|Home | About | Journals | Submit | Contact Us | Français|
Arterial oxygen partial pressure can increase during inspiration and decrease during expiration in the presence of a variable shunt fraction, such as with cyclical atelectasis, but it is generally presumed to remain constant within a respiratory cycle in the healthy lung. We measured arterial oxygen partial pressure continuously with a fast intra-vascular sensor in the carotid artery of anaesthetized, mechanically ventilated pigs, without lung injury. Here we demonstrate that arterial oxygen partial pressure shows respiratory oscillations in the uninjured pig lung, in the absence of cyclical atelectasis (as determined with dynamic computed tomography), with oscillation amplitudes that exceeded 50mmHg, depending on the conditions of mechanical ventilation. These arterial oxygen partial pressure respiratory oscillations can be modelled from a single alveolar compartment and a constant oxygen uptake, without the requirement for an increased shunt fraction during expiration. Our results are likely to contribute to the interpretation of arterial oxygen respiratory oscillations observed during mechanical ventilation in the acute respiratory distress syndrome.
In the healthy lung, arterial partial pressure of oxygen (PaO2) is thought to remain almost constant within the respiratory cycle during spontaneous breathing, via a physiological PO2 gradient between the alveoli and the pulmonary circulation1. Respiratory PaO2 oscillations smaller than 16mmHg have been detected with slow response time sensors2 in animals with uninjured lungs when abnormally large tidal volumes (VT) greater than 20mlkg−1 or even 30mlkg−1 were delivered during mechanical ventilation3, 4.
In the injured lung, especially in animal models of the acute respiratory distress syndrome (ARDS), large PaO2 oscillations have been observed by means of fast-response oxygen sensing technology5–12. Here, PaO2 oscillations have been explained by variable within-breath shunt due to the repetitive opening and collapse of lung regions (cyclical atelectasis), causing PaO2 to increase during inspiration and decrease during expiration. This phenomenon may be used to personalize mechanical ventilation based on real-time, continuous PaO2 monitoring, where the appearance of an oscillating PaO2 signal may be a diagnostic biomarker for cyclical atelectasis.
A significant limitation of the above mentioned studies is that the PaO2 oscillations observed in animal models of ARDS mostly appeared in association with a high average VT of 20mlkg−1 7, 13, 30mlkg−1 6, 14, or greater10. These large VT were delivered in experiments without application of positive end-expiratory pressure (PEEP), and imposing elevated peak end-expiratory pressures in order to provoke cyclical atelectasis in the injured lung. Even with these large VT, PaO2 oscillations only between 3 and 22mmHg were observed in the uninjured lung, at a RR lower than 10 breaths per minute, and mean PaO2 above 450mmHg6, 15.
No study employing fast oxygen-sensing technology has explored the responses of the uninjured lung to mechanical ventilation at physiological PaO2 levels. Experimental evidence for this response to mechanical ventilation could also be useful to understand response of the lung to mechanical ventilation in the presence of injury.
We developed a fibre optic oxygen sensor16, optimized its response time and linearity17–20, tested it in vitro 21, showed it to be resistant to clotting, and demonstrated it to be capable of detecting rapid PaO2 changes in an animal model of ARDS18, 34. With this new technique to study gas exchange dynamically, we explore here PaO2 responses to mechanical ventilation in a porcine model.
We hypothesised that in apnoea or expiration PaO2 declines at a rate that is inversely proportional to lung volume, as had been modelled22, 23, and that this decline would be readily measurable in the uninjured pig lung during 20s end-expiratory and end-inspiratory breath holds. We also hypothesised that, during tidal breathing, PaO2 oscillates dynamically in phase with lung volume in the uninjured lung, in the absence of cyclical atelectasis. This hypothesis led us to also explore the possibility that different ventilation strategies (pressure and volume control) and inspired-to-expired ratios could affect the mean PaO2 value, while airway pressures (end-expiratory and end-inspiratory), respiratory rate, or fraction of inspired oxygen (FIO2) were kept unvaried.
