Prolonged Prone Position Ventilation Is Associated With Reduced Mortality in Intubated COVID-19 Patients

Abstract

Background

Prone position ventilation (PPV) is resource-intensive, yet the optimal strategy for PPV in intubated patients with COVID-19 is unclear.

Research Question

Does a prolonged (24 or more h) PPV strategy improve mortality in intubated COVID-19 patients compared with intermittent (∼16 h with daily supination) PPV?

Study Design and Methods

Multicenter, retrospective cohort study of consecutively admitted intubated COVID-19 patients treated with PPV between March 11 and May 31, 2020. The primary outcome was 30-day all-cause mortality. Secondary outcomes included 90-day all-cause mortality and prone-related complications. Inverse probability treatment weights (IPTW) were used to control for potential treatment selection bias.

Results

Of the COVID-19 patients who received PPV, 157 underwent prolonged and 110 underwent intermittent PPV. Patients undergoing prolonged PPV had reduced 30-day (adjusted hazard ratio [aHR], 0.475; 95% CI, 0.336-0.670; P < .001) and 90-day (aHR, 0.638; 95% CI, 0.461-0.883; P = .006) mortality compared with intermittent PPV. In patients with Pao₂/Fio₂ ≤ 150 at the time of pronation, prolonged PPV was associated with reduced 30-day (aHR, 0.357; 95% CI, 0.213-0.597; P < .001) and 90-day mortality (aHR, 0.562; 95% CI, 0.357-0.884; P = .008). Patients treated with prolonged PPV underwent fewer pronation and supination events (median, 1; 95% CI, 1-2 vs 3; 95% CI, 1-4; P < .001). PPV strategy was not associated with overall PPV-related complications, although patients receiving prolonged PPV had increased rates of facial edema and lower rates of peri-proning hypotension.

Interpretation

Among intubated COVID-19 patients who received PPV, prolonged PPV was associated with reduced mortality. Prolonged PPV was associated with fewer pronation and supination events and a small increase in rates of facial edema. These findings suggest that prolonged PPV is a safe, effective strategy for mortality reduction in intubated COVID-19 patients.

Graphical Abstract

FOR EDITORIAL COMMENT, SEE PAGE 469

Take-home Points.

Study Question: Does prolonged prone position ventilation (PPV) improve mortality in patients intubated for COVID-19?

Results: Prolonged PPV (> 24 h) is associated with improved 30-day and 90-day survival in patients intubated for COVID-19 compared with intermittent PPV (∼16 h prone with daily supination) without an increased risk of overall complications, and reduces the total number of pronation and supination episodes.

Interpretation: Among intubated COVID-19 patients, prolonged PPV was associated with improved 30-day and 90-day survival, fewer pronation and supination episodes, and no increased risk of overall complications compared with intermittent PPV.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus that causes COVID-19, results in critical illness in up to 5% of patients. Most patients require prolonged mechanical ventilation, with mortality comparable to that of ARDS.^{1, 2, 3, 4, 5} Few effective therapies are available for reducing mortality in COVID-19 patients admitted to the ICU, highlighting the importance in identifying additional therapies for this high-mortality group.^{6, 7, 8}

Prone position ventilation (PPV) reduces mortality in ARDS.⁹ However, PPV is a resource-intensive intervention requiring multiple highly trained staff to execute each pronation and supination event.¹⁰ Additionally, the optimal duration for PPV is unknown.¹¹ Current practice tends to follow the protocol set out in the PROSEVA trial, whereby patients were randomized to receive PPV for at least 16 h per day, with daily supination events.^12,13 Meta-analyses have demonstrated a mortality benefit when patients are proned for > 12 h/day, but the physiologic benefits of PPV continue to accrue through 24 h of prone ventilation.^14,15 Furthermore, supination is frequently accompanied by a de-recruitment event, possibly increasing lung stress and strain, ventilator-induced lung injury, and mortality.¹⁶ These observations suggest that prolonged exposure to PPV may offer an additional mortality benefit. Three prior studies evaluated safety and efficacy of PPV longer than 24 h; however, no studies have compared this prolonged PPV approach with one including daily supination.^{17, 18, 19} Additionally, lengthening PPV duration would reduce the frequency of pronation and supination events, decreasing the number of staff required to implement this lifesaving treatment. Thus, determining whether a prolonged PPV strategy is efficacious may benefit both individual patients and strained hospital systems by reducing health care resource utilization.

