Introduction

The diagnostic criteria of acute respiratory distress syndrome (ARDS) in children and adults have been widely recognized, while the diagnostic criteria of neonatal acute respiratory distress syndrome (NARDS) were not developed until 2017 (Montreux criteria) [1]. NARDS is associated with meconium aspiration, sepsis and infectious pneumonia, pulmonary hemorrhage, perinatal asphyxia complicated by severe respiratory failure, and biliary pneumonia according to the Montreux criteria [1]. A study in late preterm and term infants showed that the incidence of ARDS is approximately 28.8% [2].

Invasive mechanical ventilation, including conventional mechanical ventilation (CMV) and high-frequency oscillatory ventilation (HFOV), is an effective method for the treatment of ARDS in critically ill neonates. With its combined advantages of delivering small tidal volumes with low phasic pressure changes at supraphysiologic frequencies [3], stable/higher mean airway pressure, and avoiding gas trapping, HFOV may be associated with improved lung protection. In addition, studies of HFOV in children also showed benefits in short-term oxygenation, although no improvement in clinical outcomes was observed [4]. However, the results of some previous human trials comparing the two modes of ventilation remain inconclusive [5].

To date, there are insufficient studies comparing HFOV and CMV in the treatment of NARDS, and the clinical outcomes of CMV and HFOV in the treatment of neonates with respiratory failure are inconsistent; therefore, we conducted this retrospective study to compare the mortality and incidence of bronchopulmonary dysplasia (BPD) between neonates with moderate-to-severe perinatal-onset NARDS who received HFOV and CMV.

Materials and methods

Patient selection

This retrospective study was conducted in the Department of Neonatology, Children’s Hospital of Chongqing Medical University, China. The data were collected from the medical records of the Department of Neonatology from 1 January 2014 to 31 October 2018. Ultimately, 700 neonates (gestational age of 24 weeks 0 day to 42 weeks 0 days) with perinatal-onset moderate-to-severe NARDS who received HFOV or CMV within 24 h after birth were enrolled.

Inclusion criteria

For a neonate to be included, the following four criteria had to be fulfilled: (1) gestational age between 24 + 0 and 42 + 0 weeks; (2) diagnosis of NARDS; and (3) moderate-to-severe NARDS and invasive ventilation within 24 h after birth.

Exclusion criteria

Neonates meeting at least one of the following criteria were not eligible for the study: (1) neonates who only needed noninvasive ventilation; (2) neonates with major congenital anomalies or chromosomal abnormalities; (3) neonates with upper respiratory tract abnormalities; (4) neonates with pulmonary adenomatous malformations, sequestration, congenital diaphragmatic hernia, or pulmonary hypoplasia; or (5) neonates with crossover between CMV and HFOV.

Intervention

Indications of invasive ventilation include respiratory failure (including respiratory acidosis, arterial blood gas pH <7.2 and PaCO2 >60 mmHg; hypoxemia, PaO2 <50 mmHg under auxiliary oxygen supply or under noninvasive ventilation; fraction of inspired oxygen (FiO2) >40%; severe apnea) [6].

The indications for CMV mechanical ventilation in this study were as follows [7]: (1) PaO2 <50 mmHg, (2) PaCO2 >60 mmHg, (3) intractable or recurrent apnea, (4) gasping or poor respiration, (5) O2 saturation <85% on supplemental oxygen.

The indications for HFOV mechanical ventilation in this study were as follows [8]: insufficient carbon dioxide elimination, respiratory acidosis (i.e., pH <7.20–7.25); and/or hypoxemia (PaO2 <50 mmHg)/hypercarbia (PaCO2 >60 mmHg); apnea/bradycardia; increasing respiratory distress; high fractional oxygen requirement. HFOV was for primary use not rescue in this study.

Ventilation strategies

Infants assigned to CMV received ventilation with Servo-i and SLE 5000 ventilators. Conventional ventilation was delivered by time-cycled, pressure-limited ventilators starting with a rate of 30–40 breaths per minute and an inspiratory time of 0.4 s. The aimed tidal volumes were 4–6 ml/kg, and positive end-expiratory pressure (PEEP) was 5–8 mmHg.

Infants assigned to HFOV received ventilation with the SensorMedics 3100A and SLE 5000 ventilators. The open lung strategy was used in this study. There was no crossover to HFOV or CMV in this study. Patients were ventilated with the selected ventilation modes (HFOV or CMV) through the treatment period.

