Extracorporeal membrane oxygenation (ECMO) support has been known to be beneficial in cases of severe Adult Respiratory Distress Syndrome (ARDS). The increase in such cases results in an increase in scenarios of severe hypoxemia even during an ECMO run. The purpose of this study was to evaluate the benefit of the prone position (PP) in an analytical observational retrospective cohort study. The study included adult patients with ARDS, caused as a result of SARS-CoV-2, undergoing PP during ECMO support in the period from 2020 to 2021. Thirty-five patients were placed in PP with an average of 3.6 cycles per patient. The group of patients undergoing >3 PP cycles had a significant improvement in oxygenation during PP, 𝑃𝑎𝑂₂ (60.13 vs. 66.15, mmHg p = 0.0065) and 𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ (136 vs. 155, p = 0.0026). After adjusting for confounding variables (age, RESP score, and days from the start of ECMO and the first cycle of PP), the group with >3 cycles showed a hazard ratio of 0.2 (95% confidence interval, 0.051–0.78; p = 0.02). The study outcomes confirmed the benefits of PP as a strategy against severe hypoxemia in ECMO, and evaluated variables such as the number of cycles, which may be associated with improved survival in this subgroup of critically ill patients.

Keyword: Prone position, Extracorporeal membrane oxygenation, severe   hypoxemia, survival

Prone position (PP) is a recommended support strategy in patients with moderate to severe adult respiratory distress syndrome (ARDS). In spite of reducing the distensibility of the rib cage, owing to a reduction in the abdominal expansion and to the fact that the posterior thoracic wall of the rib cage is less compliant, PP generates a more homogeneous distribution of stress and strain on the lung parenchyma. This consequently decreases the hyperinflation risk of non-dependent lung regions while reducing atelectrauma of dependent lung regions. This increase in alveolar recruitment is achieved because the dorsal lung mass is greater than the ventral one, thereby acquiring a better distribution (𝑉⁄𝑄). This is attributable to the fact that perfusion is maintained mainly in the dorsal regions when the patient is in the PP.¹

Moderate to severe ARDS was defined as a ratio of 𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ of <150, with a 𝐹𝑖𝑂₂ of at least 0.6, a positive end-expiratory pressure (PEEP) of at least 5 cm𝐻₂𝑂, and a tidal volume (VT) close to 6 mL per kilo of ideal weight. In these patients, PP for mean duration of 17 hours increased the rate of successful extubation and, most importantly, reduced the 28-day mortality from 32.8% in the supine group to 16% in the prone group (p < 0.001), despite showing no variations in the time of mechanical ventilation (MV) and ICU stay.²

Venovenous (VV) ECMO (ExtraCorporeal Membrane Oxygenation) is a rescue support against severe ARDS refractory to conventional strategies, which provides complete blood oxygenation, eliminates 𝐶𝑂₂ retention, while allowing a lung-protective and resting MV that minimizes the risk of ventilator-induced lung injury. Early initiation of this support therapy leads to a 90-day mortality reduction and to a reduced therapeutic failure when compared with conventional ventilatory support.³

Owing to disease severity and the extensive pulmonary compromise, the patient may present severe hypoxemia in spite of the support provided by the VV ECMO, even under this extracorporeal support. In the face of this scenario, there were strategies aimed at improving oxygenation, such as increasing oxygen content (increasing ECMO blood flow or hemoglobin level), reducing recirculation (distance between cannulas or change of cannulation configuration), reducing oxygen consumption (sedation, neuromuscular relaxation, or therapeutic hypothermia), reducing cardiac output (beta-blockers), or reducing intrapulmonary shunt, for which PP has been a widely used maneuver.⁴

Taking into account a significant percentage of prone patients during ECMO support in our institution, it is necessary to objectively consider the variation observed with the PP and the number of prone cycles, the gasometric parameters, ventilatory mechanics variables, and the survival.

An analytical observational retrospective cohort study was designed. Using non- probabilistic convenience sampling, a total of 35 patients with SARS-CoV-2 and ARDS, aged ≥18 years, defined under the Kigali/Berlin criteria, placed in the PP during ECMO support and treated during the period 2020–2021, were included.

