Overall, two strategies have been adopted: continuous positive airway pressure (CPAP) and BILEVEL, i.e., pressure support (PS) +: positive end expiratory pressure (PEEP) with the use of two interfaces: facemask and helmet (1, 2, 15, 28, 31, 32). Both interfaces improve oxygenation and reduce aerosolization compared with a high flow nasal cannula (HFNC) and COT.

Patients with COVID-19 may require prolonged NRS therapies (due to low oxygenation on admission) to avoid MV. In this case, there might be a need to apply high PS of up to 12 cmH₂O (15, 29, 32), PEEP values between 8 and 12 cmH₂O (2, 15, 29, 32), reaching up to 15 cmH₂O (33) and as little air leakage as possible. The helmet is one of the best interfaces to promote patient comfort (1, 2, 15, 34). It is also associated with reduced intubation rates and better correction of hypoxemia (35). However, the helmet is associated with greater rebreathing of CO₂, requiring intensive monitoring (36). Furthermore, a randomized clinical trial by Arabi et al. (34) divided 320 patients into two groups: half used a helmet and the other half received COT. No differences in mortality were observed after 28 days, but helmet mortality was observed at day 180. Compared with COT, no differences in mortality were observed (37).

The best comfort level can be achieved by adjusting ventilation parameters, such as PS and PEEP levels. The PS level is associated with the generation of V_T necessary to ensure adequate aeration; the PEEP level is responsible for ensuring oxygenation. A retrospective study reported the use of CPAP in 46 patients with PEEP ranging between 8 and 12 cmH₂O. The PEEP level was adjusted according to clinical tolerance, air leakage, and peripheral saturation of oxygen (SpO₂) (31). Only nine patients were intubated between days 7 and 14, and the authors recommended the use of CPAP to avoid intubation. Similar results were found in another study that compared HFNC with CPAP in 151 patients (98% with helmet) and NRS in 72 patients (15 with a helmet and 57 with a facemask) (1). One hundred sixty-three patients received HFNC. For the first two interfaces, the authors established a mean PEEP level of 10.2 cmH₂O during CPAP and 9.5 cmH₂O in NRS. Although all the interfaces were shown to improve oxygenation, there was no difference in intubation rates and length of stay (1). In an important randomized clinical trial, COT (low or high flow) was compared with CPAP adjusted to a mean PEEP of 8.2 cmH₂O. Only 36% of the patients in the CPAP group were intubated compared with 44% of the HFNC group. In addition, CPAP reduced mortality compared with COT (26). More recently, Colaianni-Alfonso et al. (35) studied 112 patients with moderate to severe COVID-19 who failed HFNC; the patients were divided into two groups for CPAP: one group used a facemask interface with median PEEP of 12 cmH₂O and the other a helmet interface with median PEEP of 14 cmH₂O. The groups remained on continuous CPAP for 24 h. It was observed that the helmet group had lower intubation rates and a more marked improvement in oxygenation compared with the facemask group, which had higher intubation rates and longer length of stay. Although these data favor the helmet, caution is required in the interpretation, because the PEEP value applied must be consistent with the clinical condition of the patient.

Adjustments other than PS and PEEP can be fine-tuned at the bedside. In 2009, a study (36) evaluated 13 patients after extubation and randomly performed three 20-min periods of NRS with three interfaces: facemask, helmet, and helmet with a 50% increase in PS and PEEP associated with a high rate of pressurization (rise time). Using the first two interfaces, PS had a mean level of 10 cmH₂O, PEEP of 5 cmH₂O and 0.2 s of pressurization time. In the third group, PS had a mean level of 15 cmH₂O, PEEP of 8 cmH₂O, and the shortest possible pressurization time, i.e., 0.05 s. The authors analyzed transdiaphragmatic pressure (Pdi), which is a surrogate of inspiratory effort through an esophageal catheter. Pdi was reduced in the helmet group with higher PS (15 cmH₂O) and PEEP (8 cmH₂O) and fast pressurization time (0.05 s). This highlights an important comparison between the facemask and helmet interfaces. Keeping the same cycling-off (25%), the helmet had more asynchrony events at the end of inspiration compared with the facemask, with ventilator cycling sooner or later compared with the end of the patient’s inspiratory time. In this case, the ventilator’s inspiratory time was shorter than the patient’s neural time. The authors point out that an overlap exists between the PS applied by the ventilator and the patient’s neural time. As an explanation, the authors hypothesize that cycling with the helmet seems to occur due to changes in flow caused by the mechanical characteristics of the interface and not by the characteristics of the patient. With the patient’s inspiratory time longer than that of the helmet, the interface promotes minimal reduction in ventilatory overload. Further studies on patients with COVID-19 comparing the two interfaces are necessary for a better understanding of their ventilatory repercussions.

