Outcomes and complications of invasive versus non-invasive respiratory support in extremely preterm infants: a retrospective study

Abstract

Background

Because the respiratory system of extremely/very preterm infants is immature, respiratory support is essential to their care; however, the effects of invasive and non-invasive ventilation modes on prognosis and complications remain controversial. The efficacy and safety of invasive mechanical ventilation (IMV), nasal continuous positive airway pressure (NCPAP), and non-invasive high-frequency oscillatory ventilation (NHFOV) were to be compared in extremely/very preterm infants. 

Methods

Clinical data from 220 extremely/very preterm infants admitted between January 2024 and March 2025 were retrospectively analyzed. Infants were stratified by initial ventilation modality into the invasive group (AG, n = 87), the non-invasive NCPAP group (BG, n = 82), and the non-invasive NHFOV group (CG, n = 51); treatment outcomes and complications were compared. 

Results

The CG had the shortest total ventilation time (8.5 ± 2.8 days), the lowest rate of reintubation (17.6%), and the highest weaning success rate (88.2%), all of which were superior to the other groups (P < 0.05); the AG had the highest incidence of bronchopulmonary dysplasia (37.9%) and intracranial hemorrhage (13.8%) (P < 0.05). Mortality, necrotizing enterocolitis (NEC), retinopathy of prematurity (ROP), and sepsis incidence not statistically significant among the three groups (P > 0.05).

Conclusion

Non-invasive ventilation, especially NHFOV, was found to optimize treatment outcomes and reduce complications in extremely/very preterm infants; its preferential use was recommended.

Background

Extremely/very preterm infants are regarded as the most vulnerable group in neonatal intensive care unit (NICU) because their lungs are underdeveloped and their respiratory centers are inadequately regulated [1]. Alveolarization is incomplete, pulmonary surfactant (PS) is deficient, and airway tone regulation is weak, so respiratory distress syndrome (RDS) and respiratory failure are readily induced [2], and respiratory support is considered the core guarantee of survival [3]. Statistics indicate that more than 80% of infants born at < 28 weeks require respiratory support, and the chosen mode of support is directly related to quality of survival and long-term prognosis [4, 5].

Currently, respiratory support modalities used in clinical practice are divided into invasive and non-invasive categories. Invasive mechanical ventilation (IMV) is delivered through endotracheal intubation and provides powerful support; severe hypoxaemia is rapidly corrected, and it is regarded as the rescue option for critical respiratory failure. However, invasive procedures are accompanied by significant risks: airway mucosa may be injured by tracheal intubation, positive-pressure ventilation may over-distend or collapse alveoli and activate inflammatory responses, and the risk of bronchopulmonary dysplasia is thereby increased [6]; fluctuations in intrathoracic pressure may also disturb cerebral blood flow and provoke periventricular-intraventricular hemorrhage (PVH-IVH), and difficult weaning may prolong the treatment course [7]. Non-invasive ventilation is gradually adopted because of its non-invasive advantages [8]; nasal continuous positive airway pressure (NCPAP) maintains alveolar patency through continuous positive airway pressure and reduces shear-stress injury, so it is applied in mild-to-moderate respiratory distress [9]. Nevertheless, its support level is limited, and severe cases may require conversion to invasive ventilation because of inadequate ventilation, which conversely increases risk. Non-invasive high-frequency oscillatory ventilation (NHFOV) combines high-frequency oscillation with positive-pressure ventilation; gas exchange is promoted with small tidal volumes, and carbon dioxide elimination is especially facilitated, so it is theoretically more suitable for moderate-to-severe cases [10]. However, its efficacy difference from NCPAP remains controversial [11].

Extremely/very preterm infants are highly sensitive to respiratory support; minor changes in pressure or oxygen concentration may trigger cascades via effects on cerebral blood flow, pulmonary remodelling, or intestinal perfusion [12]. Therefore, clarification of the associations between different ventilation modalities and treatment outcomes or complications is essential for optimizing decisions and reducing iatrogenic injury. This study compared the relationships of synchronized intermittent mandatory ventilation (SIMV), NCPAP, and NHFOV with outcomes and complications in extremely/very preterm infants, so that evidence for safer and more efficient respiratory support strategies can be provided.

Materials and methods

Study subjects

A retrospective cohort study was conducted. Extremely/very preterm infants who were admitted to the NICU of The Fourth Hospital of Shijiazhuang Hospital from January 2024 to March 2025 were enrolled. Demographic and clinical characteristics were collected. 220 eligible infants were stratified by initial ventilation modality into the invasive group (AG, n = 87), the non-invasive NCPAP group (BG, n = 82), and the non-invasive NHFOV group (CG, n = 51).

