The primary objective of this study was to characterize the correlation between ∆Pes and each of the following: Pocc, P0.1, PMI in mechanically ventilated children. Secondary objectives focused on finding the respective thresholds for Pocc, P0.1, and PMI that detect excessively high and low inspiratory effort.

We performed secondary analysis of physiologic data from children on pressure control (PC) or pressure support (PS) ventilation, enrolled in an ongoing randomized trial testing a lung and diaphragm protective ventilation strategy (REDvent, R01HL124666) (Clinical Trials. gov NCT03266016) that uses esophageal manometry at Children’s Hospital Los Angeles (CHLA) [22]. The protocol has been approved by the CHLA Institutional Review Board as well as an independent Data Safety and Monitoring Board. All patients were enrolled in the parent REDvent study and were between 1 month and 18 years of age, and met hypoxemia criteria for pediatric ARDS [23]. In all patients, informed consent was obtained. Detailed inclusion and exclusion criteria are provided in the supplement.

Physiologic waveforms of flow, esophageal pressure (Pes) and airway pressure (Paw) were recorded once daily. All patients were intubated with a cuffed endotracheal tube. For each day the patient remained intubated, a target of 3 inspiratory and 3 expiratory hold maneuvers were performed by using occlusion buttons on the ventilators. For analysis, all patients had to have evidence of spontaneous breathing, measured by negative deflection of the esophageal pressure waveform during inspiration. We used median values of the measures obtained from up to 3 inspiratory and expiratory occlusion maneuvers and the median ΔPes in up to 3 spontaneous PC and/or PS breaths at a time point near the occlusion maneuvers per test day. Data were selected for analysis at the waveform level, and inappropriate waveforms were excluded (Additional file 2). Ventilators included Servo I (Maquet, Solna, Sweden), NKV-550 (OrangeMed, Santa Ana, CA), or AVEA (CareFusion, Yorba Linda, CA) ventilator. All test days using the AVEA were later excluded because it was found that the AVEA ventilator does not allow airway pressure to rise above a set peak inspiratory pressure during an inspiratory hold, which invalidates PMI measurements. The elements of airway and esophageal pressure used for the analysis are described in Fig. 1.

Physiologic waveforms of airway pressure and esophageal pressure during the end-inspiratory (A) and end-expiratory hold (B). The elements of airway and esophageal pressure used for analysis are as follows. Peak pressure (Ppeak): the highest airway pressure before the start of an end-inspiratory hold. Plateau pressure (Pplat): the airway pressure that reached a plateau during an end-inspiratory hold. Respiratory muscle pressure index (PMI): Pplat minus Ppeak. Airway occlusion pressure (P0.1): the drop in airway pressure during expiratory occlusion from the beginning of the drop to 100 ms after the first drop in airway pressure. Expiratory occlusion pressure (Pocc): the difference from positive end-expiratory pressure (PEEP) to the lowest airway pressure during the first inspiratory cycle during an expiratory hold maneuver. Delta Pes (∆Pes): the difference from end-expiratory Pes to maximum negative Pes during non-occluded (PC or PS) breaths. A In airway occlusion at the end of inspiration, if the patient became relaxed during occlusion, Pplat is achieved. The difference between the plateau pressure and Ppeak is the PMI. B In airway occlusion at the end of expiration, the maximum negative pressure during the next spontaneous breath is Pocc, and the negative pressure 0.1 ms after the start of inspiration is P0.1

Airway pressure (Paw) was measured with a proximal sampling line placed just after the endotracheal tube, along with a self-calibrating pneumotachometer (Viasys Variflex 51,000–40094; Conshohocken, PA). One of the three esophageal catheters were used, based on the size of the patient (Carefusion, Avea SmartCath 6, 7, or 8 Fr). The amount of air inflated into esophageal balloon was determined before each measurement using a previously validated calibration algorithm [1] All sensors were connected to a custom-made hardware device (New Life Box, Applied Biosignals, Weener, Germany), which recorded data at a frequency of 200 Hz. The data were then post-processed with a custom-built software program for breath annotation using R (R Core Team, Vienna, Austria). All measurements of P0.1, Pocc or PMI were computed in the post-processing software.

For analyses evaluating data which were independent between groups, Kruskall-Wallis or Pearson's chi-square tests were performed with Bonferroni correction for multiple comparisons. For data that was not independent (i.e., measurements on multiple days from the same patient), we chose one observation per patient at random or used linear mixed models to first control for patient level effects. Similarly, repeated measures correlation was used with log or cubed transformation as necessary to satisfy assumptions of normality. Dose response relationships between ∆Pes and suggested cut-offs of Pocc, P0.1, and PMI are reported with box-plots [9, 21]. Subgroup analyses were performed based on mode of ventilation (Synchronized Intermittent Mandatory Ventilation Pressure Control with Pressure Support (SIMV PC-PS) or Pressure Support Ventilation (PSV)) and age (< 1 year, 1 year to < 9 years, >  = 9 years to <  = 18 years). Additional sensitivity analyses were performed based on adequacy of calibration with the esophageal catheter, using respiratory muscle pressure (Pmusc) instead of ∆Pes, and evaluating the relationship between patient effort on PC versus PS breaths when on SIMV (Additional file 2).

Several thresholds have been proposed using esophageal pressure to characterize high or low inspiratory effort [9]. For analysis, we evaluated three potential thresholds to characterize high effort (∆Pes > 12, 15, 20 cmH2O) and 2 thresholds to characterize low inspiratory effort (∆Pes < 3, 5 cmH2O). We developed regression equations for the relationship between ∆Pes and Pocc, P0.1, and PMI, respectively, and report the respective values for each parameter which correspond to the proposed high and low effort ∆Pes thresholds. We then describe the sensitivity and specificity of each of the values to detect high and low effort [9].

In addition, the following data were collected on all test days: respiratory mechanics (resistance, peak flow, tidal volume, static respiratory system compliance (CRS), chest wall compliance (CCW), lung elastance (EL)/ respiratory system elastance (ERS)); respiratory drive (pH, pCO2, respiratory rate (actual RR), State Behavioral Scale (SBS) [24], Pain scores (FLACC) [25]), cumulative opioid/benzodiazepine dose per day; and ventilator settings(ventilator rate (vent RR), positive end-expiratory pressure (PEEP) (Additional file 2)).

Analyses were performed with EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan) and R (R Core Team, Vienna, Austria). Additional details on the materials and methods are provided in the Additional file 2.