The choice of the ideal interface, as well as fine ventilator adjustments, should aim to achieve patient-ventilator synchrony, reduce the work of breathing, especially with the helmet interface and to ensure comfort. A multicenter randomized clinical trial randomized 54 patients (mean age, 66 years) to NRS and 55 to HFNC (15). The NRS group underwent therapy for at least 48 h and used a helmet interface with PS and PEEP levels ranging between 10 and 12 cmH₂O, no pressurization time (rise time), expiratory trigger between 10 and 50% to avoid double trigger, inspiratory trigger to avoid auto-triggering, and maximum inspiration time between 1 and 1.2 s. PS was titrated individually to ensure high flows to the patient. The NRS group showed lower intubation rates and more MV-free days compared with the HFNC group. This allows us to conclude that the success of NRS is based on fine ventilatory adjustments.

Typically, NIV is a therapy performed in the ICU. However, with the pandemic exceeding the capacity of available beds, this tool gained space outside the ICU. Cammarota et al. (37) conducted a systematic review with meta-analysis including 17 articles containing 3,377 patients which showed the effectiveness of NIV outside the ICU environment as an adequate tool to deal with the demand for ventilatory assistance.

The literature is not concordant regarding the therapeutic objectives of NRS. Perkins et al. (28) indicated different factors as therapeutic targets, such as: SpO₂ > 90%, respiratory rate ≤ 25 bpm, and a reduction in the work of breathing (26). Aliberti et al. (2) stated that NRS weaning can be performed if the patient’s SpO₂ > 94% with FiO₂ < 50% and PEEP ≤5 cmH₂O. Arabi et al. (34) stated that application of PEEP should target SpO₂ between 92 and 98%, and that the respiratory rate should be <25 bpm. More recently, Colaianni-Alfonso et al. (35) state that application of PEEP should target an SpO₂ between 92 and 96%.

After starting NRS, patient monitoring must be constant to assess its effectiveness or failure. Arabi et al. (34) suggest assessment every 1 to 3 h, but this may vary according to the intensive care unit (ICU). In case of therapy refractoriness, the literature indicates that MV should not be postponed. Some signs of failure mentioned in the literature are: respiratory rate > 40 bpm, respiratory acidosis with pH <7.25–7.30, use of accessory muscles, dyspnea, swallowing disturbance, SpO₂ < 88–90% for more than 5 min, PaO₂/FiO₂ ratio < 100, persistent requirement for FiO₂ > 70%, hemodynamic instability (systolic blood pressure < 90 mmHg or mean blood pressure < 65 mmHg, even with volume resuscitation), deterioration in the level of consciousness (2, 15, 34, 35). Contrary to the data, the intubation criteria in the randomized clinical study by Perkins et al. (28) are stricter. These authors compared CPAP with low and high flow oxygen therapy and considered an SpO₂ ≤ 94% with an FiO₂ of at least 40% to be a ventilatory risk. Robba et al. (27) suggest immediate intubation if the PaO₂/FiO₂ ratio does not improve and/or PaCO₂ < 30 mmHg and/or respiratory rate > 28 bpm using accessory muscles for more than 3 h. More recently, the study by Colaianni-Alfonso et al. (35), who compared helmet CPAP and helmet facemask (discussed earlier), considered pH <7.35 as a criterion for intubation, in addition to all previous signs. These data indicate that there is no clear guideline for the signs of NRS failure, allowing the use of some scales to help diagnose it (Figure 1).

