Time spent within the SpO2 target range during nasal CPAP therapy with different backup modes in extremely preterm infants: a randomised cross-over trial

Abstract

Objectives

To study the efficacy of different modes of backup nasal intermittent positive pressure ventilation (NIPPV) during nasal continuous positive airway pressure (nCPAP) therapy to keep extremely preterm (EPT) infants within the oxygen saturation (SpO₂) target range.

Design

Single-centre randomised cross-over trial.

Setting

Level 3 neonatal intensive care unit.

Participants

24 EPT infants (median gestational age of 24+1 and median postnatal age of 43 days).

Interventions

EPT infants on nCPAP therapy with backup NIPPV during apnoea >4 s were assigned in random sequence to immediate cessation of backup NIPPV or to stepwise reduction of backup NIPPV support on restart of spontaneous respiration.

Main outcome measures

Primary outcome was time spent within the SpO₂ target range (88%–95%). Secondary outcomes included frequency and duration of hypoxaemic and hyperoxaemic episodes, impact on cerebral tissue oxygenation, parameters of gas exchange and vital signs.

Results

The mean percentage of time spent within the SpO₂ target range did not differ between immediate cessation of backup NIPPV on restart of spontaneous respiration compared with the stepwise reduction of backup NIPPV (61.1% vs 64.7%, p=0.15). There were no differences in the time spent below or above the SpO₂ target range, the number of prolonged (>30 s) and severe (<80% and <70%) desaturations and in the parameters of cerebral tissue oxygenation. The stepwise reduction of backup NIPPV resulted in a significant increase in mean airway pressure (7.2 cm H₂0 vs 5.6 cm H₂0, p<0.01) and reduction in respiratory frequency (60.1 vs 63.8, p=0.02) compared with the immediate cessation of backup NIPPV on restart of spontaneous respiration, while mean fraction of inspired oxygen, transcutaneous carbon dioxide pressure and heart rate did not vary between the two interventions.

Conclusion

Our findings underpin that new sophisticated ventilation modes like stepwise reduction of backup NIPPV do not necessarily provide a clinical benefit compared with the standard mode. For this, technological advances require robust evaluation within well-designed clinical trials demonstrating the safety and efficacy before being used in everyday clinical practice.

Trial registration

German Clinical Trials Register; registration number: DRKS00031956 (https://drks.de/search/en/trial/DRKS00031956), date of registration: 26 May 2023.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Although non-invasive ventilation is the preferred ventilation mode for extremely premature babies in order to avoid long-term respiratory health sequelae, these babies are subject to frequent fluctuations in oxygenation that are associated with increased risk for most of the acute and longer-term morbidities of prematurity and should therefore be minimised.

WHAT THIS STUDY ADDS

• Stepwise reduction of backup nasal intermittent positive pressure ventilation (NIPPV) compared with immediate cessation of backup NIPPV on restart of spontaneous respiration did not improve the time spent within the SpO₂ target range or reduce the number of episodes of hypoxaemia or hyperoxia. The stepwise reduction of backup NIPPV reduced the respiratory frequency, probably mediated by the increase in mean airway pressure during the prolonged execution of NIPPV following apnoea, but did not impact carbon dioxide pressure or fraction of inspired oxygen requirements.

How this study might affect research, practice or policy

• The results of our study underline the clinical need for scientific confirmation of treatment strategies that are considered superior by subjective clinical judgement in randomised trials. Therefore, it is important to carefully evaluate the various modes of non-invasive respiratory support introduced into clinical care during recent years for their efficacy to keep extremely preterm infants within the SpO₂ target range.

Introduction

The outcome of preterm infants is significantly influenced by recurrent fluctuations in oxygen saturation (SpO₂) during their stay in the neonatal intensive care unit (NICU).^{1 2} The causes of these fluctuations are immaturity per se with recurrent apnoeas, active exhalation and the still poorly developed functional residual capacity.^{3 4} Close associations are described for bronchopulmonary dysplasia (BPD), retinopathy of prematurity (ROP) and psychomotor impairment.²⁵,⁸ These fluctuations in oxygenation are very common, and preterm infants spend only 50%–70% of the time in their SpO₂ target range, particularly when on non-invasive respiratory support (NIV).⁹,¹² The global trend to promote NIV strategies to avoid the harm caused by mechanical ventilation demands improved strategies to keep preterm infants within the SpO₂ target range.^{4 13}