Data were obtained from the registry database of patients who required ECMO support because of SARS-CoV-2 at Fundación Clínica Shaio, a high-complexity institution in Bogotá DC, and a center of reference and excellence in Colombia, recognized by the s. With protocols for initiation of VV ECMO support against severe hypoxemia (𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ < 80) or severe hypercapnia (pH < 7.20 and 𝑃𝑎𝐶𝑂₂ > 80 mmHg) with fewer than 10 days of mechanical ventilation and refractory to conventional management including: sedation for RASS of −5, neuromuscular relaxation, protective MV and PP. The standard VV ECMO configuration is percutaneous insertion of a femoral venous drainage cannula and a jugular venous cannula for venous return with blood flow parameters of 60–80 mL/kg, delivered fraction of inspired 𝑂₂ of 100% and sweep gas in a 1:1 ratio to blood flow.

Most patients on ECMO receive pressure-controlled ventilation at our institution, with the aim of ensuring protective ventilation. The decision to use the PP was made according to medical judgment in cases of severe hypoxemia with clinical implications. Prior to the PP maneuver, target RASS of −5 and adequate neuromuscular blockade are guaranteed. The procedure was performed by a multidisciplinary team of at least 6 people. During PP, postural changes are guaranteed every 2 hours to avoid skin injury and peripheral nerve damage; prone cycles are extended from 16 hours of PP to 6 hours of supine positioning. The decision to complete the PP cycles was also a clinical decision.

COVID-19 virus was detected in a nasal swab sample collected in 1.3 mL of transport medium with a 0.3 mL minimum sample volume, according to the institutional protocol for COVID-19 tests, for subsequent molecular analysis by FilmArray BioFire COVID-19.

Variables of interest

The following demographic and medical variables were collected for all patients: age, body mass index (BMI), RESP severity score (Respiratory ECMO Survival Prediction), days of MV assessed at ECMO support initiation; sex; days between ECMO initiation and the first PP cycle.

Once on VV ECMO support, programming and variables in mechanical ventilation were collected: PEEP, control pressure (CP), respiratory rate (RR), tidal volume (VT), peak pressure (PIP), plateau pressure, driving pressure and lung compliance; blood gasometry (pH,𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂, 𝐻𝐶𝑂₃, and 𝐶𝑂₂) variables collected in each patient 1 hour before and 6 hours after each PP cycle. Additionally, the number of PP cycles during ECMO was obtained.

The outcome variable of interest was time to event for in-hospital death, which was estimated with Time 0 (𝑡_#), which corresponds to the date of the first PP, and time 1 (𝑡_$), as the date of decannulation, indicating either the success of the therapy or the death of the patient.

Data analysis

Quantitative variables were expressed with measures of central tendency (mean) and dispersion (standard deviation), after verifying the normality of their distribution using the Shapiro–Wilk test; alternatively, they were described using medians and interquartile ranges. Categorical variables were reported using absolute and relative frequencies. To estimate intergroup statistical differences (for example, between survivors or between cycles), a two-proportion Z-test was used. In polytomous qualitative variables, expected cell frequencies were calculated; should any of them be ≤ 5, Fisher's exact test was used, and when they were > 5, the chi-square test was used (𝑋²) instead. In order to correct the level of significance for multiple comparisons and to maintain the type I error at 0.05, the Bonferroni method was used, when appropriate.

The in-hospital survival function for the event of interest was measured using a non- parametric model (the Kaplan–Meier estimator) by groups according to the number of PP cycles (first group for patients who received ≤ 3 cycles and the second group for those patients who received > 3 cycles). Log-rank and Wilcoxon tests were used to find differences between these survival curves.

A semiparametric Cox Proportional Hazards method was used to adjust this survival function for potential confounding and interaction variables, and bivariate and multivariate estimates of the association between the number of PP cycles and survival were obtained. The regressor variables were selected using the stepwise technique with a bringing in probability of 10% and a removing probability of 15%. Graphical methods such as log-log plots (–In[-survival function] versus In t) were used to evaluate the proportional hazard assumption in search of parallel curves, and Schoenfeld residuals for the overall assessment and for each of the variables included within the model (looking for the absence of slope in the graphs).

Statistical tests were considered significant at a value of p ≤ 0.05 and, whenever pertinent, 95% confidence intervals (95% CI) were used. Data analysis was conducted using the statistical package STATA version 15.

