Nasal continuous positive airway pressure (NCPAP) is often used as respiratory support in infants with apnoea of prematurity (AOP).1 The physiological effects of NCPAP include improved oxygenation2 3 and lung function,4 5 reduced upper airway resistance,6 7 stenting of the upper airway,8 preservation of lung volume9 and, at least with some devices, reduced work of breathing.10

Following publication of a meta-analysis in 2002,11 nasal intermittent positive pressure ventilation (NIPPV), in addition to caffeine, has been increasingly used instead of NCPAP to treat AOP. It was our clinical impression, however, that other devices delivering NCPAP or NIPPV, such as bubble NCPAP or variable flow devices, are equally effective. We therefore evaluated three NCPAP/NIPPV devices, compared to a conventional ventilator in NIPPV mode, for their effect on bradycardia and desaturation events, the most important symptoms of AOP. Our hypothesis was that the three systems were as effective in treating AOP as a conventional ventilator delivering NIPPV.

METHODS

Patients

Between June 2004 and January 2006, inborn infants admitted to the neonatal intensive care unit at Tuebingen University Hospital were screened for eligibility. Inclusion criteria were: (i) gestational age at birth <34 weeks, (ii) postmenstrual age and body weight at time of study ⩽38 weeks and >1000 g, respectively, and (iii) a requirement for NCPAP to treat AOP as judged by the attending neonatologist. Infants with congenital or chromosomal abnormalities, acute infections, intraventricular haemorrhage, requirement for additional inspired oxygen, or patent ductus arteriosus were excluded. Written informed parental consent was obtained for each infant.

Study design and protocol

A randomised crossover study with four treatment phases was conducted. Following recruitment, infants were allocated to a random sequence of four different NCPAP/NIPPV devices. The random sequence, corresponding to a 4×4 Latin square, was created by Byers’ random selection algorithm.12 Each device was applied for 6 h, yielding a total study duration of 24 h for each patient. Infants were fed at 2 h intervals and received their routine nursing care while placed in an isolette at thermoneutrality and in a prone, 15° head-up tilt position. The study protocol was approved by the Ethics Committee of Tuebingen University Hospital.

Cardio-respiratory recordings

The following signals were monitored throughout and recorded using a computerised polygraphic system (Embla N7000 and Somnologica Studio 3.0; Embla, Broomfield, CO): chest and abdominal wall movements (built-in respiratory inductance plethysmography; Embla), pulse oximeter saturation (SpO₂) and pulse waveform (Radical in 2 s averaging mode; Masimo, Irvine, CA) and electrocardiography and beat-to-beat heart rate (Embla). Digital black-and-white video frames were captured (Panasonic, Osaka, Japan).

Recordings were analysed without access to clinical data, that is, the author (JS) who analysed the recordings was not involved in clinical management, and analysis of cardio-respiratory events was carried out without access to the video frame to ensure blinding to the NCPAP/NIPPV devices used. Total and artifact-free recording times were determined. Artifact-free recording time was defined as all quiet resting periods minus nursing and feeding times. Recordings were then analysed manually for the presence of central apnoeas, desaturation events and bradycardias. A central apnoea was scored if the amplitude of the chest and abdominal wall movement channel fell to <20% of the average amplitude of the preceding breaths for at least 10 s.13 Mixed/obstructive apnoeas were not analysed because the NCPAP/NIPPV devices did not allow airflow recording. A desaturation event was defined as a fall in SpO₂ to ⩽80% for more than 1 s. Desaturation events with a distorted pulse waveform signal within 7 s prior to their onset were considered artifactual and excluded (corresponding to the signal processing time of the pulse oximeter). A bradycardia was defined as a fall in heart rate to ⩽80 beats per minute for more than one beat. To exclude spurious events caused by body movements, bradycardias with a distorted electrocardiography signal immediately prior to their onset were excluded. Baseline heart rate and SpO₂ were defined as the mean of the respective parameter within artifact-free recording time and automatically calculated using software (Somnologica Studio 3.0; Embla). Respiratory rate was manually determined during each period of regular breathing; the mean of these values was calculated to determine an infant’s baseline respiratory rate.14 Finally, event rates for central apnoeas, desaturation events and bradycardias were calculated as the number of respective events per hour of artifact-free recording time. The relative cumulative event time was calculated as the summed duration of all bradycardias and desaturation events divided by artifact-free recording time and multiplied by 100.