The administration of caffeine to treat apnoea is the only medical therapy that is safe and efficacious.¹⁴ The second important element of therapy is NIV support. The positive end-expiratory pressure (PEEP) transduction during nasal continuous positive airway pressure (nCPAP) therapy or the mean airway pressure (MAP) during nasal intermittent positive pressure ventilation (NIPPV) stabilises spontaneous respiration and gas exchange.¹⁵,¹⁷ During recent years, different forms of NIV were introduced into clinical use. Recent studies demonstrated that the type of NIV has an influence on the time spent in the SpO₂ target range; for example, sNIPPV was shown to be superior to nasal high-frequency oscillatory ventilation (nHFOV).¹²

Despite the various NIV ventilation options available, nCPAP continues to be the most commonly used. The effectiveness in maintaining functional residual capacity, reducing work of breathing, improving gas exchange and preventing upper airway collapse and lung damage has been demonstrated in many studies.¹⁸,²⁰ nCPAP therapy has been further developed in recent years so that various nCPAP modes with different backup functions are available, intended to compensate for periods of apnoea. Background ventilation can either be stopped at the child’s first spontaneous breath (immediate cessation of backup NIPPV) or slowly phased out over a period of 5 min (stepwise reduction of backup NIPPV). Both modes are also frequently used in our NICU. Since it is unclear which nCPAP mode is superior, we aimed to determine whether one of the two approaches is more effective at maintaining extremely preterms (EPTs) within their target oxygen saturation range.

Methods

Study population

The study was executed at the level 3 NICU of the perinatal centre Ulm between 10/2023 and 10/2024. It was approved by the Ethics Committee of the Medical Faculty of Ulm University (100/23) on 26 April 2023. The study was registered at the German Clinical Trials Register (registration number: DRKS00031956) prior to inclusion of the first patient and executed according to the rules of the Declaration of Helsinki. Informed parental consent was obtained prior to study inclusion.

All preterm infants born <32 weeks’ gestation, with a birth weight <1500 g and a postnatal age >120 hours on NIV with a PEEP of 4–8 cm H₂O, fraction of inspired oxygen (FiO₂) <0.60 and ≥4 events of hypoxaemia <80% and/or apnoeas >20 s duration and/or bradycardias <100/min within the last 12 hours before study entry were eligible using the same criteria as during our previous cross-over studies on different NIV modes.^{12 21 22} Infants with severe congenital malformations affecting respiratory drive, lung or cardiovascular function, scheduled blood transfusions or surgical interventions, treatment of bacterial infection ≤72 hours or active immunisation ≤48 hours prior to study entry, planned ROP examination during the study, use of automated FiO₂ control or escalation of NIV ≤12 hours before study entry were not enrolled.

Patient involvement

The ‘Förderkreis für intensivpflegebedürftige Kinder Ulm e.V.’ is the local patient and family representation to support the specific needs of the study population. The organisation was informed about the study’s aims and supported the conduct of the study.

Study design

This study consisted of two consecutive 24-hour treatment periods. After receiving the study consent, infants on nCPAP were randomly allocated to the sequence of stepwise reduction of backup NIPPV followed by immediate cessation of backup NIPPV or vice versa using sequentially numbered opaque sealed envelopes containing the two sequences of intervention in random order. The Sophie respirator (Fritz Stephan GmbH, Gackenbach, Germany) was used for the study, and the two interventions are specific to this ventilator. Apnoea duration before the onset of backup NIPPV was set to 4 s. At the child’s first spontaneous breath, background ventilation is terminated with an assisted breath. With frequency-controlled nCPAP backup ventilation, background ventilation starts identically to the other nCPAP backup ventilation mode. However, if spontaneous breathing is detected again after an apnoea, the background ventilation rate is reduced in three consecutive stages by one third of the set backup frequency, provided that no recurrent apnoea is detected. If renewed apnoea is detected, the device immediately switches back to background ventilation, and the algorithm starts again. In the fourth and fifth stages of backup ventilation, there is no further reduction in the backup frequency, but the spontaneous breathing frequency must be at least two-thirds of the backup frequency in order to jump to the next stage. If all five stages have been completed without apnoea and with a sufficient spontaneous breathing rate, backup NIPPV is ended. It takes 1 min to complete each stage.

