Article Text

Lung volume distribution in preterm infants on non-invasive high-frequency ventilation
  1. Vincent D Gaertner1,
  2. Andreas D Waldmann2,
  3. Peter G Davis3,4,5,
  4. Dirk Bassler1,
  5. Laila Springer6,
  6. Jessica Thomson4,5,
  7. David Gerald Tingay4,5,7,
  8. Christoph Martin Rüegger1
  1. 1 Newborn Research, Department of Neonatology, University Hospital and University of Zurich, Zurich, Switzerland
  2. 2 Department of Anesthesiology and Intensive Care Medicine, Rostock University Medical Center, Rostock, Germany
  3. 3 Newborn Research Centre and Neonatal Services, The Royal Women's Hospital, Melbourne, Victoria, Australia
  4. 4 Murdoch Children's Research Institute, Melbourne, Victoria, Australia
  5. 5 University of Melbourne, Melbourne, Victoria, Australia
  6. 6 Department of Neonatology, University Children's Hospital Tubingen, Tubingen, Germany
  7. 7 Department of Neonatology, The Royal Children's Hospital Melbourne, Parkville, Victoria, Australia
  1. Correspondence to Dr Christoph Martin Rüegger, Newborn Research, Department of Neonatology, University Hospital and University of Zurich, 8091 Zurich, Switzerland; christoph.rueegger{at}


Introduction Non-invasive high-frequency oscillatory ventilation (nHFOV) is an extension of nasal continuous positive airway pressure (nCPAP) support in neonates. We aimed to compare global and regional distribution of lung volumes during nHFOV versus nCPAP.

Methods In 30 preterm infants enrolled in a randomised crossover trial comparing nHFOV with nCPAP, electrical impedance tomography data were recorded in prone position. For each mode of respiratory support, four episodes of artefact-free tidal ventilation, each comprising 30 consecutive breaths, were extracted. Tidal volumes (VT) in 36 horizontal slices, indicators of ventilation homogeneity and end-expiratory lung impedance (EELI) for the whole lung and for four horizontal regions of interest (non-gravity-dependent to gravity-dependent; EELINGD, EELImidNGD, EELImidGD, EELIGD) were compared between nHFOV and nCPAP. Aeration homogeneity ratio (AHR) was determined by dividing aeration in non-gravity-dependent parts of the lung through gravity-dependent regions.

Main results Overall, 228 recordings were analysed. Relative VT was greater in all but the six most gravity-dependent lung slices during nCPAP (all p<0.05). Indicators of ventilation homogeneity were similar between nHFOV and nCPAP (all p>0.05). Aeration was increased during nHFOV (mean difference (95% CI)=0.4 (0.2 to 0.6) arbitrary units per kilogram (AU/kg), p=0.013), mainly due to an increase in non-gravity-dependent regions of the lung (∆EELINGD=6.9 (0.0 to 13.8) AU/kg, p=0.028; ∆EELImidNGD=6.8 (1.2 to 12.4) AU/kg, p=0.009). Aeration was more homogeneous during nHFOV compared with nCPAP (mean difference (95% CI) in AHR=0.01 (0.00 to 0.02), p=0.0014).

Conclusion Although regional ventilation was similar between nHFOV and nCPAP, end-expiratory lung volume was higher and aeration homogeneity was slightly improved during nHFOV. The aeration difference was greatest in non-gravity dependent regions, possibly due to the oscillatory pressure waveform. The clinical importance of these findings is still unclear.

  • intensive care units
  • neonatal
  • neonatology

Data availability statement

Data are available upon request.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • There are potential benefits of non-invasive high-frequency oscillatory ventilation (nHFOV) over nasal continuous positive airway pressure (nCPAP), but pathophysiological mechanisms remain unclear.

  • Oscillatory volumes generated during nHFOV are substantially transmitted to the lung level of preterm infants.

  • Compared with the regional distribution of tidal breaths, oscillations preferentially reach the non-gravity-dependent lung regions but the impact on aeration homogeneity is unclear.

What this study adds?

  • Tidal volumes were larger during nCPAP but distribution of tidal volumes throughout the lung was similar between nHFOV and nCPAP.

  • Overall aeration was greater during nHFOV which was attributable to an increased aeration of non-gravity-dependent lung regions.

  • Air distribution during nHFOV was more homogeneous than during nCPAP probably due to alveolar recruitment by the oscillatory waveform.