Figure 1 illustrates the tidal nature of PaO2 as observed in an anesthetised pig with uninjured lungs during volume (VC) and pressure control (PC) mechanical ventilation. PaO2 oscillations with an amplitude of about 10mmHg and the same frequency as breathing were observed at a RR of 12 breaths per minute, inspired-to-expired ratio (I:E) of 1:1, tidal volume (VT) of about 10mlkg−1, and at mean PaO2 near 130mmHg. The peak airway pressure was maintained below 13.6 cmH2O. The airway pressure remained lower than the mean pulmonary artery pressure throughout the experiment.
Figure 2 shows the rate of PaO2 decline over time during breath hold manoeuvres at end-expiratory volume (493ml; left), end-inspiratory volume (783ml; centre), and at the end of a large inspiration (1,073ml; right). The figure illustrates the slower rate of decline observed at the larger lung volume, where also a transient increase in PaO2 was observed immediately following the large inspiration. There was good agreement between measurements of VT recorded via the respiratory monitor (Datex Ultima) and estimates of the lung volume increases consequent upon inflation made by CT image analysis: the mean volume difference (i.e. bias) was −22 ml, with limits of agreement between −67 (95% CI −80 to −54) and 22 (95% CI 9 to 35). The SD of the differences was 23ml (SE±3.7ml) (Figure S1). Representative slice CT images recorded as soon as the breath hold manoeuvre began, and just before tidal breathing was restored are shown in Figure S2 for reference. Similarly, Figure S3 shows three-dimensional reconstructions of the lung obtained from CT images, highlighting regions associated with different Hounsfield Units ranges corresponding to well- or poorly-ventilated lung, and atelectasis.
Table 1 shows that the three lung volumes measured with CT during the breath hold manoeuvres were significantly different (p<0.001), as were the associated rates of PaO2 decline (p<0.01). The rates of PaO2 decline ranged from 6.9±1.3mmHgs−1 at end-expiratory lung volume to 3.4±0.6mmHgs−1 during the large inspiration (VT20). Table 1 also shows whole lung atelectatic mass fraction measured with CT as soon as the breath hold manoeuvre began and just before breathing was restored. There was no significant difference in the whole lung atelectatic mass fraction at the beginning and at the end of the breath hold manoeuvres [C.I. −1.4%–0.6%].
Figure 3 shows representative PaO2 values over time during volume (A–D) and pressure (E–H) control mechanical ventilation delivered at a RR of 12 breaths per minute, with PEEP set at 5 cmH2O. These two ventilation modes and I:E ratios of (A, E) 1:2, (B, F) 2:1, (C, G) 1:4 and (D, H) 4:1 affected both the mean PaO2 and the oscillation amplitude. Similarly, Fig. 4 shows representative PaO2 values over time for a RR of 6 breaths per minute, where greater changes in mean PaO2 levels and larger oscillations were observed.
Table 2 shows the mean PaO2 values associated with the eight combinations of mechanical ventilation settings studied at RR of 12 (VT=10mlkg−1), and the associated mean airway pressure. In the experiments considering I:E of 1:2 and 2:1, the ventilation mode (pressure or volume control) resulted in significant PaO2 changes [F(1, 5)=18.8, p=0.007], as did inverting I:E ratios (i. e. switching the duration of inspiration with that of expiration) [F(1, 5)=8.4, p=0.03]; no significant interaction effect was observed [p>0.05]. In the experiments considering I:E of 1:4 and 4:1, the ventilation mode resulted in significantly greater PaO2 changes [F(1, 5)=11.6, p=0.02], as did inverting I:E ratios [F(1, 5)=16.5, p=0.01], with a significant interaction [F(1,5)=11.5, p=0.02]. The greatest mean PaO2 change was observed when I:E was inverted from 1:4 in volume control mode to 4:1 in pressure control mode, when the mean PaO2 increased by 19mmHg, from 125mmHg (±10mmHg) to 144mmHg (±8mmHg). At RR12 the amplitude of the PaO2 oscillations decreased on average from approximately 10 to about 7mmHg when inverse-ratio ventilation was used. The reduction in amplitude of the PaO2 oscillations was particularly clear at RR6 (VT=20mlkg−1), when inverse-ratio ventilation reduced average PaO2 oscillations from 40 to 15mmHg, observed respectively in pressure control mode (I:E 1:4) and volume control mode (I:E 4:1). Table 3 shows the similar exaggerated changes observed at RR6. At this RR, the greatest mean PaO2 change was observed when I:E was inverted from 1:4 in volume control mode to 4:1 in pressure control mode, when mean PaO2 increased by 24mmHg, from 137mmHg (±18mmHg) to 161mmHg (±18mmHg).
