Introduction

In acute respiratory distress syndrome (ARDS) management, positive end-expiratory pressure (PEEP) counteracts gravity-dependent alveolar collapse, decreasing shunt and hypoxemia [1], reduces the shearing forces of cyclic alveolar opening/closing, and increases compliance [2]. Due to the heterogeneity of lung injury in ARDS, the application of PEEP can result in recruitment in some lung areas while causing overdistension in others. Suboptimal PEEP may induce ventilator-induced lung injury (VILI) [1].

Randomized controlled trials (RCTs) comparing high vs. low PEEP strategies have not consistently demonstrated the superiority of either [3]. While a network meta-analysis of 18 RCTs suggested a potential mortality benefit of higher PEEP [4], this cumulative analysis failed to consider the impact of individualized PEEP titration and the adverse effects of high PEEP on non-PEEP responders [5].

Electric impedance tomography (EIT) is a bedside imaging technique that identifies changes in lung impedance, a proxy for lung volume [6]. EIT-guided PEEP titration distinguishes PEEP-induced recruitment from overdistension [7,8,9]. Hse et al. demonstrated increased survival with EIT-guided PEEP titration, albeit with higher use of ECMO in the EIT group [10]. Another RCT failed to demonstrate such benefit [11].

Mechanical power (MP) is a physiological construct of the energy transmitted to the patient during invasive mechanical ventilation (IMV). MP integrates the major components of positive pressure ventilation that drive VILI [12]: elastic–static (related to PEEP), elastic–dynamic (related to driving pressure, [ΔP]), and resistive (related to flow and airway resistance). High MP is associated with ARDS mortality [13]. Given the conflicting data regarding the utility of EIT and the need for feasible surrogate endpoints to guide larger multicenter RCT, we performed a randomized crossover trial to explore the effects of EIT-guided PEEP titration on MP in patients with ARDS. We hypothesized that EIT-guided PEEP titration would result in lower MP, compared to the use of the High-PEEP/FiO2 table.

Materials and methods

Study design and population

In this single-center randomized crossover trial, we compared EIT-guided PEEP selection vs. High-PEEP/FiO2 tables (NCT 03793842). The University of Michigan Institutional Review Board approved this study (HUM00148126). We obtained informed consent from each patient’s legal representative. We included patients ≥ 18 years receiving IMV for ARDS management for < 72 h with a PaO2/FiO2 ratio < 150 and a PEEP > 8 cm/H2O. Exclusion criteria are provided in Additional file 1.

Study protocol

Patients were randomly assigned 1:1 to receive EIT-guided PEEP titration followed by PEEP selection via the High-PEEP/FiO2 table (EIT first) or vice versa (tables first). Randomization was achieved using opaque, sealed envelopes. At randomization, all patients received lung-protective ventilation (LPV). PEEP was selected according to High-PEEP/FiO2 tables. Due to the need to maintain PEEP, a washout period was not feasible. Patients who were proned after randomization were excluded from the analysis due to its effects on PEEP requirement and MP [14].

During the EIT-guided PEEP titration phase, patients underwent a recruitment maneuver. PEEP was then decreased by two cmH2O every 5–10 min until a 10% drop in delta end-expiratory lung impedance in dorsal regions was detected by EIT, five cmH2O PEEP was reached, or hemodynamic instability/hypoxemia developed. PEEP was then set based on the intercept between the lower overdistension and collapse percentages [15].

Patients randomized to the EIT-first group underwent EIT-guided PEEP titration, as above. This was followed by six hours of management per the University of Michigan ARDS protocol (Additional file 1), with PEEP left at the EIT-determined level. Afterward, patients crossed over to a PEEP level set using the High-PEEP/FiO2 tables, which was maintained for 14–18 h. Patients randomized to the tables-first group underwent the same interventions in the reverse order (Fig. 1). After both interventions, FiO2 and respiratory rate (RR) were adjusted for oxygenation > 90% and a pH 7.3–7.45, respectively.

Fig. 1
figure 1

Trial design and crossover. Red arrows represent the time point at which post-intervention data were collected. Center crossed arrows represent the time of crossover

Full size image

Outcomes

The primary outcome of this study was the change in MP after each PEEP selection strategy. Secondary outcomes included changes in the 4ΔP + RR index (an estimate of ventilator energy transfer to the lung) [13], elastic–static, elastic–dynamic, and resistive powers, as well as ΔP, plateau pressure (Pplat), PaO2/FiO2 ratio, and static compliance (Cstat). We calculated MP with Gattinoni’s simplified formula: 0.098 X RR X TV (Ppeak-[Pplat-PEEP/2]) [12] and analyzed its components.

Statistical analysis

We compared baseline characteristics using Fisher’s test for categorical and a two-sample t test for continuous variables. Changes in ventilator parameters with each intervention were compared using paired t tests. We fit serial linear mixed-effect regression models assessing the association between the interventions and change in MP, adjusting for randomization order and pre-intervention MP in sequential models. We repeated this to determine the association between intervention and the 4ΔP + RR index, MP components, and ΔP. Our small sample size represents a convenience sample similar in scope to other EIT studies. Statistical analyses were performed in StataMP version 17.0 (StataCorp).

