Article Text

Volume guaranteed? Accuracy of a volume-targeted ventilation mode in infants
  1. Olivia Farrell1,2,
  2. Elizabeth J Perkins1,
  3. Don Black1,
  4. Martijn Miedema1,3,
  5. Joel Don Paul1,2,
  6. Prue M Pereira-Fantini1,2,
  7. David Gerald Tingay1,2,4
  1. 1 Neonatal Research, Murdoch Children’s Research Institute, Melbourne, Australia
  2. 2 Department of Paediatrics, University of Melbourne, Melbourne, Australia
  3. 3 Neonatology, Academic Medical Centre, Amsterdam, The Netherlands
  4. 4 Neonatology, Royal Children’s Hospital, Melbourne, Australia
  1. Correspondence to Dr David Gerald Tingay, Neonatal Research, Murdoch Childrens Research Institute, Royal Children’s Hospital Flemington Rd., Parkville 3052 Victoria, Australia.; david.tingay{at}


Objectives Volume-targeted ventilation (VTV) is widely used and may reduce lung injury, but this assumes the clinically set tidal volume (VTset) is accurately delivered. This prospective observational study aimed to determine the relationship between VTset, expiratory VT (VTe) and endotracheal tube leak in a modern neonatal ­volume-targeted ventilator (VTV) and the resultant partial arterial pressure of carbon dioxide (PaCO2) relationship with and without VTV.

Design Continuous inflations were recorded for 24 hours in 100 infants, mean (SD) 34 (4) weeks gestation and 2483 (985) g birth weight, receiving synchronised mechanical ventilation (SLE5000, SLE, UK) with or without VTV and either the manufacturer’s V4 (n=50) or newer V5 (n=50) VTV algorithm. The VTset, VTe and leak were determined for each inflation (maximum 90 000/infant). If PaCO2 was sampled (maximum of 2 per infant), this was compared with the average VTe data from the preceding 15 min.

Results A total of 7 497 137 inflations were analysed. With VTV enabled (77 infants), the VTset−VTe bias (95% CI) was 0.03 (−0.12 to 0.19) mL/kg, with a median of 80% of VTe being ±1.0 mL/kg of VTset. Endotracheal tube leak up to 30% influenced VTset−VTe bias with the V4 (r2=−0.64, p<0.0001; linear regression) but not V5 algorithm (r2=0.04, p=0.21). There was an inverse linear relationship between VTe and PaCO2 without VTV (r2=0.26, p=0.004), but not with VTV (r2=0.04, p=0.10), and less PaCO2 within 40–60 mm Hg, 53% versus 72%, relative risk (95% CI) 1.7 (1.0 to 2.9).

Conclusion VTV was accurate and reliable even with moderate leak and PaCO2 more stable. VTV algorithm differences may exist in other devices.

  • infant
  • mechanical ventilation
  • tidal volume
  • volume targeted ventilation
  • lung injury

Statistics from

What is already known on this topic?

  • Volume-targeted ventilation (VTV) modes aim to stabilise tidal volume delivery and meta-analysis suggests improved outcomes when used.

  • The benefits of VTV are reputed to be due to less atelectasis and volutrauma and more stable arterial carbon dioxide levels.

  • The lung protective ability of VTV requires the ventilator’s algorithm to reliably deliver the tidal volume set by the clinician.

What this study adds?

  • In more than 7 million analysed inflations, the set tidal volume was delivered within a clinically acceptable range for most inflations using VTV.

  • This is the first study to show that VTV targeting expiratory tidal volume, with good endotracheal tube leak compensation, stabilises arterial carbon dioxide.

  • The first confirmatory evidence that a modern neonatal ventilator delivers the theoretical benefits of VTV.


The optimal application of mechanical ventilation requires minimising the risks of ventilator-induced lung injury (VILI).1 2 Both excessive and inadequate tidal volume (VT) and applied pressure have been shown to increase VILI in neonates.3 4 Volume-targeted ventilation (VTV) modes that adjust delivered pressure, aiming to maintain a constant VT set by the clinician, have been shown to reduce the risk of death, chronic lung disease and short-term morbidity in preterm infants.5 Despite this, VTV is not universally used in neonatal respiratory care.6

Critical to the lung protective potential of VTV is the accuracy and adaptiveness of the ventilators VTV algorithm and the ability of the clinicians to trust that the set V(VTset) is reliably delivered.7 8 This is particularly important during spontaneous breathing and variable endotracheal tube (ETT) leak states.8 To account for these factors, most modern ventilators target the measured expiratory VT (VTe). There are few reports addressing the reliability of specific VTV algorithms in neonates. The existing studies found a more stable VT, with lower peak inflation pressures, when used in combination with patient-triggered ventilation than without.5 9–11 However, on an inflation-to-inflation basis, inflating pressure fluctuations exceeded the manufacturer’s guidelines in one of these studies.11 Effective VTV should also result in more stable partial arterial pressure of carbon dioxide (PaCO2), with less potentially injurious hypocarbia and hypercarbia episodes.10 12 13 Despite the availability of VTV in neonatal ventilators for more than 15 years, the temporal relationship between PaCO2 has not been reported.

