Introduction

Mechanical ventilation is essential in the treatment of patients affected by acute lung injury (ALI) or acute respiratory distress syndrome (ARDS). However, if the ventilatory parameters are not titrated adequately, it may cause volutrauma and atelectrauma as well as worsening of inflammatory processes [15, 31, 38]. Positive end-expiratory pressure (PEEP) improves lung function and may reduce the duration of mechanical ventilation [30]. One goal of a protective ventilatory strategy is to choose a value of PEEP that maximises recruitment while avoiding over-distension. However, the best way to choose the optimal PEEP level is still under investigation.

Computed tomography (CT) provides an objective tool for assessing lung aeration and consequently for establishing optimal ventilation settings [17, 18, 27, 37, 45]. However, CT is not available at the bedside, it is not adequate for a continuous monitoring and it is associated with high doses of ionising radiation.

In clinical practice PEEP is usually adjusted according to oxygenation [43], but this may overlook PEEP-induced over-distension and intra-tidal recruitment/derecruitment. An alternative approach is to set PEEP in order to maximise dynamic compliance (Cdyn) [41] or similarly to minimise the elastance (Ers) of the respiratory system (Ers = 1/Cdyn) [2, 4, 5], during a decremental PEEP trial. Even though Cdyn has the advantage of being continuously provided by the ventilator, its estimation is strongly affected by non-linearities, which may be relevant in diseased lungs, and by respiratory muscle activity, requiring either deep sedation or the use of an esophageal balloon in the presence of spontaneous breathing.

We have recently shown that reactance or the oscillatory compliance (CX5) derived from reactance measured by the forced oscillation technique (FOT) at 5 Hz may be used to monitor recruitment/derecruitment during mechanical ventilation, overcoming the limitations mentioned above [9]. Reactance and CX5 are measured by analysing the response of the respiratory system to very small oscillatory pressures (~2 cmH2O amplitude) with a frequency of 5 Hz.

The aim of the present study was to evaluate the ability of CX5 to identify the minimum level of PEEP which should be applied in order to prevent volume derecruitment and to compare it to the gold standard provided by CT.

Materials and methods

The study was carried out at the Departments of Clinical Physiology and Radiology of the University Hospital of Uppsala, Sweden, and it was approved by the local animal ethics committee.

Study population

Seven pigs (weight 29.8 ± 2.1 kg) were studied. A detailed description of animal preparation can be found in the electronic supplementary material.

Experimental protocol

Lung injury was induced by broncho-alveolar lavage. PEEP was increased from 0 to 24 cmH2O in steps of 4 cmH2O and subsequently decreased from 24 to 0 in steps of 2 cmH2O. At each PEEP level, a CT scan was performed during a 10 s expiratory hold, and arterial blood was sampled for gas analysis. The duration of each step was 8 min, the total duration of the experiment was ~150 min.

Experimental set-up and measurements

FOT was applied using a set-up which has been described elsewhere [9]. Briefly, low amplitude (~2 cmH2O peak-to-peak) sinusoidal pressure oscillations at 5 Hz were generated by a loudspeaker connected to the inspiratory line of the mechanical ventilator. Flow at the airway opening (\( \dot{V}_{\text{ao}} \)) was measured by a differential pressure transducer and a mesh-type heated pneumotachograph connected at the inlet of the tracheal tube. Pressure (Ptr) was measured at the tip of the endotracheal tube by a differential pressure transducer. A detailed description of the set-up and measurements can be found in the electronic supplementary material.

Data analysis

Lung mechanics

The estimation of the respiratory system input impedance (Zrs) was obtained from the flow and pressure signals by a least-squares algorithm [14, 25]. Zrs is composed of a real part, called resistance (Rrs), and an imaginary part, called reactance (Xrs). Xrs was used to compute oscillatory compliance (CX5) using the following equation [9]:

$$ C_{\text{X5}} \, = {\frac{ - 1}{{2 \cdot \pi \cdot 5 \cdot X_{\text{rs}} }}} $$
(1)

CX5 data were averaged over the periods of expiratory hold needed to perform CT scans, providing a single data point for each CT scan.

Open lung PEEP (PEEPol) was defined as the PEEP level corresponding to a maximum in Xrs and CX5 during the decremental trial.

Cdyn was calculated by fitting the equation of motion of the respiratory system to Ptr and \( \dot{V}_{\text{ao}} \) by the least-squares method on approximately 5–10 breaths immediately preceding the CT scan. An extension of the equation of motion, including a volume-independent (C1) and a volume-dependent (C2) component of compliance, was also considered [26, 34].