Blinding of the intervention was not feasible. A Graseby capsule was used for synchronisation of backup NIPPV. Both backup NIPPV modes are standard of care in our NICU, and included infants were exposed to at least one of the modes prior to inclusion in the study. For the settings of backup ventilation, configurations of the last backup ventilation applied before study inclusion were used, whereby the frequency was set to 40 /min with an inspiratory time of 0.3. Prongs or a nasal mask (Fritz Stephan GmbH, Gackenbach, Germany) were used as an nCPAP interface at the discretion of the clinical team, but changing the device was not allowed during the study. The positioning of the children during the first study phase was noted and executed analogously in the second phase. Data were continuously recorded. The SpO₂ target range was kept at the NICU standard of 88%–95%. For consistency of the approach, guidelines for FiO₂ adjustment were placed at the bedside, as during the previous studies.^{12 13} The nurse-to-patient ratio was 1:2. All routine cointerventions remained unchanged, including kangaroo care, rounds and routine suctioning. Caffeine citrate is applied as respiratory stimulant in our unit routinely from birth to all EPT infants at a dosage of 20 mg/kg body weight split into two times per day dosages. Extra doses of caffeine and doxapram or changes in the dosage were not allowed during the study and administered at the same time on both study days. Criteria for prematurely terminating the study participation were as during a recent study a pH <7.1 due to CO₂ retention in two independent measurements, necrotising enterocolitis, respiratory deterioration measured by a clinical score (one point for heart rate <80 /min and/or SpO₂ <70% without tactile stimulation, three points with stimulation, study termination with 30 points in 8 hours) and the indication to start invasive ventilation based on clinical decision (online supplemental table 1).²²

Primary and secondary endpoints

The primary endpoint was the percentage of time in the SpO₂ target range (88%–95%) that was set as in our and nearly identical to other research groups’ previous studies on the topic.^{1112 21},²⁴ For children without supplemental oxygen (FiO₂=0.21), the SpO₂ target range was ≥88%.

Secondary endpoints included the percentage of time spent with an SpO₂<88%, <80% and <70% and >95% when on supplemental oxygen and the number of bradycardias <100 /min. In addition, the duration (defined as episodes with a duration of >30 s, >60 s and >120 s) of hypoxic (SpO₂<88%, <80% and <70%) and hyperoxic episodes (SpO₂>95% with FiO₂>0.21) and the duration (>10 s) of bradycardias was analysed as done before.^{12 21 22} Furthermore, mean SpO₂ and FiO₂, heart rate, respiratory rate, transcutaneous carbon dioxide pressure (tcpCO2), cerebral tissue oxygenation (SctO₂) and fluctuations in SctO₂ calculated from the median area under the curve (% * sec/h) below and above the individual SctO₂ range during each study period (defined as ±5% and ±10% of the individual SctO₂ median of each infant) were recorded.

Data recording and analyses

SpO₂ was recorded from the right hand using the Masimo Radical 7 pulse oximeter (Masimo Corporation, Irvine, California), with an averaging time of 2 s by preductal placement on the right hand. Additionally, the postductal SpO₂ signal was measured in addition to heart rate and respiratory rate with the Carescape B450 monitor (General Electric, Boston, USA). Sentec Digital Monitoring System with OxiVenT Sensor (Sentec, Therwil, Switzerland) was used for tcpCO₂ measurements. SctO₂ was always measured on the right forehead using the Masimo O3 tissue oximeter. SctO₂ values were not respected during clinical routine. Signals were recorded every second into separate folders of a specifically developed software with a time stamp and an option to mark events like patient rounds or dislocation of the nCPAP device.¹³ Data analyses for the primary outcome and the secondary endpoints of oxygenation, bradycardia and apnoea were executed automated using custom-made software solutions. Baseline characteristics were retrieved from the medical records and documentations by the clinical teams. BPD was categorised according to the 2001 National Institute of Child Health and Human Development (NICHD) consensus definition.²⁵ Alternative forms of respiratory support, including high-flow nasal cannula support and the fraction of oxygen applied by low-flow nasal cannula, were converted as done previously.²⁶

Sample size and statistical analysis

Sample size calculation was based on the results from a previous study displaying a 7.5% difference in the primary endpoint.¹² Assuming a comparable clinically relevant difference between stepwise reduction and immediate stop of NIPPV, a sample size of 18 patients was calculated using a two-sided paired t-test with a variance of the difference of 81, a two-sided type one error of 0.05 and a power of 0.90. To account for dropouts for insufficient data recording, parental withdrawal of consent and non-parametric parameter distribution, the sample size was increased by 33% to 24 patients. All statistical tests except for the primary endpoint were performed in an exploratory manner. Adjustment for multiple testing was not made.

In addition to the prespecified analysis of patients treated according to the protocol (main analysis according to protocol), we performed further analyses: the first additional analysis included only those patients for whom data were available from both intervention periods, independent of the duration of observation. Patients lacking any data for one of the study periods were excluded. To support our a priori assumption that there are no carry-over effects between the two study phases, we performed a seccond additional analysis. This analysis was based on the patient data from the main analysis according to protocol, excluding the first 6 hours of the respective treatment period in both study phases (analyses controlling potential carry-over effects) in addition to the calculatory assessment of potential sequence effects (figure 1).

