Based on the EIT-guided PEEP trial, PEEP was decreased in 31% of patients and increased in 32% of patients. We found a significant positive correlation between PEEPset and BMI. Patients in PEEPlower group had improved respiratory mechanics after the PEEP trial, a lower BMI, longer time between onset of symptoms and intubation, and a higher incidence of pulmonary embolism during ICU admission. In patients in the PEEPlower group, an increase in PEEP resulted in major alveolar overdistention and a small amount of recruitment on EIT. In PEEPhigher group, we observed a significant increase in plateau pressure and improved oxygenation after the PEEP trial. In addition, an increase in PEEP resulted in significant alveolar recruitment and small amounts of alveolar overdistention on EIT. Hence, the latter group should be considered as recruitable. The PEEP trial was relatively safe, as 5% of patients had a desaturation and 1% of patients was hypotensive during the PEEP trial.
PEEPset resulted in a trend toward improved respiratory mechanics in the PEEPlower group and improved oxygenation in the PEEPhigher group. Both an improved driving pressure and improved oxygenation after a change in PEEP are associated with reduced mortality rate in patients with ARDS . Therefore, we should aim to identify the patients that are likely to respond to PEEP, i.e., recruitability.
Recruitability is the amount of collapsed lung tissue that has the potential for reaeration at higher airway pressures . An increase in PEEP in the patients in PEEPlower group resulted in major alveolar overdistention and a small amount of alveolar recruitment, whereas the patients in PEEPhigher group had significant alveolar recruitment and less alveolar overdistention. In patients with COVID-19-related ARDS, alveolar recruitment does not necessarily result in an increase in static compliance . Thus, patients in PEEPlower group were considered to have low recruitability, patients in PEEPequal group had intermediate recruitability and patients in PEEPhigher group had high recruitability.
Until now, we tended to focus on the identification of patients that had high recruitability . However, it might also be beneficial to identify the patients that have low recruitability and are prone to alveolar overdistention. Patients with low recruitability had a lower BMI, a higher incidence of pulmonary embolism, and a longer time between onset of symptoms and intubation. Patients with obesity have lower transpulmonary pressures and lower end-expiratory long volumes as a result of higher pressure from the chest wall . BMI has a positive correlation with recruitability and the use of higher PEEP, as higher PEEP increases transpulmonary pressure and reduces alveolar collapse . In addition, patients in PEEPlower group had a higher incidence of pulmonary embolism during ICU admission. These findings suggest that disturbed pulmonary perfusion, resulting in a ventilation-perfusion mismatch, caused hypoxemia in these patients. Nevertheless, all patients had a reduced static compliance, possibly leading to disturbed minute ventilation or increased dead space fraction as well . Patients in PEEPlower group had a longer time between onset of symptoms and intubation. This could indicate that these patients may have had some form of patient self-inflicted lung injury or pulmonary fibrosis . Unfortunately, we had too few CT scans at the day of PEEP titration to test this hypothesis. The PEEP trial did not reach a maximum PEEP of 24 cmH2O in four (5.3%) patients because of desaturation. These four patients were assigned to the PEEPlower group and had large amounts of alveolar overdistention. Desaturation at high PEEP could be a clear indication of ventilation-perfusion mismatch, likely due to alveolar overdistention.
An observational cohort performed in the Netherlands found a median PEEP titrated by the clinician of 14.0 cmH2O (11.0–15.0) . Two small observational cohorts that used EIT to titrate PEEP found a median PEEP of 12.0 cmH2O [28, 29]. In our EIT-guided population, we found a higher median PEEP of 18.0 cmH2O (14.0–20.0) as compared to the other studies. Explanations are the relatively high BMI in our cohort and long duration of mechanical ventilation in the cohort of Sella et al. . In addition, there is no consensus on how to interpret EIT data obtained during a PEEP trial [14, 30].
In our study, total PEEP was arbitrarily set at the PEEP level above the intersection of the curves representing relative alveolar overdistention and collapse [18, 19]. We chose this method as it is an intuitive and simple approach that can be performed at the bedside, but arguably assumes that both alveolar overdistention and collapse are equally harmful . Both Perrier et al.  and Sella et al.  chose to set PEEP at the intersection of both curves itself, whereas Franchineau et al.  chose to limit alveolar collapse at 15%, independent of alveolar overdistention. The last approach favors alveolar collapse over alveolar overdistention and likely resulted in a lower set PEEP as compared to the method used in this study. Future research should focus on the best approach to titrate PEEP based on EIT data and its association with clinical outcomes.