Computed tomography analysis

Changes in lung aeration were studied by whole-body CT scans. Images were reconstructed with 8-mm slice thickness and analysed using dedicated software (Maluna 2.02). The total lung volume was subdivided into over-aerated (OA, −1,000 to −900 HU), normally aerated (A, −900 to −500 HU), poorly aerated (PA, −500 to −100 HU) and non-aerated (NA, −100 to +100 HU) volumes as previously suggested [17, 45]. Lung gas (Vgas) and tissue (Vtiss) volumes were calculated using standard equations [17] for the whole lung and for each aeration compartment. The amount of derecruited lung was quantified as the volume of tissue in the non-aerated region and expressed as a percentage of total tissue volume (VtissNA%). Similarly, the percentage amounts of aerated (VtissA%) and poorly aerated (VtissPA%) tissue were calculated.

Statistical analysis

Data are expressed as mean ± SD. Significance of differences between different PEEP levels was tested by one-way ANOVA for repeated measurements. Multiple comparison after ANOVA was performed using Holm-Sidak test. Differences between PEEPol and the corresponding PEEP on the inflation limb of the curve were tested by paired t test. Differences were considered statistically significant for p < 0.05. The sensitivity and specificity of CX5 for the detection of lung collapse was calculated using CT as a reference.

A detailed description of the methods used for data analysis can be found in the electronic supplementary material.

Results

Figure 1 shows the average behaviour of Xrs, CX5, Cdyn, VtissNA% and blood gases.

Fig. 1
figure 1

Relationship between PEEP for all pigs during the inflation-deflation PEEP trial and a oscillatory compliance (CX5), b respiratory system reactance (Xrs), c dynamic compliance (Cdyn), d percentage amount of non-aerated tissue (VtissNA%), e partial pressure of carbon dioxide (PaCO2) and f partial pressure of oxygen (PaO2). Values are mean and SD. The arrows identify the inflation and deflation limbs of the curve. The circles identify the optimum PEEP defined as the maximum CX5 value reached during the deflation limb. *p < 0.05 compared to the previous step

During inflation Xrs, CX5 and Cdyn significantly increased until a PEEP of 16 cmH2O, after which they tended to decrease. The decremental limbs of these curves were shifted upwards and to the left compared to the inflation limbs and reached maximum values at PEEP 14 cmH2O for Xrs and CX5 and at 12 cmH2O for Cdyn. C1 versus PEEP presented a similar shape compared to Cdyn, with a maximum at PEEP 12 cmH2O on the decremental limb. During inflation VtissNA% significantly decreased from PEEP of 0 to 20 cmH2O, while during deflation it did not change significantly until PEEPol, after which it significantly increased. During deflation PaCO2 was always lower than during inflation and presented a minimum at 12 cmH2O. PaO2 increased with PEEP in a sigmoidal way, deflation values being higher than inflation values at each PEEP level.

Figure 2 shows CX5, blood gases and CT data and a representative CT slice (~1 cm above the diaphragmatic dome) during deflation from one representative animal. The CT images show the regional distribution of differently aerated regions.

Fig. 2
figure 2

a Oscillatory compliance (CX5), b blood gases (PaO2 in black and PaCO2 in blue) and c percentage of the tissue volume of differently aerated regions (black over-aerated, blue normally, green poorly and red non-aerated regions) obtained during the deflation limb of the PEEP trial from a representative animal. Right A representative CT slice (selected approximately 1 cm above the diaphragmatic dome) of the same animal at the different PEEP levels. Color code for regions as above

Comparing individual animals, the shapes of the curves were similar. In particular, Xrs and CX5 always showed a dome-shaped behaviour reaching maximum values at, on average, a PEEP of 13.4 ± 1.0 cmH2O. Cdyn and C1 also displayed similar shape, but reached their maximum values at a lower PEEP compared to FOT data (12.0 ± 1.6 and 12.6 ± 1.5 cmH2O respectively).

To characterise the behaviour of various parameters around PEEPol, we averaged the data considering for each animal the differences between the values measured at PEEPol and those measured two steps immediately before and after PEEPol (Fig. 3). As expected, CX5 presented negative differences with respect to PEEPol for all protocol steps. Cdyn displayed a similar shape but with the maximum on average one step (2 cmH2O) lower than CX5. PaO2 increased slightly for PEEP levels higher than PEEPol, while it sharply decreased for PEEP levels lower than PEEPol. PaCO2 presented positive differences with respect to PEEPol. Finally, VtissNA% abruptly increased for PEEP levels lower than PEEPol, indicating the beginning of derecruitment.