Respiratory support systems

The following devices and modes were used: a conventional ventilator (Stephanie; Stephan, Gackenbach, Germany) delivering NIPPV via Hudson prongs (RCI, Temecula, CA); a variable flow device in NCPAP mode (Infant Flow; Electro Medical Equipment, Brighton, UK) and in non-synchronised NIPPV mode (Infant Flow Advance; Electro Medical Equipment) and a conventional NCPAP/underwater bubble system, generating CPAP by placing the end of the expiratory limb 6 cm under water15 and delivering it via Hudson prongs. All devices were adjusted to achieve an approximate positive end-expiratory pressure of 5–6 cm H₂O. Systems were regularly monitored and flow adjusted to keep positive end-expiratory pressure constant. Prong size was chosen to comfortably fit the infant’s nostrils. NIPPV was delivered with a peak inspiratory pressure of 15 cm H₂O at a rate of 10 l per minute and an inspiratory time of 0.4 s. Gas flow was usually 6 l per minute, but could be increased up to 15 l per minute by the conventional ventilator to compensate for air leaks. The variable flow device delivered positive end-expiratory pressure via a flow of 7–10 l per minute. For the NIPPV mode, peak inspiratory pressure was achieved by adding a peak flow of 5 l per minute above baseline, thereby achieving a peak inspiratory pressure of 10 cm H₂O at a rate of 10 l per minute.

Statistical methods

The primary study variable was the cumulative event rate of all bradycardias and desaturation events per hour of artifact-free recording time. Sample size calculations were based on a pilot study comprising five patients. This study revealed an expected overall mean of 7.5 events per hour and a variance across treatments of 6.25. Hence, 30 study participants were considered sufficient to detect a treatment effect of ±2.5 events per hour with 0.05 type I and 0.2 type II error. An interim analysis was to be performed after 16 patients had been studied. Using the O’Brien-Fleming criteria,16 the study would have been terminated if the actual p value was <0.0052.

For the analysis, we intended to compare each NCPAP/NIPPV device with all the others. Patient recruitment took longer than expected, however, due to a lack of eligible patients. Most infants on the neonatal intensive care unit either needed additional oxygen while on NCPAP or no longer needed NCPAP once they were breathing room air. Hence, the study was terminated after 16 patients had been studied. Due to this decrease in sample size, statistical power was likely lost for some comparisons. We therefore decided to reduce the number of statistical tests and compared only the conventional ventilator in NIPPV mode with the other three test devices.

Descriptive statistics as numbers and percentages as well as median, minimum and maximum were used to summarise demographic and clinical characteristics. Comparisons between treatment modalities adjusted for study phase, interaction (study phase×treatment) and random effects (ie, individuals) were carried out using univariate analysis of variance. If the latter revealed significant differences between treatment modalities, pair-wise post-hoc comparisons with the conventional ventilator in NIPPV mode as reference were performed using Dunnett’s t test. All statistical hypothesis tests on the primary study variable were done after performing a Box-Cox transformation to obtain an approximately normally distributed test variable.

Secondary study variables were the event rate of all central apnoeas, desaturation events and bradycardias per hour of artifact-free recording time, the baseline respiratory rate, heart rate and SpO₂, and the relative cumulative event time. For secondary study variables, non-parametric tests for multiple related samples were used. Pair-wise comparisons using Wilcoxon’s test were performed if a global test (ie, Friedman’s test) revealed significant differences between treatment modalities. A statistical test result was considered significant if the corresponding p value was <0.05. No adjustment for multiple testing was performed for secondary study variables. All analyses were carried out with statistical software packages: sample size calculations were done using nQuery Advisor 4.0 (Statistical Solutions, Saugus, MA), while the remaining analyses were done using SPSS 12.0 (SPSS, Chicago, IL).

RESULTS

Within the study period, 22 infants met inclusion criteria and 16 parents gave consent. The demographic and clinical characteristics of enrolled infants (10 boys, six girls) are presented in table 1. All infants were receiving caffeine as a respiratory stimulant and mean artifact-free recording time (4.7–5.0 h) did not differ between treatment modalities.

Concerning the primary study variable, there was a significant difference between treatments (f value 4.2; degrees of freedom 3; p value 0.012). During treatment with the conventional ventilator in NIPPV mode, the median cumulative event rate was more than twice as high as with the variable flow NCPAP device, and 50% higher than with the NIPPV/variable flow NCPAP device. There was no relevant difference compared to the conventional NCPAP/underwater bubble system (table 2).

Concerning secondary study variables (table 3), baseline heart rate was significantly higher with the standard ventilator than with either the NIPPV/variable flow NCPAP devices or the conventional NCPAP/underwater bubble system, and baseline SpO₂ was significantly lower with the NIPPV/conventional NCPAP device compared to any other NCPAP/NIPPV device. The total duration of bradycardia and desaturation was approximately halved with either variable flow device compared to the NIPPV/conventional NCPAP device, but this difference did not reach statistical significance. All other secondary study variables (ie, rate of central apnoeas, desaturation events or bradycardias, and baseline respiratory rate) were not significantly different between the devices. No severe adverse event (eg, pneumothorax) were observed with any system.