Figure 1. The flowchart details inclusion and exclusion of patients into the study, the main analysis according to protocol sample and further data analyses. NIPPV, nasal intermittent positive pressure ventilation.

Continuous variables were described as mean and SD or median and range. Categorical variables were reported as absolute and relative frequencies. Fisher’s exact test was used for group comparisons of categorical variables. For continuous variables, the paired t-test or Wilcoxon signed-rank test, as appropriate, was used for group comparisons. To account for potential sequence effects of the interventions, we additionally assessed between-treatment effects by means of a two-sample t-test for independent samples using the intra-individual differences between the outcomes in both periods as the raw data. A two-sample t-test for period sums was applied as a preliminary test to check the assumption of negligible carry-over effects for the two different sequences, to control for effects of unbalanced group allocation and time period effects.²⁷ Additionally, a two-sided 95% CI was provided for the mean difference in the primary endpoint between stepwise reduction and immediate cessation of backup NIPPV. A two-sided p value ≤0.05 was considered statistically significant. All statistical analyses were performed using Spyder Integrated Development Environment software (GitHub, San Francisco, California) V.5.5.0; Excel version Office 2019 (Microsoft, Redmond, Washington); SAS, V.9.4 (The SAS Institute, Cary, North Carolina, USA) and R, V.4.5.0 (R Foundation for Statistical Computing, Vienna, Austria).

Results

24 EPT infants were enrolled. Median gestational age was 24+1 weeks (range: 22+4 to 27+6), and the median birth weight 625 g (range: 360–1210). Further patient baseline characteristics are summarised in table 1. Five patients were excluded from the study because parental consent was withdrawn or because the predefined study termination criteria were met, including SpO_₂ instability or clinical routine care interventions affecting respiratory stability during the study (figure 1). On average, infants were included in the study at a postnatal age of 43 days with a wide range from 23 to 95 days (corresponding to a corrected age of 30+6 weeks and range 27+3 to 37+0) reflecting the variable respiratory course after delivery and the manifestation of respiratory instability over the course in the NICU.^{28 29} The median body weight at study inclusion was 1135 g (range 565–2470). Prestudy respiratory settings were on average 6 cm H₂0 for the PEEP and a median FiO₂ of 0.28 (range: 0.21–0.35) (further details in table 1). Allocation to the sequence of interventions was almost balanced with eight infants starting with stepwise reduction of backup NIPPV, while 11 infants were first allocated to immediate cessation of backup NIPPV. There was no difference in clinical routine measures during the two intervention periods, including selection of the nCPAP device, nurse-to-patient ratio and kangaroo care.

Table 1. Baseline characteristics of included patients and characteristics at study entry (n=19).

Characteristics	Median and range or N (%)
Maternal and infant baseline characteristics
Gestational age (weeks), median (IQR)	24+1 (22+4 to 27+6)
Birth weight, g	625 (360–1210)
Small for gestational age (<10th percentile)	1 (5%)
Male sex	7 (37%)
Singleton birth	15 (79%)
Maternal age, years	31 (20–38)
Antenatal steroids before birth	19 (100%)
Pre-eclampsia/HELLP	2 (11%)
Chorioamnionitis	11 (58%)
PPROM	11 (61%), na=1
Mode of delivery: C-section	18 (95%)
Umbilical cord pH	7.35 (7.16–7.46), na=1
Apgar 1 min	6 (1–8), na=1
Apgar 5 min	7 (4–9)
Apgar 10 min	9 (8–10)
Surfactant therapy	19 (100%)
Doses of surfactant	1 (1–4)
Postnatal corticosteroids <7 days after birth	0 (0%)
Postnatal corticosteroids ≥7 days after birth	5 (26%)
Invasive mechanical ventilation	15 (79%)
No BPD	3 (16%)
Mild BPD	8 (42%)
Moderate BPD	4 (21%)
Severe BPD	4 (21%)
Intraventricular haemorrhage (any grade)	5 (26%)
Focal intestinal perforation	2 (11%)
Necrotising enterocolitis	1 (5%)
Retinopathy of prematurity (any grade)	16 (84%)
Retinopathy of prematurity (therapy)	3 (16%)
Infant characteristics at study entry
Gestational age at study entry, weeks	30+6 (27+3 to 37+0)
Postnatal age at study entry, days	43 (23–95)
Weight at study entry, g	1135 (565–2470)
Prestudy PEEP, cm H2O	6 (3–8)
Prestudy FiO2, cm H2O	0.28 (0.21–0.35)
Caffeine therapy, n (%)	19 (100%)
Caffeine dosage, mg/kg/d	20 (20–20)
Doxapram therapy, n (%)	1 (5%)
Doxapram dosage, mg/kg/h	0.00 (0.00–0.90)

Data are displayed as median and range or n (%).