Fig. 3
figure 3

Differences in a oscillatory compliance (CX5), b dynamic compliance (Cdyn), c partial pressure of oxygen (PaO2), d partial pressure of carbon dioxide (PaCO2) and e percentage amount of non-aerated tissue (VtissNA%) with respect to open lung PEEP (PEEPol) at the two steps preceding and following PEEPol. *p < 0.05, **p < 0.01 with respect to PEEPol

Haemodynamic variables, CT volumes and respiratory mechanics parameters are reported in Table 1.

Table 1 Haemodynamic variables, CT volumes and respiratory mechanics parameters at open lung PEEP (PEEPol), and two steps before and two steps after PEEPol

FOT was able to detect lung collapse with 100% sensitivity and 92% specificity compared to CT.

Figure 4 shows the differences in CX5, PaO2 and VtissNA% between PEEPol and the corresponding value on the incremental limb of the curve. CX5 at PEEPol was significantly higher than at the corresponding PEEP on the incremental limb of the curve; PaO2 was significantly higher and VtissNA% was significantly lower.

Fig. 4
figure 4

Oscillatory compliance (CX5), oxygenation (PaO2) and percentage amount of non-aerated tissue (VtissNA%) at open lung PEEP (PEEPol) (grey bars) and at the same PEEP value on the inflation limb of the curve (black bars). **p < 0.01

Discussion

In the present study we evaluated the use of FOT for bedside identification of the lowest PEEP able to prevent lung derecruitment after a recruitment manoeuvre in a surfactant-depletion model of ALI. We compared FOT data with the distribution of lung aeration as measured by whole-lung CT scan.

The main finding of this study was that FOT could identify PEEPol in good agreement with CT. In particular, CX5 identified the minimum PEEP level required to maintain lung recruitment with high sensitivity and specificity.

Comparison with other studies

The use of lung mechanics in dynamic conditions to optimise PEEP, originally suggested more than 30 years ago by Suter et al. [42], has recently been re-evaluated [2, 4, 5, 41], providing promising results for the bedside optimisation of mechanical ventilation.

We have recently shown that CX5 derived from Xrs, measured by single frequency FOT at 5 Hz, is very sensitive to lung volume recruitment and derecruitment [9]. This approach offers three main advantages over dynamic respiratory mechanics estimated by multi-linear regression analysis. First, the stimulus applied during FOT induces very small lung volume changes, minimising the artefacts due to non-linearities of the respiratory system. In contrast, to calculate dynamic compliance multi-linear regression analysis is performed over a whole breath. In this large volume range, the lung does not behave linearly and compliance is not constant. To compensate for this, it is possible to assess compliance at different volume increments by the SLICE method [22] or to add a volume-dependent term in the equation of motion [26, 34], but in this way the convergence properties of the fitting are reduced. Second, since CX5 is not affected by the spontaneous respiratory activity of the patients, it does not require either sedation or the use of an esophageal balloon. The current need for deep sedation or muscle paralysis to perform accurate PEEP titration guided by static or dynamic compliance is the major impediment for a more widespread use of lung mechanics to select PEEP. Third, since CX5 may be measured with a high time resolution (0.2 s at 5 Hz), it allows the assessment of within-breath changes in lung mechanics. In this way it could be possible to monitor intra-tidal recruitment and over-distension, both of which have a role in ventilator-associated lung injury [21, 29, 30]. It is possible to account for intra-tidal recruitment or over-distension also by introducing a volume-dependent component of compliance in the equation of motion [26, 34] or by analysing the shape of the dynamic pressure-time curve [20, 34] during constant-flow ventilation. The disadvantage of all these approaches is that they require sedation.

We found a systematic difference between PEEP corresponding to the maximum Cdyn and PEEP corresponding to the maximum CX5, the latter being on average 2 cmH2O greater than the former (Fig. 3). Interestingly, this result is in good agreement with a mathematical modelling study [24] reporting that open lung PEEP is greater than the PEEP providing maximum tidal PV slope and therefore maximum Cdyn. In fact, PEEP is currently set one step higher than the PEEP at which Cdyn achieves a maximum. However, this is an empirical rule which is not based on a clearly defined physiological reason. The use of a volume-dependent component of compliance, even if it provided slightly greater values for optimum PEEP compared to Cdyn, was not able to provide the same optimum PEEP as that identified by CX5.

Another result of this work is that it gives further support for the significance of titrating PEEP during decremental PEEP trials. This is clearly shown by the higher values of Xrs found on the deflation limb compared to the ones measured during deflation at PEEPol. In a mathematical model of ARDS, Hickling [24] showed that, with incremental PEEP, the PEEP level associated with the maximum mean tidal PV slope did not coincide with open lung PEEP. The interpretation supported by our data is that during incremental PEEP the lung is recruited by the peak pressure, while during deflation, PEEP prevents the collapse of the lung that has already been fully recruited.