The aim of this prospective observational study was to determine 1) the relationship between VTset and measured VTe and 2) the relationship between VTe and PaCO2 with and without VTV.


The study was performed in the Neonatal Intensive Care Unit (NICU) at the Royal Children’s Hospital, Melbourne, Australia, a referral tertiary unit, and approved by the Institutes Ethics in Human Research Committee. The study was performed in accordance with STROBE guidelines.14

The SLE5000 infant ventilator (SLE UK, South Croydon, UK), a neonatal ventilator (accelerating/decelerating flow waveform device) that measures pressure and flow (hot-wire anemometer) at the airway opening, was studied, being the only conventional ventilator used in our NICU. The study was performed in two epochs, as the manufacturer introduced a new VTV algorithm (with reported improved functionality, during the study) with a 2-year gap to allow for software updates and staff training of at least 1 year of clinical use with the new software: the first from April 2013 to March 2014 (V4 algorithm) and the second from March 2016 to August 2016 (V5 algorithm). Both algorithms target VTe and offer leak compensation up to 20% (V4) and 50% (V5).

Intubated neonates less than 44 weeks corrected gestational age were studied if they were receiving mechanical ventilation in a patient-triggered modality with or without VTV enabled. Neonates were not studied if they were receiving high-frequency ventilation, muscle relaxants, had a major congenital abnormality or if extubation was anticipated within 24 hours. 

At the commencement of the study, a custom-built data acquisition box (see online supplementary) was connected to the digital RS232 output port of the ventilator. From this port, the SLE5000 continuously outputs 40 parameters, including modality, VTV status, VTset, VTe and measured ETT leak. These data were recorded for every inflation (text file onto an SD (Secure Digital) card) for up to 24 hours unless there was a clinical reason to remove it earlier. Clinicians could provide respiratory support at their discretion, including performing arterial blood gas analysis, changing modes and activating/deactivating VTV.

Supplementary material 1

At the completion of each study period, the SD card was removed, data were extracted and the first 90 000 inflations per infant (or maximum within the 24 hours period if less) were exported into Prism V.7.02 for Windows (GraphPad Software, San Diego, California, USA) for analysis. For inflations with VTV enabled, matched VTset, VTe and ETT leak data were extracted for each inflation and the VTset−VTe difference was calculated. PaCO2 from any arterial blood gas analyses (maximum of 2 per infant) during the study period was manually recorded. The mean VTe and minute ventilation for the 15 min prior to the PaCO2 were calculated.

A convenience sample size of 100 (50 for each algorithm) was chosen to allow sufficient number of individual inflations within VTV permutations and a representative population of the broad spectrum of respiratory disease and ventilator settings. Descriptive statistics for the measured parameters were calculated. The Bland-Altman technique (using VTset as the comparator) and linear regression analysis were used to describe the inflation-to-inflation VTset−VTe differences and ETT leak relationships. Linear regression was also used to analyse the ETT leak−VTe and VTe−PaCO2 relationships. Fisher’s exact test was used to determine whether VTV influenced the rates of hypocarbia or hypercarbia. All statistical analyses were performed using Prism, and p<0.05 was considered significant.


The demographic characteristics of the 100 infants studied (7 497 137 inflations) are summarised in table 1. There was no difference between the V4 and V5 groups, or the infants managed with and without VTV (data not shown). There was no difference in the use of synchronised intermittent mandatory ventilation and fully synchronised modes across all groups. Seventy-seven infants (5 477 530 inflations) had VTV enabled at some stage during the recording period with a median (range) 80 399 (352, 90 000) VTV inflations analysed per infant. The most common indication for respiratory support was primary parenchymal lung disease (n=55). At study commencement, five infants had either evolving or established chronic lung disease. Fifty-seven infants had at least one arterial blood gas analysis during the study period, with 42 infants contributing two sets of matched PaCO2–VTe/minute ventilation data (99 data triplicates). Forty of these infants were ventilated with VTV and 17 infants without VTV at the time of blood gas analysis. Infants receiving VTV were a mean (95% CI) 2 (0 to 4) completed weeks gestation less mature in the cohort who had PaCO2 samples (Welch’s t-test). Demographic factors did not influence the VTV accuracy results (data not shown).

