In this single-center, retrospective, observational, cohort study, we recruited eligible patients admitted to the medical intensive care unit (MICU) and center intensive care unit (CICU) of West China Hospital of Sichuan University, a large tertiary teaching hospital, between January 1, 2017, and December 31, 2021.

Patients were eligible to be included in the study if they were between 18–85 years of age, were diagnosed with PCP by etiological examination (fungal smear/cultures or metagenomics next generation sequencing) of sputum or bronchoalveolar lavage fluid, and received invasive mechanical ventilation targeted at moderate–severe ARDS for less than 72 h prior to admission to the ICU.

Exclusion criteria were pregnancy, advanced carcinoma or massive hemorrhage, and invasive mechanical ventilation less than 72 h after admission to the ICU.

Patients who received PPV treatment were enrolled in the prone cohort, and those who received no PPV were enrolled in the supine cohort.

The PPV procedure conducted in our study was based on the protocol published as part of the PROSEVA trial [15]. The durations mostly occurred for 12 h [11, 16] and were extended to 16–18 h when necessary [15, 17]. Patients were eligible to receive PPV if the PaO2/FiO2 ratio was lower than 200 mmHg (typically lower than 150 mmHg), upon evaluation by their attending physician. Complications leading to immediate termination of PPV included unexpected artificial airway extubation or obstruction, hemoptysis, oxygen saturation less than 85% on pulse oximetry for more than 2 min when the FiO2 was 1.0, cardiac arrest, a heart rate of less than 50 beats or more than 160 beats per minute for more than 30 s, a systolic blood pressure of less than 90 mm Hg for more than 5 min, or any other life-threatening incidents.

Mechanical ventilation was delivered according to the ALVEOLI study [18], with end-inspiratory plateau pressure (Pplat) maintained below 30 cm H2O, and volumes guaranteed between 6–8 mL/kg PBW or 4–6 mL/kg PBW during extracorporeal membrane oxygenation (ECMO) therapy. Permissive hypercapnia was flexibly performed when lung-protective ventilation was impeded, with pH values of 7.30–7.35 and 7.25–7.35 in patients with chronic obstructive pulmonary disease.

The primary end point was mortality at 28 days. The secondary endpoints were mortality at 90 days, the rate of successful extubation on day 28, the length of invasive mechanical ventilation days and ventilator-free days at 28 days, the length of ICU stay on day 90, the pneumothorax incidence rate, and the tracheotomy rate at 90 days.

Successful extubation was defined as no reintubation for 48 h after extubation. In patients who had undergone tracheotomy, successful weaning from the ventilator was defined as the ability to breathe through the tracheostomy cannula for at least 48 h with oxygen therapy or high-flow oxygen therapy.

We collected data via the hospital electronic health record system, including age, sex, body mass index, Sequential Organ Failure Assessment (SOFA) score, Simplified Acute Physiology Score II (SAPS II), smoking history, patient origin, coexisting conditions, coexisting germs in the respiratory system, anti-PCP drug before ICU admission, cointerventions, ventilator settings and arterial blood gas (ABG) parameters, laboratory tests and radiological characteristics.

Chronic obstructive pulmonary disease, asthma, and interstitial lung disease were recorded as chronic bronchopulmonary disease, while patients using immuno-suppressive drugs, such as cyclosporin A, protopic, mycophenolate mofetil, azathioprine, and prednisone were considered receiving immuno-suppressive therapies.

For the prone cohort, ventilator settings, ABG values and adverse events were collected for every single PPV session at three time points: 0–2 h before prone positioning (SPV1), 0–2 h after shifting to prone positioning (PPV) and 1–2 h after supervised repositioning (SPV2).

Efficacy analyses were performed using the full analysis set, and patients who received at least one PPV session were assigned to the prone cohort. To investigate the effect of PPV on oxygenation, we compared patient PaO2/FiO2 ratios between the SPV1 and PPV time points in the first PPV session and compared the PaO2/FiO2 ratio between cohorts during the first 3 ICU days. We compared the PaO2/FiO2 of the PPV versus SPV1 time points. The effects of prone positioning were classified into three degrees according to the change in the PaO2/FiO2 ratio between PPV and SPV1 as follows: type A represented conditions for which the PaO2/FiO2 ratio increased by over 15%, type B represented conditions for which the PaO2/FiO2 ratio changed between − 15% and + 15%, and type C represented conditions for which the PaO2/FiO2 ratio decreased by over 15%. Adverse events occurring in the first 3 ICU days in both cohorts were compared. We chose patients who received at least 3 consecutive PPV sessions in the prone cohort and all patients from the supine cohort to compare PaO2/FiO2 trends between cohorts, and PaO2/FiO2 values were gathered corresponding to the supine position after PPV. The analysis was performed with repeated-measures ANOVA and the ΔPaO2/FiO2 from D2 to D1 and D3 to D1 were compared by Student’s t test.

Continuous variables are expressed as means with standard deviations (SD) or quartiles (upper quartile Q1, median Q2, lower quartile Q3), according to the distribution, and categorical variables are described as percentages. Continuous variables were compared between groups with Student’s t test or Mann–Whitney U tests, based on the distribution, and categorical variables were compared by Chi-square test or Fisher’s exact test. Patient survival was analyzed by the Kaplan–Meier method and compared between groups with the use of the log-rank test. Factors associated with mortality were identified with Cox proportional hazards regression, and the results were expressed as hazard ratios (HRs) with 95% CI. The proportional hazard assumption was verified using the Schoenfeld test.

Missing data were appended by mean completer or regression completer according to their randomization. For patients who were transferred to other hospitals, we contacted their family members by phone to obtain information on their outcomes.

The statistical analysis was performed using R software (R for Mac, version 4.2.3) and Prism (for Mac, version 9.4.1). All reported P values were two-sided, and a P value of less than 0.05 was considered to indicate statistical significance.