Limitations of the study

The total Zrs measures the mechanical properties of the whole respiratory system, which comprises both lung and chest wall. During a PEEP trial, changes in end-expiratory lung volumes are shared between the lung and the chest wall. Since chest wall compliance (Ccw) changes with volume, it will affect Xrs and CX5, potentially impairing the accuracy of Xrs regarding the evaluation of lung volume recruitment. However, from a mathematical simulation (see Appendix 1), we estimated that the contribution of Ccw changes to Xrs changes was negligible compared to the effect of recruitment/derecruitment.

The thresholds separating different aeration regions used in this study are those internationally recommended [16, 17, 32, 33, 36]. However, these thresholds are arbitrary, and they may influence the estimation of the amount of recruited and hyperinflated volume. We used a cut-off density between the aerated and the over-aerated region of −900 HU, similar to Vieira et al. [45] while other groups used cut-off levels of −800 HU [7]. The absolute cut-off may differ due to differences in CT calibration [1] and CT resolution [44].

Compared to other studies using similar animal models [6, 19], we found analogous results regarding CT data but not for the amount of over-aerated tissue, which was negligible in our study. The disparity may be explained by the higher slice thickness used in the present study. It has been demonstrated that low spatial resolution CT provides adequate data for most of the aeration partitions, but it may significantly underestimate lung over-inflation in patients with acute lung injury [44]. Since the aim of the present study was to correlate the amount of lung volume derecruitment with FOT data, the CT analysis had to be extended to the whole lung with the lung perimeter segmented manually to avoid errors in the separation of collapsed lung with chest wall tissues. We accomplished this task by analysing more than 3,300 8-mm slices; the choice of thinner slices would have required manual processing of almost 20,000 images without significantly contributing to the main message of the study.

Moreover, over-aerated regions identified by CT represent lung tissue overfilled with gas. Conversely, over-distension is defined as the condition of excessive mechanical stress applied to lung tissue, and it is related to alveolar wall tension. In severe ARDS, in which the whole lung has an increased mass, high levels of PEEP may result in local over-distension even if the average density of the region is greater than −900 HU [17]. Therefore, the occurrence of mechanical distension that is detected by FOT at end-expiration may not be directly related to the amount of over-aerated volume measured by CT.

In this study we used a model of ALI where natural surfactant is removed from the alveolar space by repetitive sequences of bronchial lavage/drainage [3]. This model is characterised by atelectasis resulting from distal airway collapse with much less contribution of inflammation and oedema [23, 39, 40]. The lung is easily recruitable, and the model may thus more closely mimic the clinical conditions of surfactant deficiency in premature neonates rather than the typical ARDS seen in adult intensive care units. In human ARDS, PEEP-induced lung re-aeration probably results from the displacement of the gas-liquid interface distally in the alveolar space, and it is unlikely that PEEP acts by exceeding hypothetical ‘threshold opening pressures’ [28, 29]. Even if these mechanisms of lung volume recruitment are quite different, the mechanical consequences on input impendance (Zin) measurements are likely similar.

Nevertheless, an important advantage of FOT at 5 Hz is that it is not affected by the spontaneous breathing of the patient, as confirmed by the results of several studies [11, 12, 14]. Moreover, FOT at 5 Hz has been successfully applied during continuous positive airway pressure (CPAP) [13] and non-invasive mechanical ventilation in COPD patients. Future investigations should address PEEP titration by FOT in patients on pressure support.

Another attractive feature of this technique is that it allows measurements of within-breath changes of respiratory mechanics [13, 14]. This is potentially useful during mechanical ventilation to assess intra-tidal recruitment and over-distension. In the present study we reported only values of Xrs measured at end-expiration because they are more appropriate to the aim of detecting the minimum level of PEEP able to prevent lung volume derecruitment. Future studies are required in order to validate FOT measurements at end-inspiration and to investigate whether intra-tidal changes in FOT parameters can be used to identify cyclic recruitment and over-distension to guide the clinician in the choice of ventilatory settings.

Finally, even if there is clear evidence that an appropriate ventilation optimisation strategy based on measurement of lung mechanics reduced inflammatory markers in ARDS patients compared to oxygenation-based ventilation tailoring [35], future studies are needed to evaluate if an Xrs-based optimisation strategy performs better than the other approaches in terms of inflammatory response.

Conclusion

This study demonstrates that the measurement of Xrs by FOT during a decreasing PEEP trial identifies the lowest PEEP able to maintain the lung recruited as confirmed by whole-lung CT scans in a surfactant-depletion model of ALI.

Considering that FOT is non-invasive, that it can be easily integrated in commercial mechanical ventilators and that the computation of Xrs can be automated, this technology is potentially useful for optimising PEEP during mechanical ventilation of ALI/ARDS patients.