Early in the pandemic, uncertainty about appropriate treatment of critically ill COVID-19 patients led to heterogeneity in clinical practice, providing a unique opportunity to study which PPV parameters contribute to improved mortality. In this study, we assessed the association between prolonged (> 24 h prone) or intermittent (∼16 h prone with daily supination) PPV strategy and mortality in intubated COVID-19 patients.

Study Design and Methods

This is a retrospective cohort study of all patients with confirmed COVID-19 and acute respiratory failure consecutively admitted to a medical or surgical ICU at three Mass General Brigham hospitals in the Boston, Massachusetts, metropolitan area, from March 11, 2020, through May 31, 2020. These hospitals included both academic medical centers and community hospitals. Each hospital employed dedicated proning teams comprising individuals with prior experience with PPV. Positive end-expiratory pressure (PEEP) was set at each institution either based on ARDSNet PEEP/Fio₂ tables²⁰ or best compliance during a decremental PEEP titration trial. Remdesivir was given as a 200-mg single dose on day 1 followed by 100 mg daily for 5 days. Tocilizumab was the anti-IL-6 monoclonal antibody administered most often in this cohort, typically given as a 400-mg single dose. Most patients did not receive high-dose steroids early in the course of the disease. Inclusion criteria were age 18 years or older, positive SARS-CoV-2 reverse transcription-quantitative polymerase chain reaction test, need for mechanical ventilation, and at least one PPV episode.

Prone strategy was classified based on exact dates and times of pronation and supination events. Prolonged PPV was defined as a prone duration lasting at least 24 h before supination. Intermittent PPV was defined as daily pronation and supination events. Patients were classified based upon the prone strategy employed for their first proning episode. Patients who underwent a single PPV event during their ICU stay, or who had prone sessions on nonconsecutive days, were classified as intermittent if the longest prone session length was less than 24 h. Hospitalwide guidelines were not available during the study for choice of prone strategy, but physicians typically used clinical improvement and subsequent stability as a means for determining whether patients should receive intermittent or prolonged PPV.

Study data were collected and managed using the REDCap electronic data capture tool hosted at Mass General Brigham.^21,22 Data were abstracted through May 31, 2021, by board-certified or board-eligible physicians. The study protocol was approved by the Institutional Review Board at Massachusetts General Hospital (2020P001119). The need for informed consent was waived.

Outcomes

The primary outcome was 30-day all-cause mortality. Secondary outcomes included 90-day mortality, hospital length of stay (LOS), ICU LOS, ventilator-free days at 28 days, and prespecified complications of PPV. Prespecified subgroup analysis included patients with Pao₂/Fio₂ ratio ≤ 150 immediately before PPV, based on the PROSEVA trial and existing clinical guidelines.^9,12

Statistical Analyses

To control for potential treatment selection bias between the intermittent and prolonged prone strategies, we carried out a propensity score based inverse probability treatment weighted (IPTW) analysis. Propensity scores were calculated through a multivariable logistic regression model that included admission hospital, time-to-intubation in days, PEEP at time of intubation, and need for dialysis at time of proning. Variables were chosen based on an a priori assessment of the likelihood that they would contribute to the clinical decision to repeatedly pronate or supinate a patient. Standardized mean (or proportional) difference less than 0.1 for each variable was used to verify balance between treatment groups. A sensitivity analysis was performed, incorporating additional variables in the propensity score model.

To determine the association between PPV strategy and mortality, a parametric outcome survival model was fit using the Weibull distribution with the survreg package (version 3.2-13) incorporating stabilized IPTW weights to adjust for potential treatment selection bias. This outcome model adjusted for potential confounders including age, sex, Charlson comorbidity index (CCI), sequential organ failure assessment (SOFA) score at hospital admission, Pao₂/Fio₂ ratio immediately before pronation, BMI, treatment with anti-IL-6 therapy, and treatment with paralytics, while stratifying on admission hospital. Survival time was calculated from the time of PPV initiation to death. We performed a prespecified subgroup analysis focused only on patients with Pao₂/Fio₂ ≤ 150 just before pronation, as described. Propensity score weights were recalculated for this subgroup analysis and the outcome models adjusted for the same set of covariates in the primary analyses. Only four patients (1.5%) were excluded from the analysis because of missing IPTW covariates; thus, we did not adjust for missingness in the final outcome models.

