Fluctuations in Oxygen Saturation during Synchronized Nasal Intermittent Positive Pressure Ventilation and Nasal High-Frequency Oscillatory Ventilation in Very Low Birth Weight Infants: A Randomized Crossover Trial Open Access

Very preterm (VLBW) infants get exposed to a hyperoxic environment ex utero, but the fraction of inspired oxygen (FiO₂), as well as the exposure time vary greatly, which dramatically impacts the extent of injuries in preclinical models [1, 2]. Reactive oxygen species production is boosted by hyperoxia but also by hypoxemia with subsequent reoxygenation [3, 4]. This latter finding is of utmost importance as preterm infants experience a magnitude of hypoxemic events, particularly when breathing spontaneously, and their antioxidative defense mechanisms are poorly developed [5]. Association studies linked the amount and severity of fluctuations in oxygen saturation (SpO₂) to all important outcomes, including mortality, bronchopulmonary dysplasia (BPD), retinopathy of prematurity (ROP), and impaired psychomotor development [3, 4, 6]. Hence, SpO₂ targeting has become a research priority.

With the improvements in noninvasive ventilation (NIV) and surfactant application, fewer VLBW infants require invasive mechanical ventilation. These advances aim to reduce the BPD burden, but hyperoxic and hypoxemic episodes pose a therapeutic challenge that cannot be solely tackled by medication treating central apneas [1, 5]. Continuous positive airway pressure (CPAP) is still the best-studied form of NIV and experimental studies demonstrated that CPAP efficiently reduces the oxidative burden as well as lung injuries [7]. Newer modes of NIV, like nasal intermittent positive pressure ventilation (NIPPV) or nasal high-frequency oscillatory ventilation (nHFOV), were introduced into daily clinical practice to improve its efficacy. NIPPV was shown to be superior to CPAP as primary ventilatory support to avoid mechanical ventilation or as the first-line mode after extubation. NIPPV is more efficient when applied in a synchronized (sNIPPV) manner [8]. nHFOV decreases the risk of intubation compared to CPAP when used as primary respiratory support and post-extubation [9, 10]. But neither sNIPPV nor nHFOV had a positive effect on the important outcomes of prematurity compared to CPAP [8, 9]. One study in more mature infants demonstrated no difference in extubation failure between NIPPV and nHFOV, but nHFOV was more efficient in CO₂ elimination [11]. Based on these studies, proponents of nHFOV argue its application in NICU during the weaning process.

Primary aim of this randomized crossover study was to establish whether sNIPPV or nHFOV was better suited to maintain SpO₂ within a preset target range in respiratory unstable VLBW infants on NIV with supplemental oxygen. Further analyses were focused on the number and duration of hypoxic and hyperoxic episodes.

All preterm infants <32 weeks with a birth weight <1,500 g and postnatal age >72 h at the tertiary level NICU of the perinatal center Giessen on sNIPPV without: severe congenital malformations, scheduled blood transfusion or surgical interventions, treatment for bacterial infection for the last 72 h before study entry and escalation of NIV for desaturations <70% for >1 min or bradycardias <100/min for >30 s within 12 h prior to study entry were considered eligible for this randomized crossover trial. Prerogative for study inclusion was the persistent dependency on NIV with a positive end-expiratory pressure (PEEP) ≤8 cmH₂O, need for supplemental oxygen ≤60% and at least four hypoxemic episodes with SpO₂ <80% for >30 s within the last 12 h before study entry documented in the electronic monitoring system.

Using sealed envelopes, infants were randomly assigned to start with sNIPPV or nHFOV using a Sophie respirator (Fritz Stephan GmbH, Gackenbach, Germany) for 8 h. Blinding of the interventions was not feasible. The second observation period was started the next day at around the equivalent time as during the first study phase. CPAP was used prior and between the two interventional periods to minimize carryover effects with the use of sNIPPV or nHFOV (shown in Fig. 1). The sNIPPV settings during the study were as during sNIPPV before CPAP wash-out with an inspiratory time of 0.3 s and a mandatory synchronized rate of 40/min (SIMV-mode). The Graseby capsule was placed in subxiphoid position as recommended to allow synchronization during NIPPV. Reaction time of the Grasby capsule is 50 msec and synchronization rate >80% in comparable settings [12]. Mean airway pressure (MAP) during nHFOV was set as during sNIPPV before study entry or during the first intervention period depending on the allocated sequence of NIV. Oscillation amplitude was started at 20 cmH₂O, 40% inspiration:expiration ratio and a frequency of 10 Hz. nHFOV amplitude was adapted to reach identical transcutaneous pCO₂ as during sNIPPV. Either prongs or mask were used as CPAP interface and infants were kept identically in prone or supine position during both study phases at the discretion of the attending nurse. The mouth was not closed with a chin strap. Parents were allowed to kangaroo care for up to 50% of the observation period. Data were continuously recorded without interruption during infant handling. The SpO₂ target was kept at the NICU standard of 88–95%. Physicians and nurses were advised to a standard protocol of FiO₂ adjustment as published [13]. The nurse-to-patient ratio was at least 1:2. All routine co-interventions including feeding intervals and medical therapy of apnea with caffeine citrate and doxapram remained unchanged during the total study period. The criteria for study discontinuation (start of invasive mechanical ventilation at the discretion of the attending physician, respiratory acidosis with a pH <7.1 in two subsequent blood gas analyses or necrotizing enterocolitis) were not fulfilled in any participant.

