Kunstmatige beademing in buikligging wordt al vele jaren toegepast bij patiënten met ARDS. De fysiologische mechanismen waardoor patiënten baat kunnen hebben bij buikligging zijn een verbeterde ventilatie-perfusie match, een vergroting van het eind-expiratoire longvolume met daarbij een meer homogene distributie van teugvolumina door long recruitment en veranderingen in thoraxwand mechanica. De LUNG SAFE studie (Bellani, 2016) en de APRONET-studie (Guérin, 2018) hebben laten zien dat in het pre-COVID tijdperk buikligging slechts in beperkte mate werd toegepast bij ernstige ARDS. De APRONET-studie laat tevens zien dat de belangrijkste reden hiervoor is dat in de praktijk buikligging vaak nog als rescue-therapie wordt gezien en de ernst van de hypoxemie vaak niet ernstig genoeg wordt geacht voor het toepassen hiervan. Dit suggereert dat de identificatie van patiënten met ernstig ARDS door middel van de PaO₂/FiO₂ ratio weinig wordt gebruikt voor het starten van buikligging of dat er onvoldoende awareness is dat buikligging het risico op (of het verergeren van) ventilator-induced lung injury (VILI) vermindert (Guerin, 2018). Het beoordelen van het nut van buikligging met het oog op herhaling van toepassing van buikligging wordt verschillend gedaan. In de PROSEVA trial werd vooraf aan een nieuwe sessie de indicatie beoordeeld op basis van de P/F-ratio in rugligging conform de situatie vooraf aan de eerste poging (Guerin 2013). De APRONET studie laat zien dat ook andere argumenten gebruikt worden zoals verbetering van oxygenatie in buikligging ten opzichte van rugligging (Guerin 2018). Ook de duur van buikligging is variabel tussen verschillende centra.

Deze module geeft een samenvatting van het bewijs van de mogelijke klinische voordelen van beademing in buikligging, en wanneer en hoe lang buikligging wordt aanbevolen. Deze module focust zich op matig-ernstige ARDS-patiënten die gecontroleerd beademd worden.

Summary of literature

Description of studies

Bloomfield (2015) conducted a Cochrane systematic literature review and meta-analysis regarding the effect of mechanical ventilation in the prone positioning compared with mechanical ventilation in the supine position for adult patients with acute respiratory failure, but most included studies focused on adult patients with ARDS. The search of Bloomfield (2015) was updated in 2014 using CENTRAL, Ovid MEDLINE, EMBASE, CINAHL and LILACS in Ovid MEDLINE for eligible randomized controlled trials. In total 9 RCTs and a study reporting follow-up subgroup analyses of one of the RCTs were included. For the current review we excluded one RCT (Leal, 1997) as the intervention period was less than two days and the study was published prior to 2000 (being two of our exclusion criteria) and we excluded the subgroup analysis (Chiumello, 2012). Table 1 shows an overview of the RCTs that were included in the current review.

Note that Bloomfield (2015) mentioned three studies reporting (additional) results of RCTs who were waiting for classification. Two of these studies could not be included in the current literature review (see the table with reasons for exclusion under the tab Methods), but data of Girard (2014), reporting a secondary outcome measure (pressure sores) of Guerins (2013) RCT, will be included in the current literature review.

Munshi (2017) conducted a systematic literature review with the same papers as included in Bloomfield (2015). In the current review, Munshi (2017) was only used for the outcome measure unplanned catheter removal as this measure was not reported by Bloomfield (2015).

Table 1. Study characteristics of included studies

^‡median; ARDS = acute respiratory distress syndrome; ARF = acute respiratory failure; CAP = community-acquired pneumonia; NR=Not Reported

*ARDS severity according to Berlin definition: PaO₂/FiO₂ 200–300 mmHg as mild, PaO₂/FiO₂ 100–200 mmHg as moderate, and PaO₂/FiO₂ <100 mmHg as severe ARDS.

^# Lung-protective ventilation <8 ml/kg predicted body weight.

^¶ All “yes” results were due to slow enrollment.

Results

Mortality, short term (10-30 days or ICU-mortality)

Seven of the included RCTs (Chan, 2007; Fernandez, 2008; Gattinoni, 2001; Guerin, 2004; Guerin, 2013; Mancebo, 2006; Taccone, 2009) in Bloomfield (2015) that were selected for the current review reported short term mortality. Because of heterogeneity within studies, results in the current review are reported for subgroups (duration of proning, disease severity and hours after meeting entry criteria or starting ventilation, see figure 1).