Predictors of complications due to PPV were identified using generalized linear models. Penalized splines were initially used to test for nonlinear relationships between continuous predictors, such as duration of PPV, and outcomes. These models showed a linear relationship between duration of PPV and complications; thus, the final model incorporated prone duration as a linear covariate.

Secondary analyses focused on understanding the role of change in Pao₂/Fio₂ ratio with proning (ΔPF). Patients were determined to have an improved ΔPF if the Pao₂/Fio₂ ratio after proning was greater than before proning (ie, ΔPF > 0 mm Hg). Parametric survival models using the Weibull distribution were used to estimate the association between mortality and improvement in Pao₂/Fio₂ ratio alone and after adjusting for potential confounders, including age, sex, CCI, SOFA score, Pao₂/Fio₂ ratio before PPV, BMI, treatment with anti-IL-6 therapy, and treatment with paralytics, while stratifying on admission hospital.

Two-sided P-values less than .05 were considered significant. All data analysis was performed using R (version 4.2.1). Continuous variables are presented as median (interquartile range), and categorical data are presented as number (%). The R code used for this analysis is shown in e-Appendix 1.

Results

Patient Characteristics

A total of 267 patients received PPV while mechanically ventilated for confirmed COVID-19; 157 of 267 (58.8%) received prolonged PPV (Fig 1). The median age was 62 years; 64.4% were male; and 55.4% were White (Table 1). Prolonged PPV patients were less likely to be treated with paralysis (27.4% vs 43.6%; P = .009) and less likely to receive anti-IL-6 therapy (9.6% vs 32.7%, P < .001) than intermittent PPV patients. No group differences were seen in severity of illness at admission assessed using SOFA, simplified acute physiology score II, and Pao₂/Fio₂ ratio. Similar trends were seen between PPV strategies in the subgroup of patients with Pao₂/Fio₂ ≤ 150 just before pronation (e-Table 1). Low tidal volume ventilation was achieved in the cohort, with an average tidal volume of 6.01 (5.84-6.27) cc/kg ideal body weight, and an average driving pressure of 11.0 (9.5-13.0) cm H₂O before pronation. Of the patients initially receiving intermittent PPV, 24 of 110 (21.8%) transitioned to a prolonged strategy (defined as subsequent prone sessions lasting 24 or more hours), whereas of patients initially receiving prolonged PPV, 5 of 157 (3.2%) transitioned to an intermittent strategy (defined as daily supination events).

Figure 1.

Study flow diagram.

Table 1.

Demographics and Characteristics of Patients Admitted to the ICU With COVID-19

Characteristics	Overall	Prolonged Prone	Intermittent Prone	Pa
	(n = 267)	(n = 157)	(n = 110)
Demographics
Age	62 (51-72)	63 (52-70)	60 (51-74)	.991
Male	172 (64.4)	103 (65.6)	69 (62.7)	.724
White	148 (55.4)	89 (56.7)	59 (53.6)	.712
Hispanic or Latino/a	117 (46.8)	70 (48.6)	47 (44.3)	.589
BMI	29.8 (26.5-34.8)	29.6 (26.7-34.5)	30.1 (26.0-35.7)	.743
CCI	3 (1-5)	3 (1-4)	4 (2-5)	.145
SAPS IIb	32 (25-41)	33 (26-43)	32 (24-39)	.299
SOFA Scoreb	6 (4-8)	7 (4-8)	6 (4-8)	.555
Pao2/Fio2 on admissionc	157 (99-211)	158 (105-211)	149 (99-211)	.623
Vasopressor use within 24 h of admissiond	138 (52.5)	88 (56.8)	50 (46.3)	.122
Days of COVID-19 symptomsc	6 (4-9)	7 (4-10)	6 (3-7)	.004
COVID-19 therapies
Remdesivir	28 (10.5)	16 (10.2)	12 (10.9)	1
Anti-IL-6	51 (19.1)	15 (9.6)	36 (32.7)	<.001
Steroids	68 (25.5)	41 (26.1)	27 (24.5)	.883
Admission hospital				<.001
MGH	175 (65.5)	126 (80.3)	49 (44.5)
NWH	27 (10.1)	16 (10.2)	11 (10.0)
SH	65 (24.3)	15 (9.6)	50 (45.5)
Intubation characteristics
Time to intubation, d	1 (0-2)	1 (0-2)	1 (0-3)	.046
Ventilator mode				.305
Volume control	254 (95.1)	152 (96.8)	102 (92.7)
Pressure control	10 (3.7)	4 (2.5)	6 (5.5)
Pressure support	3 (1.1)	1 (0.6)	2 (1.8)
Vt/kg IBW,e mL/kg	6.01 (5.84-6.27)	6.00 (5.84-6.22)	6.04 (5.79-6.35)	.397
PEEP, cm H2O	12 (10-14)	12 (10-14)	12 (10-14)	.276
Additional ICU therapies
Paralysis	91 (34.1)	43 (27.4)	48 (43.6)	.009
Pulmonary vasodilator	85 (31.8)	53 (33.8)	32 (29.1)	.501
ECMO	14 (5.2)	4 (2.5)	10 (9.1)	.037
Tracheostomy	89 (33.3)	61 (38.9)	28 (25.5)	.031

