPPV was nonsynchronized. The primary outcome measured was need for intubation in the first 72 hours of life. Secondary outcomes included total duration of ETT ventilation, total duration on NCPAP, total duration on supplemental oxygen, pneumothorax, BPD, PDA, necrotizing enterocolitis, IVH (grade III/IV), retinopathy of prematurity (ROP) (≥ stage 3), time to full feeds, and length of hospital stay.

Ramanathan 2012 performed a randomized controlled multicenter trial. Within 120 minutes of delivery, they enrolled 110 infants born between 26 + 0 and 29 + 6 weeks who required intubation for respiratory distress. They randomized infants at the time they received surfactant to NIPPV, nonsynchronized, via a ventilator or a bilevel device, or to NCPAP (bubble CPAP, synchronized inspiratory positive airway pressure (SiPAP), or ventilator). The primary outcome was the need for mechanical ventilation at seven days. Secondary outcomes included number of doses of surfactant, days on a ventilator, days on CPAP, days on oxygen, mortality, air leak, pulmonary hemorrhage, PDA, IVH grade III/IV, spontaneous intestinal perforation, ROP stage 3 or higher, length of stay, and CLD.

Salama 2015 performed a quasi‐randomized trial that enrolled 60 infants (28 to 34 weeks with RDS). The population was mixed at randomization: Some infants had received prophylactic surfactant (those < 29 weeks) and others had not. Researchers randomized 30 infants to NIPPV, provided by a Neoport E100M ventilator (DRE Medical, Louisville, KY, USA) in the synchronized mode, and 30 infants to NCPAP (bubble CPAP). Randomization was based on admission numbers. The primary outcome was failure of noninvasive respiratory support with defined criteria. Secondary outcomes included duration of mechanical ventilation, nasal injury, abdominal distention, gastrointestinal perforation, pneumothorax, CLD, sepsis, and IVH.

Wood 2013 (abstract) performed a two‐center RCT that enrolled 120 infants at 28 + 0 to 31 + 6 weeks and less than 6 hours of life with respiratory distress. Infants did not receive surfactant before randomization. Study authors provided NIPPV (n = 60) using a SiPAP in the synchronized mode (Trigger); the control group received CPAP (n = 60) without additional settings. The primary outcome was failure of noninvasive support.

Excluded studies

See Characteristics of excluded studies.

Studies that evaluated infants with apnea included Gizzi 2015,Lin 1998,Lin 2011,Pantalitschka 2009, and Ryan 1989.

Studies that evaluated infants after extubation included Barrington 2001,Friedlich 1999,Gao 2010,Jasani 2016, Kahramaner 2014,Khalaf 2001,Khorana 2008,Moretti 2008, and O'Brien 2012.

Studies that were not RCTs (or quasi‐randomized trials) included Aghai 2006,Herber‐Jonat 2006,Liu 2003,Manzar 2004, and Migliori 2005.

Other studies included Bhandari 2007 (compared synchronized NIPPV vs mechanical ventilation), Baneshi 2014 (no report on primary outcome), Chen 2015 (enrolled twins only), Kugelman 2014a (compared high‐flow nasal cannula vs NIPPV), Salvo 2015 (compared two methods of providing NIPPV), Santin 2004 (compared NIPPV vs conventional ventilation), Shi 2010 (included term infants), Shi 2014 (included term infants), and Zhou 2015 (examined DuoPAP (Hamilton Medical, Bonaduz, Switzerland)).

Ongoing studies and studies awaiting classification included Chen 2013,Fu 2014,Gao 2014,Sasi 2013, and Silveira 2015.

NIPPV versus NCPAP (by population) (Comparison 1)

PRIMARY OUTCOME

Respiratory failure: defined by respiratory acidosis, increased oxygen requirement, or apnea that was frequent or severe, leading to additional ventilatory support during the first week of life (Outcome 1.1)

Overall, included trials enrolled 1061 preterm infants. Three of the 10 trials (Ramanathan 2012; Kugelman 2007; Sai Sunil Kishore 2009 showed statistically significant benefit for infants initially treated with NIPPV in terms of respiratory failure in the first week of life. Meta‐analysis showed that the effect was clinically important (typical risk ratio (RR) 0.65, 95% confidence interval (CI) 0.51 to 0.82; typical risk difference (RD) ‐0.09, 95% CI ‐0.13 to ‐0.04), with 11 infants (95% CI 8 to 25) needing to be treated with NIPPV to prevent one respiratory failure. We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.1; Figure 3). Heterogeneity was low (for all outcomes).

Figure 3

---

---

Forest plot of comparison: 1 NIPPV vs NCPAP (by population), outcome: 1.1 Respiratory failure.

Two studies (Lista 2009; Ramanathan 2012) included infants who received surfactant before randomization via INSURE. Six studies (Armanian 2014; Bisceglia 2007; Kirpalani 2013; Kugelman 2007; Meneses 2011; Wood 2013) included infants who did not receive surfactant, and two studies (Sai Sunil Kishore 2009; Salama 2015) included a mixed population of infants. The effect on respiratory failure was statistically significant in the three populations of infants studied.