A total of 35 patients were included in the study, 17 survived and 18 died. This corresponds to a cumulative incidence of mortality of 51.42%. Table 1 describes the sociodemographic and clinical characteristics of the patients who were pronated, overall and according to group. No significant differences were found in any of these variables between individuals who survived during the hospital stay versus those who did not. Notably, the time elapsed between the ECMO initiation, and the first PP cycle was 5.7 days for all patients, 5.76 for the survivors and 5.66 for the deceased; the average number of cycles was 3.68 per patient in general, 3.94 for the group of survivors and 3.44 for the group of deceased patients, however, no significant differences could be observed (Table 1).

Additionally, a comparison of the same characteristics was performed. In this case, patients were classified by PP cycles (≤ and > 3 cycles). Table 2 shows no statistical significance in this comparison. However, gasometric and ventilatory variables were compared before and after PP. A non-significant increase in 𝑃𝑎𝑂₂ (61.32 mmHg vs. 64.15 mmHg, p = 0.09) and a non- significant decrease in 𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ (141 vs. 139, p = 0.09) can be observed during PP. The variables in MV did not yield significant differences in terms of PP. There was a non-significant increase in compliance (9.44 vs. 10.75 mL/cm𝐻₂𝑂, p = 0.28) and tidal volume (191 vs. 203 mL, p = 0.14) (Table 3). The total estimated number of PP cycles performed on the 35 patients was 129 cycles.

Considering the average of three PP cycles per patient, these were classified into those who received ≤ 3 cycles (19 patients) and >3 cycles (16 patients). In the group with ≤ 3 cycles, no differences could be observed in gasometric or ventilatory variables before and during PP. In the group who received >3 cycles PP, a significant increase in the values is reflected during the PP of 𝑃𝑎𝑂₂ (60.13 vs. 66.15, p = 0.0065) and of 𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ (136 vs. 155, p = 0.0026). When comparing values before PP in this same group, the variables in MV showed a non-significant increase in the tidal volume during PP (192.90 vs. 212.75, p = 0.07), as well as in terms of lung compliance (10.29 vs. 13.03, p = 0.46) (Table 4).

In-hospital survival estimation via Kaplan–Meier showed differences, although these differences were statistically nonsignificant, for the group with >3 cycles PP (p = 0.072) (Figure 1 and Table 5). A crude estimate showed no association between the number of cycles and survival; however, when using a multivariate method to adjust for confounding variables, PP in more than 3 cycles reduces mortality risk by 80% compared with subjects who underwent ≤ 3 PP cycles (hazard ratio: 0.200, CI 95%: 0.0517–0.7800, p = 0.020) (Table 6). This model showed an adequate fit and compliance with the assumption of proportional risks (Figure 2).

Figure 1 Kaplan-Meier survival estimate in pronated patients during ECMO.

Figure 2 Evaluation of the proportional hazards assumption of the multivariate model in Table 6.

Evidence regarding PP in patients with ECMO support has increased over time. Several studies, mostly retrospective cohort studies, have reported improved oxygenation and respiratory compliance with a concomitant decrease in 𝐶𝑂₂, especially in patients with a high BMI and whose etiology was viral pneumonia. Severe hypoxemia (𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ < 85), a plateau pressure > 32 cm𝐻₂𝑂, and ECMO weaning failure after 7–10 days of support, were among the PP indications.⁵

Our retrospective cohort of 35 patients with severe hypoxemia on ECMO placed on PP represented a 28.92% of the total number of patients (121 patients on ECMO for SARS- CoV-2 in the period 2020–2021) with a slightly lower survival, possibly attributable to the greatest pulmonary compromise. Among those patients who were in PP, there were no major complications associated with this procedure, and a significant improvement in oxygenation and a tendency toward improvement in ventilatory mechanics could be observed in those patients undergoing more than three PP cycles.