Mild-severe ARDS, <16 hours proning, starting after 48 hours:

211 out of 565 (37.3%) proned patients died compared to 192 out of 530 (36.2%) patients who were treated in supine position (RR 1.04; 95%CI 0.89 to 1.21). This difference was not clinically relevant.

Moderate-severe ARDS, >16 hours proning, starting within 48 hours:

143 out of 502 (28.5%) proned patients died compared to 193 out of 482 (40.0%) patients who were treated in supine position (RR 0.71; 95%CI 0.52 to 0.95). This difference was clinically relevant in favor of proned patients.

Moderate-severe ARDS, >16 hours proning, starting after 48 hours:

4 out of 11 (36.4%) proned patients died compared to 4 out of 11 (36.4%) patients who were treated in supine position (RR 1.00; 95%CI 0.33 to 3.02). This difference was not clinically relevant.

Figure 1. Forest plot: Short-term mortality#

# Short-term mortality: 10-30-day mortality or IC mortality.

Mortality, long term (>30 days)

Eight of the included RCTs (Chan, 2007; Fernandez, 2008; Gattinoni, 2001; Guerin, 2004; Guerin, 2013; Mancebo, 2006; Taccone, 2009; Voggenreiter, 2005) in Bloomfield (2015) that were selected for the current review reported long term mortality. Because of heterogeneity within studies, results in the current review are reported for subgroups (duration of proning, disease severity and hours after meeting entry criteria or starting ventilation, see figure 2).

Mild-severe ARDS, <16 hours proning, starting within 48 hours:

1 out of 21 (4.8%) proned patients died compared to 3 out of 19 (15.8%) patients who were treated in supine position (RR 0.30; 95%CI 0.03 to 2.66). This difference was clinically relevant in favor of proned patients.

Mild-severe ARDS, <16 hours proning, starting after 48 hours:

275 out of 565 (48.7%) proned patients died compared to 249 out of 529 (47.1%) patients who were treated in supine position (RR 1.04; 95%CI 0.93 to 1.18). This difference was not clinically relevant.

Moderate-severe ARDS, >16 hours proning, starting within 48 hours:

181 out of 502 (36.1%) proned patients died compared to 232 out of 482 (48.3%) patients who were treated in supine position (RR 0.75; 95%CI 0.60 to 0.95). This difference was clinically relevant in favor of proned patients.

Moderate-severe ARDS, >16 hours proning, starting after 48 hours:

5 out of 11 (45.5%) proned patients died compared to 6 out of 11 (54.5%) patients who were treated in supine position (RR 0.83; 95%CI 0.36 to 1.94). This difference was clinically relevant in favor of proned patients.

Figure 2. Forest plot: Long-term mortality#

# Long-term mortality: >30-day mortality or in-hospital mortality

PaO₂/FiO₂ ratio at day 7 to 10

Four of the included RCTs (Guerin, 2004; Guerin, 2013; Mancebo, 2006) reported the PaO₂/FiO₂ ratio at day 7 or 10. Note that Taccone (2009) also reported this quotient, but baseline PaO₂/FiO₂ ratio was measured when the patient was already in prone position. Therefore, this study was not included in this analysis. Two of the included RCTs (Gattinoni, 2001; Voggenreiter, 2005) in Bloomfield (2015) that were selected for the current review reported the change in PaO₂/FiO₂ ratio. 1169 patients (prone, n=592; supine, n=577) were included in the meta-analysis of change data (figure 3). Change data from Guerin (2004) and Mancebo (2006) was calculated as the difference between first day/baseline data and day 7 or 10 data. For variance, the standard deviation at day 7 or 10 was used. Bloomfield (2015) mentioned that inflating the entered standard deviations with approximately 50% had no impact on interferences. Guerin (2013) was not included in the meta-analysis as formally mandated reduction in PEEP during proning would negatively impact oxygenation and would confound the results.

The PaO₂/FiO₂ ratio was more increased (MD 24.03; 95%CI 13.35 to 34.71) in patients with ARDS who were mechanically ventilated in prone position compared to patients with ARDS who were mechanically ventilated in supine position. This difference was clinically relevant.

Figure 3. Forest plot: Mean increase in PaO₂/FIO₂ ratio (mmHg) at 7 or 10 days

Quality of Life

No studies reporting quality of life were included in the summary of literature.