Tube diameters were chosen according to the patient’s weight. Ventilator settings were adjusted at the discretion of the attending clinician to maintain arterial oxygen saturation, as measured by pulse oximetry, between 88 and 96%, a PaO2 between 50 and 80 mm Hg, a PaCO2 between 35 and 55 mm Hg, and a pH between 7.20 and 7.45. Analysis of the first blood gas was performed at admission, monitored every 6–18 h for 24 h, and then monitored daily until the infant was extubated. Chest radiography was performed immediately at admission and was reviewed daily for 2 to 3 days and then once every 5 to 7 days until extubation. Chest ultrasound was performed after admission and was reviewed daily for 2 to 3 days and then once every 5 to 7 days until extubation. Echocardiography was performed within 24 h after admission. If required, echocardiography was performed again at any time.

Extubation criteria are as follows: pH >7.20, PaCO2 <55 mmHg, Paw 6 cmH2O; FiO2 0.30; sufficient spontaneous breathing effort without any clinical and radiologic sign of respiratory distress, per clinical evaluation [9].

Definition of the important diagnoses and concepts

The diagnosis of NARDS was made according to the criteria established by the Montreux conference in 2017 [1]: acute onset; oxygenation impairment with reduced end expiratory lung volume requiring positive pressure to recruit the alveoli; respiratory failure not fully explained by lung edema due to heart failure; and diffuse bilateral opacities with loss of aeration on chest radiographs. The following neonatal respiratory disorders qualify for a diagnosis of ARDS: meconium, milk, bile, blood, or water aspiration; lung hemorrhage; and infectious or biliary pneumonia. Acute processes occurring in extrapulmonary organ systems that might trigger acute respiratory failure via systemic inflammatory responses and thus meet ARDS criteria, such as early- or late-onset sepsis, fetal inflammation, perinatal asphyxia, and necrotizing enterocolitis.

Moderate-to-severe perinatal-onset NARDS was determined by the use of the oxygenation index (OI) and triggers, including perinatal asphyxia, early-onset sepsis, and other perinatal factors that occurred in the first 24h of life. Severe NARDS was determined by an OI of ≥16, and moderate NARDS was determined by an OI between 8 and 16 [1].

Diagnostic criteria for pneumonia, MAS, and aspiration were according to the references [[1, 10]].

Diagnostic criteria for the following outcomes were according to the literature: bronchopulmonary dysplasia (BPD) [11], retinopathy of prematurity (ROP) [12], NEC [13], intraventricular hemorrhage (IVH) [14], and ventilation-free days (VFDs) [15].

Data collection

Data collected included demographics (i.e., gestational age, sex, birth weight, age at admission, and age at NARDS diagnosis), perinatal factors (premature rupture of membranes [PROM] and triggers sepsis, pulmonary hemorrhage, MAS, pneumonia, aspiration of blood, and perinatal asphyxia), elements of NARDS (i.e., disease severity and type of ventilation), 5-min Apgar score, lactate at admission, strategies during hospitalization (i.e., surfactant use), primary outcomes (i.e., death rate and BPD), and secondary outcomes (i.e., IVH ≥ 3rd stage, ROP ≥ 2nd, NEC ≥ 2nd stage, air leak, duration of invasive ventilation, ventilation-free days, and duration of hospitalization).

Data were obtained from the clinical records and were collected daily on web-based case report systems provided by Open CDMS. All data were collected from the electronic medical record.

Primary outcomes were neonatal intensive care unit (NICU) mortality and the incidence of BPD. Secondary outcomes included complications during hospitalization, such as IVH ≥ 3rd stage, ROP ≥ 2nd stage, NEC ≥ 2nd stage, air leakage, duration of invasive ventilation, ventilation-free days, and duration of hospitalization.

Statistical analysis

All data were analyzed using SPSS 23.0 software (Chicago, IL, USA). The results are presented as the median (interquartile range [IQR]) for continuous variables and the number (%) for categorical variables. The normality of the data distribution was tested with the Kolmogorov-Smirnov test. Independent continuous variables were analyzed using T tests, and categorical variables were analyzed using chi-square or Fisher’s exact tests. A p value <0.05 was considered statistically significant.