Bronchopulmonary dysplasia (BPD) was categorised according to the 2001 NICHD consensus definition.²⁵ Alternative forms of respiratory support, including high-flow nasal cannula support and the fraction of oxygen applied by low-flow nasal cannula, were converted as done previously.²⁶

BPD, bronchopulmonary dysplasia; C-section, caesarian section; FiO₂, fraction of inspired oxygen; HELLP, haemolysis, elevated liver enzymes and low platelet count; na, cases with missing data; PEEP, positive end-expiratory pressure; PPROM, preterm premature rupture of membranes; ROP, retinopathy of prematurity.

With a mean difference: 3.6±10.5 (95% CI −1.5 to 8.7; p=0.15), the primary outcome of time spent within the SpO₂ target range did not differ significantly between the two interventions for our main analysis sample (main analysis according to the protocol): 64.7±10.8 (mean for stepwise reduction of backup NIPPV) and 61.1±13.5 (mean for immediate cessation of backup NIPPV), (figure 2a, table 2a). Adjusting the analyses for potential sequence effects of the interventions did not alter the results. The mean time spent within the SpO_₂ target range was 64.7±10.8 during stepwise reduction of backup NIPPV and 61.1±13.5 during immediate cessation of backup NIPPV, yielding a mean difference of 3.3±5.3 (95% CI −1.4 to 8.0; p=0.15). Similar results were obtained when SpO_₂ was analysed using recordings from the Carescape monitor (72.4±11.6 vs. 70.0±12.0; p=0.13). The findings also remained consistent when including all patients with available recordings from both intervention periods (analysis including all patients for whom data were available in both intervention periods, independent of the observation interval), with mean values of 63.7±10.9 for stepwise reduction and 59.8±13.2 for immediate cessation of backup NIPPV (p=0.08; online supplemental table 2). To exclude potential carry-over effects, the primary outcome was reanalysed after excluding the first 6 hours of recordings for both interventions (main analysis controlling potential carry-over effects). The mean time within the SpO_₂ target range remained comparable, at 64.4±11.2 for stepwise reduction and 60.9±14.2 for immediate cessation of backup NIPPV (p=0.26; online supplemental table 3). The data show no evidence for possible carry-over effects for our main analysis according to the protocol (non-significant result from the two-sample t-test applied to the period sums (p=0.13)).

Figure 2. Time spent within the SpO₂ target range during various NIPPV backup modes. (a) Distribution of time spent within the SpO₂ target range (88%–95% or ≥88 with FiO₂=0.21) during stepwise reduction of backup NIPPV following apnoea for >4 s during nCPAP backup mode (red boxplot) and for immediate cessation of backup NIPPV on restart of spontaneous respiration (orange boxplot). MAP data are given as mean and SD. Statistical analysis was executed with paired t-tests, *p<0.05. (b) Patients included in the main analyses from (a) were segregated for the time spent within the SpO₂ target range during the two NIPPV backup modes. For n=9 patients, the time with SpO₂ within the target range was higher for the stepwise reduction of backup NIPPV (green lines), while in n=10 patients, immediate cessation of backup NIPPV (blue lines) was superior. Data are given as percentages. 19: patient with an outstanding percentage of time spent above the SpO₂ target range. (c) Patients included in the main analyses from (a) were furthermore segregated for the treatment sequence. FiO₂, fraction of inspired oxygen; MAP, mean airway pressure; nCPAP, nasal continuous positive airway pressure; NIPPV, nasal intermittent positive pressure ventilation; SpO2, oxygen saturation.

Table 2. Outcome data for immediate cessation of backup NIPPV on restart of spontaneous respiration or stepwise reduction of backup NIPPV for the main analysis sample (main analysis according to the protocol) (n=19).

