Introduction
Acute respiratory distress syndrome (ARDS) is a severe condition that affects around 1 in 10,000 people every year with life-threatening consequences [
1]. The pathophysiology of ARDS is characterized by bronchoalveolar injury and alveolar collapse (atelectasis) [
2-
5]. The use of recruitment maneuvers (RMs) in ARDS to open up unstable, collapsed alveoli using a brief increase in transpulmonary pressure has become common practice in intensive care units [
3], and a large variety of RMs has been proposed in the literature [
3,
6-
12]. However, there remains a great deal of confusion regarding the optimal way to achieve and maintain alveolar recruitment in ARDS and, in many cases, the precise mode of action of particular RMs is not well understood [
7,
8,
13,
14].
The most frequently used recruitment maneuver in ARDS treatment is sustained inflation (SI) [
8]. Studies have shown varying degrees of success, with several reporting post RM improvement in oxygenation [
15,
16] and reduction in lung atelectasis [
17]. However, SI has also been shown to result in increased risk of hypotension [
14,
16] and barotrauma [
18], decline in oxygenation [
19] and has even been reported to be ineffective [
20].
An alternative recruitment strategy that has been recently proposed is the prolonged recruitment maneuver (PRM) [
21], in which positive end-expiratory pressure (PEEP) is fixed to a higher than baseline level and the positive inspiratory pressure is progressively increased. When PRM was compared with SI in an experimental model of mild acute lung injury (ALI) induced in a rat lung, it showed improved alveolar recruitment, gas exchange and a reduced level of lung damage. To date, however, no further evidence is available to support the use of PRM in ARDS patients.
The third RM considered in this paper is the maximal recruitment strategy (MRS), which was evaluated via patient trials in [
6,
22]. When this strategy was followed by ventilation with low tidal volumes and titrated PEEP, a median of 45% relative lung tissue recruitment was observed in quantitative computed tomography (CT) scan analysis. However, as noted in [
13], the final PEEP levels applied at the end of the titration phase in the MRS resulted in inspiratory plateau pressures in the patient population of approximately 40 cm H
2O, on average. This far exceeds the 28 cm H
2O safety limit, which had been associated with increased inflammatory response in a previous study [
23], and even the 30 cm H
2O cutoff proposed by the ARDS Network. As noted in [
6], it seems very likely that the MRS caused high degrees of alveolar stress and strain in some patients.
In this study, we employ a high-fidelity computational simulator that reproduces the static and dynamic characteristics of several ARDS patients, to (a) compare the efficacy of the three RMs described above in improving key patient parameters describing oxygenation, carbon dioxide (CO2) retention and dynamic compliance and (b) investigate the effects of different PEEP settings in maintaining effective lung recruitment across a representative ARDS patient spectrum. Our central hypothesis is that computational simulation can be used to evaluate and understand the mode of operation of RMs for ARDS patients.
Discussion
The marked improvement in PaO
2/F
IO
2 seen in all five patients following the application of the MRS is striking, and might be explained as follows. During inspiration, a normal lung increases its volume uniformly, as almost all of its compartments have the same dynamic elastance and homogenous structure. This is not true in the case of diseased lungs, however. Gattinoni and colleagues [
32] have shown that the ARDS lung is characterized by a small functional volume termed the ‘baby lung’, which stays open throughout the respiratory cycle. At end-expiration, an adequate PEEP can maintain some of the recruited lung regions open for the next respiratory cycle to take place. Therefore, selecting an appropriate recruitment strategy and modifying the PEEP applied at the end of an intensive ventilation period should increase and maintain the size of the baby lung and contribute to improving the gas exchange. As shown in Figure
9, in the simulator almost 100% of the alveolar compartments are recruited during a maximal recruitment maneuver, while 80% of the lung is recruited during a PRM or a SI maneuver. However, the last two maneuvers are followed by an immediate derecruitment, whereas the MRS is able to maintain alveolar recruitment for a significant period of time.
It should be noted that for each patient, apart from the variation in the ventilator pressure during the RM, no other parameters such as the ventilator settings of IE ratio, ventilation rate (VR), or F
IO
2 were changed. This explains the difference in outputs (Figure
9) between patients who each had most of their alveolar units recruited (Figure
8).
The results also display the presence of several interesting phenomena within the lung during the implementation of the various RM. For all RMs, a PEEP level of 10 cm H2O with identical driving pressure at the end of recruitment yielded an improved PaO2/FIO2 in comparison to that achieved with a PEEP level of 10 cm H2O before the RM was instigated. This implies that a higher level of oxygenation could be maintained with the same PEEP if an RM is utilized. In some cases (Patients A, B and E), a lower level of post-RM PEEP (5 cm H2O), was enough to maintain the PaO2/FIO2 at similar levels to that achieved with a pre-RM PEEP of 10 cm H2O.
During the PRM, large fluctuations were clearly noticeable in PaCO2 values in all patients. As expected, these coincided with the change in driving pressure that the PRM produced. A reduction in driving pressure increased the PaCO2 levels while an increase in driving pressure reduced PaCO2 levels. Furthermore, using MRS with a final PEEP of 10 cm H2O showed a post RM improvement in PaCO2 in all cases. It is highly likely that the above is due to the improved gas exchange following the recruitment of previously de-recruited units.