Non-invasive high-frequency oscillatory ventilation (nHFOV) is increasingly used as respiratory support for preterm infants.1–3 There are potential benefits over nasal continuous positive airway pressure (nCPAP), including reduced rates of post-extubation failure and improved respiratory stability.4–7 The mechanisms of these potential benefits remain unclear. Recently, our group provided evidence of substantial transmission of oscillatory volumes into the lung of preterm infants on nHFOV. We demonstrated that compared with the regional distribution of tidal breaths, oscillations preferentially reached the right and non-gravity-dependent (NGD) lung regions.8 However, the effects of these differences on overall ventilation and lung aeration remain unclear.

Spatiotemporal changes in lung volumes can be measured using electrical impedance tomography (EIT). EIT is a non-invasive and radiation-free breath-by-breath imaging method that can be used continuously at the bedside and is thus ideally suited to preterm infants.9 However, comparative EIT data of preterm infants on nHFOV and nCPAP support are sparse.

In the current study, EIT data from a randomised crossover trial comparing nHFOV with nCPAP in preterm infants were analysed.7 The aim was to determine differences in regional ventilation characteristics and end-expiratory lung volumes between nHFOV and nCPAP.

Materials and methods

The original trial was registered with the Australian and New Zealand Clinical Trials Registry (ACTRN12616001516471).

Population and intervention

The setup of the original study has been described previously.7 8 Infants were eligible if they were (1) born <30 weeks’ gestation, (2) older than 7 days, (3) between 26 and 34 weeks’ postmenstrual age, (4) extubated for more than 24 hours and (5) clinically stable while receiving nCPAP support. They received nHFOV and nCPAP in a randomised crossover design, each for 120 min with an initial washout period of 30 min on the respective therapy. Infants were managed on a Babylog VN500 ventilator (Dräger Medical System, Lübeck, Germany) and short binasal prongs (Hudson Respiratory Care, Temecula, California, USA) for both intervention periods. The positive end-expiratory pressure (PEEP; during nCPAP) and the mean airway pressure (MAP; during nHFOV) were set to the PEEP level before study commencement and the applied pressures were equal during both therapies. Frequency (8 hertz (Hz)), PEEP, MAP and inspiratory to expiratory ratio (1:1) were not adjusted. The smallest amplitude to achieve visible chest wall vibration was used and modified to maintain transcutaneous carbon dioxide (CO2) levels between 40 and 60 mm Hg. Infants were nursed in a prone position. Thus, dorsal lung regions were considered non-gravity-dependent.8

Data collection

Complete methods of data collection and extraction have been described previously.8 A textile electrode belt with 32 electrodes was fastened at nipple level.10 During each intervention period, four 10-minute EIT sequences were recorded using the SenTec BB2 EIT device (SenTec, Landquart, Switzerland) at a frame rate of 48 Hz in a custom-built infant imaging package.10 11 The EIT belt remained on the infant’s thorax between recordings and throughout both modes of ventilation. For each recording, the first 30 stable consecutive breaths of artefact-free tidal ventilation were identified and data were extracted using ibeX (V.1.1, SenTec, Landquart, Switzerland). Recordings were excluded from analysis if more than three electrodes had insufficient skin contact or if less than 30 consecutive breaths could be identified. Data preparation included the following steps: predefined anatomical lung regions based on the vendor-provided human model chest atlas were projected into the EIT image and EIT signals outside of these regions excluded.12 13 Then, EIT signals were extracted and normalised for body weight. No frequency filter was applied to assess overall differences including spontaneous breathing as well as oscillation signals.

Data analysis

First, tidal volume (VT) distribution was evaluated in 36 horizontal slices across the lung,14 and compared between the two modes of ventilation. Second, overall changes in ventilation including silent spaces (SS), the deviation from the ideal centre of ventilation (ideal minus actual centre of ventilation; CoVi−a) and the coefficient of variation (CV) were calculated (see figure 1A).9 15 SS correspond to areas of the lung with little or no ventilation,16 and were calculated in the gravity-dependent (GD) and NGD hemithorax separately (SSGD and SSNGD , respectively). The CoVi−a indicates a difference in ventilation in respective lung areas (eg, GD or NGD). SS and CoVi−a are percentages and thus, they were not normalised for body weight. Finally, the SD of impedance changes in all pixels was divided by the mean value of impedance to calculate the CV.17 The CV correlates with ventilation homogeneity and lower values indicate improved homogeneity (figure 1A).

Figure 1

Explanation of ventilation (A) and aeration parameters (B) derived from electrical impedance tomography (EIT) recordings. Functional EIT images shown here are derived from a typical recording to illustrate the methods. The deviation from the ideal centre of ventilation (CoVi−a) is calculated by subtraction of the two variables. AU, arbitrary units; CoVa, actual centre of ventilation; CoVi, ideal centre of ventilation; EELI, end-expiratory lung impedance; EILI, end-inspiratory lung impedance; GD, gravity dependent; NGD, non-gravity dependent; VT, tidal volume; ∆EELI, change in end-expiratory lung impedance.