The literature presents some scales/indices that help in the diagnosis of therapeutic failure during NRS:

At the beginning of the pandemic, the initial recommendation was early intubation to protect the lungs (3, 49). Due to the urgency caused by the pandemic, limited staff, and few noninvasive ventilatory resources to meet the demand, many patients experienced worsening respiration and consequent intubation.

With the reduction in the number of cases, noninvasive ventilatory support such as NRS and HFNC was introduced with the aim of reducing ventilatory effort and avoiding intubation (29). However, when instituting noninvasive therapy, adequate monitoring is required to avoid P-SILI (8, 27, 50). In patients with COVID-19, reduced lung compliance and heterogeneous distribution of inspired V_T are observed due to the presence of areas of low V/Q ratio that are distributed irregularly throughout the lung (4, 8, 24, 25, 51). In spontaneous ventilation, during the inspiratory phase, sufficient diaphragmatic contraction is required to counteract pulmonary elastic recoil forces (8, 24, 25, 27, 50). This generates large variations in transpulmonary pressure (P_L).

In the early stages of COVID-19 (L phenotype), when there are no large pulmonary consolidations, no variations in transpulmonary pressure are observed, which allows the application of NRS with greater safety (6, 25, 29). Under normal conditions of spontaneous breathing, during the inspiratory phase, pleural pressure decreases uniformly, whereas PL increases uniformly (50). In situations of increased ventilatory drive, greater inspiratory efforts are observed to generate a given V_T caused by greater negative pleural pressure, increasing P_L, which reflects inspiratory efforts. This allows non-homogeneous distribution of lung pressures and volumes, leading to P-SILI (8, 25, 50). Expiratory efforts can also cause P-SILI (50). During intense expiratory activity, pleural pressure increases, drastically reducing P_L, with alveolar collapse occurring in most dependent lung regions and peripheral airways. Battaglini et al. (50) suggest a study of stress/strain for a better understanding of the disease. Stress is the distribution of force applied per unit of lung area, and strain evaluates the stretching of this alveolar unit and is directly proportional to stress (25, 50). During the development of COVID-19, ventilatory efforts become more vigorous, leading to regional hyperdistention, especially in non-dependent regions, and further compromising dependent regions (8, 24, 25, 50). Therefore, both stress and strain are increased. This is associated with the development of pneumothorax and pneumomediastinum (50).

PP is a non-pharmacologic strategy widely adopted in moderate/severe cases of ARDS with inadequate gas exchange (i.e., PaO₂/FiO₂ ratio < 150, with FiO₂ > 60%) even with PEEP optimized within the concept of LPV. In ARDS, PP redistributes air volume from ventral to dorsal areas, promoting lung homogeneity (19, 20) because lung ventilation is dependent on gravity (20). PP also reduces regional lung stress/tension by displacing non-ventilated areas ventrally (20, 62, 63). Recruitment of the dorsal region of the lung is observed with subsequent increase in regional oxygenation and de-recruitment of the ventral region, leading to a decrease of the hyperinflated tissue (63, 64). In this case, a reduction of the dorsal shunt is observed, improving oxygenation (19, 20). Grasselli et al. (62) state that oxygenation can improve between 60 and 80%.

Final PaO₂ is a weighted average of the PaO₂ of blood flowing from different lung units. This means that the number of atelectatic units in the dependent lung regions is proportional to the severity of hypoxemia (63). In a supine position, with an angle of 0°, approximately 60% of the total lung mass is dependent. In COVID-19, perfusion irregularity promotes greater perfusion in these regions, leading to a decrease in the V/Q ratio (20, 62). During PP, however, only 40% are in the dependent position; i.e., fewer lung units are hyperperfused, resulting in better oxygenation (63). The consequence, in terms of ventilatory mechanics to the PP, is a decrease in total compliance of the chest wall, due to the functional stiffening of the anterior chest wall (63, 64). Thus, an improvement in lung compliance values and a more homogeneous V/Q distribution are expected (64). This also reduces VILI, resulting in improved parameters of ventilatory mechanics (49).