To ensure the standardization and traceability of clinical decision-making, the selection of the initial ventilation mode in this study strictly adhered to the following standardized criteria to mitigate potential selection bias inherent in retrospective studies:

Criteria for IMV intubation were as follows: IMV was initiated as the primary mode if any of the following conditions were met: (1) Severe respiratory failure indicated by arterial blood gas analysis within the first hour of life, defined as PaO₂ < 50 mmHg (with FiO₂ ≥ 60%) or PaCO₂ >65 mmHg (with pH < 7.20); (2) Frequent apnea (≥ 3 episodes per hour, each lasting >20 s and accompanied by a heart rate < 100 bpm, requiring bag-mask ventilation); (3) Inability to maintain SpO₂ within the target range of 89%−95% despite nasal cannula oxygen therapy with FiO₂ ≥ 50%; (4) Clinical diagnosis of severe RDS, evidenced by chest X-ray showing bilateral diffuse ground-glass opacities with air bronchograms, and requiring urgent PS administration (as PS delivery necessitated endotracheal intubation).

Criteria for non-invasive ventilation mode selection were as follows: (1) NCPAP (BG): NCPAP was initiated as the primary mode if the following criteria were met: mild to moderate respiratory distress, defined as the ability to maintain SpO₂ at 85%–89% with nasal cannula oxygen (FiO₂ 30%–40%), arterial blood gas showing PaCO₂ 45–60 mmHg (pH 7.25–7.35), and absence of frequent apnea (< 1 episode per hour). (2) NHFOV (CG): NHFOV was initiated as the primary mode if the following criteria were met: moderate respiratory distress, defined as SpO₂ maintained at 85%–89% with nasal cannula oxygen (FiO₂ 40%–50%), or persistent PaCO₂ 55–65 mmHg (pH 7.20–7.25) under initial NCPAP settings (positive end-expiratory pressure (PEEP) 5 cmH₂O, FiO₂ 40%), without meeting the criteria for invasive intubation and in the absence of severe nasal structural abnormalities (e.g., nasal malformations).

Inclusion criteria: gestational age < 32 weeks, confirmation by obstetric records or first-trimester ultrasound; requirement for respiratory support within 72 h after birth, with the initial modality being SIMV, NCPAP, or NHFOV; complete clinical data, including perinatal history, laboratory results, and follow-up records until discharge or death.

Exclusion criteria: presence of major congenital malformations requiring surgical intervention (e.g., congenital heart disease, tracheoesophageal fistula); confirmed genetic or metabolic disorders (e.g., phenylketonuria); maternal infectious diseases transmissible to the infant (e.g., active tuberculosis, HIV, untreated syphilis); transfer to another hospital or loss to follow-up within 72 h after admission; other types of initial respiratory support (e.g., nasal high-flow oxygen therapy).

Invasive respiratory support

In the AG (invasive ventilation group), all intubated infants initially received SIMV. Neither assist-control ventilation nor high-frequency oscillatory ventilation (HFOV) was utilized during the study period. This approach was based on the inclusion of extremely preterm infants with a gestational age < 32 weeks, for whom SIMV provides effective respiratory support while preserving some capacity for spontaneous breathing, thereby reducing ventilator-induced lung injury in the developing lungs, consistent with our center’s standardized initial protocol for invasive respiratory support in this population.

The AG adopted oral endotracheal intubation connected to a servo-controlled ventilator (20163544759, Bird, USA). Pressure control mode was used for ventilation (rather than volume control mode). Initial settings were as follows: FiO₂ 30–50%, respiratory rate (RR) 30–60 breaths/min, peak inspiratory pressure (PIP) 15–25 cmH₂O, PEEP 4–6 cmH₂O, and inspiratory time (Ti) 0.3–0.5 s. Arterial blood gases were monitored every 4–6 h via radial artery sampling, and parameters were adjusted dynamically according to target values (pH 7.25–7.45, PaO₂ 50–80 mmHg, PaCO₂ 45–55 mmHg). If PaO₂ was < 50 mmHg, FiO₂ was increased in increments ≤ 5% or PEEP was increased in increments of 0.5–1.5 cmH₂O. If PaCO₂ was > 60 mmHg, RR or PIP was increased in increments ≤ 2 breaths/min or 1–2 cmH₂O, respectively.

Weaning criteria were based on the 2021 Neonatal Mechanical Ventilation Guidelines: stable spontaneous breathing with RR < 60 breaths/min; FiO₂ ≤ 30% and PEEP ≤ 4 cmH₂O, with PaO₂ ≥ 50 mmHg and PaCO₂ ≤ 55 mmHg; absence of frequent apnoea (≤ 1 episode/h without resuscitation); controlled infection and stable haemodynamics. After these criteria were met, transition to SIMV was performed, and parameters were progressively reduced until weaning.