The primary outcome was time spent within the SpO₂ target (SpO₂ 88–95%). Time spent above the SpO₂ target was not taken into account for hyperoxemia calculation when the infant was breathing room air. Sample size calculation was based on the findings from a previous study displaying a 7.5% difference [13]. Assuming comparable clinically relevant differences between sNIPPV and nHFOV, a sample size of 18 patients was calculated using a two-sided paired t test with an inner subject variance of 81, a two-sided type one error of 0.05 and α power of 0.90. To account for dropouts due to insufficient data recording quality, patient consent withdrawal and nonparametric parameter distribution, the sample size was increased by 33% to 24 patients. All statistical tests with exception for the primary endpoint were performed in an exploratory manner and have to interpreted as hypothesis generating only and not as confirmatory. An adjustment for multiple testing was not made.

Secondary outcomes included the time spent in hypoxemia (SpO₂ <88%), in severe hypoxemia (SpO₂ <80%), and in iatrogenic hyperoxemia (SpO₂ >95% with FiO₂ >0.21). Additionally, the number of hypoxemic (SpO₂ <88%) and severe hypoxemic (SpO₂ <80%) and of bradycardia episodes (<100/min) for ≥30 s, ≥60 s, and ≥120 s, of FiO₂ adjustments, mean cerebral tissue oxygen saturation (SctO₂) and time of SctO₂ <65% and <60% were evaluated. Furthermore, the mean SpO₂, respiratory frequency, heart rate, transcutaneous pCO₂ and FiO₂ during each study phase were calculated.

Baseline demographic parameters were recorded and definitions used as described [14]. Silverman score and abdominal circumference measured 1 cm above the navel were recorded before and after the intervention. The SpO₂ sensor was attached to the right upper extremity. SpO₂, heart rate, and respiratory frequency were recorded from an IntelliVue MP50 monitor (Philipps, Amsterdam, The Netherlands, SpO₂ averaging time 8 s). Ventilation parameters and FiO₂ were retrieved from the respirator and any adjustment documented. Continuous measurement of SctO₂ by near-infrared spectroscopy was executed using a Root kit 3.0 device (Massimo, CA). Signals were recorded at 2-s intervals into separate data folders of a specifically developed software solution with a time stamp and the option to mark events including kangaroo care, patient rounds, and dislocation of the CPAP device.

Patients with data acquisition for <75% of recording time in either of the study arms were excluded from the analyses. Continuous variables were described as mean and standard deviation or median and quartiles. Categorical variables were reported as absolute and relative frequencies. For continuous variables, the paired t test or Wilcoxon signed-rank test, as appropriate, was used for statistical comparisons. A two-sided p value ≤0.05 was considered to be statistically significant. Statistical analyses were performed using Sigma Plot (Systat Software Inc., CA) version 12.3 and R, version 4.1.2.

Data sets from 22 VLBW infants with a median gestational age of 25+4 weeks (range 22+3 to 28+0) at birth and a median postnatal age of 26.5 days (range 10.0–84.0) at study inclusion were available for analysis. Records from 2 patients did not fulfill the prespecified criteria. Further patient baseline characteristics are presented in Table 1. The mean pre-study PEEP was 5.8 (5.0–8.0) cmH₂O and FiO₂ 0.26 (0.23–0.45) (Table 2). Allocation to the intervention sequences was balanced with 11 infants each first assigned to sNIPPV or nHFOV. Overall, 14 of 22 caregivers made use of the option for kangaroo care and time periods did not differ significantly between both study phases (Table 3).