The correct indication for PP is directly correlated with the duration of the disease and the patient’s clinical status. The L phenotype is characterized by moderate to severe hypoxemia, even with normal lung compliance (6). This phenotype is considered unresponsive to PP, and the observed improvement in oxygenation is due to the redistribution of blood flow from dorsal to ventral areas, without any alveolar recruitment, as seen in ARDS (42). This, PP in this phenotype does not bring great benefits, because this phenotype has no or little recruitment capacity. However, better aeration of dorsal regions is noted, reducing the chances of atelectrauma (64). Furthermore, COVID-19 is progressive, evolving to the H phenotype, which is more recruitable (6, 25). In this phenotype, there may be a worsening of lung compliance, without any relationship with the conduct. It is at this point that PP becomes more indicated. There is also an improvement in oxygenation, but at the expense of directing blood flow to dorsal regions with alveolar recruitment between patients (62). The ventilatory difficulty of the F phenotype contraindicates PP, because the benefits will be few. This is due to organizing pulmonary fibrosis (25, 27, 63). At this time, protective ventilation is prioritized (25).

The study by Fossali et al. (64) provides information relevant to the topic. The authors studied 21 patients with a mean age of 67 years. They performed chest computed tomography in a supine position and PP. Afterward, within the ICU, the authors performed electric impedance tomography (EIT) to verify distribution and ventilation and perfusion. All were protectively ventilated, without adjustments, in pressure regulated volume-controlled mode with PEEP maintained at 10 cmH₂O. The authors described that there was no difference in the compliance of the respiratory system in both decubitus positions. The authors hypothesizes that in supine position, there may be alveolar units subject to cyclic openings and closings, which would be reduced in PP. In addition, another possible reason is that there was a decrease in lung elastance associated with increased chest wall rigidity. In addition, there was recruitment of dorsal regions, with perfusion improvement in these regions and de-recruitment of ventral regions. This allowed reduction in barotrauma and atelectrauma, reduction of areas with dead space, reducing the number of alveolar units with low V/Q, which improved V/Q matching. This dorsal de-recruitment is called spongelung (65) and is characterized by a reduction in dorsal pulmonary tension and ventral hyperdistention. The authors also point out that there was a reduction in the dead space/shunt ratio in PP and that this is also a marker of lung protection. However, the patients included in this study had been ill for an average of 8 days. Considering that COVID-19 is a progressive disease, it can be inferred that the patients were in phenotype transition to H and F, when the PP has few benefits.

The retrospective study by Langer et al. (66) divided 1,057 patients ventilating protectively into two groups (PP and supine position) with a mean age of 63 years. The average time to perform the first PP was 2 days. The authors observed that there was no difference in oxygenation and ventilatory mechanics between the groups. This can be explained by the high compliance at baseline. Therefore, the effect of PP may not work solely by recruitability, but through the redistribution of pulmonary blood flow.

Weiss et al. (19) studied 42 patients with a mean age of 59 years, but with significant obesity (body mass index (BMI) > 34 kg/m²), also under LPV. The researchers performed three PP sessions. In contrast to the article mentioned earlier, there was improvement in oxygenation after the first PP session, but a similar effect was not observed during the second and third PP sessions. This can be attributed to disease progression.

Recently, the COVID-19 Veneto ICU Network research group developed the PROVENT-C19 Registry, a large multicenter protocol specifically for patients with COVID-19 with the aim of describing the population that most benefits from PP (67). On admission, anthropometric data, data on comorbidities, and the type of ventilatory support used before EIT will be collected. The outcomes to be analyzed include differences in gas exchange and the PaO₂/FiO₂ ratio and ventilatory parameters before and after PP, prone duration, and ICU and hospital mortality. Considering the expected large population of this study, there will be an important improvement in clinical practice.

The recommended duration of PP is at least 16 h (61, 68, 69). However, some studies have reported durations longer than 16 h of PP with different outcomes. The prospective study by Engerström et al. (70) evaluated 1,714 patients with a mean age of 64 years. The mean time between intubation and first PP session was 20.4 h. No association between early PP and survival was observed. Protti et al. (20) studied 15 patients with a mean age of 69 years and a mean BMI of 29 kg/m². Patients were intubated within 2 days and were placed in PP within 3 days. There was a reduction in the volume of non-aerated gas and hyperventilated areas, indicating a lower possibility of VILI and an increase in respiratory compliance. An important point in this study is that the patients did not experience delayed intubation. This certainly has effects on the outcomes.