The need for respiratory support was determined if any of the following criteria were met: (1) Oxygen dependency: In room air (FiO₂ = 21%), persistent SpO₂ < 88% (or < 85% for extremely preterm infants with gestational age < 28 weeks), requiring supplemental oxygen via nasal cannula or mask (FiO₂ >21%) to maintain target SpO₂ (90%−94%); (2) Ventilatory impairment: Tachypnea (RR >60 breaths/min) with retractions/grunting, or arterial blood gas indicating mild respiratory failure (pH < 7.30 and PaCO₂ >50 mmHg), requiring non-invasive ventilation (NCPAP/NHFOV); (3) Apnea risk: ≥1 apnea episode requiring tactile stimulation or positive pressure ventilation (duration ≥ 20 s, accompanied by SpO₂ drop >10% or heart rate < 100 bpm).

Non-invasive respiratory support

1. BG: Binasal prongs were connected to a continuous positive airway pressure system (AD-I, Guangdong Pigeon Medical Device Co., Ltd.). Initial settings were as follows: FiO₂ 30–40%, PEEP 5–7 cmH₂O, and gas flow 8–10 L/min. Parameters were titrated according to transcutaneous SpO₂ (target 89–95%): if SpO₂ remained < 89% for >5 min, FiO₂ was increased (in 5% increments) or PEEP was increased (in 1 cmH₂O increments); if SpO₂ remained >95% for >10 min, parameters were gradually reduced.

2. CG: A high-frequency oscillatory ventilator (ST228A, Asycomfy) was employed. Initial settings were as follows: FiO₂ 30–40%, mean airway pressure (MAP) 6–8 cmH₂O, oscillatory frequency 8–10 Hz, amplitude 15–25 cmH₂O, and bias flow 10–12 L/min. Adjustments were made according to blood gas results: if PaCO₂ was >60 mmHg, amplitude was increased (in 2–3 cmH₂O increments) or frequency was increased (in 1 Hz increments); if PaO₂ was < 50 mmHg, MAP was increased (in 0.5–1.5 cmH₂O increments).

Criteria for non-invasive ventilation failure were defined as follows: inability to maintain SpO₂ ≥85% despite FiO₂ ≥ 60% after parameter adjustment; persistent PaCO₂ > 65 mmHg with pH < 7.20 for > 2 h; ≥3 episodes of apnoea requiring bag-valve-mask (BVM) ventilation within 24 h (duration > 20 s with heart rate < 100 beats/min); development of pneumothorax or severe nasal injury precluding continued use of the non-invasive device. Upon meeting any failure criterion, immediate conversion to SIMV was performed.

Adjuvant therapies

With the exception of PS administration, adjunctive therapies including anti-infective treatment, nutritional support, and management of thermoregulation and internal homeostasis constituted standard basic interventions for extremely preterm infants. Their standardized implementation helped minimize confounding influences from variations in baseline care when comparing the efficacy of different respiratory support modalities, thereby ensuring the reliability of the study findings.

1. PS was administered: after RDS was diagnosed, poractant alfa injection (3 mL:0.24 g; Chiesi, Italy) was given to all infants within 12 h via the endotracheal tube at 100–200 mg/kg; if FiO₂ ≥ 40% was still required after 48 h, an additional 100 mg/kg dose was administered.

2. Anti-infective therapy was provided: broad-spectrum antibiotics were selected according to blood-culture results and infection markers; ampicillin combined with ceftazidime was routinely used and adjustments were made based on individual conditions.

3. Nutritional support was initiated: enteral feeding was started within 24 h after birth; breast milk was prioritized and preterm formula was supplemented if necessary, beginning at 10–20 mL/(kg·d) and increased daily by 10–20 mL/kg; amino-acid and lipid emulsions were concomitantly administered to maintain a caloric intake of 80–120 kcal/(kg·d).

4. Additional measures were applied: incubator care was used to maintain 36.5–37.5℃ body temperature; blood glucose was kept between 40 and 150 mg/dL; electrolyte disturbances and anaemia were corrected; Red blood cells were transfused when hemoglobin was < 100 g/L.