Primary outcome
Variables	Stepwise reduction of backup NIPPV	Immediate cessation of backup NIPPV	Estimated treatment difference	95% CI	P value
Time (%) within the SpO2 target range (88%–95% or ≥88% with FiO2=0.21)	64.7 (10.8)	61.1 (13.5)	3.6 (10.5)	−1.5 to 8.7	0.15

Secondary outcomes
Variables	Stepwise reduction of backup NIPPV	Immediate cessation of backup NIPPV	P value
Time (%) below the SpO2 target range (88%)	22.6 (8.9)	22.4 (10.6)	0.91
Time (%) above the SpO2 target range (95%)	12.7 (7.8)	16.6 (12.0)	0.18
Time (%) with SpO2<80%	5.3 (4.2)	6.0 (4.4)	0.46
Time (%) with SpO2<70%	1.2 (1.7)	1.3 (1.5)	0.79
Hypoxaemias (<88%) ≥30 s/h	6.0 (2.8)	5.9 (3.0)	0.92
Hypoxaemias (<88%) ≥60 s/h	2.3 (1.3)	2.4 (1.7)	0.87
Hypoxaemias (<88%) ≥120 s/h	0.6 (0.4)	0.7 (0.8)	0.5
Hypoxaemias (<80%) ≥30 s/h	0.9 (0.8)	1.2 (1.1)	0.26
Hypoxaemias (<80%) ≥60 s/h	0.2 (0.2)	0.3 (0.4)	0.28
Hypoxaemias (<80%) ≥120 s/h	0.0 (0.1)	0.1 (0.1)	0.8
Hypoxaemias (<70%) ≥30 s/h	0.2 (0.2)	0.2 (0.3)	0.79
Hypoxaemias (<70%) ≥60 s/h	0.0 (0.0)	0.0 (0.1)	0.76
Hypoxaemias (<70%) ≥120 s/h	0.0 (0.0)	0.0 (0.0)	0.96
Mean SpO2	90.5 (1.6)	90.8 (2.1)	0.49
Mean FiO2	26.3 (5.1)	26.3 (4.6)	0.91
MAP	7.2 (1.2)	5.6 (1.4)	<0.01*
Pulse rate	165.8 (8.2)	166.3 (8.9)	0.57
Bradycardia (<100 /min) ≥10 sec/h	0.3 (0.4)	0.3 (0.4)	0.81
tcpCO2	46.1 (5.6)	47.4 (3.6)	0.23
Respiratory rate	60.1 (7.9)	63.8 (10.4)	0.02*
Mean SctO2	64.45 (38.92–71.45)	63.45 (50.73–73.05)	0.86
SctO2 above 5% threshold ((%*s)/h)	511.71 (0.21–9466.33)	560.80 (0.00–8253.40)	0.98
SctO2 below 5% threshold ((%*s)/h)	847.98 (1.58–6520.44)	946.20 (50.60–5309.90)	0.57
SctO2 above 10% threshold ((%*s)/h)	12.10 (0.00–4731.28)	24.82 (0.00–4186.55)	0.7
SctO2 below 10% threshold ((%*s)/h)	50.10 (0.00–2398.62)	88.76 (0.00–2088.80)	0.42

Data are presented as mean (SD) or n (%). SctO₂ data are presented as median and range.

Group comparison of continuous data: paired t-test or Wilcoxon signed rank test (SctO₂ data).

^*p<0.05.

FiO₂, fraction of inspired oxygen; MAP, mean airway pressure; NIPPV, nasal intermittent positive pressure ventilation; SctO₂, cerebral tissues oxygen saturation; SpO₂, oxygen saturation; tcpCO₂, transcutaneous carbon dioxide pressure.

On an individual patient basis, the time spent within the SpO₂ target range was higher during stepwise reduction of backup NIPPV for nine patients, while for the other 10 patients this was higher with immediate cessation on restart of spontaneous breathing (figure 2b). The sequence of interventions for the individual patient is additionally displayed in figure 2c. When comparing infant baseline characteristics separated for the intervention with a higher percentage of time spent within the SpO₂ target range, there were no significant differences apart from birth weight (table 3). For the secondary outcomes of oxygenation, there was no difference for the time spent below or above the SpO₂ target range and for the number of prolonged and severe hypoxaemic episodes and hyperoxaemias (table 2b). As well, there was no variation in the parameters of cerebral tissue oxygenation, including mean SctO₂, the proportion of time spent with SctO₂ above and below the 5% and 10% thresholds, the mean SpO₂, the mean heart rate, the mean tcpCO₂ and the mean applied FiO₂. However, the prolonged backup NIPPV support resulted in a significant increase in MAP during stepwise reduction of backup NIPPV (7.2±1.2 cm H₂0 vs 5.6±1.4 cm H₂0, p<0.01) and a significant reduction of the spontaneous respiratory rate (60.1±7.9 vs 63.8±10.4, p=0.02) (table 2b).