CCI = Charlson comorbidity index; ECMO = extracorporeal membrane oxygenation; IBW = ideal body weight; MGH = Massachusetts General Hospital; NWH = Newton-Wellesley Hospital; SAPS = simplified acute physiology score; SH = Salem Hospital; SOFA = sequential organ failure assessment.

Data missing for ^b3, ^c1, ^d4, ^e6 patients.

^aTwo-tailed P value based on Pearson χ² test for categorical data and Mann-Whitney U test for continuous data.

IPTW weights were constructed separately for all patients, and those with Pao₂/Fio₂ ≤ 150 before pronation. Of 267 patients receiving PPV, the IPTW cohort included 263 patients (Fig 1, study flow diagram). Demographics of the four patients excluded because of missing covariates are available in e-Table 2. After adjusting for potential treatment selection bias with stabilized IPTW in the overall cohort, no differences were found in key variables between PPV strategies (e-Table 3, e-Fig 1). In patients with Pao₂/Fio₂ ≤ 150 before pronation, the IPTW cohort included 166 patients; no differences in key variables between PPV strategies were found (e-Table 4, e-Fig 2).

Primary Outcome

Kaplan-Meier curves for 90-day survival are shown in Figure 2. In multivariable Weibull survival models, prolonged PPV was associated with reduced 30-day mortality (aHR, 0.475; 95% CI, 0.336-0.670, P < 0.001) and 90-day mortality (aHR, 0.638; 95% CI, 0.461-0.883; P = .006) compared with intermittent PPV in the overall cohort. In prespecified subgroup analyses focused on patients with Pao₂/Fio₂ ≤ 150 before pronation, the protective effect of prolonged PPV on mortality was stronger. Prolonged PPV was associated with both reduced 30-day (aHR, 0.357; 95% CI, 0.213-0.597; P < .001) and 90-day mortality (aHR, 0.562; 95% CI, 0.357-0.884; P = .008) compared with intermittent PPV. Sensitivity analysis incorporating additional covariates in the IPTW model showed similar results (e-Table 5). Hospital LOS, ICU LOS, and ventilator-free days at 28 days differed between survivors and nonsurvivors but not between PPV strategies (e-Table 6).

Figure 2.

Kaplan-Meier plot of the probability of survival from time of proning to 90 days in (A) the entire cohort and (B) patients with Pao₂/Fio₂ ≤ 150 at time of pronation.

Proning Characteristics and Complications

No differences in time from intubation to proning between PPV strategies were found (Table 2). Patients receiving prolonged PPV had longer cumulative length of time in the prone position (median, 68 h; interquartile range [IQR], 46-120, vs 48 h; IQR 19-80, P < .001) and fewer pronation and supination events compared with the intermittent PPV group (median 1, IQR 1-2, vs 3, IQR 1-4, P < .001). Duration of first prone session was longer in patients receiving prolonged PPV (median, 40 h; IQR, 27-55, vs 17 h, IQR 14-20, P < .001). The empirical cumulative distribution functions of PPV session duration are available in e-Figure 3.

Table 2.

