In the present systematic review and meta-analysis, we aimed to evaluate whether NHFOV would reduce intubation rate as compared with NCPAP as the primary respiratory supporting modes in preterm infants with RDS. As a result, we found that NHFOV was superior to NCPAP in reducing the rate of intubation, and no heterogeneity was found. Similar result also appeared in sensitivity analysis after excluding one study with significant difference. And it was consistent with the subgroup analysis of the multicenter study by Shi et al.  The result suggests that NHFOV is a more reasonable selection to reduce the risk of intubation as compared with NCPAP in preterm neonate with RDS.
In the past twenty years, several studies have been enforced to compare the beneficial effects between NHFOV and NCPAP in preterm infants, and the results were encouraging. The study by van der Hoeven M et al. in 1998 demonstrated that, comparing with the NCPAP group, the PCO2 level was lower in the NHFOV group in preterm and term neonates and the using criteria of NHFOV was deterioration on NCPAP.  A multicenter study also indicated the beneficial effects of NHFOV for preterm infants as a remedial measure after failing to other noninvasive modes in the mean number of apneas, bradycardias, or desaturations (3.2±0.4 vs. 1.2±0.3; P< 0.001), FiO2 (0.48±0.03 vs. 0.40±0.02; P<0.001) and the levels of PCO2 (74±6 vs. 62 ±4; P= 0.025).  Otherwise, a study by Wang et al. in 2017 also indicated that, as a method of rescuing treatment after failure of other noninvasive respiratory supporting strategies, NHFOV significantly reduced the numbers of apnea(1.2±1.1 vs. 6.3±2.1, P<0.01) and SpO2<0.85(1.1±1.2 vs. 4.3±1.5, P<0.01), the levels of PCO2(43±8 vs. 56±10, P<0.01) and FiO2(0.30±0.07 vs. 0.39±0.11, P<0.01). And it was consistent with the meta-analysis by Li et al.,  in which NHFOV was demonstrated to increase the removal of carbon dioxide and reduce the risk of intubation and IV as compared with NCPAP/BP-CPAP in preterm infants. In contrast, NHFOV did not improve the oxygenation (0.42±0.12 vs. 0.40±0.10, P>0.05) and reduced the levels of PCO2 (49±8 vs. 48±7, P>0.05) as a prophylactically used mode in neonates after extubation,  and which was consistent with another randomized controlled cross-over study by Klotz D et al.  and the result did not show difference in levels of PCO2(54.8±14.6 vs. 52.7±9.3, P=0.44; 49.0±8.1 vs. 47.7±9.5, P=0.55) between the NHFOV-NCPAP and NCPAP-NHFOV periods.
The first reasonable cause to explain the differences among the studies might be the diagnosis, and NHFOV might be more suitable in newborn infants with more severe respiratory distress. Previous studies were mainly enforced in the pre-neonatal acute respiratory distress syndrome(ARDS) era. In 2017, the first consensus definition for neonatal acute respiratory distress syndrome(ARDS) was provided,  and RDS and ARDS should be therefore diagnosed and compared independently. Our previous randomized controlled study also compared the beneficial effects between NHFOV with NCPAP on the need for IV in preterm infants with RDS and ARDS after extubation, and the results indicated that NHFOV did reduce significantly the need for endotracheal ventilation and levels of the PCO2 as compared with NCPAP, especially in the subgroup of infants diagnosed with ARDS.  Among the trials included in the present study, the diagnosis of "the included criteria" were no completely consistent and ARDS was not included. The diagnosis of "the included criteria" was "RDS" in the studies by Shi et al. , Malakian et al.  and ranpour et al. . But it was "moderate-severe RDS" in the study by Zhu et al.  And it was consistent with the previous reports by Mukerji et al.  and Wang et al.,  in which NHFOV was successfully used in reduced PCO2 and/or intubation rate as rescuing treatment after failure of other noninvasive ventilation. In contrast, there was similar intubation rate between NHFOV and NCPAP when the infants was stable on NCPAP after extubation. 
The second cause to explain the inconsistence might be the observation time of "need for mechanical ventilation". Among the trials included, the observation time of "need for mechanical ventilation" were different. The time of observation of "failure of NIPPV or NCPAP" was "7 days after birth" in the study by Shi et al.  and "within the first 72hr of life" in the studies by Malakian et al.  and Zhu et al.  and Iranpour et al.  did not limit the observation time of "failed nasal support".
The third cause might be the initial setting and subsequent adjustment of respiratory parameters of NHFOV. Among the four trials included, the parameters were different.(table 2) To further verify the results, a summary of all clinical studies referring to NHFOV in newborn infants was made and it was shown in table 4. [35-38] According to the summary, four reasons were accessible. Firstly, according to the report of Mukerji et al.,  visible chest oscillation was not necessary because of elimination of CO2 during NHFOV also occurring in the upper respiratory airway deadspace; Secondly, the mean tidal volume was higher with I:E at 50% than at 33% (2.4 ml vs. 1.4 ml; P < 0.001);  Thirdly, the setting of respiratory parameters should also be adjusted according to diagnosis and purpose, and an example was that Luca et al. suggested different parameters boundaries for NHFOV use in BPD-risk and postextubation newborn infants.  Last, the classifications of respiratory failure should be considered. Insufficient removal of carbon dioxide was usually one of the most important causes to induce higher incidence of intubation. Therefore, besides amplitude firstly up-regulated, frequency should be adjusted within the reported ranges to avoid hypercarbia. [11,39]
Besides efficacy, safety is another important focus when NHFOV was used. Similar to other noninvasive ventilation, NHFOV could also result in side effects. In the four included studies, Malakian et al.  reported that there were no differences in traumatization of nasal skin and mucosa, air leaks, IVH, feed intolerance and time to full feeds between the two groups. And there were no relevant reports in the studies by ranpour et al.  and Zhu et al. . Shi et al.  indicated that the rate of thick secretions causing an airway obstruction was higher in the NHFOV group than in the NCPAP group(13.8% vs 5.3%; 95% CI of risk difference, 1.218 to 6.645; P = 0.018).
and it was consistent with the study by Fischer HS et al..  Other than, air-trapping and NEC were shown in the study of Czernik Cet al. 
The major limitations of the present study: 1) The included studies were mainly performed in the pre-ARDS era, and no ARDS was included. 2) The initial respiratory parameters of NHFOV were different among the four trials. 3) The assessment for secondary outcomes were not enough in the small sample size. They might induce potential bias, including restricted application scope. These problems could be overcome in additional studies according to the present data. Recently, we have organized a multi-centers, randomized controlled trials regarding comparing NHFOV and NCPAP as the respiratory support modes after extubation in preterm infants (NCT03099694), and the results could give us more reasonable explanations.
In summary, among preterm infants with RDS, NHFOV was superior to NCPAP with respect to reducing the risk of intubation as the primary respiratory support strategies in the early life. Larger trials are needed to verify the beneficial effects.