Table 3. Characteristics of patients in favour of immediate cessation of backup NIPPV on restart of spontaneous respiration or stepwise reduction of backup NIPPV from the main analysis sample (main analysis according to the protocol).

	Patients in favour of stepwise reduction of backup NIPPV (n=9)	Patients in favour of immediate cessation of backup NIPPV (n=10)	P value
Gestational age at birth, weeks	23+3 (22+4 to 27+3)	24+3 (23+2 to 27+6)	0.14
Birth weight, g*	460 (360–835)	695 (530–1210)	0.02*
Gestational age at study entry, weeks	30+0 (28+0 to 37+0)	31+0 (29+0 to 34+6)	0.55
Postnatal age at study entry, days	34 (23–95)	43 (25–53)	0.78
Weight at study entry, g	1120 (565–2290)	1140 (1005–2470)	0.49
Small for gestational age, n (%)	1 (11%)	0 (0%)	0.47
Male gender, n (%)	2 (22%)	5 (50%)	0.35
Singleton birth, n (%)	8 (89%)	7 (70%)	0.58
Maternal age, years	31 (26–37)	31 (20–38)	0.81
Antenatal steroids before birth, n (%)	9 (100%)	10 (100%)	–
Pre-eclampsia/HELLP, n (%)	1 (11%)	1 (10%)	1.00
Chorioamnionitis, n (%)	3 (33%)	8 (80%)	0.07
PPROM, n (%)	4 (50%)	7 (70%)	0.63
Mode of delivery: C-section, n (%)	9 (100%)	9 (90%)	1.00
Umbilical cord pH	7.34 (7.16–7.44)	7.36 (7.28–7.46)	0.28
Apgar 1 min	4 (1–7)	6 (1–8)	0.38
Apgar 5 min	7 (5–9)	8 (4–9)	0.39
Apgar 10 min	9 (8–10)	9 (8–10)	0.47
Surfactant therapy, n (%)	9 (100%)	10 (100%)	–
Doses of surfactant	1 (1–4)	1 (2–3)	0.72
Postnatal corticosteroids <7 days after birth, n (%)	0 (0%)	0 (0%)	–
Postnatal corticosteroids ≥7 days after birth, n (%)	3 (33%)	2 (20%)	0.63
Invasive mechanical ventilation, n (%)	8 (89%)	7 (70%)	0.58
No BPD, n (%)	2 (22%)	1 (10%)	0.92
Mild BPD, n (%)	3 (33%)	5 (50%)
Moderate BPD, n (%)	2 (22%)	2 (20%)
Severe BPD, n (%)	2 (22%)	2 (20%)
Intraventricular haemorrhage (any grade), n (%)	1 (11%)	4 (40%)	0.30
Focal intestinal perforation, n (%)	0 (0%)	2 (20%)	0.47
Necrotising enterocolitis, n (%)	1 (11%)	0 (0%)	0.47
Retinopathy of prematurity (any grade), n (%)	8 (89%)	8 (80%)	1.00
Retinopathy of prematurity (therapy), n (%)	1 (11%)	2 (20%)	1.00

Data are presented as median and range or n (%).

Group comparison of continuous data: Wilcoxon signed rank test. Group comparison of categorical data: Fisher’s exact test.

^* p<0.05.

BPD, bronchopulmonary dysplasia; C-section, caesarian section; HELLP, haemolysis, elevated liver enzymes and low platelet count; NIPPV, nasal intermittent positive pressure ventilation; PPROM, preterm premature rupture of membranes; ROP, retinopathy of prematurity.

Discussion

This study contributes to our limited understanding of the effects of varying the mode of NIV and stability of oxygenation in EPT infants. This is a research area of utmost importance, as these infants are at particular risk for frequent fluctuations in SpO₂ that are associated with an increased risk of mortality and the severe outcomes of premature delivery.^{2 6 8} We show here that stepwise reduction of backup NIPPV following apnoea was not superior to immediate cessation on restart of spontaneous respiration in EPT infants on nCPAP. At first sight, the 3.6% difference in mean time spent within the SpO₂ target range between the two backup modes might suggest that we have potentially missed a benefit due to the sample size. However, patients were evenly distributed in terms of time spent within the target range across both modes. Moreover, when the single outlier patient—who spent an unusually high proportion of time above the SpO_₂ target—was excluded, the observed difference was reduced by half (49.9%). Finally, our observed difference between the two interventions is minor and appears negligible in comparison to the more pronounced, statistically significant effects reported in other studies with similar design and sample size.^{11 12 21} Moreover, the primary outcome remained unchanged when patients who met the study’s predefined termination criteria were included in the analysis (analysis including all patients for whom data were available in both intervention periods, independent of the observation interval) (figure 1 and online supplemental table 2). Grouping patients according to the intervention in which they spent more time within the SpO_₂ target range did not reveal any differences in baseline characteristics or clinical parameters at inclusion, except for birth weight. This makes it unlikely that a benefit is limited to a specific patient subgroup or related to postnatal age. Due to our inclusion criteria and eligibility of patients, the postnatal age range at inclusion (day 23–95) was large. Lastly, we did not observe carry-over effects between the time spent below and above the SpO₂ target range by the interventions.