The most representative multicenter study included six centers specialized in ECMO and 240 patients, of which, 107 were managed with PP for 15 hours on average, suffering minor and reversible complications in 6% of cases. An improvement in oxygenation could be observed and, despite a longer duration on ECMO (16 vs. 10 days), a lower hospital mortality was observed in the PP group on ECMO (30% vs. 53%, p = 0.0241).⁶

The use of ECMO support in patients with ARDS induced by SARS-CoV-2 has increased, and owing to the severe pulmonary compromise, refractory hypoxemia under this support has also increased. Within the clinical approach, PP in a retrospective study was considered in 56% of ECMO cases with severe hypoxemia (𝑃𝑎𝑂₂⁄𝐹𝑖𝑂₂ < 80 mmHg) despite 𝐹𝑖𝑂₂ and 𝐹𝐷𝑂₂ of 100%, and in cases of extensive pulmonary consolidation on chest images (>50% of the lung volume). Oxygenation improved without changes neither in 𝐶𝑂₂ nor in lung compliance, and with an increase in mortality in the PP group (78.6% vs. 27.3%) explained by the greater severity and extent of the pulmonary consolidation, as well as through the possible exposure to greater mechanical power.⁷

In the present study, PP for >3 cycles was a protective factor for survival. A possible explanation for the clinical and survival improvement with prolonged PP cycles may be the slow alveolar recruitment process that occurs with position change in patients with severely compromised lung compliance.

A recent meta-analysis consolidated data from 13 studies as follows: 12 observational studies, 4 of them with match control. PP was started after ECMO initiation with a total of 1836 patients. The reason specified for PP only in a few studies was severe hypoxemia. PP had a prolonged duration of >12 hours with an average 2–3 sessions. PP was associated with a significant improvement in survival at 28 days (74% vs. 58%; RR 1.31 with 95% CI of 1.21–1.41; p < 0.001). Hospital mortality in the ICU at 60 and 90 days also improved. Conversely, the duration of MV increased in prone patients by 11.38 days on average. There were no major complications reported in any of the studies.⁸

Among the strengths of our study, we not only assessed PP but also examined the impact of PP based on the number of cycles, regardless of the initial results in oxygenation and ventilatory mechanics. These outcomes proved a reduction in mortality in patients who received >3 PP cycles.

Our study has some limitations. The time variable from VV ECMO initiation and PP may have influenced the results, which in our study began approximately 5.7 days from VV ECMO initiation. This time may be late when compared with other studies that have shown benefits in the survival rates.

Finally, the sample size in our study was limited to prove an absolute reduction in the risk of mortality. Although a benefit in the survival rates of those patients with refractory hypoxemia on VV ECMO who received >3 PP cycles could be demonstrated, further studies with larger samples to improve the level of evidence would be necessary to confirm these results.

The benefits of VV ECMO support against ARDS lead to an increase in cases, and consequently, to a proportional increase in more severe cases of hypoxemia, even when patients are on VV ECMO. PP is a strategy that improves the parameters of mechanical ventilation, oxygenation, and survival. The duration and number of PP cycles is a relevant variable to consider when deciding to start PP in patients with severe hypoxemia on VV ECMO.

None.

None.

1. Gattinoni L, Taccone P, Carlesso E, et al. Prone position in acute respiratory distress syndrome. Rationale, indications, and limits. Am J Respir Crit Care Med. 2013;188:1286–1293.
2. Guérin C, Reignier J, Richard JC, et al. Prone positioning in severe acute respiratory distress syndrome. N Engl J Med. 2013;368:2159–2168.
3. Combes A, Peek GJ, Hajage D, et al. ECMO for severe ARDS: systematic review and individual patient data meta–analysis. Intensive Care Med. 2020;46(11):2048–2057.
4. Montisci A, Maj G, Zangrillo A, et al. Management of refractory hypoxemia during venovenous extracorporeal membrane oxygenation for ARDS. ASAIO J. 2015;61(3):227–236.
5. Roca O, Pacheco A, García–de–Acilu M. To prone or not to prone ARDS patients on ECMO. Crit Care. 2021;25(1):315.
6. Giani M, Martucci G, Madotto F, et al. Prone Positioning during Venovenous Extracorporeal Membrane Oxygenation in Acute Respiratory Distress Syndrome. A Multicenter Cohort Study and Propensity–matched Analysis. Ann Am Thorac Soc. 2021;18(3):495–501.
7. Garcia B, Cousin N, Bourel C, et al. Lille Intensive Care COVID–19 group. Prone positioning under VV–ECMO in SARS–CoV–2–induced acute respiratory distress syndrome. Crit Care. 2020;24(1):428.
8. Papazian L, Schmidt M, Hajage D, et al. Effect of prone positioning on survival in adult patients receiving venovenous extracorporeal membrane oxygenation for acute respiratory distress syndrome: a systematic review and meta–analysis. Intensive Care Med. 2022;48(3):270–280.