It is interesting that, in all cases, an increase in ventilator pressure caused a rise in PaCO2 initially. This is possibly due to the increasing pressure initially increasing pulmonary dead space, such that for the same minute ventilation, more ventilation was wasted, and consequently, less CO2 was eliminated.
The two patients with severe ARDS (Patients B and E) exhibited different responses to the same RM. For example, PaO
2/F
IO
2 dropped significantly post-RM in Patient B for MRS-10, whereas in Patient E PaO
2/F
IO
2 was maintained at a higher level at the end of the maneuver. This reflects the variations that can exist within the ARDS population and provides an example of different pathologies (for example varying distributions of TOPs in ARDS patient [
33]), presenting with similar symptoms (in this case similar initial PaO
2/F
IO
2 values). However, as seen in Figures
8 and
9, sufficient oxygen and recruitment could be attained in Patient E with lower inspiratory pressures than those required in Patient B. These results strongly motivate the development of patient-specific ventilation strategies, evaluated by individual PaO
2 changes, rather than focusing solely on general algorithms.
The implemented model has a number of limitations. The model does not individually consider attributes such as superimposed pressure, the vertical gravitational affect; or surface tension changes on the alveolar and airway walls. Their effects have instead been lumped into the governing equation for the pressure volume relationship of individual alveolar units via the parameter P
ext, which was determined individually for each alveolar unit within each individual patient during the model configuration stage (see Additional file
1). Effects of overdistension of alveoli have not been modeled and the model also assumes a fixed cardiac output. Therefore, attributes that may be associated with alveolar overdistension, namely, right ventricular impairment, reduced oxygen delivery and increased impact of venous shunting are presently not considered. This explains the smaller than expected rise in PaCO
2 values [
22] observed in our simulation during the recruitment phase of the MRS. Damage to the alveolar-capillary membrane, which can introduce and increase the bacterial and cytokine presence in the systemic circulation [
34], and cause further inflammatory responses, is also not currently included in the model. Although the model can determine pulmonary function outcomes (corresponding to respiratory mechanics), information about clinical outcomes associated with RM, such as mortality, cannot be acquired. However, the model does allow for observations to changes in hemodynamic parameters (see Additional file
1 for relevant model equations and Additional file
3 for some examples), and also includes hysteresis (through the inclusion of parameters affecting the alveolar compliance directly, and through time-varying and pressure-dependant changes in the airway resistances (see Additional file
1)), hypoxic pulmonary vasoconstriction, ventilation perfusion mismatch and cyclical collapse-reopening.
The results presented here show the potential of computational simulation to offer an alternative to large-scale clinical trials which have, to date, failed to answer many key questions, such as:
-
What level of PEEP should be applied to maintain recruitment in newly opened alveoli?
-
What inspiratory pressures need to be maintained to recruit alveoli?
-
What are the values and distributions of critical opening times?
A significant limitation with conducting clinical trials is that it is very difficult to compare the results of different studies due to the different cohort of patients included in the trials and other interstudy variations. It should also be noted that there is no evidence of whether the development of atelectasis itself has an adverse affect on the patient. Permissive atelectasis with lower PEEP may be a less deleterious option than risking lung injury using higher PEEP and/or higher tidal volumes in some patients [
35]. The optimal PEEP for many RMs is still to be determined and the most recent study in this area by Chiumello
et al. [
30] demonstrated the importance of considering RM timings with applying an RM [
30,
36].
Apart from the commonly administered RMs considered here, a number of other RMs have recently been proposed that should theoretically improve alveolar function but cannot be tested due to their experimental nature and the lack of patient data that would lend support for their introduction into practice [
37-
39]. The role of RMs also extends beyond ARDS patients; the administration of RMs has been shown to reduce markers of lung stress in patients following general anesthesia [
40-
43]. In both cases, computational simulation could play a key role in establishing the potential benefits of RMs and in the design of optimized patient- and disease-specific protocols.
An important question that needs to be addressed in future work in this area is the effect of RMs on the other organs [
44]. Mortality linked to ALI and ARDS often involves multiple organ failure, as the disease is not limited to the lungs. RMs also alter the working physiology of surrounding organs, and their impact on the heart and circulation cannot currently be accurately measured [
45,
46]. Changes in intrathoracic and transpulmonary pressures have a secondary effect of decreasing venous return and cardiac preload, so patients also suffer the additional stress of a reduction in cardiac output during the procedure [
47,
48]. It is difficult to compare the risks associated with different RMs, although a stepwise RM has been shown to have a smaller effect on cardiac output than the more widely used SI [
44,
49]. In this study, we focused on the effect of recruitment maneuvers on the pulmonary system. However, we are developing hemodynamic computer simulations of the complete cardiac system that also integrate the pulmonary and systemic circulation in order to better understand these issues. Such research tools could offer an invaluable alternative or complement to time-consuming and expensive clinical trials, and could provide answers to many key unanswered questions about the clinical efficacy of RMs in health and disease.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AD, DGB and JGH conceived, designed and coordinated the study. AD, WW, DGB and JGH developed the computational simulator and implemented the computer simulations. AD, OC, MC, TA, MH and JGH contributed to the acquisition, analysis, and interpretation of data and protocols. OC, MC, TA and JGH contributed to clinical evaluation of the results of the study. All authors contributed to drafting, revising and finalizing the manuscript. All authors read and approved the final manuscript.