Table 1

Subject characteristics at time of enrolment

Volume-targeted ventilation accuracy

The mean (SD) VTset was 4.2 (0.7) mL/kg and VTe was 4.2 (0.8) mL/kg, with a resultant VTset–VTe bias (95% CI) of 0.03 (–0.12, 0.19) mL/kg (figure 1). There was a significant relationship between VTset and VTe (r=0.34, p<0.0001; linear regression). There was no difference in the V4 (n=34; 2 305 076 inflations) and V5 (n=43; 3 172 454 inflations) algorithms, with the mean (95% CI) VTset–VTe difference being 0.4 (0.2, 0.7) mL/kg and 0.3 (0.1, 0.4) mL/kg, respectively (paired t-tests; data not shown). VTV accuracy was not influenced by postnatal age (p=0.53, linear regression), birth weight (p=0.39) or dose of opiate infusion (p=0.96).

Figure 1

(A) Bland-Altman plot of VTset and expiratory tidal volume (VTe) . Solid black line denotes the bias, dashed black lines denote the 95% CI of the limits of agreement. (B) Relationship between VTset and VTe; y=0.736x+1.073 (r=0.34, p<0.0001; linear regression). Solid black line represents the line of best fit and dotted black lines represent 95% CI bands. To ease visual interpretation of figures, and after seeking statistical advice, symbols represent the average values for each infant rather than all values analysed (maximum 90 000/infant). 

Figure 2 shows the percentage of VTe that was equal to and within ±1.0, ±2.0 and ±3.0 mL/kg of the VTset for each infant. Overall VTe equalled VTset in a median (range) of 4 (0, 28)% of inflations for each infant. Overall, 80 (15, 100)% of inflations were within ±1.0 mL/kg, and 73 (15, 100)% and 86 (30, 100)% of V4 and V5 inflations, respectively. Ninety-eight (30, 100)% and 99 (41, 100)% of inflations were within ±2.0 and ±3.0 mL/kg of VTset.

Figure 2

Percentage of inflations in each infant that expiratory tidal volume (VTe) equalled to or was within ±1.0, ±2.0 and ±3.0 mL/kg of VTset using volume-targeted ventilation (n=77 infants). Box plots represent median and interquartile range with error bars showing minimum to maximum values.

Overall, 93% of VTset were within 3.5–8.0 mL/kg and for these inflations, a median (range) of 12 (0, 99)% and 0 (0, 22)% of VTe were less or greater than that range. V5 resulted in 92 (56, 100)% of VTe being within the set 3.5–8.0 mL/kg, while this was 74 (1, 100)% with V4 and 26 (0, 99)% due to VTe <3.5 mL/kg.

Endotracheal tube leak compensation during volume-targeted ventilation

Figure 3 shows the relationship between measured ETT leak and VTset–VTe difference for those infants receiving VTV with the V4 and V5 algorithms. Using the older V4 algorithm, there was a significant linear relationship (p<0.0001, linear regression). The relationship was non-significant for the V5 algorithm (p=0.2062) suggesting better ETT leak compensation within the range of ETT leaks measured.

Figure 3

Relationship between endotracheal tube (ETT) leak and the VTset–VTe difference for infants receiving VTV using the V4 (n=34; A) and V5 (n=43; B) algorithms. There was a significant relationship between ETT leak and VTset–VTe difference using the V4 (p<0.0001, r=0.64; linear regression) but not the V5 (p=0.2062, r=0.04) algorithms. Solid black line represents the line of best fit and the dotted black lines 95% CI. Linear regression equation shown in figure panel. As per the rationale detailed in figure 1, symbols represent the average values for each infant rather than all values analysed. VTe, expiratory tidal volume.

Relationship between volume-targeted ventilation and partial arterial pressure of carbon dioxide

There was no significant correlation between VTe or minute ventilation and PaCO2 in the 40 infants managed with VTV at the time of an arterial blood gas (figure 4). In contrast, there was a significant inverse relationship between VTe and PaCO2 (but not minute ventilation) in the 17 infants ventilated without VTV. The use of VTV was also more likely to be associated with a PaCO2 between 40 and 60 mm Hg, although this did not reach significance; 72% versus 53%, relative risk 1.7 (1.0, 2.9); p=0.1 (Fisher’s exact test).