Outcome measures

1. Treatment-outcome indices.

   • a
     Duration of ventilation: 
     Invasive-ventilation duration was calculated from intubation to successful extubation (re-intubation periods were added); 
     Non-invasive-ventilation duration was defined as: BG: from nasal-prong connection to NCPAP discontinuation; CG: from initiation of high-frequency oscillation to switch to another modality.
   • b
     Weaning success rate was defined as the proportion of infants who, after the first attempt to discontinue invasive or non-invasive ventilation, did not require the same modality again within 72 h.
   • c
     Rescue intubation rate was calculated for the non-invasive ventilation groups (BG and CG), defined as the proportion of infants who transitioned to IMV after meeting the failure criteria for non-invasive ventilation specified in Sect. Non-invasive respiratory support.
   • d
     Re-intubation rate was defined for the invasive ventilation group (AG) as the proportion of infants who required reinsertion of an endotracheal tube within 72 h after successful extubation (indicating failure of liberation from invasive ventilation), consistent with the standard clinical definition of re-intubation.
   • e
     Mortality rate was defined as the proportion of deaths during hospitalization, including deaths due to treatment failure and post-withdrawal mortality (after discontinuation of care by family decision).
     1) Total duration of ventilation support was defined as the cumulative time, recorded in days, from the initiation of any ventilation support (including non-invasive or invasive ventilation) until complete weaning from all forms of ventilation support (without requiring non-invasive ventilation, oxygen therapy, or invasive ventilation). Invasive ventilation duration was recorded as the cumulative duration, recorded in days, during which the infant received IMV (including endotracheal intubation with SIMV mode). In cases of reintubation after extubation, the durations of both invasive ventilation episodes were combined. Non-invasive ventilation duration was recorded as the cumulative duration, recorded in days, during which the infant received non-invasive ventilation (including NCPAP and NHFOV modes), excluding periods of oxygen therapy alone.
2. Complication indices

   • a
     Respiratory complications [13]:
     BPD was diagnosed if FiO₂ > 21% or mechanical ventilation was required at 36 weeks (corrected gestational age); severity was graded as mild (FiO₂ < 30%), moderate (FiO₂ 30–50%), or severe (FiO₂ > 50% or need for mechanical ventilation).
     Pneumothorax: Cases with intrapleural free air confirmed by chest X-ray, requiring tube thoracostomy or spontaneous resolution.
     Pulmonary infection was defined by meeting ≥ 2 of the following criteria: temperature > 38℃ or < 36.5℃; leukocyte count > 20 × 10⁹/L or < 5 × 10⁹/L; positive sputum or tracheal-aspirate culture; new infiltrates on chest X-ray.
   • b
     Neurological complications:
     PVH-IVH was graded regarding Papile criteria [14] by cranial ultrasound: grade I (sub-ependymal hemorrhage), grade II (intraventricular hemorrhage, without ventricular dilatation), grade III (intraventricular hemorrhage, with ventricular dilatation), grade IV (intraventricular hemorrhage, with periventricular infarction); the proportion with grade ≥ III was recorded.
   • c
     Digestive complications:
     Necrotizing enterocolitis (NEC) was diagnosed at grade ≥ II regarding modified Bell criteria [15], combining clinical features (abdominal distension, vomiting, bloody stools) with radiographic evidence (pneumatosis intestinalis or portal venous gas on abdominal X-ray).
   • d
     Other complications:
     Retinopathy of prematurity (ROP) was detected by fundus examination regarding International Classification of ROP (ICROP) [16]. Cases meeting ≥ stage 2 in zone II or with Plus disease were recorded.
     Sepsis was defined as a positive blood culture accompanied by systemic inflammatory response (abnormal temperature, leukocyte or platelet counts) and categorized as early-onset (≤ 72 h) or late-onset (> 72 h).
     All indices were independently assessed by two attending physicians (or higher-ranking physicians); disagreements were resolved by departmental consensus.

Statistical analysis

SPSS 27.0 was utilized to indicate measurement data as (±s), compared by one-way ANOVA (F value); If ANOVA indicated significant differences among groups, post-hoc pairwise comparisons were conducted using Tukey’s HSD test. Compared to the LSD test, Tukey’s HSD method provides enhanced robustness in pairwise comparisons among three or more groups by adjusting the Type I error rate for multiple comparisons, thereby effectively reducing the risk of false-positive results and better aligning with the statistical requirements of this three-group (AG, BG, CG) ventilation strategy comparison. Count data were expressed as n (%) and compared by χ² test; pairwise comparisons were performed with Bonferroni correction. P < 0.05 was statistically meaningful.

Results

Basic data

No visible distinctions were found among the three groups in gestational age, birth weight, gender distribution, mode of delivery, 1- and 5-minute Apgar scores, incidence of maternal complications (preeclampsia, chorioamnionitis, premature rupture of membranes), admission temperature, arterial blood gas parameters (pH, PaO₂, PaCO₂, lactate), proportion of PS use, or SNAPPE-II score (P > 0.05), confirming comparability. The rates of PS administration were 80 cases (92.0%) in the AG, 74 cases (90.2%) in the BG, and 46 cases (90.2%) in the CG, with no statistically significant difference among the three groups (χ²=0.168, P = 0.92). Further analysis of the timing of PS administration (within 12 h after birth) and the need for additional doses revealed that a second dose of PS was required in 18.8% (15/80) of the AG due to severe RDS, 14.9% (11/74) in the BG, and 13.0% (6/46) in the CG, with no statistically significant difference in the rates of additional dosing among the groups (χ²=1.032, P = 0.597). These findings indicate no significant difference in the overall severity of RDS among the three groups, thereby excluding the potential confounding effect of PS usage on treatment outcomes (Table 1).

