This pilot physiological study was conducted in two tertiary university hospitals in France: one pediatric intensive care unit (Necker universitary hospital, Paris) and one adult intensive care unit (Ambroise Paré universitary hospital, Boulogne-Billancourt). The study was approved by an institutional review board (CPP Sud-Ouest et Outre-Mer 3-n°2019-A02814-53) and registered on the clinicaltrials database (NCT 04184674). Written informed consent from the patient or their guardians was obtained before inclusion.

Patients suffering from new-onset (within 48 h following admission) ARDS [11, 12] were included if invasively ventilated. Exclusion criteria comprised: neonates ≤ 37 weeks of corrected gestational age and/or less than 72 h of age, pregnant women, contra-indication to esophageal catheter, extracorporeal membrane oxygenation, congenital heart defect affecting RV function, and lack of social coverage.

Patients received ventilation using cuffed endotracheal tubes connected to an ICU ventilator (V500 or Evita XL, Draeger, Netherlands, for children. Hamilton S1, USA, for adults). The ventilator settings and adjuvant therapies were left at the physician’s discretion, following guidelines for lung-protective ventilation [11, 13, 14]. Briefly, the ventilation strategies used in both units were defined to deliver a tidal volume around 6 mL/kg of predicted body weight, with a maximal airway plateau pressure at 28–30 cmH2O, and a minimal positive end-expiratory pressure (PEEP) level set at 5 cmH2O, then adapted depending on severity and respiratory mechanics. Patients could be ventilated in volumetric or barometric mode, although volume-(assist) control ventilation was prioritized in the two centers. Esophageal pressure monitoring was not considered for ventilator settings modifications. Critical care echocardiography (CCE), followed by static respiratory mechanics measurements were performed at day one, then, in case of PEEP level modification, up to 3 days.

CCE (transthoracic echocardiography for all, followed by transesophageal echography in adults) was performed by the same investigator along the procedure, according to a standard protocol [15]. RVI was defined as either acute cor pulmonale (ACP) or RV systolic dysfunction (RVD). The former was defined as a dilated RV (RV/left ventricle (LV) end-diastolic area ratio > 0.6) associated with a septal dyskinesia, [16] and the latter as a fractional area change (FAC) < 35%, and/or a tricuspid annular plane systolic excursion (TAPSE) < 16 mm in adults or a z-score < − 2 in children [17], and/or a S wave velocity < 10 cm/s [18]. Each CCE could be classified as ACP only, RVD only, or both ACP and RVD. All the ultrasound images and loops were stored and exported to dedicated software (Echopac™, version 201, GE Healthcare Systems, Chicago, USA). A blinded analysis was performed off-line (Echopac) by two cardiopediatricians for children, and two intensivists for adults, all having expertise in RV assessment.

PES was measured using an esophageal balloon catheter (Marquat, France, for children; Nutrivent, Italy, for adults) inserted as recommended [19, 20], and filled with the appropriate volume of air [21, 22]. Airflow and airway pressure (PAW) were measured using a proximal calibrated pneumotachometer (Hans Rudolph, USA, for children; Hamilton, USA, for adults). All signals were displayed on the ventilator screen in adults, and recorded at 200 Hz using an analogical/numeric acquisition system (MP150, Biopac systems, USA) run on a PC computer and displayed using dedicated software (AcqKnowledge, version 4.2, Biopac systems) in children. Tidal volume was obtained by integrating the airflow signal over time.

Gas leaks were carefully excluded before measurements. After one minute of stable breathing, the investigator performed 3 consecutive inspiratory and expiratory holds. Airway and esophageal plateau pressures were obtained at the end of the inspiratory hold, while airway and esophageal PEEP were obtained at the end of a prolonged expiratory hold (4 s in adults, 3 in children) to account for intrinsic PEEP [23]. PL was calculated online by subtracting PES from PAW. The subsequent transpulmonary driving pressure (ΔPL) and compliances of the respiratory system (CRS), lung (CL), and chest wall (CCW) were calculated and normalized on actual (children) or predicted (adults) body weight. Expiratory-PL (PL-EXP) was computed directly from the measured end-expiratory PL, while inspiratory-PL (PL-INSP) was computed using the elastance ratio method [24].

CCE data were first described at T0 at the patient level according to the presence or not of RVI, in adults, and children. Each measure of respiratory mechanics was coupled to simultaneous CCE to assess the impact of PL on RVI using mixed-effect logistic regression models (one for ACP, one for RVD). For each model, the patient was treated as a random effect, to adjust for the correlation between repeated measures within the same patient. Based on the ACP prevalence in adults, and the physiological differences between children and adults, we anticipated RVI prevalence of 30% in the total population. Assuming the inclusion of 45 patients with at least 2 measurements, 4 variables could be included in the models. Variables of interest were selected a priori based on the physiologic rationale and the current literature evidence regarding ARDS-associated RVI, with variables exhibiting collinearity excluded. Four variables of interest were finally included: ΔPL, partial pressure of carbon dioxide (pCO2), lung injury severity, and age [1, 25, 26]. ΔPL was used as a continuous variable as a surrogate for the tidal lung stress. PCO2 was measured using arterial blood gas in adult patients, and in either arterial or venous blood gas in children. Because of the differences between pediatric and adult guidelines regarding the estimation of hypoxemia and lung injury severity, we considered the level of the arterial partial pressure of oxygen to fraction of inspired oxygen ratio (PaO2/FiO2) and the oxygenation index (OI) in children. OI was either measured or calculated from the oxygen saturation index (OSI) (using the equation OI = 0.0745 + 1.7830 * OSI [27]) in the absence of arterial blood gas. Lung-injury severity was then categorized in mild/moderate or severe ARDS following the recommended thresholds of PaO2/FiO2 (severe ARDS if PaO2/FiO2 ≤ 100) or OI (severe ARDS if OI ≥ 16) [12, 28]. This transformation of the variable allowed to minimize the number of missing data, and to achieve the greatest homogeneity in the population. Because of a highly skewed distribution and to distinguish adults and children, age was also categorized in two classes: adults (≥ 18 years old) and children (< 18 years old). Because echocardiographic assessment of RV dilation might be uncertain in infants [29], children less than 1 year old were excluded from all the analyses using the ACP definition.

Respiratory mechanics were then compared in CCE studies with or without RVI, in adults and children separately, using the t-test or Wilcoxon rank sum test depending on the variables’ normality. Continuous variables were presented as means ± standard deviations (SD) or medians and interquartile ranges (IQR) depending on the sample size and their distribution, and categorical variables as numbers and percentages. We also described the relationship between ΔPL and driving airway pressure (ΔPAW) using linear regression model. Data were analyzed using R programming software. A p value < 0.05 defined statistical significance.