Propensity score matching (PSM) was estimated using a logistic regression model in which the following covariates were included: gestational age, sex, birth weight, age at admission, age at NARDS diagnosis, perinatal factors (PROM and triggers sepsis, pulmonary hemorrhage, MAS, pneumonia, aspiration of blood, and perinatal asphyxia), OI, surfactant use, prenatal glucocorticoid, postnatal glucocorticoid, 5-min Apgar score, and lactate at admission. Those covariates were selected based on the previous literature in which an association between those indicators and mortality and the incidence of BPD in adults and PARDS was reported [16,17,18]. A 1:1 “nearest neighbor” case-control match without replacement was used [19]. Each neonate who received CMV was matched with the neonate who received HFOV and who had the closest estimated propensity scores. Post-match variables were compared between the groups using univariate and multivariate analysis. The variables that were considered to have a univariable association (p<0.2) with mortality were included in multivariate logistic regression analysis to determine the independent factors of mortality (p<0.05). After each step of the multivariate analysis, we evaluated the collinearity by starting with a correlation matrix if the addition of the new variable had a large effect on the point estimate for another parameter in the model. When variables were highly correlated, the variable with the highest univariable association with the outcome in the multivariate analysis was retained. All analyses were verified by three statisticians (LC, JX, and SQX).

Results

Clinical characteristics of neonates with NARDS

Between 1 Jan 2014 and 31 Oct 2018, 40,176 infants were admitted to the Department of Neonatology, Children’s Hospital of Chongqing Medical University. Among them, 3302 infants suffered from respiratory failure and needed invasive mechanical ventilation, and the total number of infants who needed assisted ventilation, including noninvasive and invasive mechanical ventilation, was 5691. Ultimately, 700 infants met the criteria for entry (199 cases were assigned to HFOV and 501 cases to CMV) (Fig. 1).

Fig. 1
figure 1

Flowchart of study enrollment

Full size image

The characteristics of the two groups before and after matching are shown in Tables 1 and 2. Before matching, 8 out of the 17 covariates, including sex, age at admission, prenatal glucocorticoid use, postnatal glucocorticoid use, 5-mi Apgar score, OI, pulmonary hemorrhage, and pneumonia, were significantly different between the two groups. After matching, 254 cases were included in the PSM model. All 17 covariates were well balanced, and no significant differences were observed (Tables 1 and 2).

Table 1 Baseline characteristics of neonates in CMV and HFOV groups
Full size table
Table 2 Triggers of NARDS in CMV and HFOV groups
Full size table

Mortality and BPD

Before matching, the overall mortality of neonates with NARDS was 18.5% (130 of 700), and the incidence of BPD was 19% (133 of 700). After matching, the mortality of neonates with NARDS was 25.2% (64 of 254), and the incidence of BPD was 19.2% (49 of 254). There were no significant differences in mortality or the incidence of BPD between the two groups (25.2 [32 of 127] vs 25.2% [32 of 127]; p=1.000 and 16.5 [21 of 127] vs 22.0% [28 of 127]; p=0.266) (Table 3). In particular, the death rate of neonates with severe NARDS was 40.47% (51/126), while that of neonates with moderate NARDS was 10.23% (13/127) after matching.

Table 3 Primary and secondary outcomes of neonates in CMV and HFOV groups after propensity score matching
Full size table

Secondary outcomes

There were no significant differences between the two groups regarding the other secondary outcomes, including air leakage, ROP, NEC, surfactant treatment, PDA treatment, duration of invasive ventilation, and duration of hospitalization (Table 3). The incidence of IVH (≥ 3rd stage) was significantly lower in the HFOV group than in the CMV group (3.9 vs 11.8%, p=0.020). The VFDs in the CMV group was significantly higher than that in the HFOV group (20.96 vs 15.22 days, p=0.01).

Univariate and multivariate analyses of neonate mortality at NARDS diagnosis

Univariate and multivariate analyses were performed to identify the risk factors that were associated with neonate mortality at NARDS diagnosis, and the results are shown in Tables 4 and 5.

