Heterogeneity of regional inflection points from pressure-volume curves assessed by electrical impedance tomography

Although mechanical ventilation (MV) might be necessary to maintain adequate gas exchanges in patients with acute respiratory failure, ventilator-induced lung injury (VILI) can amplify local and systemic inflammation and contribute to disease progression [1‐3]. VILI is caused by multiple mechanisms such as high volumes, high pressures, and cyclic opening and closing of the peripheral airways [4]. Moreover, lungs with patchy infiltrates are heterogeneous and have considerable parenchyma loss of volume available for ventilation [5]. The gravitational increasing weight of the parenchyma and the inhomogeneous inflammation process create differences between lung units located in different regions of the lungs, with different distending pressures, dynamic behavior, and therefore physiologic needs [6, 7]. Information coming from mechanical respiratory properties assessed through airway pressure (Paw) characterize the global behavior of the lung but possibly cannot identify regional peculiarities.

The pressure-volume (P-V) curve is a respiratory monitoring technique that explores changes in the respiratory system compliance along a wide range of Paw (e.g., between 0 and 40 cmH₂O). A P-V curve usually has sigmoidal shape [8] with two inflection points—the lower (LIP) and the upper (UIP)—and an almost linear part in between. The physiological interpretation of the P-V curve classically considers the region between the boundaries of the LIP and UIP as safe for mechanical ventilation (i.e., a PEEP level higher than LIP and plateau pressure lower than UIP) to prevent both atelectrauma and barotrauma [9, 10]. However, one can question whether that this approach is able to prevent VILI since it considers the respiratory system as a whole and does not take into account the local mechanical behavior which is influenced by gravity and super-imposed pressures [11]. In fact, it is possible that the same airway pressure might overinflate parts of the lung while being harmless for others or that a certain level of positive end-expiratory pressure (PEEP) is able to keep parts of the lung ventilated not being enough for others.

Electrical impedance tomography (EIT) is a radiation-free lung imaging technique, which allows continuous bedside monitoring of the regional mechanical properties by measuring changes in impedance associated with ventilation [12‐15]. In the present study, we hypothesized that regional P-V curves would provide different information from those obtained from global P-V curves, both in terms of upper and lower inflection points. To confirm this hypothesis, we constructed pressure-volume curves for each pixel row from non-dependent to dependent lung regions of patients affected by acute hypoxemic respiratory failure (AHRF) and acute respiratory distress syndrome (ARDS) [16].

The main results of this study are as follows: (1) The global P-V and the regional P-V curves do not provide the same information, and inflection points of the global P-V curve are different from those of the PVr (i.e., higher LIP and lower UIP); (2) the analysis of the regional P-V curves allows to detect the heterogeneity of the inflection points of the different lung regions; regional inflection points are gravity dependent with UIPr_MIN located in the most non-dependent lung and LIPr_MAX in the most dependent; (3) the “safe” limit of pressure between LIPr_MAX and UIPr_MIN was reduced compared to the one obtained by global P-V curve (Table 2).

Different attempts have been made in the last 30 years for use of the P-V curve of the respiratory system to set mechanical ventilation [9, 10]. The inflection points of this curve express changes in compliance, reflecting a rise in ventilated alveolar units—due to alveolar recruitment or airway opening or both, or a pressure that exceed intrinsic PEEP—(for LIP) and the overstretching of already ventilated lung (for UIP). Hence, setting mechanical ventilation in the linear part of the curve between these two points could be protective against both intra-tidal alveolar opening and closing (atelectrauma) and overinflation (barotrauma/volutrauma). Despite the physiological background, this approach considers the lung as a homogenous mono-compartimental system without taking the entity of regional inflammation [23] and the gravitational superimposed pressure on each region into account [11].

On the one hand, the LIP of the global P-V curve has been classically considered the critical opening pressure at which alveoli start to open or to be fully recruited but this concept has recently been questioned. A mathematical model of ARDS [24] and CT scans on patients affected by ALI/ARDS [25] showed that recruitment of new units occurs also beyond the LIP, influencing the slope of the P-V curve. Moreover, alveolar opening pressure (i.e., recruitment) is not a fixed threshold but a continuous range of pressure, progressively increasing from the non-dependent to the dependent lung. Finally, LIP could represent reversal of airway closure rather than alveolar recruitment [26]. We confirmed these findings since we found a gradient of regional lower inflection points from the non-dependent to the dependent lung: moving toward the vertebral-dependent lung, the LIPr increases, probably as an effect of the highest alveolar (or small airways pressure) pressure needed to overcome the resultant superimposed pressure (Fig. 3). Moreover, we found a significant difference between the LIPg, the regional minimum LIP (LIPr_MIN), and the LIPr in the most non-dependent lung (ROIs 1, 2, 3). Therefore, the LIP on the global P-V curve may express the behavior of the uppermost compartment that receives air at the beginning of inflation, as already suggested by Hickling using a mathematical model [24]. This would, however, imply that information about the LIPr in the central and dependent lung is not adequately reflected by the global P-V curve. Since the dependent lung is the region most affected by intra-tidal recruitment and airway closure due to the higher pressure needed to start inflation, using the global LIP—that expresses the behavior of the non-dependent lung—to set PEEP would leave large dependent regions prone to atelectrauma. Our data show that the minimal pressure needed to ventilate above all regional LIP (LIPr_MAX) was 15.8 [9.25–21.1] cmH₂O, a value much higher than that derived from the LIPg. If we hypothesize that the ideal ventilation should occur on the linear part of each regional P-V curve and that the starting point of the respiratory cycle is PEEP, then LIPg would underestimate the value to achieve this target.

