Prone position versus usual care in hypoxemic COVID-19 patients in medical wards: a randomised controlled trial

Abstract

Background: Benefit of early awake prone positioning for COVID-19 patients hospitalised in medical wards and who need oxygen therapy remains to be demonstrated. The question was considered at the time of COVID-19 pandemic to avoid overloading the intensive care units. We aimed to determine whether prone position plus usual care could reduce the rate of non-invasive ventilation (NIV) or intubation or death as compared to usual care alone.

Methods: In this multicentre randomised clinical trial, 268 patients were randomly assigned to awake prone position plus usual care (N = 135) or usual care alone (N = 132). The primary outcome was the proportion of patients who underwent NIV or intubation or died within 28 days. Main secondary outcomes included the rates of NIV, of intubation or death, within 28 days.

Results: Median time spent each day in the prone position within 72 h of randomisation was 90 min (IQR 30-133). The proportion of NIV or intubation or death within 28 days was 14.1% (19/135) in the prone position group and 12.9% (17/132) in the usual care group [odds ratio adjusted for stratification (aOR) 0.43; 95% confidence interval (CI) 0.14-1.35]. The probability of intubation, or intubation or death (secondary outcomes) was lower in the prone position group than in the usual care group (aOR 0.11; 95% CI 0.01-0.89 and aOR 0.09; 95% CI 0.01-0.76, respectively) in the whole study population and in the prespecified subgroup of patients with SpO₂ ≥ 95% on inclusion (aOR 0.11; 95% CI 0.01-0.90, and aOR 0.09; 95% CI 0.03-0.27, respectively).

Conclusions: Awake prone position plus usual care in COVID-19 patients in medical wards did not decrease the composite outcome of need for NIV or intubation or death. Trial registration ClinicalTrials.gov Identifier: NCT04363463 . Registered 27 April 2020.

Keywords: Awake prone position; COVID-19; Hypoxemia; Medical wards; Non-intubated; Respiratory failure; Spontaneous breathing.

Fig. 1

Patient recruitment flowchart. Abbreviation ICU, intensive care unit. ^aPatients discharged from the ICU after non-invasive ventilation or intubation for COVID-19 pneumonia. ^bThirty-one patients had oxygen therapy at home; 24 had severe chronic obstructive pulmonary disease Gold stage 3 or 4; 9 had known chronic diffuse interstitial lung disease. ^cRecent thoracic trauma, pneumothorax, unstable spine or pelvis fractures. ^dSix patients had respiratory frequency > 40/min on screening; two showed excessive use of accessory respiratory muscles as judged by the clinician in charge; one had an indication for non-invasive ventilation for acute pulmonary oedema. ^eThree patients had chronic neuromuscular disease; one had an intestinal occlusive syndrome on screening; one had no social insurance

Fig. 2

Between-group comparison of outcomes within 28 days in prespecified subgroups; results of multivariable logistic regression analysis. Abbreviations OR, odds ratios; CI, confidence interval; SpO₂, oxygen saturation measured by pulse oximetry; BMI, body mass index. ^aThere was no three-way interaction (P = .72). Among the three possible two-way interactions, only the interaction “intervention by initial SpO₂ < or ≥ 95%” was significant (P = .021) and was kept in the model. ^bThere was no three-way interaction (P > .99). Among the three possible two-way interactions, only the interaction “intervention by BMI < or ≥ 30 kg/m²” was significant (P = .049) and was kept in the model. ^cThere was no three-way interaction (P = .97). Among the three possible two-way interactions, only the interaction “intervention by initial SpO₂ < or ≥ 95%” was significant (P = 0.02) and was kept in the model. ^dThere was no three-way interaction (P = .93). Among the three possible two-way interactions, only the interaction “intervention by initial SpO₂ < or ≥ 95%” was significant (P = .01) and was kept in the model. ^eThere was no three-way interaction (P = .79). There was no significant interaction among the three possible two-way interactions (all P > .05). None of those interaction terms were introduced in the logistic model

Fig. 3

Probability of intubation or death using time-to-event analysis. Abbreviations HR, hazard ratio; SpO₂, oxygen saturation measured by pulse oximetry. Note that the y axis is truncated at 0.6 to make the curves easily distinguishable. Groups were compared by Cox proportional hazards analysis (see Additional file 2, page 8)