Characteristics and Respiratory Parameters of Proning

Characteristic	Overall	Prolonged Prone	Intermittent Prone	Pa
	(n = 267)	(n = 157)	(n = 110)
Characteristics of Proning
# Pronation and supinationb events	2 (1-3)	1 (1-2)	3 (1-4)	< .001
Time to proning, days	2 (1-4)	2 (1-4)	2 (1-4)	.844
Prone duration (total),c,d hours	59.4 (34.3-102.5)	67.8 (45.5-120.0)	47.7 (19.4-79.9)	< .001
Prone duration (first session),e hours	25 (17-46)	40 (27-55)	17 (14-20)	< .001
Respiratory parametersf
Pao2/Fio2c
Before proning	132 (103-162)	138 (106-164)	126 (96-159)	.298
After proningg	212 (161-280)	220 (162-278)	204 (158-283)	.471
Change with proning	79 (28-136)	79 (28-139)	80 (28-134)	.773
PEEP,c cm H2O
Before proning	12 (10-14)	12 (10-14)	12 (10-14)	.405
After proningf	12 (11-15)	13 (12-14)	12 (11-15)	.849
Driving pressure, cm H2O
Before proningh	11.0 (9.5-13.0)	11.0 (9.0-13.0)	12.0 (10.0-14.0)	.088
After proningf,i	11.0 (9.0-13.0)	11.0 (9.0-13.0)	11.0 (10.0-14.0)	.171
Change with proningj	0.0 (-1.0-1.0)	0.0 (-2.0-1.0)	0.0 (-1.0-1.0)	.397
Compliance, mL/cm H2O
Before proningk	32.3 (25.4-40.0)	33.3 (26.2-41.7)	30.8 (25.0-37.3)	.235
After proningf,l	33.3 (26.1-40.0)	33.0 (26.0-40.0)	33.8 (26.9-39.4)	.678
Change with proningm	0 (-3.9-4.0)	0 (-4.1-4.9)	0 (-2.7-3.1)	.561
Ventilatory ratio
Before proningj	1.53 (1.26-1.87)	1.47 (1.26-1.82)	1.59 (1.34-2.03)	.103
After proningf,n	1.58 (1.32-1.93)	1.55 (1.30-1.83)	1.63 (1.36-2.06)	.170
Change with proningo	0.00 (-0.12-0.12)	0.00 (-0.11-0.09)	0.00 (-0.12-0.19)	.574

Data missing for ^c2, ^h68, ⁱ59, ^j95, ^k72, ^l63, ^m99, ⁿ25, ^o106 patients.

^aTwo-tailed P value based on Pearson χ² test for categorical data and Mann-Whitney U test for continuous data.

^bOne event defined as a pair of pronation and then supination.

^dCalculated as the time spent in the prone position excluding any time spent supine.

^eCalculated for the first discrete pronation/supination episode.

^fMeasured and calculated at the time of the first PPV session.

^gCollected within 6 h of pronation.

When comparing the effect of PPV strategy on changes in Pao₂/Fio₂ ratio with the first proning event, there was no overall difference between groups in the Pao₂/Fio₂ ratio before or after proning, nor did the magnitude of change in Pao₂/Fio₂ ratio with the first proning event differ (Table 2). In univariable analysis, improvement in Pao₂/Fio₂ ratio (ie, change in Pao₂/Fio₂ ratio > 0 mm Hg) with proning was associated with improved 30-day (hazard ratio [HR], 0.576; 95% CI, 0.382-0.868; P = .009) and 90-day mortality (HR, 0.617; 95% CI, 0.414-0.919; P = .02). Similarly, for patients with Pao₂/Fio₂ ≤ 150 at the time of proning, improvement in Pao₂/Fio₂ ratio with proning was associated with 30-day (HR, 0.467; 95% CI, 0.277-0.789; P = .005) and 90-day (HR, 0.547; 95% CI, 0.326-0.917; P = .02) mortality. In the multivariable Weibull model in the overall cohort, improvement in Pao₂/Fio₂ ratio remained associated with improved 30-day (aHR, 0.329; 95% CI, 0.211-0.512; P < .001) and 90-day mortality (aHR, 0.447; 95% CI, 0.287-0.694; P < .001). Inclusion of improvement in Pao₂/Fio₂ ratio did not alter the association between prolonged PPV and 30-day (aHR, 0.409; 95% CI, 0.289-0.579; P < .001) or 90-day mortality (aHR, 0.577; 95% CI, 0.414-0.805; P < .001).