Previous randomized controlled trials in patients with ARDS compared PEEP titrated using EIT to conventional methods. In patients with mild-to-severe ARDS, He et al.  showed EIT resulted in a similar PEEP compared to the PEEP/FiO2 table, but was decoupled from FiO2. In patients with moderate-to-severe ARDS, Hsu et al.  showed PEEP and mortality rate were lower using EIT compared to pressure–volume loops, but mortality rate was high overall (31% in de EIT group and 56% in the control group, versus 21–27% in the study by He et al.  and 29% in the current study). In our study, PEEP was not changed on average for the entire cohort after titration using EIT, but was changed with ≥ 2 cmH2O in the majority of patients.
This study has several limitations. First, this retrospective analysis was not prespecified in the study protocol and results should be considered hypothesis-generating. The main purpose of this EIT-guided PEEP trial protocol was to improve clinical practice. As a consequence, mechanical ventilation parameters were only recorded at PEEPbase and PEEPset, limiting a more accurate retrospective analysis of the PEEP trials and EIT data at every PEEP step. A major limitation of this study is the lack of randomization and of the sequence of interventions. All patients received PEEP set by the clinician using the PEEP-FiO2 table first and then the EIT-guided PEEP trial. A part of the improvements in oxygenation and respiratory mechanics may be due to the PEEP trial itself, instead of the titration of PEEPset. This is noticeable in the changes in respiratory mechanics for the PEEPequal group. Second, only patients with COVID-19-related ARDS were included in this study. Although respiratory mechanics in non-COVID-19-related ARDS and typical ARDS seem to be similar, it is uncertain whether results can be generalized to the non-COVID-19-related ARDS population [8, 9]. Third, maximum and minimum PEEP reached in all trials varied. The estimation of the amount of collapse and overdistention is based on the maximum compliance for each EIT pixel. It is probable or even likely maximum compliance is not reached for all pixels, e.g., due to residual collapse in the dependent lung at the highest PEEP level. Therefore, approximately 0% alveolar collapse at PEEP 24 cmH2O does not necessarily mean that application of higher or prolonged airway pressures cannot result in additional alveolar recruitment. Fourth, we performed a PEEP trial with small steps of 2 cmH2O and a short step duration of 30 s. Some other studies report larger steps and longer duration for similar PEEP trials [15, 28, 32]. There is a tradeoff between step size, step duration and the time it takes to complete the protocol. After a change in PEEP, respiratory mechanics can change in multiple ways with different time frames. By rapidly changing PEEP, we did not allow for slow effects like slow derecruitment, morphological changes to the abdomen and diaphragm, changing hemodynamics and changes in pO2 and pCO2. In addition, as a result of the large numbers of patients with COVID-19, we chose a time-efficient study protocol. Fifth, hemodynamic monitoring was limited to continuous measurement of blood pressure and heart rate. PEEP titration is more than balancing alveolar overdistention and collapse, as PEEP influences cardiac output as well . Although the PEEP trials had limited effects on blood pressure and heart rate, we cannot exclude a decrease in cardiac output. In addition, we did not assess pulmonary perfusion with EIT. Hence, EIT-guided PEEP titration might have resulted in optimal ventilation, but not necessarily in an optimal ventilation-perfusion match. Sixth, ventilation distribution assessed by EIT is measured in only a small cross-sectional slice of the lung. Ventilation distribution changes when the EIT belt is placed more cranially or caudally, further complicating EIT-guided PEEP titration . Seventh, we used devices from two manufacturers to perform the EIT measurements. Although the devices apply the same algorithm by Costa et al.  to derive the relative collapse and overdistention, results could vary due to differences in belts, reconstruction models and algorithms. Additional file 1: Tables S5–12 in the supplementary materials show the results presented in Tables 1, 2, 3, 4 split by EIT device. Considering the limited data, it seems possible the Timpel Enlight 1800 gives higher values overdistention at high PEEP compared to the Dräger Pulmovista 500. Due to the small amount of measurements with the Timpel device (n = 7), we were not able to properly compare the devices. Overall, considering only the measurements with the Dräger device (n = 68) does not change our interpretation or conclusions.
In conclusion, a PEEP trial guided by EIT as compared to PEEP titration based on the PEEP-FiO2 table resulted in a clinically relevant change in PEEP in 63% of patients with COVID-19-related ARDS. We found a significant positive correlation between set PEEP and BMI. Patients in whom PEEP was decreased had a lower BMI, a longer time between onset of symptoms and intubation, and a higher incidence of pulmonary embolism. Our results support the hypothesis that PEEP should be personalized in patients with COVID-19-related ARDS in order to reduce the total amount of alveolar overdistention and collapse, i.e., too low or too high PEEP.