Figure 4

Relationship between expiratory tidal volume (VTe) and partial arterial pressure of carbon dioxide (PaCO2) (A and B) and minute ventilation (MV) and PaCO2 (C and D) for infants managed with volume-targeted ventilation (VTV) (69 arterial gases, n=40 infants; A and C) and without VTV (30 arterial gases, n=17 infants; B and D). There was no relationship between PaCO2 and VTe (p=0.10, r=0.04; linear regression) and MV (p=0.08, r=0.05) with VTV. Without VTV, there was a significant relationship between PaCO2 and VTe (p=0.004, r=0.26) but not MV (p=0.281, r=0.04). Linear regression equations are shown in each panel. Solid black line represents the line of best fit and dashed lines 95% CI. Dotted black lines represent the normocapnia (40–60 mm Hg) range.


Within the era of antenatal corticosteroids and exogenous surfactant therapy, VTV is one of the few neonatal ventilator modes that has been shown to reduce lung injury in preterm infants.5 However, achieving tight VT control during the dynamic conditions of neonatal lung disease is complex and little has been reported regarding the accuracy of the different VTV options. In our pragmatic prospective observational study of >7 million inflations in a diverse population of 100 infants, we found that the VTV algorithms of a modern neonatal ventilator (SLE5000) were able to accurately and reliably deliver VT and adapt to ETT leak and were associated with more stable PaCO2 than without VTV.

For the ventilator used in our study, overall VTe was delivered within ±1.0 mL/kg of the VT the clinician intended (VTset) in 80% of inflations with VTV enabled, although the intersubject and intrasubject variability suggests a clinically acceptable degree of reliability. The accuracy of VTV has been reported before in a benchtop study7 and in small sample size neonatal studies limited to a relatively small number of inflations or reporting of averaged VT over time.8–11 15 In contrast, by accessing the large amount of data generated by modern ventilators, we were able to extract inflation-to-inflation data for >7 million inflations from 100 infants, the largest to date. Understanding the inflation-to-inflation variability is important when considering VTV accuracy, providing insight as to why VTset–VTe discrepancies might occur, as even short periods of inappropriately delivered ventilation may cause VILI.16 17

Discrepancy between VTset and VTe may have occurred due to factors beyond control of the ventilator algorithm, such as highly variable patient respiratory effort or the contribution of unsupported and therefore unregulated patient breaths.11 No VTV mode can be 100% accurate due to inherent errors in the flow sensors18 and the fact that the ventilator cannot prevent an infant from generating volumes higher than VTset or anticipate a lower respiratory effort. In such situations, the algorithm will need a few inflations to readjust the delivered pressure and may overshoot and undershoot while doing so. We could not address these concerns in the current study, due to the high level of computer processing needed for the large number of inflations. This limited the ability to individually assess patient effort and algorithm reaction.

To our knowledge, this is only the second report of inflation-to-inflation accuracy during VTV in vivo. McCallion and coworkers reported the accuracy of the Dräger BabyLog 8000+ to deliver VTV using a VTe-based algorithm over 10 min of patient-triggered ventilation in 10 preterm infants (6540 inflations).11 Although unable to determine long-term patterns, the authors found that, overall, VTV was accurate but they observed that there were large variations in VT between adjacent inflations by up to 2.2 mL/kg. Our results confirm the observation that inflation-to-inflation VT delivery can be highly variable in an infant. Notwithstanding the different ventilator and study design, together, the similar results are reassuring and indicate that the lung protective features of VTV identified in randomised controlled trials appear being delivered.

Fundamental to ensuring the VT the lung receives is consistent with the VTset during VTV is the ability to compensate for ETT leak. Most neonatal ventilators now offer leak compensation during VTV,7 and the manufacturers of the SLE5000 report that the older V4 and newer V5 algorithms can adapt VT delivery to ETT leaks up to 20% and 50%, respectively. We found that VTset–VTe discrepancy increased linearly with increasing leak using the V4 algorithm, but not the V5. This suggests that the V5 algorithm was more effectively adapting to ETT leak. To our knowledge, this is the first time ETT leak compensation has been reported using VTV and suggests that other ventilators warrant investigation. Such investigation should consider differences in ventilator flow characteristics as this will influence the impact of leak on VT generation.