The change in the mass of the slice that was atelectatic was less than 2% [C.I. 1.5–2.0%] during tidal breathing. Figure 5 shows the proportions of the lung slice mass associated with atelectasis, poor and normal aeration over the period of the respiratory cycle, for each ventilatory condition studied. The atelectatic mass was not different between inspiration and expiration in any of the conditions studied. In contrast, the fraction of normally aerated lung mass increased during inspiration, with a corresponding decrease in the fraction of poorly aerated lung mass. The portion of overinflated lung slice never exceeded 0.1% of the total lung mass.
Figure 6 shows the predicted values of dPA(t)/dt according to the proposed model, against the measured values of dPA(t)/dt. The predicted values overestimated the measured values by 31%.
Figure 7 shows results from the simulations of PaO2. These simulations showed changes in mean PaO2, and PaO2 oscillation amplitudes that were similar to those observed in vivo under the various conditions imposed.
In this experimental study in mechanically ventilated pigs with uninjured lungs, we showed that PaO2 declined at different rates during breath hold manoeuvres depending on lung volume in vivo, that significant dynamic PaO2 oscillations occur, and that these PaO2 oscillations are determined by alveolar oxygen changes within the respiratory cycle.
During tidal breathing with mechanical ventilation, PaO2 oscillations were observed at respiratory rates of 12 and 6 breaths per minute, at mean PaO2 values between ~120 and ~150mmHg, at any I:E ratio and ventilation mode considered in this study. These PaO2 oscillations achieved maximum amplitudes greater than 50mmHg in the absence of cyclical atelectasis as measured by CT image analysis during ventilation in pressure control mode at a RR of 6 breaths per minute, I:E 1:4, VT of 20mlkg−1, and with 5 cmH2O PEEP. In this sense, any shunt should be of the same magnitude throughout the respiratory cycle, unless blood flow is redistributed within the breath. The observed increase in PaO2 during inspiration suggests that this redistribution of blood flow to non- or poorly ventilated regions did not occur, indicating that it was not a determinant of respiratory PaO2 oscillations.
In agreement with the results obtained during breath holds, the weighted-average lung volume was a strong determinant of mean PaO2 value, as observed by altering I:E ratios and ventilation control mode. In particular, mean PaO2 was higher when inspiration lasted longer than expiration, which is when lung volume is greater across the respiratory cycle. Moreover, the amplitude of the PaO2 oscillations was smaller at greater mean lung volume, suggesting that there was a slower PaO2 decline during the expiratory phase at I:E of 2:1 and 4:1 than when these I:E ratios were inverted. Overall, these respiratory PaO2 oscillations could be predicted from the oscillations in arterial partial pressure of carbon dioxide24.
Respiratory PaO2 oscillations have previously been interpreted primarily as being caused by cyclical atelectasis associated with tidal recruitment and derecruitment during ventilation in animal models of the acute respiratory distress syndrome (ARDS)5, 6, 9, 25. Differences between our work and the published literature include animal species, smaller tidal volumes, greater RR, reduced airway pressures, physiological mean PaO2, and crucially that in our work we studied uninjured lung as opposed to animal models of lung injury. It is possible that the use of much larger tidal volumes (as it appears in the literature) might have resulted in larger PaO2 oscillations in our uninjured lung model; we did not investigate this possibility as tidal volumes greater than those studied in our experiments are neither physiological, nor of clinical interest26. It is intriguing to notice that the PaO2 oscillation amplitude of ~22mmHg observed at I:E of 4:1, RR 6, VT 20mlkg−1, mean PaO2 value of ~161mmHg in our work is similar to that of ~25mmHg reported in a pig saline lavage model of ARDS in rather similar conditions (I:E 4:1, RR 5, VT 20mlkg−1,mean PaO2 value of 425mmHg)9, 27. In this respect, it is possible that the oscillations reported in the literature were caused in part by oxygen mass balance as described in equations 4 and 5 (ventilatory delivery vs pulmonary uptake) rather than cyclical recruitment and derecruitment.