Third, the net signal at end-expiration (end-expiratory lung impedance; EELI) was isolated as the impedance signal at the end of each expiration in arbitrary units to assess overall lung aeration. This was determined in the whole lung (EELItotal) and in four regions of interest (horizontal quantiles ranging from NGD to GD; EELINGD, EELImidNGD, EELImidGD, EELIGD) to assess global and regional aeration (figure 1B). Then, the difference in aeration from nCPAP to nHFOV was calculated overall and for each quantile (∆EELI=EELInHFOV−EELInCPAP). Finally, homogeneity of lung aeration during both modes of ventilation was measured. To do this, signals in each quantile of regional aeration were first weighted to the known pixel contribution of each region to normalise for differences in lung size.18–20 Then, the NGD half of the EIT signal was divided by the GD half:

Embedded Image

Subsequently, only the most NGD quantile was divided by the most GD quantile to assess homogeneity in the outermost parts of the lung:

Embedded Image

For both ratios, a value of 1 represents equal distribution of air and a value <1 or >1 describes an aeration which favours the GD or NGD lung, respectively (see figure 1B).

Statistical analysis

Averages of each EIT recording were computed for subsequent analyses. For assessment of ventilation and aeration parameters, four individual recordings were used per infant and mode of ventilation, and comparisons between nHFOV and nCPAP were performed using a mixed-model analysis of variance controlling for within-subjects variance (using the ‘afex’ package in R statistics, V.3.6.2).21 This means that statistical findings from this analysis relate to within-infants differences and thereby, to differences between the two modes of respiratory support. Normally distributed data are presented as mean with SD or 95% CI. Non-parametric data are presented as median and IQR. Comparisons between nHFOV and nCPAP were corrected for multiple testing using the Bonferroni-Holm method. Adjusted p values of <0.05 were considered statistically significant.



Among 30 infants, 228 EIT recordings containing 6840 breaths were analysed, 112 recordings during nHFOV and 116 during nCPAP (figure 2). Demographic and clinical characteristics of the included infants are provided in table 1.

Figure 2

Flow chart of infants and recordings analysed during this study. EIT, electrical impedance tomography; nCPAP, nasal continuous positive airway pressure; nHFOV, non-invasive high-frequency oscillatory ventilation.

Table 1

Baseline demographics and clinical characteristics (N=30 infants)

Comparison of ventilation distribution during spontaneous breathing

Ventilation distribution across 36 slices of the lung during nHFOV and nCPAP is provided in figure 3. Overall, relative VT was significantly greater in all but the six most GD lung slices during nCPAP (figure 3).

Figure 3

Ventilation distribution during nHFOV and nCPAP. (A) Tidal volume (VT) distribution across 36 slices from gravity-dependent to non-gravity-dependent lung regions. (B) Difference in VT distribution between nHFOV and nCPAP. (C) Ratio of VT (ie, nCPAP divided by nHFOV). Infants were nursed in a prone position throughout the study and ventral lung regions were considered gravity dependent. *VT during nCPAP were significantly higher than during nHFOV in all slices marked by the asterisk (all p<0.05). All comparisons were performed using a mixed-model analysis of variance where presented differences relate to within-infant differences. AU/kg, arbitrary units per kilogram body weight; nCPAP, nasal continuous positive airway pressure; nHFOV, non-invasive high-frequency oscillatory ventilation.

Homogeneity of the net signal

Ventilation homogeneity, SS in both hemithoraces as well as CoVi−a were comparable between nHFOV and nCPAP (table 2A).

Table 2

Differences between nHFOV and nCPAP in ventilation parameters (A), distribution of lung aeration (B) and aeration homogeneity (C); N=30 infants

Distribution of lung aeration

Overall, EELI was higher during nHFOV than during nCPAP, particularly in both NGD quantiles, but EELI in the mid-GD quantile was lower during nHFOV (table 2B and figure 4A). Aeration homogeneity was slightly increased during nHFOV when compared with nCPAP and this effect was more pronounced when we assessed the outermost parts of the lung (table 2C). In fact, AHRouter is close to 1 during nHFOV, indicating an almost equal distribution of aeration in the outermost parts of the lung (figure 4B).