Adverse Events

Unplanned central catheter removal

Three of the included RCTs (Chan, 2007; Gattinoni, 2001; Taccone, 2009) in Munshi (2017) that were selected for the current review reported unplanned central catheter removal. 657 patients were included in the meta-analysis (figure 4).

60 out of 325 (18.5%) proned patients had unplanned central catheter removal compared to 41 out of 332 (12.3%) patients who were treated in supine position (RD 0.03; 95%CI -0.12 to 0.18). This difference was not clinically relevant.

Figure 4. Forest plot: Unplanned central catheter removal

Unplanned extubation

Eight of the included RCTs (Chan, 2007; Fernandez, 2008, Gattinoni, 2001; Guerin, 2004; Guerin, 2013; Mancebo, 2006; Taccone, 2009; Voggenreiter, 2005) in Bloomfield (2015) and Munshi (2017) that were selected for the current review reported unplanned endotracheal tube extubation. 2141 patients were included in the meta-analysis (figure 5).

113 out of 1099 (10.3%) proned patients had unplanned endotracheal tube extubation compared to 98 out of 1042 (9.4%) patients who were treated in supine position (RD 0.02; 95%CI -0.01 to 0.04). This difference was not clinically relevant.

Figure 5. Forest plot: Unplanned extubation

Endotracheal tube obstruction

Three of the included RCTs (Guerin, 2004; Guerin, 2013; Taccone, 2009) in Bloomfield (2015) that were selected for the current review reported endotracheal tube obstruction. 1597 patients were included in the meta-analysis (figure 6).

130 out of 818 (15.9%) proned patients had endotracheal tube obstruction compared to 76 out of 779 (9.8%) patients who were treated in supine position (RR 1.78; 95%CI 1.22 to 2.57). This difference was clinically relevant in favor of the supine group.

Figure 6. Forest plot: Endotracheal tube obstruction

Ventilator-associated pneumonia

Four of the included RCTs (Fernandez, 2008, Guerin, 2004; Mancebo, 2006; Voggenreiter, 2005) in Bloomfield (2015) and Munshi (2017) that were selected for the current review reported ventilator-associated pneumonia. 1007 patients were included in the meta-analysis (figure 7).

115 out of 531 (21.7%) proned patients had ventilator-associated pneumonia compared to 118 out of 476 (24.8%) patients who were treated in supine position (RR 0.83; 95%CI 0.66 to 1.06). This difference was clinically relevant in favor of the prone group.

Figure 7. Forest plot: Ventilator-associated pneumonia

Pressure sores

Three of the included RCTs (Chan, 2007; Gattinoni, 2001; Voggenreiter, 2005) in Bloomfield (2015) that were selected for the current review and Girard (2014) reported pressure sores. Guerin (2004) was not included as this study reported events per day instead of the number of patients with pressure sores. 823 patients were included in the meta-analysis (figure 8).

179 out of 416 (43.0%) proned patients had pressure sores compared to 139 out of 407 (34.1%) patients who were treated in supine position (RR 1.26; 95%CI 1.07 to 1.49). This difference was clinically relevant in favor of the supine group.

Figure 8. Forest plot: Pressure sores

Level of evidence of the literature

The majority of studies were terminated early due to slow enrollment. We decided not to lower the level of evidence because of slow enrollment on itself, since this is not a reason for probable bias.

Mortality (short term)

For the subgroup “mild-severe ARDS, <16 hours proning, starting after 48 hours”, the level of evidence started at high and was downgraded to very low level because of risk of bias (one level, many patients crossed over from intervention groups) and imprecision (two levels, 95% CI crossing both borders of clinical relevance).

For the subgroup “moderate-severe ARDS, >16 hours proning, starting within 48 hours”, the level of evidence started at high and was downgraded to a moderate level because of imprecision (one level, 95% CI crossing border of clinical relevance).

For the subgroup “moderate-severe ARDS, >16 hours proning, starting after 48 hours”, the level of evidence started at high and was downgraded to a very low level because of risk of bias (one level, allocation bias) and imprecision (two levels, 95% CI crossing both borders of clinical relevance).

Mortality (long term)

For the subgroup “mild-severe ARDS, <16 hours proning, starting within 48 hours”, the level of evidence started at high and was downgraded to very low level because of risk of bias (one level, differential use of co-intervention) and imprecision (two levels, 95% CI crossing both borders of clinical relevance).