Results

Sixteen patients were enrolled in this study. One patient was withdrawn due to hemodynamic instability; three were proned after randomization and excluded from the analysis (Additional file 2: Fig S1). Baseline characteristics are shown in Table 1. The median baseline MP was 20 J/min (IQR: 19, 28). EIT led to a significant change in PEEP compared to tables (mean difference of change: − 2 cmH2O, 95% CI − 3.95, − 0.05, p = 0.046),

Table 1 Baseline characteristics of participants across study arms
Full size table

EIT decreased MP compared to PEEP/FiO2 tables (− 4.36 J/min, 95% CI − 6.7, − 1.95, p = 0.002). (Table 2). This difference persisted after adjusting for randomization order and pre-intervention MP. (Additional file 3: Tables S1-2). EIT led to a decrease in the 4ΔP + RR index (− 11.42 J/min, 95% CI − 19.01, − 3.82, p = 0.007) mainly through a decrease in elastic–dynamic power (− 1.61 J/min, 95% CI: − 2.99, − 0.22, p = 0.027), and driving pressure (− 2.92 J/min, 95% CI: − 4.59, − 1.23, p = 0.003) (Table 2 and Fig. 2). These differences persisted after adjusting for randomization order and baseline MP (Additional file 3: Table S3). Elastic–static and resistive powers were unchanged across both interventions.

Table 2 Comparison of changes in ventilator parameters with EIT vs tables, for all participants, n = 12
Full size table
Fig. 2
figure 2

Changes in mechanical power and its components after each intervention. Changes in PEEP (A and B), MP by the Gattinoni’s Simplified formula (C and D), 4∆PxRR index (E and F), elastic–dynamic power (G and H), elastic–static power (I and J), and resistive power (K and L). Asterisk indicates a statistically significant difference in the change with the EIT versus High-PEEP tables interventions based on p value < 0.05

Full size image

EIT led to changes in peak pressures (− 6.25 cmH2O, p = 0.003), Pplat (− 4.53 cmH2O, p = 0.006), and Cstat (+ 7.93 ml/cmH2O, p = 0.008) (Additional file 4: Fig S2). There was no significant change in RR or PaO2/FiO2 ratio.

After the EIT phase, one patient developed pneumomediastinum, which did not require additional intervention. Three patients developed hypotension during the RM. In one patient, the protocol was stopped due to persistent hemodynamic instability.

Discussion

In this randomized crossover trial, we found a significant decrease in MP using EIT-guided PEEP titration compared to High-PEEP/FiO2 tables in mechanically ventilated patients with moderate–severe ARDS. This difference persisted after sensitivity analysis and adjustment for randomization order and pre-intervention MP. A reduction in the elastic–dynamic MP mainly drove the decrease in MP.

Zhao and colleagues reported that EIT-guided PEEP titration was associated with improved respiratory mechanics [15]. Similarly, a RCT by Hsu and colleagues reported improvement in ΔP, Cstat, and survival rates with EIT-guided PEEP titration compared to the pressure–volume curves through a decrease in PEEP [11].

He and colleagues compared the effects of EIT-guided PEEP titration vs. a low PEEP/FiO2 table [10] without finding differences in survival, ventilator-free days, or ICU stay. However, this study was limited by using similar PEEP between groups and including mild ARDS. In our study, EIT-guided PEEP titration led to significant changes in PEEP, and we only enrolled patients with moderate–severe ARDS. Using a crossover design allowed us to analyze the effects of each intervention on an individual level by using each patient as their own control. Using the High-PEEP/FiO2 tables as the control intervention permitted comparison with the strategy associated with better ventilation/perfusion matching [16] and outcomes in severe ARDS [4].

In patients with ARDS, persistent elevation of MP > 17 J/min is associated with higher mortality [13, 17]. Although patients in our study received standard LPV per protocol at baseline, MP was elevated (median 20 J/min). EIT-guided PEEP titration led to a mean reduction of MP by 4.36 J/min, a reduction associated with decreased mortality [17, 18], particularly when achieved during the initial hours of IMV [18]. The reduction in MP was achieved through decreased elastic–dynamic power, but not the elastic–static or resistive powers. Although changes in RR were allowed to achieve pH > 7.3, they were not different across groups.

Our findings are consistent with previous observations [13] that propose oscillating mechanical stresses as the main injurious mechanism for VILI. In our study, EIT decreased PEEP levels despite ∆Ps believed to be LPV, suggesting EIT-directed PEEP titration to be a more effective means of optimizing LPV than PEEP/FiO2 tables via a reduction in MP. In addition, observational studies have shown that lung recruitment is not systematically associated with detectable improvements in Cstat nor ∆P, therefore precluding accurate titration of PEEP based exclusively on these parameters [19, 20].

Our study has several limitations; 1) excluding subjects on prone positioning could have introduced post-randomization selection bias. 2) Sample was small, but the effects in MP reduction were significant and occurred despite optimal LPV at baseline. This suggests a strong effect of EIT in reducing MP. 3) We did not include a washout phase. However, our analysis considered the order of interventions to assess for carryover effects. 4) We did not assess for recruitability before the recruitment maneuver. This could have impacted sample’s enrichment. 5) Our intervention focused on titrating PEEP during the initial 24 h. However, PEEP/FiO2 tables are meant to guide continuous changes in PEEP based on FiO2 responses, dead space fraction, and mechanics. This was not assessed. 6) MP was calculated using airway not transpulmonary ΔP rather which could have introduced measurement bias.

Conclusion

This study shows that EIT-guided PEEP titration decreases MP in patients with moderate–severe ARDS compared to a high-PEEP/FiO2 table. A decrease in the dynamic–elastic component primarily drives the reduction in MP. The clinical impact of EIT-guided PEEP titration should be tested in large multicenter trials.