On the other hand, alveolar overdistension leads to compression of pulmonary vessels and capillaries, disruption of alveolar epithelium, and physical breaks in endothelial plasma membranes [27] triggering the proinflammatory signaling cascade which results in inflammation, edema, and cell death [28]. Titrate tidal volume to keep a plateau pressure below the “overdistension threshold” can be useful to avoid lung inflammation.

Amato et al. [10] found a massive reduction of 28-day mortality (38% vs 71%) in patients ventilated with low tidal volumes (6 ml/kg/PBW); this finding was later confirmed by large-scale RCTs [29] and meta-analysis [30] proposing low tidal volume ventilation as the standard of care. Despite this potential advantage, PBW is not an accurate index of the actual lung size since it does not express the dimension of the baby lung [5, 31]. The upper inflection point is considered the beginning of lung overstretch and, ideally, could be used to titrate indirectly TV on the lung size. Roupie et al. [9] found that, in a ventilation based on a TV = 10 ml/kg, in 80% of the studied population the TV needed to be reduced to 7.8 ± 0.9 ml/kg to obtain a Pplat below the UIP. We found that UIPg was not able to highlight the regional overstretch since UIPg differed from UIPr_MIN, a value that corresponds to the lowest pressure at which regional overinflation starts. Moreover, UIPg was different from UIPr in the non-dependent lung (Fig. 3); using the UIPg to limit pressure can therefore underestimate the potential damage on this region.

In our population, UIPr_MIN was 30.1 [23.5–37.6] cmH₂O, not far from the suggested as protective Pplat threshold found in AHRF patients affected by sepsis [32, 33]. In addition, a recent multilevel mediation analysis on 3562 ARDS patients showed that a reduction of Pplat < 30 H₂O could not further improve survival rate. It is important to underline that in our population the range was wide, from a minimum of 13.9 cmH₂O to a maximum of 49.7 cmH₂O, being below 30 cmH₂O in 44% of measurements. A fixed value, 30 cmH₂O in this case, cannot represent therefore the universal safe number, suggesting that a tailored safe threshold should be set based on the patient’s characteristics.

In this analysis, we introduced two variables: LIPr_MAX—the minimal pressure able to overcome all regional lower inflection points—and UIPr_MIN—the maximal pressure to avoid a ventilation above all regional upper inflection points. We hypothesize that LIPr_MAX can be helpful in setting PEEP while UIPr_MIN can represent a pressure limit to avoid regional overstretch.

Recently, respiratory driving pressure (DP) has been found to predict mortality in ARDS patients [34] and a titration of TV on DP has been suggested to reduce VILI [35, 36]. The connection between DP and overdistension has been furthermore confirmed by a CT scan study on humans [37]. In the current analysis, we introduced a new variable: ΔP_LIN. Its value reflects the maximal pressure range which avoids regional overinflation and derecruitment, thereby allowing a ventilation simultaneously above every regional LIP and below every regional UIP. ΔP_LIN expresses, therefore, individual threshold pressures, above which VILI might occur. In our study population, although mean ΔP_LIN was 12.6 [7.4–20.8] cmH₂O and therefore close to the value defined as protective in a recent prospective study [38], in 56% of the cases, it was below 14 cmH₂O, the limit value of driving pressure suggested by Amato et al. [34]. This indicates that, in some patients, a driving pressure threshold of 14 cmH₂O, despite generally considered protective, would be slightly too high in order to prevent regional damage (Fig. 4).

Our results underline that the traditional P-V curve could give partial information on the way patients should be ventilated. The linear part of pressure-volume relationship does not pertain to all lung regions, being instead a compromise between the mechanical characteristics of the dependent and not dependent part of the lung. Since it is fundamental to select the correct amount of PEEP to limit atelectrauma, while avoiding overcoming the UIP to limit overdistension, patients could be ventilated in the linear portion of the P-V curves taking into account all lung regions. We were able to find this linearity in the entire lung regions by simply using the maximal LIP and the minimal UIP. LIP and UIP exhibited a huge variability among patients, underlining that mechanical ventilation must be personalized. It is tempting to say that the analysis of the regional P-V curves could help in achieving this goal. In particular, it remains to be elucidated whether the VT associated with linear regional DP is able to control CO₂ and if this new approach could lead to a novel indication for extracorporeal lung assist [39].

Our study has several limitations. Firstly, we performed only inspiratory pressure-volume curve, as it is the most used for evaluating the variability of the compliance of the respiratory system. Alternatively, a physician could examine the expiratory limb of the curve to evaluate closure pressure. However, we decided to use the same methodology used in a RCT to set PEEP using the P-V curve [10]. Secondly, the study population consisted of a limited number of patients with heterogeneous diseases; therefore, no inferences on the patient outcome can be derived from our data. Thirdly, we did not measure transpulmonary pressure: this tool would have been useful to distinguish between the respiratory system and lung regional inflection points. Fourthly, the lack of randomization of PEEP levels and the return to zero pressure at the end of each step could have influenced the results of this study [18].

Regional inflection points derived by EIT show high variability, reflecting the heterogeneity of the lung. Regional P-V curves obtained by EIT could convey more sensitive information than global lung mechanics on pressures within which all lung regions express linear compliance. Future studies will explore if setting mechanical ventilation based on regional P-V curves could impact on patients’ outcome.