Figure 4

Changes in lung aeration between nHFOV and nCPAP. (A) Mean and 95% CI of ∆EELI over the whole lung and for four quantiles of the lung (NGD to GD) separately. (B) Mean and 95% CI of the aeration homogeneity ratio (AHR) for nHFOV and nCPAP separately. (B1) The AHR comparing the two hemithoraces, (B2) the AHR considering only the outermost regions of the lung. Colours correspond to colours in figure 1. All comparisons were performed using a mixed-model analysis of variance where presented differences relate to within-infant differences. Asterisks indicate significant differences: *padj <0.05, **padj <0.01, ***padj <0.001. AHRouter, AHR of the outermost regions of the lung; AU/kg, arbitrary units per kilogram body weight; GD, gravity dependent; nCPAP, nasal continuous positive airway pressure, NGD, non-gravity dependent; nHFOV, non-invasive high-frequency oscillatory ventilation; ∆EELI, change in end-expiratory lung impedance.


The primary purpose of this study was to determine differences in regional ventilation characteristics and end-expiratory lung volumes between nHFOV and nCPAP. While VT were distributed evenly throughout the lung during both modes of non-invasive respiratory support, aeration was higher during nHFOV and preferentially distributed toward the NGD lung regions. This effect resulted in a more uniform aeration across the lungs during nHFOV.

In our study, spontaneously generated VT were consistently larger during nCPAP compared with nHFOV in all but the most GD lung slices which is in line with our previous finding that overall VT are larger during nCPAP.8 However, VT were distributed similarly, and the homogeneity of ventilation was not affected by the mode of non-invasive respiratory support. We have shown previously that the oscillations generated during nHFOV are more likely to reach the NGD areas of the lung.8 The spontaneous tidal breaths, by contrast, are generated by the infant itself and are thus less affected by the mode of ventilation as long as the positive distending pressure is maintained at the same level and the infant remains in the same position. Moreover, the spontaneous breathing signal contributes much more signal change to each breathing cycle than the oscillations.8 Both factors may explain the similarities in the distribution of the net signal over one breathing cycle and may indicate that clinical differences between nHFOV and nCPAP may be mostly due to mechanisms pertaining to the superimposed oscillations.

Both modes of ventilation showed only a very small deviation from the ideal centre of ventilation and similarly, only a small percentage of SS, corresponding to areas of little or no ventilation. It is reassuring that despite an increase of aeration in the NGD and a decrease in the GD lung regions, there was neither an increase in overdistension (corresponding to SSNGD) nor in atelectasis (corresponding to SSGD) during nHFOV. We speculate that the use of a higher MAP during nHFOV may only need to be used during specific situations with a high risk of atelectasis.

In our study, overall lung aeration, corresponding to functional residual capacity (FRC), was higher during nHFOV than during nCPAP. While spontaneous breaths are mainly transmitted by laminar flow, the gas flow mechanisms during nHFOV may include more complex phenomena such as bulk flow, radial mixing, convective dispersion and pendelluft.22–25 These mechanisms may enable additional alveolar recruitment, thus contributing to an increased FRC during nHFOV.26 This is supported by our previous finding that oscillatory volumes during nHFOV are transmitted to the lung level and make up approximately one-fifth of VT during nHFOV.8 However, it has to be noted that the overall increase in EELI during nHFOV in this population of stable infants was only small and the pathophysiological factors contributing to additional aeration need to be investigated further.

Previously, we saw that oscillations during nHFOV in preterm infants are preferentially transmitted to the NGD lung.8 This finding is now reflected in an increased overall aeration in the NGD regions of the lung during nHFOV compared with nCPAP and a decreased aeration in the mid-GD lung quantile. This may be due to various frequency-dependent effects: (1) in sheep, an increasing oscillatory frequency improved the ventilation specifically in the NGD lung.27 (2) While the GD lung is compressed due to gravitational forces, the NGD lung has a higher compliance and volume delivery to NGD regions is increased for higher frequencies.28 29 (3) Complex gas flow mechanisms during nHFOV are frequency dependent and may thus contribute to an increased homogeneity.30 As there was no difference in ventilation distribution, increased aeration in the NGD lung could be mainly attributable to oscillatory volumes.8

Most importantly, we demonstrated that the increased aeration in the NGD lung during nHFOV led to an improved homogeneity in air distribution. This effect was more prominent in the outermost parts of the lung, where air was distributed almost perfectly even. Heterogeneity in lung aeration is associated with increased mechanical friction20 31 32 and subsequent pulmonary tissue injury,19 20 32 which is of particular relevance in infants with evolving bronchopulmonary dysplasia (BPD).33 34 However, the absolute differences in homogeneity were small. Whether the slightly improved homogeneity during nHFOV is indeed associated with important clinical outcomes such as a decrease in the rate of BPD or longer-term lung function needs to be investigated in adequately powered trials of longer-term use of nHFOV.