Table 1.

Baseline characteristics

Index	AG (n = 87)	BG (n = 82)	CG (n = 51)	Statistic	P
Perinatal indicators
Gestational age (weeks)	28.6 ± 1.8	28.9 ± 1.7	28.7 ± 1.9	F = 0.327	0.721
Birth weight (g)	1290 ± 280	1320 ± 260	1300 ± 270	F = 0.412	0.663
Gender (male), n (%)	49 (56.3%)	46 (56.1%)	29 (56.9%)	χ²=0.011	0.995
Mode of delivery (cesarean section), n (%)	59 (67.8%)	57 (69.5%)	35 (68.6%)	χ²=0.084	0.958
1-minute Apgar score (points)	6.5 ± 1.2	6.7 ± 1.1	6.8 ± 1.3	F = 0.385	0.681
5-minute Apgar score (points)	7.8 ± 0.9	7.9 ± 0.8	7.8 ± 1.0	F = 0.156	0.856
Maternal complications
Preeclampsia, n (%)	19 (21.8%)	18 (22.0%)	11 (21.6%)	χ²=0.052	0.974
Chorioamnionitis, n (%)	12 (13.8%)	12 (14.6%)	7 (13.7%)	χ²=0.036	0.982
Premature rupture of membranes, n (%)	31 (35.6%)	29 (35.4%)	18 (35.3%)	χ²=0.001	0.999
Admission indicators
Body temperature (°c)	36.3 ± 0.4	36.4 ± 0.3	36.3 ± 0.4	F = 0.279	0.757
Arterial blood gas
pH	7.23 ± 0.08	7.25 ± 0.07	7.24 ± 0.08	F = 0.512	0.6
PaO2 (mmHg)	58.6 ± 8.2	60.2 ± 7.9	59.3 ± 8.5	F = 0.637	0.529
PaCO2 (mmHg)	52.4 ± 6.5	51.8 ± 6.3	52.1 ± 6.7	F = 0.115	0.891
Lactate (mmol/L)	3.2 ± 0.9	3.1 ± 0.8	3.2 ± 0.9	F = 0.178	0.837
PS use, n (%)	80(92.0%)	74(90.2%)	46(90.2%)	χ²=0.168	0.92
SNAPPE-II scores (points)	28.6 ± 7.5	27.9 ± 7.2	28.3 ± 7.4	F = 0.142	0.868

Therapeutic outcomes

The shortest total ventilation duration was recorded in the CG (8.5 ± 2.8 days), followed by the BG (9.2 ± 3.0 days), and the longest duration was observed in the AG (15.8 ± 5.1 days). Pairwise comparisons showed that the AG was markedly longer than both non-invasive groups (P < 0.001), and the CG was shorter than the BG (P = 0.032). Total duration of ventilation support was shortest in the CG (8.5 ± 2.8 days), significantly shorter than that in the AG (15.8 ± 5.1 days, P < 0.001) and the BG (9.2 ± 3.0 days, P = 0.023). Duration of invasive ventilation was only recorded in the AG (10.2 ± 4.5 days), while no invasive ventilation was administered in the BG or CG, rendering statistical comparison unnecessary. Duration of non-invasive ventilation showed no significant difference between the CG (8.5 ± 2.8 days) and the BG (9.2 ± 3.0 days, P = 0.215), and both were shorter than that in the AG (5.6 ± 3.1 days, P < 0.001).

The rescue intubation rate was lowest in the CG (17.6%, 9/51), significantly lower than that in the BG (29.3%, 24/82, P = 0.036), suggesting a potential advantage of NHFOV in reducing the failure rate of non-invasive ventilation. Additionally, the reintubation rate in the invasive ventilation group (AG), defined as the need for reintubation within 72 h after extubation, was 14.9% (13/87). Although not a primary comparison metric in this study, it is reported here to align with standard clinical terminology. The CG had the shortest length of hospital stay (40.5 ± 9.8 days), followed by the BG (42.8 ± 10.7 days), while the AG had the longest (56.3 ± 12.5 days). The differences between the AG and the two noninvasive groups were substantial (both P < 0.001), but two noninvasive groups showed negligible differences (P = 0.215).

The weaning success rate was the highest in the CG (88.2%, 45/51), followed by the BG (82.9%, 68/82), and the AG had the lowest rate (71.3%, 62/87). The AG was significantly lower than the two noninvasive groups (both P < 0.05), while two noninvasive groups showed negligible differences(P = 0.312). In addition, the mortality rates in the three groups were 11.5% (10/87) in the AG, 9.8% (8/82) in the BG, and 7.8% (4/51) in the CG, with no significant differences (P = 0.708) (Table 2; Fig. 1).