No differences were found in ventilator mode, settings, or mechanics between patients receiving either PPV strategy (Table 2, e-Table 7). A total of 48.3% of patients had a complication associated with PPV (Table 3), the most common of which were pressure injuries (29.2%) and facial edema (11.6%). Patients receiving prolonged PPV had a higher rate of facial edema (15.3% vs 6.4%; P = .04) and a lower rate of peri-proning hypotension (1.3% vs 7.3%; P = .03). PPV strategy was not associated with overall increased risk of complications related to prone position (prolonged vs intermittent, aOR, 0.658; 95% CI, 0.388-1.106; P = .116; e-Table 8).

Table 3.

Complication in Prone Position

Complication	Overall	Prolonged Prone	Intermittent Prone	Pa
	(n = 267)	(n = 157)	(n = 110)
Any complication of proning	129 (48.3)	73 (46.5)	56 (50.9)	.558
Early cessation of proning	30 (23.3)	13 (17.8)	17 (30.4)	.144
Arrhythmias	6 (2.2)	5 (3.2)	1 (0.9)	.415
Hypotension	10 (3.7)	2 (1.3)	8 (7.3)	.027
Loss of vascular access	5 (1.9)	1 (0.6)	4 (3.6)	.187
Chest tube displacement	1 (0.4)	0 (0.0)	1 (0.9)	.858
OG/NG tube displacement	2 (0.7)	0 (0.0)	2 (1.8)	.330
Accidental extubation	2 (0.7)	2 (1.3)	0 (0.0)	.640
Endotracheal tube displacement	9 (3.4)	6 (3.8)	3 (2.7)	.886
Worsening ventilator mechanics	8 (3.0)	5 (3.2)	3 (2.7)	1
Facial edema	31 (11.6)	24 (15.3)	7 (6.4)	.041
Pressure injuries	78 (29.2)	48 (30.6)	30 (27.3)	.655
Conjunctival hemorrhage	1 (0.4)	1 (0.6)	0 (0.0)	1
Vomiting	6 (2.2)	1 (0.6)	5 (4.5)	.089
Oropharyngeal injury	7 (2.6)	4 (2.5)	3 (2.7)	1
Rhabdomyolysis	2 (0.7)	2 (1.3)	0 (0.0)	.640

OG = orogastric; NG = nasogastric.

^aTwo-tailed P value based on Pearson χ² test.

Discussion

In this study of intubated COVID-19 patients, prolonged PPV was associated with higher 30-day and 90-day survival compared with intermittent PPV in both the overall cohort and in a prespecified subgroup of patients with Pao₂/Fio₂ ≤ 150 before pronation. PPV strategy was not associated with an overall higher rate of proning-related complications and resulted in significantly fewer pronation and supination events. These findings suggest that intermittent supination, especially in patients with more severe disease, may be injurious. Prolonged PPV has a favorable safety profile and reduces resource utilization because of the need for fewer pronation and supination episodes.

This study is an important addition to the literature on PPV in acute respiratory failure. First, although prior studies have characterized safety and efficacy of PPV for greater than 24 h, this is the first to directly compare a prolonged PPV approach with the intermittent strategy laid out in the PROSEVA trial, addressing a long-standing question on whether lengthening the duration of PPV improves outcomes.^{9, 10, 11,17, 18, 19,23} Second, this study contributes detailed information on the number of pronation or supination episodes with implications for resource utilization, and further extends our knowledge of the effects of PPV on oxygenation and lung mechanics.²⁴ Third, this study provides a comprehensive look at complications of PPV, highlighting the areas to focus development of preventative measures.¹⁰

There are several potential benefits to choosing prolonged rather than intermittent PPV in patients intubated for COVID-19. First, the physiological benefits of PPV, including improved compliance of the respiratory system and reduced lung strain, improve continuously over 24 h of prone ventilation.¹⁵ The de-recruitment associated with repeated supination may lead to increased atelectotrauma and ventilator-induced lung injury, possibly contributing to mortality.^25,26 In addition, meta-analyses have indicated that PPV reduces mortality only if its duration is greater than 12 h per day^14,27,28 and in patients receiving low tidal volume ventilation (LTVV; ≤ 8 mL/kg ideal body weight).²⁹ Because the PROSEVA trial showed that LTVV plus 17 h of PPV is superior to LTVV while supine, our findings that prolonged PPV is associated with improved mortality compared with intermittent PPV logically follows. Second, from a resource utilization standpoint, prolonged PPV requires significantly fewer health care personnel to implement because of fewer pronation and supination episodes. Third, although our data suggest that the safety profile of prolonged PPV is comparable to that of traditional intermittent PPV, it will be important to conduct additional research into methods to reduce the risk of potential complications related to prolonged PPV. A randomized controlled trial comparing prolonged with intermittent PPV will be important to verify these findings.