MAP and transcutaneous pCO₂ did not differ between both interventions creating comparable conditions of respiratory support. The primary outcome of SpO₂ within the target was significantly higher during sNIPPV compared to nHFOV (Table 3, shown in Fig. 2). An individual patient analysis revealed that 16 out of 22 spent more time in the SpO₂ target during sNIPPV.

For the secondary outcomes, the percentage of time spent in hypoxemia and the mean FiO₂ were assessed and found to be significantly lower during sNIPPV than nHFOV, while the mean SpO₂, the number and duration of severe hypoxemic episodes (SpO₂ <80%) per hour of recording using subcategorizing criteria, the total duration in hyperoxemia, number of bradycardic episodes, and the number of manual FiO₂ adjustments did not differ between both interventions (Table 3; shown in Fig. 2). As the time spent with SpO₂ <70% was very low in our study population (1.5 ± 0.3% for sNIPPV vs. 1.2 ± 0.3% for nHFOV), we did not detail these analyses. In contrast to the SpO₂ measurements, there was no significant difference in the mean SctO₂ and the proportion of time spent with SctO₂ <65% and <60%. SctO₂ <55% represented a rare event in both intervention arms (<0.2% of the time). The mean respiratory frequency was significantly lower during nHFOV, while mean heart rate, Silverman score and abdominal circumference did not differ between the groups (Table 3).

The association of fluctuations in SpO₂ with impaired neurodevelopmental outcome, ROP and BPD is supported by large multicenter studies [3, 4, 6]. Preterm infants are particularly vulnerable for desaturations followed by reoxygenation when on NIV [15]. This sequence boosts ROS production with increased risk for ROS-related diseases [5, 6]. In this randomized crossover study, sNIPPV proved superior in maintaining SpO₂ within a predefined target range when compared to nHFOV. Furthermore, our data indicate a tendency that sNIPPV better prevents the outcome relevant events of prolonged and severe hypoxemias [4, 6]. This efficacy was not achieved at the expense of increased hyperoxemia, higher FiO₂, or more manual FiO₂ adjustments. The increased remaining within the SpO₂ target range during sNIPPV even enabled reduced mean FiO₂ requirements while higher mean FiO₂ during nHFOV might be due to the tolerance of the NICU staff when faced with more time spent outside the oxygen saturation target range [1, 5].

NIV is the primary method of respiratory support in the delivery room and standard of care during weaning after extubation. Therefore, every preterm infant is exposed to at least one mode of NIV for a shorter or longer period during the stay in the NICU. For the established NIV modes, actual meta-analyses favor sNIPPV compared to CPAP [16, 17]. During recent years, several new NIV modes have entered the clinical arena. Scientific data on their short-term advantages remain limited [17, 18]. Proponents of nHFOV argue that it provides more efficient pCO₂ elimination [10]. In this study, two widely used NIV modes were studied at equivalent respiratory settings for MAP and tcCO₂ to compare their efficacy in maintaining the SpO₂ within the target range. The results show clinically relevant variations in the primary outcome in an extent equivalent to the differences observed between manual and automated FiO₂ control [13, 19]. We did not observe significant differences in mean SctO₂ and below predefined limits which are in line with previous comparable studies where variations of SpO₂ within the target did not result in differences in SctO₂ measurements [13, 20, 21]. Although no impact on SctO₂ measurements was detected, the desaturations episodes are of major concern with respect to ROS production and their impact on the acute and long-term morbidities in preterm infants [1, 5]. The higher respiratory rates during sNIPPV need to be interpreted with caution as the recording was done by ECG-based impedance measurements and the sNIPPV mode with 40 mandatory breaths may have impacted the detected respiratory frequency. Otherwise, less respiratory effort and frequency might have been required during nHFOV to keep ventilation constant at the expense of higher FiO₂ requirements. Another explanation might be the suppression of the respiratory drive during nHFOV as previously observed in newborn lambs [22]. Abdominal distension, another concern associated with respiratory instability, did not differ between the groups. Our results support an intensified research focus on SpO₂ targeting to reduce the ROS burden which is driven not only by the constant exposure to oxygen but also by the fluctuations in SpO₂.

The major strength of our analysis is the comprehensive approach to multimodal signal recording of a magnitude of relevant patient parameters within a clinical routine setting. As SpO₂ targeting is particularly difficult in respiratory unstable infants on NIV, our data are of high clinical importance [13, 19]. As no other differences of such high relevance were observed when comparing these two NIV modes, sNIPPV should be the first priority in respiratory unstable infants when either sNIPPV or nHFOV is indicated.