To emphasize the observation that PaO2 rate of decline was associated with lung volume, we performed a series of breath hold experiments at EELV with 5 cmH2O, end-inspiration and at the end of a large inspiration. We showed that PaO2 declined more rapidly at smaller lung volumes than it did at larger lung volumes. This result is in agreement with an original observation performed with slower oxygen saturation sensors during apnea, when the rate of saturation decline was “inversely proportional to the volume of air in the lung during apnea and closely related to the oxygen saturation at the beginning of apnea”28.
This difference in the rate of PaO2 decline could be partly explained by lung recruitment and derecruitment in anaesthetized animals. Lung recruitment could be observed during end-inspiratory breath holds, especially during a breath hold performed after a large inspiration, and derecruitment could be observed during an end-expiratory breath hold. Lung recruitment during an end-inspiratory breath hold could reduce the rate of PaO2 decline, and vice versa during an end-expiratory breath hold. However, in this study, we were unable to detect atelectatic mass changes during breath holds using CT image analysis, which did not provide any evidence for recruitment or derecruitment during end-inspiratory or end-expiratory breath holds respectively. Indeed, this result was partly expected given the short duration of the breath hold manoeuvres, the relatively low pressures used during the inspiratory breath holds, the application of 5 cmH2O PEEP during the end-expiratory breath holds, and the lack of known lung injury or gross regional ventilation-perfusion mismatch.
Ventilation and perfusion changes during a breath hold could determine real time changes in PaO2; in particular, elevated airway pressures applied to the lung during inspiration could cause a prevalence of zone 1 aeration29, 30, where the non-dependent lung regions receive little or no perfusion also due to changes in lung blood volume31. This phenomenon per se would increase the rate of PaO2 decline during an end-inspiratory breath hold, due to a reduced oxygen uptake in the zone 1 regions of the pulmonary circulation. The highest airway pressure applied in our experiments was 24 cmH2O during breath holds at the end of a large inspiration, when the rate of PaO2 decline was slower than during end-inspiratory and end-expiratory breath holds. Importantly, the airway pressure did not exceed the pulmonary artery pressure at any time during breath holds, and overinflated lung mass was never greater than 0.1%, suggesting that zones 1 were not dominant, hence did not affect the rate of PaO2 decline.
Two main limitations of this technology are that it is unable to demonstrate whether a respiratory PaO2 change is caused by alterations in either ventilation, or perfusion, or both, and that it does not provide evidence for their spatial distribution within the lung. In our study, the dynamic CT image analysis partly compensated for this limitation by providing an index of lung aeration and atelectasis. Although this information was anatomically limited to a single slice during tidal breathing, the antero-posterior orientation of the slice could be representative for the effect of the gravitational gradient on pulmonary aeration32.
Our study demonstrates that cyclical atelectasis is not necessary for respiratory PaO2 oscillations to appear during mechanical ventilation. We conclude that the mechanism determining these respiratory PaO2 oscillations is the variation of alveolar oxygen tension within the breath, with oxygen entering the alveoli during each inspiration, and that these cyclic variations in alveolar oxygen tension are transmitted all the way to the systemic arteries. Our results in an anaesthetized, ventilated pig model, in the absence of lung injury, may have implications for the interpretation of the results from conditions of lung disease33, in particular from patients with lung injury, and from animal models of ARDS.
The study was approved by the Animal Research Ethics Committee at Uppsala University and by the UK Home Office. All the animal experiments conformed to the National Institutes of Health Guidelines for the Use of Laboratory Animals. Relevant sections of the ARRIVE guidelines were followed. A total of 8 domestic pigs were studied (average±SD: 29±2kg). Details regarding the anaesthesia, mechanical ventilation, instrumentation and measurements are reported in the supplementary information.
The intravascular, fibre optic oxygen sensor was inserted in a carotid artery via a standard arterial catheter, and PaO2 was recorded continuously, simultaneously with analogue signals for cardiovascular and respiratory parameters from patient monitors.