Table 2.

Comparison of therapeutic outcomes

Index	AG (n = 87)	BG (n = 82)	CG (n = 51)	Statistic	P
Rescue intubation rate, n (%) (for non-invasive ventilation groups)	/	24 (29.3%)	9 (17.6%)	χ²=4.372	0.036
Re-intubation rate, n (%) (for invasive ventilation group, post-extubation)	13 (14.9%)	/	/	/	/
Length of hospital stay (days)	56.3 ± 12.5	42.8 ± 10.7	40.5 ± 9.8	F = 47.293	< 0.001
Weaning success rate, n (%)	62 (71.3%)	68 (82.9%)	45 (88.2%)	χ²=8.745	0.013
Mortality rate, n (%)	10 (11.5%)	8 (9.8%)	4 (7.8%)	χ²=0.689	0.708
Total duration of ventilation support (days, x̄ ± s)	15.8 ± 5.1	9.2 ± 3.0	8.5 ± 2.8	F = 28.631	< 0.001
Invasive ventilation duration (days, x̄ ± s)	10.2 ± 4.5	/	/	/	/
Non-invasive ventilation duration (days, x̄ ± s)	5.6 ± 3.1	9.2 ± 3.0	8.5 ± 2.8	F = 32.157	< 0.001

Fig. 1.

Comparison of ventilation duration. Note: * vs. AG, # vs. BG, P<0.05

Complications

The comparison results of complications among the three groups of patients are shown in Figs. 2, 3, and 4. The statistical analysis of respiratory complications was as follows: bronchopulmonary dysplasia incidence was highest in the AG (37.9%, 33/87) and was superior to BG (24.4%, 20/82) and CG (17.6%, 9/51) (both P < 0.05); the incidence in the CG was inferior to BG (P = 0.362).

Fig. 2.

Comparison of respiratory complications. Note: * vs. AG, # vs. BG, P<0.05

Fig. 3.

Comparison of neurological and digestive complications. Note: * as against the AG, P<0.05

Fig. 4.

Comparison of other complications. Note: * as against the AG, P<0.05

Regarding pneumothorax, the incidence in the AG (11.5%, 10/87) was superior to two noninvasive groups (BG 4.9%, 4/82; CG 3.9%, 2/51) (both P < 0.05), and there was no significant difference between the two noninvasive groups (P = 0.821). For pulmonary infection, the incidence in the AG (28.7%, 25/87) was superior to CG (15.7%, 8/51, P = 0.028), but not statistically significant versus BG (24.4%, 20/82, P = 0.215); and not statistically significant between the BG and the CG (P = 0.347) (Fig. 2).

In terms of neurological complications, the incidence of intracranial hemorrhage (≥ grade III PVH-IVH) was highest in the AG (13.8%, 12/87) and was superior to BG (6.1%, 5/82) and CG (5.9%, 3/51) (both P < 0.05); the two noninvasive groups showed no considerable differences (P = 0.896).

Regarding the digestive system, the incidence of NEC (≥ grade II) in the three groups was 8.0% (7/87) in the AG, 6.1% (5/82) in the BG, and 5.9% (3/51) in the CG (P = 0.934) (Fig. 3).

Regarding other complications, the incidence of ROP (≥ stage 2 in zone II) in the three groups was 11.5% (10/87) in the AG, 9.8% (8/82) in the BG, and 9.8% (5/51) in the CG(P = 0.942).

In addition, the incidence of sepsis in the three groups was 18.4% (16/87) in the AG, 14.6% (12/82) in the BG, and 13.7% (7/51) in the CG (P = 0.643) (Fig. 4).