Encouraging results were also found in the study by Page et al. (60). The authors studied 52 obese patients (BMI >32 kg/m²) with a mean age of 62 years. They were randomized between conventional prone (16 h) and extended prone (24 h). There was no change in respiratory mechanics, but patients who remained prone longer had more ventilator-free days.

A longer time in the prone position was reported by Rezoagli et al. (71). The standard PP group lasted for 16 h, while extended PP consisted of 40 h. Although the extended PP group was younger than standard PP group, extended PP was feasible and was able to reduce the workload of health professionals. Taking into account the oppressive condition during pandemic, the reduction in workload is an important issue to consider. Furthermore, no benefits or harm in terms of gas exchange or respiratory mechanics were found when extended PP was compared to the standard PP group.

On returning to supine position, some patients may experience a decrease and loss of oxygenation gain (59, 72), further favoring extended PP. Recently, the retrospective study by Okin et al. (72) compared 267 patients with a mean age of 62 years who were subjected to 16 h and 24 h of PP in terms of mortality; 157 patients underwent extended PP (>24 h) and 110 underwent conventional PP (up to 16 h). The authors observed that mortality at 30 and 90 days was lower in the extended PP group. In addition, the study highlights that extended PP is safe, because it reduces the number of supine sessions that are associated with alveolar de-recruitment, increased atelectasis, and VILI, contributing to mortality. It also reduces the amount of neuromuscular blockers, reducing diaphragmatic dysfunction (72).

Thus, there is no limit on the number of PP sessions as long as they are recommended. For example, Walter et al. (73) reported that some patients underwent PP 22 times. The same study also suggests that PP should be interrupted when the FiO₂ requirement is ≤60%, when the PaO₂/FiO₂ ratio is >150, and when the PEEP is ≤12 cmH₂O.

Delaying PP is associated with higher mortality. The study by Mathews et al. (74) included 2,338 patients; 702 were placed in PP within 2 days of MV and the other 1,636 within 2 days of MV with a P/F ratio < 200. The authors observed that the early PP group had greater chance of developing shock and use of corticosteroids. However, the risk of death was lower. COVID-19 is a heterogeneous disease, therefore it is not possible to define a suitable time to implement PP. One suggestion is to use the same reasoning used to determine the need to transition from NRS to intubation: the worsening of compliance and the need for high FiO₂ fractions to maintain adequate SpO₂. In this case, it is possible to infer a change from the L to the H phenotype, which has a greater possibility of recruitment, benefiting from PP (25, 63).

With regard to the objectives of prone decubitus, associated articles describe the physiologic and ventilatory changes of the position. However, when analyzing the effects of PP, the increase in survival must be considered. Directing the therapeutic target only to improve oxygenation can be a scientific limitation. To guide the understanding of this topic, Table 2 contains a summary of the included studies and outcomes found.

The use of PEEP tables, widely used for ARDS, is an easy alternative to titrate PEEP and sustain ARM (75–78). However, this strategy fails to optimize oxygenation; the PEEP response in patients with COVID-19 is highly heterogeneous due to the facts mentioned earlier (79–82). In this context, some studies chose the recruitment to inflation (R/I) ratio developed by Chen et al. (78) to assess the potential for recruitability in patients with ARDS. It ranges from 0 to 2. R/I < 0.5 indicates low potential for recruitability, increasing the risk of pulmonary overdistension without any benefit. R/I > 0.5 indicates high recruitability (79, 82, 83). After assessing recruitability, the choice of PEEP is based on ARM with decremental PEEP titration. Some studies have used only decremental PEEP titration (8, 78). Briefly, this strategy consists of gradually increasing the airway opening pressure up to 45 cmH₂O and then performing PEEP titration (in steps of 2–3 cmH₂O), maintaining the stability of hemodynamic and airway ΔP and allowing P_L to increase (80–83).