Table 4 Univariate logistic regression analysis for mortality of neonates at NARDS diagnosis
Full size table
Table 5 Multivariate logistic regression analysis for mortality of neonates at NARDS diagnosis
Full size table

In univariate analysis, variables at the diagnosis of NARDS that were considered to have a univariable association with mortality included gestational age, birth weight, age at admission, 5-min Apgar score, OI, lactate at admission, PROM, pneumonia, aspiration of blood, and perinatal asphyxia, with a p value of less than 0.2 (Table 4). Furthermore, in multivariate analysis, variables at diagnosis of NARDS that retained an independent association with mortality included birth weight and OI. The factors associated with higher mortality were lower birth weight and higher OI (Table 5).

Discussion

Our study is the first to compare the mortality, incidence of BPD, and other complications between neonates with NARDS who received either CMV or HFOV. Our data suggest that there were no significant differences in mortality or the incidence of BPD between the CMV and HFOV groups.

PARDS occurred in approximately 6% of all mechanically ventilated children in pediatric intensive care units (PICUs) internationally [20], while in our study, NARDS occurred in approximately 12% (700/5691) of all mechanically ventilated infants in the NICU. The relatively higher incidence may be associated with unique triggers for respiratory failure (e.g., MAS, perinatal asphyxia) during the neonatal period that might differ from pediatric ARDS and variations in specific aspects of developmental lung biology and maturation and neonatal immune status [1]. In our study, the mortality of neonates with NARDS before matching was 18.5%, and the mortality of neonates with NARDS was 25.2% after matching. However, data on the mortality of neonates with NARDS are lacking, and more studies reporting on the mortality of neonates with NARDS are needed. In this study, severe NARDS (OI ≥16) was associated with higher mortality than moderate NARDS (8 ≤ OI < 16). In the Montreux definition, OI is recommended to determine hypoxemia severity [1], and hypoxemia severity is closely associated with mortality. A meta-analysis comparing HFOV and CMV revealed that there is no difference in mortality for acute pulmonary dysfunction in preterm infants. Likewise, in our study, no significant difference was observed in mortality between the CMV and HFOV groups after matching.

In our study, after matching, the incidence of BPD was 16.5% and 22.0% in the CMV and HFOV groups, respectively. Although BPD is mostly associated with premature birth, it can also occur in infants born at late preterm or term infants who need aggressive ventilator therapy for severe acute lung disease and in infants with risk factors for BPD, including infection, maternal hypertension or preeclampsia, patent ductus arteriosus (PDA) [21], and antenatal corticosteroid rate. In our study, the median gestational ages of neonates after matching were 34.86 weeks in the CMV group and 34.86 weeks in the HFOV group. The reasons for the relatively high rate of BPD might be that all the neonates enrolled in our study received invasive ventilation; among them, the incidence of early-onset sepsis was high, with rates of 34.6% in the CMV group and 33.1% in the HFOV group. However, the administration rate of prenatal glucocorticoids was low, with rates of 17.3% in the CMV group and 18.1% in the HFOV group.

The effects of HFOV or CMV on neurological outcomes are controversial. Some RCTs reported similar brain injury in neonates who received HFOV or CMV [22]. Cools et al. [5] reported that although increased risks of severe grade IVH and periventricular leukomalacia (PVL) were found in some studies, the overall meta-analysis revealed no significant differences in those effects between the HFOV and CMV groups. In our study, the incidence of IVH was higher in the CMV group than in the HFOV group. However, we did not show a long-term neurodevelopmental outcome in our study, and a long-term follow-up of neurodevelopmental outcomes is necessary. In our study, there were no significant differences in the incidence of air leakage, ROP, NEC, duration of invasive ventilation, or duration of hospitalization between the two groups, which was in agreement with the findings of previous studies reporting acute pulmonary dysfunction or respiratory distress syndrome (RDS) in preterm infants [5]. Our results also showed that there were significantly more VFDs in the CMV group than in the HFOV group. However, comprehensively considering the ability to reduce ventilator-dependent time and mortality, this result failed to show a benefit of HFOV in this respect.