While the clinical importance of our results remains unclear, our data extend the pathophysiological knowledge underlining previous clinical findings: (1) in the original trial, we found reduced rates of desaturation and bradycardia during nHFOV,7 which may be partly explained by the improved aeration homogeneity shown in this study. However, other mechanisms may also be important (eg, transmission of oscillations during periods of apnoea or increased agitation). (2) In the original trial, infants required a higher fractional inspired oxygen during nHFOV,7 which may be explained by a decreased aeration in the mid-GD part of the lung, where lung recruitment may be particularly relevant. These findings need to be considered when escalating infants from nCPAP to nHFOV and may inform the discussion on the use of a higher MAP during nHFOV. (3) The initial trial did not show any effect of nHFOV on CO2 clearance,7 even though oscillatory volumes during nHFOV are transmitted into the lung.8 We speculate that the effect of the oscillations may be counteracted by slightly smaller VT during nHFOV, ultimately leading to the same amount of CO2 removal during the two modes of respiratory support. (4) In selected situations, for example, post-extubation or as primary respiratory support after birth, the small benefits detected in this study may be sufficient to explain the superiority of nHFOV seen in smaller clinical trials.35–38 Finally, we speculate that an increased aeration homogeneity may be associated with a reduced rate of lung injury after long-term use of nHFOV, as reported by a small preclinical study.39 However, this finding needs confirmation in larger clinical studies.

This study has various limitations: First, due to the crossover nature of the study, we were not able to evaluate any long-term clinical effects of nHFOV. In the original trial, we showed a reduced rate of desaturation and bradycardia but an increased oxygen requirement and higher heart rates.7 By demonstrating an increased aeration homogeneity during nHFOV, our findings now provide a potential pathophysiological explanation of these clinical benefits of nHFOV and introduce new avenues of intervention which need to be addressed in clinical trials. Second, EIT measurements can only provide relative changes and no absolute volumes. However, changes in distribution during nHFOV compared with nCPAP are independent of absolute values and may be the decisive factor for potential clinical benefits. Third, we studied a small sample of 30 preterm infants and data were analysed using a mixed-model analysis. It is unclear whether this effect would be present in a different set of infants as well, even though the direction of results was consistent across recordings and infants. Fourth, this was a single-centre study with a set positive distending pressure and a set frequency during nHFOV, and all infants remained in prone position throughout the study. The use of different ventilators, nasal interfaces, ventilator settings or body positions may lead to slightly divergent findings. Finally, the general benefits and limitations of EIT have been reported in detail before.9 Of these, differences in the chest model used to reconstruct raw EIT data may impact comparability with different studies.


During nHFOV and nCPAP, overall VT were distributed evenly throughout the lung. Aeration was greater and more uniform during nHFOV which was attributable to an increased aeration of NGD lung regions. This may be due to the differential distribution of oscillatory volumes during nHFOV and introduces new avenues of further research in this complex and emerging new mode of non-invasive respiratory support.

Data availability statement

Data are available upon request.

Ethics statements

Patient consent for publication

Ethics approval

The original trial was approved by the local ethics committee of the Royal Women’s Hospital Melbourne (Ref No 16/32). All parents provided written informed consent prior to commencement of the study.


We thank all the parents and infants who participated in the study and the staff at the neonatal intensive care unit of the Royal Women’s Hospital, Melbourne, Australia.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors PGD, DB, LS, DGT and CMR developed the concept and design of the initial study. VDG, ADW and CMR conceptualised and designed this post-hoc analysis. LS, JT and CMR were involved in patient recruitment and conducted the electrical impedance tomography (EIT) measurements. ADW developed the EIT analysis software. VDG, ADW and CMR performed EIT data analysis. All authors participated in data interpretation. VDG and CMR wrote the first draft and all authors contributed to redrafting the manuscript and revising it for intellectual input. CMR is acting as the guarantor of the overall content.

  • Funding Supported by the Victorian Government Operational Infrastructure Support Programme (Melbourne, Australia); the National Health and Medical Research Council (Practitioner Fellowship GNT 1059111 (to PGD)); the German Research Society (DFG-grant number: LO 2162/1-1 (to LS)); the TÜFF Habilitation Program (TÜFF 2459-0-0 (to LS)); Career Development Fellowships GNT 11123859 and 1057514 (to DGT)); the Swiss National Science Foundation (Early Postdoctoral Mobility fellowship P2ZHP3_161749 (to CMR)); the Swiss Society of Neonatology (Milupa Fellowship Award (to CMR)); and the Endeavour Research Fellowship by the Australian Government (ERF_RDDH_5276_2016 (to VDG)).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.