The issue of heterogeneity of oxygenation targets is a topic of discussion. Zerbib et al. (80) states that an SpO₂ between 88 and 92% is satisfactory. Ball et al. (8) suggested that the best PEEP is the one in which PaO₂ remains >60 mmHg. These two studies were less rigid about oxygenation, in contrast to previous studies dealing with ARM and PEEP titration. Randomized studies are needed to confirm whether these oxygenation targets are suitable for COVID-19.

For lungs with low recruitability (L-type phenotype; i.e., high static compliance), low levels of PEEP are sufficient to optimize PaO₂ and reduce hyperdistended areas, P_plat and airway ΔP.

When high PEEP is applied to the L-type phenotype, it is expected to increase lung volume and reduce lung heterogeneity, at the cost of increased overinflated areas compared with low PEEP (8). Usually, airway pressure increases followed by impairment in respiratory system compliance. In the L phenotype, high PEEP values are not recommended, because this is a poorly recruitable phenotype (6, 8, 25). Increasing PEEP in this phenotype contributes to worsening lung compliance.

Pan et al. (76) studied 12 patients who were protectively ventilated; mean age was 59 years and the mean R/I ratio was 0.21, indicating low pulmonary recruitability. They showed that after applying high PEEP using the PEEP table (>15 cmH₂O), P_plat remained high, with a low response in oxygenation. In addition, the authors reported that this patient profile may not respond to high PEEP in the supine position, but that recruitability seems to increase after PP. It can be inferred that this gain is due to displacement of poorly ventilated areas. This reinforces the fact that the PEEP table has partial applicability and seems to suggest that ARM should be performed together with PP.

When PEEP is applied incases with the H phenotype, the response is an improvement in lung compliance, with a reduction in P_plat and in poorly ventilated or non-ventilated areas, reducing intrapulmonary shunt (8, 25).

Protti et al. (20) studied 40 patients with early COVID-19 in the supine position and performed ARM plus decremental PEEP at three levels: 15, 10, and 5 cmH₂O. With PEEP of 15 cmH₂O, oxygenation improved in 36% of patients, but respiratory compliance improved in only 11%. There was also a reduction in non-ventilated areas and an increase in hyperventilated areas. Furthermore, two different responses were observed as PEEP increased. With an increase in PEEP from 5 to 10 cmH₂O, recruitment was predominantly dorsal, reducing non-aerated tissue, with an improvement in the PaO₂/FiO₂ ratio and an increase in respiratory compliance. However, when PEEP was increased from 10 to 15 cmH₂O, the recruitment obtained previously overlapped with the appearance of hyperventilated areas, predominantly ventral, and a decline in respiratory compliance. Furthermore, the improvement in oxygenation at high PEEP cannot be explained by recruitability, but rather by the improvement in left ventricular function, which decreases cardiac output (50).

–Ball et al. (8) studied a group of 42 recruitable and non-recruitable patients with a mean age of 63 years using LPV. The authors evaluated lung mechanics and oxygenation at two PEEP levels (8 and 16 cmH₂O). The first group benefited from high PEEP by reducing the percentage of non-aerated lung units. However, only the non-recruitable group had a reduction in poorly aerated areas. Both groups showed improved oxygenation via increased hyperaerated areas, with consequent worsening of respiratory compliance. In practical terms, this led to an increase in ΔP, P_plat, mechanical power, variables associated with VILI. The authors explained that the improvement in the P/F ratio should be interpreted as redistribution of the V′/Q′ ratio, prioritizing areas with low ventilation, and not as recruitment, even in so-called recruitable patients.

Some of the studies discussed in this review analyzed obese patients, represented by BMI >30 kg/m² (63, 64). Obese patients have a high recruitment potential and can tolerate high PEEP values, as long as the P_plat remains up to 30 cmH₂O. Usually, studies have pointed out two main reasons for the need for high PEEP in this population: (1) decreased P_L (79); and (2) predominantly ventral ventilation with a tendency to dorsal alveolar collapse under low PEEP. This scenario can be prone to VILI due to low static compliance. After application of PEEP, the studies have highlighted decreased airway ΔP and dead space, with improvement in static lung compliance, P_L, the PaO₂/FiO₂ ratio, and redistribution of pulmonary blood flow with subsequent reduction of intrapulmonary shunt (81).