Discussion

In the respiratory support treatment of extremely/very preterm infants, IMV and the two types of noninvasive ventilation (NCPAP, NHFOV) show significant differences in treatment outcomes and complications. It should be noted that the use of PS, as a core intervention for RDS treatment, reflects the severity of RDS. In this study, no statistically significant differences were observed in the rates of PS administration (AG 92.0%, BG 90.2%, CG 90.2%) or in the rates of additional PS dosing (AG 18.8%, BG 14.9%, CG 13.0%) (all P >0.05), indicating comparable baseline severity of RDS across the three groups and excluding potential confounding effects of PS use on respiratory support efficacy and complications, thereby ensuring the reliability of comparisons among different ventilation modalities. Regarding treatment outcomes, a segmented analysis of ventilation duration provided more detailed insights: the CG demonstrated the shortest total duration of ventilation support (8.5 ± 2.8 days), significantly shorter than that of the AG (15.8 ± 5.1 days, P < 0.001) and the BG (9.2 ± 3.0 days, P = 0.023). However, when comparing the duration of non-invasive ventilation alone, no significant difference was observed between the CG (8.5 ± 2.8 days) and the BG (9.2 ± 3.0 days, P = 0.215), while the AG had a shorter non-invasive ventilation duration (5.6 ± 3.1 days, P < 0.001) due to its primary reliance on invasive ventilation. This indicates that the advantage of NHFOV (CG) in reducing the total duration of ventilation support stems not from shortening the duration of non-invasive ventilation itself, but from avoiding the need for invasive ventilation (unlike the AG) and reducing the failure of non-invasive ventilation (unlike the BG, which had a rescue intubation rate of 29.3%). Separating the statistics of invasive and non-invasive ventilation durations avoids the ambiguity of mixed duration reporting and more accurately reflects the true value of different ventilation strategies in reducing invasive ventilation exposure and improving the success rate of non-invasive ventilation [17, 18]. It is essential to highlight the limitations of this study, particularly the selection bias and indication confounding inherent in its retrospective observational design. As clearly outlined in the grouping criteria, infants in the AG were assigned to SIMV due to more severe baseline conditions (e.g., higher FiO₂ requirements, more frequent apnea), which are key confounding factors influencing outcomes such as BPD and ventilation duration. Although these factors were adjusted for using multivariate regression models, residual confounding from unmeasured variables (such as subtle differences in lung maturity) cannot be entirely ruled out. Therefore, the higher incidence of BPD in the AG (37.9%) may be partially attributable to the severity of baseline conditions rather than solely resulting from invasive ventilation. This may be related to the continuous positive airway pressure and gas convection effect generated by the high-frequency oscillation of NHFOV, which can more effectively maintain alveolar expansion, improve gas exchange, reduce the risk of atelectasis, thereby decreasing the dependence on high oxygen concentration and high airway pressure, and reducing the need for reintubation due to inadequate ventilation or poor oxygenation [19]. Meanwhile, the CG had the shortest length of hospital stay (40.5 ± 9.8 days) and the highest weaning success rate (88.2%), further confirming the advantage of noninvasive ventilation, especially NHFOV, in optimizing treatment efficiency. In contrast, the AG had the longest ventilation time, the longest length of hospital stay, and the lowest weaning success rate (71.3%), which is related to the invasive damage to the airway by invasive ventilation, the increased risk of ventilator-associated infections, and the stress response of the infants to traumatic procedures, and these factors may prolong the recovery cycle and reduce the weaning success rate.

Regarding complications, the incidence of BPD in the AG (37.9%) was superior to two noninvasive groups (24.4% in the BG and 17.6% in the CG), which is closely related to airway pressure injury, oxygen toxicity, and inflammatory response caused by invasive ventilation. Overdistension or atelectasis of alveoli during mechanical ventilation can activate inflammatory pathways, leading to lung tissue remodeling and fibrosis [20], while noninvasive ventilation reduces alveolar damage through gentler pressure support, thereby lowering the risk of BPD. The incidences of pneumothorax and pulmonary infection were also significantly higher in the AG (11.5% for pneumothorax and 28.7% for pulmonary infection), suggesting that invasive procedures may compromise airway integrity and increase the chances of gas leakage and pathogen invasion [21]. Noninvasive ventilation reduces such mechanical damage and infection risks by avoiding endotracheal intubation (4.9% for pneumothorax and 24.4% for pulmonary infection in the BG; 3.9% for pneumothorax and 15.7% for pulmonary infection in the CG). Among neurological complications, the incidence of intracranial hemorrhage ≥ grade III in the AG (13.8%) was superior to noninvasive groups (6.1% in the BG and 5.9% in the CG), which is speculated to be related to hemodynamic fluctuations during invasive ventilation. Changes in intrathoracic pressure caused by positive pressure ventilation may affect cerebral blood flow perfusion, especially in extremely/very preterm infants with underdeveloped cerebrovascular system, and are prone to induce rupture and bleeding of periventricular vessels. It should be specifically noted that the optimized target SpO₂ setting for infants with BPD may have potential impacts on clinical outcomes. In a pilot randomized trial by DeMauro et al. (2025) [22] involving BPD infants born at < 30 weeks gestation and requiring respiratory support at 36 weeks postmenstrual age (PMA), a comparison between high SpO₂ targets (≥ 96%) and low targets (90–94%) revealed no significant differences in primary outcomes, such as the number of intermittent hypoxemia (IH) events with SpO₂ < 80% lasting ≥ 30 s or the proportion of time spent in hypoxemia, over the entire study period from 36 to 44 weeks PMA to 6 months of age. However, post hoc analysis indicated that during the phase from enrollment to 48 weeks PMA, the low-target group exhibited higher rates of IH events with SpO₂ < 80% and < 90%, and their discharge PMA was significantly later than that of the high-target group (median 48.0 weeks vs. 45.0 weeks, P = 0.05). This suggests that while high SpO₂ targets may not reduce long-term hypoxemia, they may decrease IH burden and provide slight clinical benefits in the early disease stage (before 48 weeks PMA). These findings complement the present study’s focus on the impact of respiratory support modalities on BPD incidence: our study aimed to reduce BPD risk by optimizing initial ventilation strategies to minimize lung injury, whereas the study by DeMauro et al. offers insights into improving outcomes through SpO₂ target adjustment after BPD is established. Together, they highlight the synergistic value of prevention and intervention in the respiratory management of BPD infants and underscore the importance of dynamically optimizing respiratory support strategies according to the disease stage in BPD care.