In order to determine the rate of PaO2 decline during apnoea, breath hold manoeuvres lasting at least 20s were performed at three lung volumes: end-expiration (end-expiratory lung volume, EELV, defined as functional residual capacity with 5 cmH2O PEEP), end of an inspiration corresponding to a VT of 10mlkg−1 (VT10), and at the end of a large inspiration, corresponding to a VT of 20mlkg−1 (VT20). Breath hold manoeuvres associated with these three lung volumes were performed six times each, in a sequence that covered each possible sequential effect, for a total of 18 breath hold manoeuvres per animal. The sequence in which breath hold manoeuvres were performed is presented in Table S1. PaO2 decline was normalized to account for small differences in body weight between animals. The steepest rate of decline measured continuously over 5s during each breath hold manoeuvre was considered for analysis. This period of 5s was used to maximize the interval time while avoiding the inclusion of PaO2 values below 100mmHg, when the rate of PaO2 decline decreases due to haemoglobin desaturation, which is beyond the scope of our study.
We aimed to establish the effect of altering I:E ratios (1:2 vs 2:1, and 1:4 vs 4:1) on mean PaO2 and PaO2 oscillation amplitudes, and the potential effect of cyclical atelectasis, if present, in determining PaO2 during tidal breathing. We studied the effects of ventilation in pressure control mode, and also explored the effect of delivering the same VT gradually in volume control mode, recording PaO2 changes and determining oscillation amplitudes at the four I:E ratios. Airway pressure was monitored and controlled throughout the experiments. The two sets of tests started with either I:E of 1:2 or 1:4 in volume control mode achieving a mean PaO2 of about 130mmHg, followed by a randomised sequence of any possible combination of I:E and pressure or volume control mode. Respiratory rates (RR) studied included 12 and 6 breaths per minute, with VT of 10mlkg−1 and 20mlkg−1 respectively. In total, we included 16 conditions (four I:E ratios, each in two control modes, and at two respiratory rates). During tidal breathing, once steady state had been achieved, data collected over a period of 2min were considered for analysis. Steady state was defined as mean PaO2 not changing by more than 5mmHgmin−1.
PEEP of 5 cmH2O was used throughout the experiments in order to prevent, or at least reduce the likelihood of collapse of dependent regions at end expiration.
We used computed tomography (CT) to measure lung volumes as well as to estimate pulmonary atelectasis and its changes during the breath hold manoeuvres and dynamically during tidal breathing in two animals. Details of the CT images acquisition, analysis procedure, calculation of air volume and tissue mass are presented in the supplementary information.
We hypothesised that the uninjured lung could be modelled as a single alveolar compartment with a constant oxygen uptake. The rate of change of oxygen in this compartment is equal to the difference in the rates of uptake by the pulmonary circulation and input via ventilation during inspiration, or elimination during expiration. Details of the mathematical modelling are presented in the supplementary information.
PaO2 data recorded during the breath hold manoeuvres were analysed by means of univariate repeated measures ANOVA with lung volumes as fixed factors (IBM SPSS Statistics, Version 22.0; Armonk, NY, USA), and post hoc analysis with the Bonferroni method. A two-way ANOVA was conducted to compare the main effects of ventilation mode and I:E ratio, and their interaction on mean PaO2 and its oscillations. Ventilation mode included two levels (volume and pressure control mode), and I:E ratio consisted of two levels (either 1:2 and 2:1, or 1:4 and 4:1). Values presented are mean±standard deviation, unless otherwise stated. Significance level was accepted at p<0.05.
Data are available upon request.
We are grateful to Lindsay Duckett, Agneta Roneus, Anders Nordgren, Kerstin Ahlgren, Monica Segelsjo for technical support, to Dr Luigi Camporota for help with CT image analysis, and Professors Keith Dorrington, Peter Robbins and Pierre Foex for helpful discussions. This work was supported by a Wellcome Trust Translation Award to ADF and CEWH, Swedish Heart and Lung Foundation, Swedish Research Council (K2015-99X -22731-01-4) to AL, an Oxford University Medical Research Fund to FF, and a Whitaker International Fellow Grant to NB.
F.F., G.H., C.E.W.H., A.L., A.D.F. designed the experiments. F.F., N.B., R.C., H.M., D.H. performed the experiments. F.F., N.B., J.C. analysed the data. F.F., G.H., C.E.W.H., A.L., A.D.F. interpreted the data. F.F., N.B., C.E.W.H., A.L., A.D.F. contributed to financial support. F.F. wrote the manuscript. All authors read and critically revised the manuscript.
The authors declare that they have no competing interests.
Electronic supplementary material
Supplementary information accompanies this paper at doi:10.1038/s41598-017-06975-6
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.