Need for endotracheal tube ventilation (Outcome 1.2)

Nine trials reported on this outcome (n = 950), which could not be ascertained in one trial (Ramanathan 2012). Meta‐analysis showed statistically significant benefit for infants initially treated with NIPPV (typical RR 0.78, 95% CI 0.64 to 0.94; typical RD ‐0.07, 95% CI ‐0.12 to ‐0.02), with 17 infants (95% CI 8 to 50) needing to be treated with NIPPV to prevent one respiratory failure. We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.2; Figure 4).

Figure 4

---

---

Forest plot of comparison: 1 NIPPV vs NCPAP (by population), outcome: 1.2 Need for intubation.

SECONDARY OUTCOMES

Mortality during study period (Outcome 1.3)

All trials reported this outcome. Overall, investigators noted no statistically significant reduction in mortality during neonatal intensive care unit (NICU) admission (typical RR 0.77, 95% CI 0.51 to 1.15). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.3; Figure 5).

Figure 5

---

---

Forest plot of comparison: 1 NIPPV vs NCPAP (by population), outcome: 1.3 Mortality during study period.

Chronic lung disease (need for oxygen at 36 weeks in surviving infants) (Outcome 1.4)

All trials except Armanian 2014 reported oxygen need at 36 weeks' corrected gestational age (CLD). Only one trial (Ramanathan 2012), in which infants received surfactant before randomization, reported a decrease in CLD. Meta‐analysis did not show a reduction in CLD (typical RR 0.78, 95% CI 0.58 to 1.06). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.4; Figure 6).

Figure 6

---

---

Forest plot of comparison: 1 NIPPV vs NCPAP (by population), outcome: 1.4 Chronic lung disease.

Pneumothorax (Outcome 1.5)

All 10 trials reported this outcome. Regardless of the population (surfactant or not before randomization), results showed no difference in the incidence of pneumothorax between infants randomized to NIPPV and those randomized to NCPAP (typical RR 0.79, 95% CI 0.42 to 1.48). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.5).

Intraventricular hemorrhage, all grades (Outcome 1.6)

Five trials (Armanian 2014; Bisceglia 2007; Kugelman 2007; Lista 2009; Salama 2015) reported on this outcome. Sai Sunil Kishore 2009 reported a combined outcome of IVH and periventricular leukomalacia. No trial showed a difference in IVH between treatment groups (typical RR 0.79, 95% CI 0.54 to 1.16; typical RD ‐0.05, 95% CI ‐0.13 to 0.03). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.6).

Severe intraventricular hemorrhage, grade III/IV (Outcome 1.7)

Only four trials (n = 430) (Bisceglia 2007 (study authors provided on request); Kugelman 2007; Meneses 2011; Ramanathan 2012) reported on this outcome. No trial showed a reduction in IVH in one treatment group compared with the other (typical RR 1.26, 95% CI 0.53 to 3.01; typical RD 0.01, 95% CI ‐0.03 to 0.05). We graded evidence for this outcome as low quality (unblinded intervention and wide CI) (Analysis 1.7).

Necrotizing enterocolitis ≥ Bell's stage 2 (Outcome 1.8)

Seven trials (Bisceglia 2007; Sai Sunil Kishore 2009; Kugelman 2007; Lista 2009; Meneses 2011; Ramanathan 2012; Wood 2013) reported on this outcome. No trial showed a reduction in necrotizing enterocolitis stage 2 or greater in one treatment group compared with the other (typical RR 0.67, 95% CI 0.34 to 1.31; typical RD ‐0.02, 95% CI ‐0.05 to 0.01). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.8).

Sepsis (Outcome 1.9)

Only two studies (Sai Sunil Kishore 2009; Salama 2015) (n = 136) reported on this outcome. Results of meta‐analysis showed no difference between groups (typical RR 0.78, 95% CI 0.36 to 1.70; typical RD ‐0.04, 95% CI ‐0.16 to 0.08). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.9).

Retinopathy of prematurity (≥ stage 3) (Outcome 1.10)

Only two studies (Meneses 2011; Ramanathan 2012) (n = 245) reported on this outcome. Meta‐analysis showed no difference between groups (typical RR 1.50, 95% CI 0.65 to 3.44; typical RD 0.03, 95% CI ‐0.04 to 0.10). We graded evidence for this outcome as moderate quality (unblinded intervention) (Analysis 1.10).

Duration of endotracheal tube intubation

Three studies reported on this outcome; all included infants who did not receive surfactant before randomization. Duration of intubation and mechanical ventilation varied significantly between trials, with one trial reporting mean duration of seven to 12 hours (Bisceglia 2007) and the other two trials reporting mean duration between 10 and 13 days (Kugelman 2007; Meneses 2011). This heterogeneity precludes a meaningful meta‐analysis. Overall, no trial reported a clinically significant reduction in time required for mechanical ventilation in infants who received NIPPV. We graded evidence for this outcome as low quality (unblinded intervention and inconsistency).