More stable PaCO2 has been proposed as one of the beneficial effects of VTV.5 We observed that PaCO2 was more stable in infants ventilated with VTV compared with infants ventilated without, with PaCO20% more likely to be within 40–60 mm Hg (the commonly targeted range on our NICU). These findings are similar to reductions in hypocarbia reported in an earlier small randomised cross-over studies using the BabyLog 8000+.10 12 We did not record the clinician’s intended target PaCO2, which may have differed from 40 to 60 mm Hg, especially in infants with chronic lung disease, who also have different VTV needs due to higher respiratory deadspace volumes. Thus, conclusions regarding the physiological outcomes of the arterial blood gases, including whether the use of VTV altered the prevalence of normocarbia and the suitability of VT choice, cannot be made.

CO2 clearance is determined by minute ventilation (product of rate and VT), so the negative direct relationship between PaCO2 and VTe when VTV was not enabled was expected. The lack of this relationship using VTV suggests that the mode was operating appropriately as the clinician determines the VT based on a desired PaCO2. Dawson and Davies also observed a lack of correlation between PaCO2 and VT using VTV (BabyLog 8000+), although VTset was reported, and there was little variability in the VT values reported.13 There was no definitive relationship between PaCO2 and minute ventilation with and without VTV, suggesting that infants may have been contributing to CO2 clearance independent of delivered rate and measured VT.

This study has some limitations not previously mentioned. Our study was observational, resulting in unbalanced subgroups, and clinical care and ventilator settings were not a priori dictated. Conclusions regarding whether VTV was achieving lung protection cannot be assumed, especially as we did not standardise the mode of triggering. Only one ventilator was studied, the SLE5000, and we relied solely on the ventilators measured data, therefore caution is required regarding the generalisation of our findings to other ventilators.7 Our choice of the SLE5000 was pragmatic and practical as it was the only ventilator in use on our NICU and the introduction of alternative ventilators would have required additional staff training to ensure accurate clinical use and avoid bias. In addition, the SLE5000 has an open source data output, allowing our unique data collection system to be used. We would encourage all neonatal ventilator manufacturers to allow open access to ventilator data. Our study was not limited to preterm infants with acute disease but rather a diverse range of neonatal conditions, including term infants and some without primary respiratory failure. Furthermore, our study population had a high rate of analgesia use, potentially decreasing patient effort. This may increase algorithm accuracy. Modern neonatal respiratory practices are changing, with non-invasive support predominating in early preterm care.2 We contend that our intentionally large convenience sample size was robust enough to allow meaningful clinical conclusions.


A representative modern neonatal VTV algorithm, with good ETT leak compensation, was accurate, reliable and effective, delivering the VTset within a clinically acceptable range. Reassuringly, the use of VTV resulted in more stable PaCO2 than without VTV. This study provides the first confirmation that modern neonatal ventilators are able to deliver the theoretical lung protective benefits of VT targeting.


The authors wish to thank Associate Professor Susan Donath of the Centre of Epidemiology and Biostatistics Unit (Murdoch Childrens Research Institute) for statistical advice and Professor Peter Davis (Royal Women’s Hospital, Melbourne, Australia) for clinical interpretation of the data.


View Abstract


  • Contributors DGT developed the concept and designed the experiment. OF, EJP, DB and JDP enrolled and studied all infants. DB designed and built the data collection system used in the study. OF, DB, MM and EJP were involved in data analysis. OF, PP-F, DB, MM and DGT interpreted the data. OF and DGT wrote the first draft of the manuscript and all authors contributed to redrafting the manuscript.

  • Funding This study is supported by the Victorian Government Operational Infrastructure Support Program (Melbourne, Australia). DGT is supported by a National Health and Medical Research Council Clinical Career Development Fellowship (grant ID 1053889).

  • Competing interests DGT has presented at conferences and workshops in which SLE, UK, has paid unrestricted travel costs or sponsorship for the meeting. DGT has also presented at conferences and workshops supported (including speaker travel costs) by other ventilator manufacturers, including Carefusion, Dräger, Fisher Paykel and Acutronic. No author received an honorarium, grant or other forms of payment to produce the manuscript. The study was not commissioned and no commercial agencies were involved in any aspect of this study. The authors have no other competing interests to declare.

  • Ethics approval The Royal Children’s Hospital Human Research and Ethics Committee. The study was approved as a low or negligible risk study meeting audit criteria and prospective parental consent was not deemed necessary.

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

  • Data sharing statement De-identified raw complete data (7 million inflations) for all 100 study subjects are available upon request from the corresponding author.

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.