Highly specialized centers have introduced EIT to expand understanding of the effect of PEEP levels (75, 81). EIT consists of placing a belt with electrodes between the fourth and fifth ribs to verify the ventilatory distribution (whether predominantly dorsal or ventral) in real time and macroscopically assess the effect of PEEP. The use of EIT during ARM and PEEP titration may guarantee the most adequate value for the patient, which may be two values below or above the values suggested by the PEEP table (75, 81). EIT shows the percentage of well-ventilated, poorly ventilated, collapsed, and hyperinflated areas; the latter two are of interest to the professional at the bedside to avoid VILI (49, 59).

Oxygenation is a therapeutic target, as is the assessment of ventilatory mechanics. Both need to be evaluated together and systematically. This review recommends that the search for the ideal P/F ratio, as well as optimal SpO₂/PaO₂ values, can lead to dangerous maneuvers of alveolar recruitment, exceeding protection limits, with the risk of P-SILI.

Beloncle et al. (10) studied 25 patients with COVID-19, 16 of whom were considered highly recruitable and 9 were considered poorly recruitable. Two PEEPs were applied: 5 and 15 cmH₂O. At high PEEP, the recruitable group showed the same mean compliance for both PEEP levels. However, oxygenation in the recruitable group was higher than in the non-recruitable group. Ball et al. (11) studied 42 patients, 32 non-recruitable and 10 recruitable. The researchers applied two levels of PEEP (8 and 16 cmH₂O). All patients then underwent computed tomography. They found that there was no percentage difference in recruitable areas despite the increase in PaO₂. Therefore, it can be concluded that the compliance of the respiratory system can mitigate oxygenation. The articles of this topic were organized in Table 3, to direct the understanding of ARM.

Considering the inclusion criteria for ECMO, it is pertinent to consider its rapid start after detection of the disorder. There is no consensus in the literature about when to start therapy, because this depends on the availability of equipment and trained staff (85). Even so, the prospective study Mustafa et al. (88) studied 160 patients with a mean age of 49 years. The researchers divided them into two groups: (1) ECMO + LPV; (2) Only MVA. The first group progressed to ECMO within 3.8 days. ECMO + LPV was associated with 68% survival, whereas LPV only was associated with 26% survival. Karagiannidis et al. (89) stated that ECMO should start within 3 days because it is associated with longer patient survival. The multicenter study by Lorusso et al. (90) analyzed 1,215 patients ventilated with a V_T < 3 mL/PBW and concluded that age > 60 years and a time longer than 4 days between the start of MV and the start of ECMO was associated with higher mortality.

Recently, Hajage et al. (91) studied 2,858 patients; 269 (mean age, 53 years) received ECMO within 14 days of hospitalization. Patients were intubated within 1 day of hospitalization, the average time to start ECMO was 6 days, and 89 and 97% of patients received PP and neuromuscular blockers, respectively, before ECMO. All patients were ventilated ultraprotectively, i.e., V_T < 4 mL/kg. It was observed that eligible patients had poor ventilatory mechanics, with a mean ΔP of 18 and a mean P_plat of 30 cmH₂O. The results showed that there was a significant improvement in ΔP to 12 cmH₂O and P_plat to 18 cmH₂O within 48 h of ECMO. These results are encouraging and reinforce the recommendations for successful ECMO: young age, few days of MV, and few comorbidities.

ECMO is a high-cost strategy that requires highly trained staff (85, 87), which may limit its widespread application and/or late start, when the patient may not benefit from the therapy.