For the subgroup “mild-severe ARDS, <16 hours proning, starting after 48 hours”, the level of evidence started at high and was downgraded to very low level because of risk of bias (one level, many patients crossed over from intervention groups) and imprecision (two levels, 95% CI crossing both borders of clinical relevance).

For the subgroup “moderate-severe ARDS, >16 hours proning, starting within 48 hours”, the level of evidence started at high and was downgraded to moderate level because of imprecision (one levels, 95% CI crossing border of clinical relevance).

For the subgroup “moderate-severe ARDS, >16 hours proning, starting after 48 hours”, the level of evidence started at high and was downgraded to very low level because of risk of bias (one level, allocation bias) and imprecision (two levels, 95% CI crossing both borders of clinical relevance).

PaO₂/FiO₂ ratio or quotient at day 7 to 10

The level of evidence started at high and was downgraded to low level because of risk of bias (one level, see risk of bias table) and inconsistency (one level, clinical heterogeneity between patients).

Quality of Life

As Chiumello (2012) did not meet inclusion criteria of the current review, results were not graded.

Adverse events

Unplanned central catheter removal

The level of evidence started at high and was downgraded to very low level because of risk of bias (one level, see risk of bias table), inconsistency (one level, clinical heterogeneity between patients) and imprecision (one level, low number of events).

Unplanned extubation

The level of evidence started at high and was downgraded to moderate level because of risk of bias (one level, see risk of bias table).

Endotracheal tube obstruction

The level of evidence started at high and was downgraded to low level because of risk of bias (one level, see risk of bias table) and imprecision (low number of events).

(Ventilator-associated) pneumonia

The level of evidence started at high and was downgraded to very low level because of risk of bias (one level, see risk of bias table), inconsistency (one level) and imprecision (low number of events).

Pressure sores

The level of evidence started at high and was downgraded to low level because of risk of bias (one level, see risk of bias table) and imprecision (one level, crossing the border of clinical relevance).

Search and select

A systematic review of the literature was performed to answer the following question: What is the effectivity of mechanical ventilation in the prone position in comparison to mechanical ventilation in supine position on mortality, PaO₂/FiO₂ ratio, Quality of Life (QoL) and adverse events in patients with ARDS?

P: Adult patients with ARDS receiving invasive mechanical ventilation at the ICU

I: Mechanical ventilation in the prone position

C: Mechanical ventilation in the supine position

O: Mortality, PaO₂/FIO₂ ratio on day 7 to 10, QoL, adverse events (unplanned central catheter removal, unplanned extubation, endotracheal tube obstruction, ventilator-associated pneumonia, and pressure sores)

Relevant outcome measures

The guideline development group considered mortality as a critical outcome measure for decision making; and PaO₂/FiO₂ ratio, QoL and adverse events as important outcome measures for clinical decision making.

The working group defined the outcome measures as follows:

Short-term mortality: 10-30-day mortality or IC mortality (according to Bloomfield 2015)

Long-term mortality: 31-180-day mortality or in-hospital mortality (according to Bloomfield 2015)

PaO₂/FiO₂ ratio on day 7 to 10: according to definitions used in the studies

QoL: according to definitions used in the studies.

Adverse events: according to definitions used in the studies.

The working group defined the following values as a minimal clinically (patient) important differences:

Mortality: 0.95≥RR≥1.05 -0.05≥RD≥0.05

PaO₂/FiO₂ ratio: MD≥10%

QoL: MD≥10% of maximum score from a validated questionnaire

Adverse events: 0.91≥RR≥1.10, or if RR is not available: -0.10≥RD≥0.10

Search and select (Methods)

The databases Medline (via OVID) and Embase (via Embase.com) were searched with relevant search terms until 17-11-2020. The detailed search strategy is depicted under the tab Methods. The systematic literature search resulted in 360 articles. Studies were selected based on the following criteria:

- Systematic review or RCT

- Published ≥ 2000

- Mechanically ventilated patients with ARDS admitted to the ICU

- Comparison between prone and supine position

- Intervention period for more than two days

- Outcome measures as described in PICO

23 studies were initially selected based on title and abstract screening. After reading the full text, 21 studies were excluded (see the table with reasons for exclusion under the tab Methods), and two studies were included.

Results

One systematic review (Bloomfield, 2015) was included in the analysis of the literature. A second systematic review (Munshi, 2017) was included for one outcome measure.

Important study characteristics and results are summarized in the evidence tables. The assessment of the risk of bias is summarized in the risk of bias tables.