Due to the limited duration of study intervals and the inclusion of patients with varying baseline characteristics, we are not able to specify the benefit to a precise gestational and postnatal age. But due to the strict inclusion criteria, pre-study respiratory parameters displayed a high inter-patient degree of correspondence within a patient population of very immature and respiratory unstable infants who have the highest risks for long-term morbidities. In line, the time spent within the SpO₂ target during sNIPPV with 59.9% was lower than in the majority of studies on this topic but within the expected range of comparable cohorts [23, 24]. This might have been overlooked in a more stable study population with less severe respiratory course. We need to acknowledge that an increase in mean airway pressure during nHFOV might have changed the outcome. Furthermore, the translation of our results to non-synchronized NIPPV that is more widely available in neonatal intensive care units is not permitted [16, 17]. Another limitation is the decreased lower limit of the SpO₂ target compared to the recommended 91% that was used during clinical routine. This probably has aggravated the time spent in hypoxemia [19, 20]. We used a standardized protocol for FiO₂ titration that had been tested in such crossover settings before. However, it is not validated for reaction accuracy compared to FiO₂ adjustments exerted by experienced NICU staff based on the individual situation estimate that might have exaggerated the duration of time spent in hypoxemia [15]. Furthermore, adherence to the protocol was not monitored. On the other hand, the averaging time of 8 s of the used SpO₂ monitor might have led to an underestimation of SpO₂ fluctuations. Lastly, we did not detect an effect on SctO₂. Compensation mechanisms for maintaining cerebral oxygenation in situations of impaired SpO₂ can explain this disparity [13, 21, 25]. Nonetheless, the data from the Canadian oxygen trial argue toward a relevant impact of prolonged fluctuations in SpO₂ [4, 6].

Even more than 50 years after the first description of NIV in preterm infants, the evidence for superiority of one mode over the other remains limited [6, 16]. The results from the large oxygen targeting trials highlighted this scientific need [26]. The NICU was littered with a variety of novel NIV modes during the recent decade and their implementation was based on theoretical considerations and best individual outcome measures. SpO₂ targeting during nHFOV was not in the focus of research. The presented data disclose its inferiority in that aspect compared to sNIPPV. In more general terms, our results underpin the need and opportunity to put research priorities on the optimal mode of NIV to retain oxygen targeting.

We highly appreciate the continuous and excellent assistance of the medical staff of the Neonatal Intensive Care Unit at the Perinatal Center Giessen. We thank all the parents of the infants for their consent to participate in the study. This work is part of the MD thesis of S.A.

The study was reviewed and approved by the Local Ethics Committee of the medical faculty of the Justus-Liebig-University Giessen (Az 149/19) and was executed according to the Declaration of Helsinki. The study was registered at the German Clinical Trials Register (DRKS00023438, https://drks.de/search/en/trial/DRKS00023438). Written parental informed consent was acquired prior to study inclusion. Patients were recruited between September 2020 and October 2021.

The authors have no conflicts of interest to declare.

This research received funding from Menschen fuer Kinder, Solms-Albshausen, Germany (to H.E.), and from a research grant of the University Medical Center Giessen and Marburg (UKGM KOOPV #07/2018 GI to M.W.). The funders had no role in the design or conduct of the study, data acquisition, data management, data analysis and interpretation of the data, manuscript preparation, revision, and final approval for submission.

Svilen Atanasov: conceptualization (supporting); data acquisition (lead); data analysis; and revising the article for important intellectual content (supporting). Constanze Dippel and Rahel Schuler: data acquisition (supporting); data analysis; and revising the article for important intellectual content (supporting). Dupleix Takoulegha: development data acquisition tool (lead) and writing and revising the article for important intellectual content (supporting). Anita Windhorst, Inez Frerichs, and Claas Strodthoff: data analysis and revising the article for important intellectual content (supporting). Jens Dreyhaupt: data analysis (lead); revising the article for important intellectual content (supporting). Markus Waitz: conceptualization (supporting); writing manuscript; and revising the article for important intellectual content (supporting). Keywan Sohrabi: conceptualization (lead); development data acquisition tool (lead); and writing and revising the article for important intellectual content (supporting). Harald Ehrhardt: conceptualization (lead); writing manuscript and revising the article for important intellectual content (lead). All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Deidentified individual participant data including data dictionaries will be made available upon publication to researchers who provide a reasonable and methodologically sound proposal for use in achieving the goals of the approved proposal. Proposals should be submitted to the corresponding author.