Duration of oxygen dependence

Four studies reported this outcome (Sai Sunil Kishore 2009; Lista 2009; Meneses 2011; Ramanathan 2012) for 412 participants. These studies showed a high degree of heterogeneity, precluding meaningful meta‐analysis. Only one study (Ramanathan 2012), in which all infants received surfactant before randomization, demonstrated a reduction in the number of days on oxygen among infants who received NIPPV (29 days vs 38 days). We graded evidence for this outcome as low quaity (unblinded intervention and inconsistency).

Duration of hospital stay

Five trials (Armanian 2014; Sai Sunil Kishore 2009Kugelman 2007; Meneses 2011; Ramanathan 2012) reported this outcome (n = 554). Hospital stay ranged between 21 and 71 days with a high degree of heterogeneity between studies; therefore, we did not perform a meta‐analysis. Only one study (Armanian 2014) demonstrated a reduction in duration of stay (22 days in the NIPPV group vs 29 days in the CPAP group). We graded evidence for this outcome as low quality (unblinded intervention and inconsistency).

Upper airway injury (Outcome 1.11)

Sai Sunil Kishore 2009 reported two infants in each group with local upper airway injury. Study authors did not specify the nature of the injury and did not state whether it was nasal septal injury. Salama 2015 reported 20 cases of nasal injury in the NCPAP group (out of 30 infants) and none in the NIPPV group, without further description. We graded evidence for this outcome as low quality (unblinded intervention and imprecision) (Analysis 1.11).

NIPPV versus NCPAP (by device) (Comparison 2)

Six trials used NIPPV delivered via ventilator (Bisceglia 2007; Sai Sunil Kishore 2009; Kugelman 2007; Meneses 2011; Armanian 2014; Salama 2015), two used bilevel devices (Lista 2009; Wood 2013), and two used both ventilator‐driven and bilevel devices (Kirpalani 2013; Ramanathan 2012). Two of the six trials using a ventilator to generate NIPPV showed benefit of NIPPV in preventing respiratory failure post extubation. Meta‐analysis of these six trials showed a reduction in rate of respiratory failure in the NIPPV group (typical RR 0.63, 95% CI 0.47 to 0.86) (Analysis 2.1). Results showed no evidence of benefit in the two trials using bilevel devices. One trial using both ventilators and bilevel devices (Ramanathan 2012) showed a reduction in respiratory failure at seven days (typical RR 0.31, 95% CI 0.11 to 0.87). Study authors provided no information on the relative proportion of infants who received NIPPV via ventilator or bilevel device (Analysis 2.1). Infants who received NIPPV via ventilator were intubated less often than those who received CPAP (typical RR 0.77, 95% CI 0.63 to 0.95; typical RD ‐0.08, 95% CI ‐0.14 to ‐0.02; number needed to treat for an additional beneficial outcome (NNTB) 13 (95% CI 7 to 50)) (Analysis 2.2).

When review authors examined CLD by device delivering NIPPV, we observed no reduction in CLD in any of the subgroups (Analysis 2.4). One trial using mixed devices showed a possible reduction in CLD in the NIPPV group, with the confidence interval including 1.0 (typical RR 0.54, 95% CI 0.29 to 1.0). We noted no difference in the rate of pneumothoraces between groups (Analysis 2.5) and no difference in mortality or severe IVH within subgroups (Analysis 2.3; Analysis 2.6).

Synchronized versus nonsynchronized NIPPV (Comparison 3)

Four studies used synchronized NIPPV (Kugelman 2007; Lista 2009; Salama 2015; Wood 2013), and five did not (Armanian 2014; Bisceglia 2007; Sai Sunil Kishore 2009; Meneses 2011; Ramanathan 2012). One study allowed both methods (Kirpalani 2013). Nonsynchronized studies (n = 572) showed overall benefit in reducing respiratory failure (typical RR 0.60, 95% CI 0.44 to 0.83), and synchronized studies (n = 304) did not demonstrate benefit (typical RR 0.65, 95% CI 0.41 to 1.02) (Analysis 3.1). Findings were similar when we examined the need for intubation: Nonsynchronized studies showed benefit (typical RR 0.74, 95% CI 0.60 to 0.92), and synchronized studies showed a trend toward benefit (typical RR 0.67, 95% CI 0.42 to 1.06) (Analysis 3.2). However, the subgroup difference was not statistically significant.

We noted no difference in mortality within subgroups (Analysis 3.3). When we examined CLD by synchronization, we found that no method showed statistically significant benefit (Analysis 3.4). We observed no difference in the rate of pneumothoraces between groups (Analysis 3.5). No study using synchronized NIPPV reported on severe IVH (Analysis 3.6).

Post hoc analysis