Preliminary prospective results from Kon et al. (92) highlighted important issues. The authors studied 27 obese patients with a mean age of 40 years. The authors chose to include only functional independent patients without comorbidities in the study. Before being eligible for ECMO, patients were LPV with a mean PEEP of 14 cmH₂O and FiO₂ > 90%. The primary endpoint of the study was survival during hospitalization and lung recovery (defined by the authors as ECMO weaning). They reported that 11 patients fully recovered on ECMO, and 13 were still on ECMO. The recovered group was successfully decannulated. All patients were tracheostomized with a median time of 24 h, allowing for lower rates of sedation and neuromuscular blockers, in addition to reducing the possibility of nosocomial infections, in contrast to other studies that reported at least 2 days from intubation to ECMO. All patients were ventilated in volume-controlled ventilation mode with 5 mL/kg PBW, with a mean PEEP of 10 cmH₂O. The patients had median low compliance (22 mL/cmH₂O) and ΔP ranging from 14 to 18 cmH₂O. In addition, patients had a mean PaCO₂ of 80 mmHg, pH <7.25, and mean serum lactate levels of 2.45 mmol/L. The primary endpoint of the study was 90-day mortality. All these factors were associated with mortality, which was 38.8%. Therefore, the main success factor for ECMO is young age, indicating the need for correct selection of patients (89).

The study by Schmidt et al. (21) included 83 patients, 30 of whom died. Forty-eight patients survived and were discharged from the ICU. The average age was 48 years. Interestingly, the surviving group had higher mean d-dimer values than the group who died. Moreover, 88% of the patients were ventilated in airway pressure release ventilation (APRV) mode, known to ensure alveolar stability and allow for greater pressurization with reduced occurrence of VILI (93). This ventilation mode was not used in almost all of the articles cited that opted for volume-controlled ventilation or pressure-controlled ventilation. Mortality was 31%. There was no information about the association of APRV and the effects of ECMO. But given the purpose of the APRV, it is possible to infer that the association would behave as double lung protection. Other studies associating ECMO and APRV are needed to confirm the positive relationship between them.

Although there are few studies reporting the use of PP in ECMO, recent evidence points to a good response from the combined therapies. Garcia et al. (94) studied 25 patients with COVID-19 that required V/V ECMO. 14 were placed on PP at least once for 16 h on average. All of them were protectively ventilated. In terms of lung mechanics, there were no statistical differences between PP and non-PP patients. However, there was an improvement in oxygenation in the PP group. Massart et al. (95) evaluated 517 patients with a mean age of 55 years on ECMO; 364 were prone during therapy and 153 were not prone. All were protectively ventilated. Lower mortality rates were observed in the PP group. There was no statistical difference between lung compliance and gas exchange values. As with PP, the outcome that should guide clinical practice is mortality. Only randomized studies will be able to confirm if the improvement in oxygenation is due to ECMO or PP or to joint therapy.

Despite its beneficial effects, the articles cited here highlight that ECMO presents a high risk of bleeding requiring anticoagulation, and many patients progress to hemodialysis (53, 85–88, 94–98) These facts, added to the fibrotic evolution of COVID, increase the chances of mortality and therapeutic failure with ECMO. However, these factors may have less impact on young patients and/or those with few or no comorbidities. The positive and negative outcomes of the ECMO studies are shown in Table 4.

Some studies compared ECMO in non-COVID-ARDS patients and COVID-ARDS patients. Although similar results were gathered about oxygenation (99, 100), the treatment time and complications were different. Chandel et al. (99) analyzed 9,271 patients who required ECMO between 2017 and 2021. Authors showed that COVID patients remained longer on ECMO when compared to non-COVID patients (19.6 days versus 10 days). Additionally, COVID patients had higher rates of developing kidney failure, requiring hemodialysis. Furthermore, COVID patients remained longer on mechanical ventilation before starting ECMO. This condition may lead to increased diaphragmatic dysfunction and mortality in COVID compared with non-COVID group. Other complications also observed in the COVID group included pneumothorax and intracranial hypertension. This can be explained by the high inflammatory cascade due to COVID.

Similar results were found in the retrospective study by Dave et al. (100). The authors studied 89 patients who used V/V ECMO, divided in two groups: 35 COVID patients and 54 non-COVID patients. COVID patients had higher in-hospital mortality rates (49% versus 24%), longer ECMO and mechanical ventilation time before ECMO (654 h versus 394 h; 3 versus 1 day, respectively) than non-COVID patients.