DISCUSSION

This study compared the effect of a conventional ventilator in NIPPV mode, a variable flow device in NCPAP and NIPPV mode (both non-synchronised), and a conventional NCPAP/underwater bubble system on bradycardia and desaturation in infants with AOP. We found that symptoms related to AOP were most effectively treated with a variable flow NCPAP device, and that non-synchronised NIPPV offered no advantage over NCPAP.

This finding is in contrast to the results from a recent meta-analysis, which concluded that NIPPV appears to reduce the frequency of apnoeas more effectively than NCPAP (delivered with conventional ventilators and via Hudson prongs).11 However, only two studies (enrolling a total of 54 infants) were included in this meta-analysis, and one reported no significant difference in apnoea rate. Neither focused on bradycardia or desaturation events or compared NIPPV against NCPAP delivered via a variable flow device. Most importantly, NIPPV in both studies was delivered in synchronised mode, precluding comparability with our study. The fact that non-synchronised NIPPV offered no advantage over NCPAP in the present study provides support for the recent conclusion that synchronising the infant’s own respiratory efforts during NIPPV is important “to ensure that the lungs are opened up and kept open” when applying nasal positive airway pressure support.17

Several studies have compared various aspects of NCPAP delivery via conventional ventilators to the variable flow device used in this study,18^–22 and showed reduced oxygen requirement,20 22 23 reduced work of breathing24 and less extubation failure with a variable flow device.20 Other investigators, however, could not show such differences,18 21 and no study differentiated between effects related to the mode of application (double versus single prongs) or the mode of flow generation (conventional flow versus variable flow).

One study compared conventional NCPAP in intubated infants with the underwater bubble system and found a significant reduction in minute ventilation with the latter, but no change in arterial CO₂ or SpO₂.25 Another study measured work of breathing with an underwater bubble system compared to a variable flow NCPAP device26 and found significantly lower resistive work of breathing and less thoraco-abdominal asynchrony with the latter, although the difference between the variable flow device and the underwater bubble system was not as large as between the variable flow device and a conventional NCPAP system.

Baseline SpO₂ was lower with the NIPPV/conventional NCPAP device than with any of the other systems. Although an increase in SpO₂ from 95.6% to 97.0%, the maximum difference observed here, corresponds to an increase of 13 mm Hg in arterial O₂ pressure,27 it is questionable whether this is sufficient to affect respiratory drive. It is more likely that NCPAP affects respiratory control via activation of pulmonary stretch receptors, as already suggested by one of the first reports in this field,28 and/or via reduced work of breathing.1 24

Limitations

Our study design did not allow for differentiation between obstructive and central apnoeas. A growing body of evidence suggests, however, that both types of apnoea may have a similar underlying pathophysiology, that is, they represent extremes in the interplay of forces regulating upper airway patency and diaphragmatic activity.1 However, the hypoxia and bradycardia potentially resulting from either type of apnoea are clinically relevant, which is why we chose their combined rate of occurrence as our primary outcome. Exclusion of distorted pulse waveforms and electrocardiography signals preceding an event may have concealed desaturation or bradycardia events during obstructive apnoeas but was considered essential to ensure exclusion of spurious events. We also abstained from analysing hypopneas, as this would have required quantification of nasal airflow or the use of calibrated induction plethysmography. Our small sample size may have prevented us from identifying differences between devices concerning secondary outcome parameters, such as the cumulative duration of desaturation and bradycardia. We did not choose cumulative event duration but frequency of events as our primary outcome parameter because we consider the frequency with which fluctuations in SpO₂ occur to be at least as relevant to the development of oxygen-related diseases (such as retinopathy of prematurity) as the time an infant spends below a certain SpO₂ threshold.29

The NIPPV/variable flow NCPAP device was used in a non-synchronised manner because this is the standard procedure in our neonatal intensive care unit. However, this may have put the system at a disadvantage compared to the variable flow device in NCPAP mode (see above). Only infants on room air were studied. This was considered necessary to separate out the effects of O₂ from those of the devices used to treat AOP, but made recruitment rather difficult (ultimately resulting in early study termination) and may have led to the exclusion of infants with more severe residual lung disease. Infants with minimal lung disease, however, may also benefit from NCPAP, and this treatment is therefore recommended, irrespective of an infant’s oxygen requirement, in current guidelines for the management of AOP.30

CONCLUSION

This study has shown marked differences in the effect of four nasal respiratory support systems on symptoms of AOP. A variable flow device in NCPAP and NIPPV mode may be more effective in the treatment of AOP than a conventional ventilator in NIPPV mode. Further studies are needed to investigate whether synchronised NIPPV performs even better.