There were no differences in the triggers of NARDS between the two groups after matching. In this study, the main triggers were sepsis and pulmonary hemorrhage. The rates of sepsis were 34.6% in the CMV group and 33.1% in the HFOV group after matching. Randolph et al. [23] reported that the most common trigger was infection in children with ARDS. Goh et al. [24] reported that the rate of sepsis in children’s ARDS triggers was 43%, while Dahlem et al. [25] reported a rate of 34%. The reported incidence of neonatal sepsis varies from 7.1 to 38 per 1000 live births in Asia [26, 27]. By comparison, rates reported in the USA and Australasia range from 1.5 to 3.5 per 1000 for early-onset sepsis [28]. This reminds us that although neonatal perinatal ARDS has its own unique trigger factors, such as MAS and asphyxia, similar to pediatric ARDS, sepsis is still an important trigger factor, especially in Asia. More attention should be paid to sepsis by clinical neonatal physicians. In addition, the rates of pulmonary hemorrhage were 36.2% in the CMV group and 34.6% in the HFOV group after matching. The exact pathogenesis of pulmonary hemorrhage remains unknown; some studies have described that it is associated with hypoxemia, sepsis, and PDA [29]. Further exploration is still needed to define the triggers of NARDS.

Multivariate analysis found that birth weight and OI trend were the most significant factors associated with mortality. An RCT on ARDS in adults conducted by Derdark et al. [30] also reported that the OI trend was the most significant post-treatment predictor of survival irrespective of the assigned ventilator. In a recent randomized trial for pediatric ARDS, the severity of hypoxia was measured by the OI, and the average mortality in children with an OI ≥ 13 at study entry was 36 vs 20% in those with an OI ≤12 [23]. In this study, the mortality of patients with severe NARDS (OI ≥16) was 40.47%, which was higher than that of patients with moderate NARDS (8 ≤ OI < 16), which was 10.23% after matching.

Many factors, related to both the intervention itself and the individuals, could interact in complex ways. The disease severity, clinical management, ventilator properties, and clinical interventions warrant further exploration of outcomes with a multivariate modeling approach. Considering the patient’s clinical features, physiological status, and response to ventilatory support to determine how to optimally ventilate the patient is of great importance [31]. A study showed that the use of a reduced tidal volume and avoidance of high airway pressures were associated with improved mortality when compared with a conventional ventilatory approach [32]. Adequate use of either mode of ventilation is crucial for neonatologists, and the importance of applying an optimal lung recruitment strategy during HFOV and monitoring and controlling tidal volume with the VG during conventional ventilation should be considered [32, 33]. A new ventilator strategy, HFOV combined with VG, has also been demonstrated to be a potential strategy to prevent lung injury [34].

There are several limitations in this study, including the following: (1) the relatively small number of patients; (2) this was a single-center study in Chinese newborns; (3) it was a retrospective paper, and Chan et al. [35] concluded that many questions remained unanswered, and HFOV should be compared to CMV in a large randomized, controlled trial; (4) propensity score matching (PSM) can balance more variables at the same time, which can only be limited to the known confounding variables, and some unknown confounding variables may still have an impact on the final result; (5) there was a lack of assessment of long-term neurodevelopmental outcomes; and (6) patients enrolled in our study who received invasive ventilation and patients with NARDS who were undergoing noninvasive ventilation were excluded.

According to the Montreux criteria, early-onset sepsis and NARDS perinatal triggers are defined by those occurring in the first 72h of life. In this study, we aimed to explore whether HFOV was superior to CMV in decreasing mortality and the incidence of BPD in infants with NARDS. To reduce various types of heterogeneity and confounding factors, we targeted neonates with moderate-to-severe NARDS that occurred in the first 24h of life. Late-onset sepsis was defined as a positive result on one or more blood cultures obtained after 72h [36]. The onset time of NEC usually occurs on the 3rd or 4th day of life in term neonates, and the onset time may be delayed for up to three or 4 weeks in preterm neonates [37]. NARDS triggered by late-onset sepsis or NEC may be ventilated after 24h. Therefore, those with NARDS triggered by late-onset sepsis or NEC were excluded. Thus, our study results could not apply to NARDS triggered by late-onset sepsis or NEC and other types of NARDS. Taken these limitations into account, this retrospective study may have some implications for clinical practice and future research on HFOV vs CMV in infants with perinatal-onset NARDS.

In summary, in this retrospective study, our results showed that HFOV was associated with reduced IVH in neonates with moderate-to-severe perinatal-onset NARDS compared with CMV. However, VFDs in the CMV group were significantly higher than those in the HFOV group and did not appear to be superior to CMV for decreasing mortality and the incidence of BPD in neonates with moderate-to-severe perinatal-onset NARDS.