Prior work has indicated that improvement in Pao₂/Fio₂ ratio with proning is associated with reduced mortality in COVID-19, although this is less clear for other causes of ARDS.^{30, 31, 32, 33} We find no differences in the change in Pao₂/Fio₂ ratio with PPV based on strategy, nor does the change in Pao₂/Fio₂ ratio explain the association with improved mortality seen with prolonged PPV. Evaluating the Pao₂/Fio₂ ratio over time in patients receiving either PPV strategy may reveal a difference in oxygenation response to prolonged PPV, as has previously been seen with PPV and recruitment maneuvers.^34,35 Additionally, the relative balance between resolution of dorsal atelectasis and development of ventral atelectasis may contribute to initial changes in Pao₂/Fio₂ ratio with PPV or contribute to a diminished response with prolonged PPV.³⁶ Further research will be necessary to determine how prolonged PPV affects ventilation inhomogeneity during ARDS and the contribution of this effect to the associated improvement in mortality.

Ventilator mechanics are known to influence outcomes in severe ARDS. First, higher PEEP has been associated with improved mortality in patients with ARDS and Pao₂/Fio₂ ratio < 200 mm Hg.³⁷ Although we found no difference in PEEP before or after pronation based on PPV strategy, we only measured mechanics up to 6 h after the first pronation episode. Future work evaluating the change in PEEP over time may reveal an important interaction between PEEP and PPV strategy, as patients who did not manifest improved Pao₂/Fio₂ ratio with proning may have undergone another PEEP titration to increase recruitment. Second, lower driving pressure is associated with reduced mortality and is hypothesized to be one pathophysiologic rationale by which PPV improves outcomes in ARDS.^16,38 We observed a slightly higher driving pressure in patients receiving intermittent PPV but found no change in driving pressure or compliance after PPV. Information on driving pressure was missing in 35.6% patients, limiting the ability to evaluate whether a change in driving pressure is associated with mortality or interacts with PPV strategy. Future work examining changes in respiratory mechanics will be necessary to assess whether changes in driving pressure are the mechanism underpinning improved mortality with prolonged PPV.

Our study has several strengths. We included all patients consecutively admitted to three Boston area community hospitals and academic centers, thus minimizing selection bias. Chart review was performed by board-certified or board-eligible physicians. Although our study has identified a benefit of PPV strategy on mortality in intubated COVID-19 patients, it is plausible that the effects of PPV strategy may extend to other causes of acute respiratory failure and merits further study.

There are several limitations to our study. First, although this was a multicenter study, it was conducted within the same hospital system. Second, the study was retrospective in nature. Although we adjusted for potential treatment selection bias and confounders selected based on prior known biology, unmeasured confounding is always a possibility in observational studies, and thus these findings should be confirmed in future randomized controlled trials. Third, patients in the prolonged PPV group benefitted from a greater cumulative duration of PPV. Although this is probably because their clinical state was more severe, we cannot completely exclude that the effect on mortality lies in this difference. Finally, the patient population represents the first wave of the COVID-19 pandemic in Boston, when there were no proven COVID-19 specific treatments. However, our overall 30-day mortality was 30.0%, comparable to pre-pandemic patients intubated for acute respiratory failure,³⁹ and patients received low tidal volume ventilation after intubation, suggesting that our results were not influenced by an overwhelmed hospital system. Current literature suggests that ICU mortality has not improved over successive COVID-19 waves,⁴⁰ highlighting the importance of identifying effective treatments in this vulnerable patient population.

Interpretation

Among intubated COVID-19 patients, prolonged PPV was associated with improved 30-day and 90-day survival, fewer pronation and supination episodes, and no increased risk of overall complications compared with intermittent PPV.

Funding/Support

Dr Okin is supported by the National Institutes of Health (NIH [grant T32HL1167275]). Dr Alba is supported by the NIH (grant 5KL2TR002542-02).

Financial/Nonfinancial Disclosures

None declared.