Another important aspect is that the only differences detected were for MAP and respiratory rate. When putting these findings into the context of the published literature, this is consistent with what the literature reports: The nearly two-point increase in MAP during stepwise backup reduction is most likely due to the prolonged NIPPV backup, as studies show that the major benefit compared with nCPAP with a lower distending airway pressure is the stabilisation of respiration.¹⁶ This is also reflected in the lower spontaneous respiratory rate during this intervention. That the mean tcpCO₂ remained unchanged is not surprising, as NIPPV does not lead to changes in lung tidal volumes and does not augment CO₂ elimination as during nHFOV.^{30 31} The unchanged mean FiO₂ applied between the two backup modes is another argument that the NIPPV support does not reach the alveolar level and therefore does not improve alveolar gas exchange but instead primarily affects breathing mechanics.^{1630 32},³⁴

The major strength of our study is the cross-over design and adequate power to detect clinically meaningful differences for SpO₂ targeting.¹² We excluded bias by applying all further routine clinical care interventions that could impact respiratory stability, like positioning of the infant and kangaroo time, in the same way during both study periods.³⁵,³⁷ For the two patients withdrawn based on clinicians’ subjective assessment of increased instability, the objective recordings did not confirm any such instability. Hereby, our study underlines the importance of objective verification of respiratory instability instead of individual judgement that can differ within and between observers. We included infants with a median gestational age of 24+1 weeks who were mostly still receiving supplemental oxygen at study entry and spent, on average, <65% of the time within the SpO_₂ target range. By focusing on this population, we studied the most vulnerable and unstable patients—those who are likely to gain the greatest benefit if time within the SpO_₂ target range can be increased. The time infants spent within the SpO₂ target range was comparable to that of similar cohorts and of our previous cross-over studies analysing different modes of NIV.^{12 21 22 38 39} Moreover, our study was executed within a real-life scenario and included situations and interventions during a full 24-hour period. Lastly, the automated software-based analyses allowed the unbiased and observer-independent evaluation of the recordings over a prolonged period of 24 hours that covers the total circadian cycle. We did not detect carry-over effects between the two periods for our main analysis (main analysis according to the protocol).

However, our study has limitations. First, we were not able to distinguish between the different causes for fluctuations outside the SpO₂ target range, as our recordings do not allow separation into central apnoea, inspiration hold and active exhalation.⁴⁰ Second, infants were not eligible during their first week of life, and the majority of infants were included beyond the first 4 weeks of life. Therefore, our results are not applicable to the initial stabilisation after preterm delivery and the early postnatal phase. Third, our study was undertaken at a single centre, and we used one type of ventilator throughout. Moreover, we used a random pressure gradient for selecting the peak inspiratory pressure based on the last prestudy settings but a standardised inspiratory time during NIPPV. However, it is unlikely that minor changes in the settings would have altered the outcomes, as pressure transmission cannot even approximate the efficacy as during invasive mechanical ventilation due to the large air leakage described during nCPAP.³⁵ Last, we are aware that our sample size calculation was based on another trial comparing other types of NIV ventilation, and the sample size (n=19 patients) was small. Thus, a type 2 error cannot categorically be excluded.¹²

Conclusion

Clinical priority on NIV respiratory support of preterm infants requires focus on keeping them stably within the SpO₂ target range to avoid the harm arising from frequent prolonged and severe desaturations and hyperoxia episodes.^{2 6 9 41} Our study shows that the use of new sophisticated modes of NIV respiratory support, such as stepwise reduction of backup NIPPV, is not necessarily associated with an improvement in the time spent in the SpO_₂ target range in our study population. Based on this, the intensification of backup ventilation in cases of respiratory instability and repeated desaturation events, which is frequently used in clinical practice, should be reconsidered. Although we did not observe a significant difference in this study, our recent work comparing sNIPPV and nHFOV and investigating prolonged inspiratory time during sNIPPV underscores the ongoing potential for research to further optimise NIV strategies.^{12 21}

Data availability statement

Data are available upon reasonable request.