Differences existed between the two types of noninvasive ventilation: NHFOV is superior to NCPAP in shortening ventilation time and reducing the rate of reintubation (17.6% vs. 29.3%), which may be attributed to its high-frequency oscillation characteristics, that is, achieving effective gas exchange through smaller pressure fluctuations, reducing alveolar collapse and carbon dioxide retention, thereby lowering the work of breathing and oxygen demand [23]. NCPAP mainly relies on continuous positive airway pressure to maintain alveolar patency, and its support may be insufficient in the face of severe atelectasis or hypercapnia, leading to some cases needing to be converted to invasive ventilation. The advantage of NHFOV in reducing the failure rate of non-invasive ventilation observed in this study is highly consistent with the findings of a prospective randomized controlled trial by Zhu et al. (2017) [24], which involved preterm infants at 28–34 weeks gestational age with moderate to severe RDS managed using the INSURE strategy. Their study demonstrated that the proportion of infants requiring transition to IMV was significantly lower in the NHFOV group (24.3%) than in the nCPAP group (56.4%, P < 0.01), with no statistically significant differences in safety outcomes such as bronchopulmonary dysplasia (BPD), intracranial hemorrhage (IVH), air leak syndrome, or mortality between the two groups. This further supports the generalizability of our findings, indicating that NHFOV, by superimposing oscillatory function to enhance airway pressure stability and reduce repeated alveolar collapse, improves oxygenation and ventilation efficiency while effectively decreasing the need for invasive ventilation, without increasing the risk of complications. Together, these studies provide clinical evidence supporting the preferential use of NHFOV as the initial non-invasive respiratory support for preterm infants with moderate to severe RDS, particularly after intervention with the INSURE strategy, to optimize respiratory outcomes. In addition, the incidence of BPD in the CG is slightly lower than that in the BG (17.6% vs. 24.4%), and although there is no statistical difference, it may suggest its potential advantage in lung protection, which is consistent with the mechanism of high-frequency ventilation reducing alveolar shear stress injury. For complications such as NEC (8.0% in the AG, 6.1% in the BG, and 5.9% in the CG), ROP (11.5% in the AG, 9.8% in the BG, and 9.8% in the CG), and sepsis (18.4% in the AG, 14.6% in the BG, and 13.7% in the CG), negligible differences existed among the three groups, indicating that these complications are more related to the immaturity of preterm infants themselves rather than being directly caused by the mode of ventilation. This also indicates that in the management of extremely/very preterm infants, the choice of ventilation mode needs to be optimized in coordination with other supportive therapies (such as nutrition and infection control) to comprehensively improve the prognosis.

Conclusion

This study is designed to compare treatment outcomes and complications among SIMV, NCPAP, and NHFOV in extremely/very preterm infants. The results demonstrated that non-invasive ventilation-especially NHFOV-was markedly superior to invasive ventilation in shortening ventilation duration, reducing re-intubation rate, increasing weaning success rate, and lowering the incidence of complications such as BPD and intracranial hemorrhage. Between the two non-invasive modes, NHFOV was found to be more advantageous in improving ventilation efficiency and decreasing failure rates.

Study limitations

Limitations of this study were recognized: the single-center retrospective design, the limited sample size, potential selection bias, and the absence of long-term neurodevelopmental outcomes. Multi-center randomized controlled trials with larger sample sizes and extended follow-up periods are required to further verify the impact of different ventilation modes on long-term outcomes in extremely/very preterm infants and to provide higher-level evidence for optimizing respiratory support strategies.

Authors’ contributions

HC, MZ, DW, SZ, and MZ is responsible for study design, data collection and collation, and the drafting of the initial manuscript. XB participates in the review of clinical data, verification of results, and revision of the manuscript. YQ provides guidance on the research direction and reviews the scientific nature and ethical compliance of the study design.

Funding

The research is supported by: Clinical analysis of different respiratory support and short-term outcomes in extremely/ultra premature infants, (No. 20231680).

Data availability

All data generated reanalyzed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

The study was approved by the local ethics committee of the fourth hospital of Shijiazhuang. All experiments were performed in accordance with relevant guidelines and regulations and was performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from the parents of the patients for collection of samples and publication of medical data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.