The main finding of this prospective clinical cross-over study was that adjustment of PEEP and VT according to the EIT-based protocol led to individualized ventilator settings with improved oxygenation and reduced alveolar cycling without causing excessive lung stress and strain. Global lung stress remained below 27 mbar in all patients, while release-derived strain was below 2.0 in 19 out of 20 patients. We chose 27 mbar as upper threshold for stress and 2.0 as upper threshold for strain because these values represent the upper limit of the physiological range postulated for human patients in previous publications [2, 20].
In one patient, we found an unphysiologically high value of release-derived strain of 3.4 after adjusting mechanical ventilation with the EIT-based protocol. However, this patient had relatively moderate Paw,plat of 29 mbar and lung stress of 21 mbar. Recruitment-adjusted strain amounted to 1.5, which is still within the physiological range. These findings support the assumption that the unphysiologically high strain observed in this patient was largely due to derecruitment during the release maneuver that was performed for calculating strain and FRC from the measured end-expiratory lung volume (EELV).
For assessing global lung stress in this study, we used the concept of elastance-based transpulmonary pressure which is based on the assumption that at zero airway pressure, transpulmonary pressure also equals zero and that transpulmonary pressure can therefore be calculated using the ratio between the elastance of the lung and the respiratory system multiplied with airway plateau pressure. Obviously, this assumption may not always be valid, especially in patients with increased lung weight and ARDS. Indeed, it has been found that elastance-based methods for estimating transpulmonary pressure yield very different results when compared to the more widely applied method that uses absolute values of esophageal pressure for directly calculating transpulmonary pressure . However, the assumption of transpulmonary pressure close to zero at zero airway pressure may be an acceptably accurate approximation for the non-dependent lung regions , where overdistension primarily occurs. Therefore, our results of stress < 27 mbar in all patients indicate that adjusting PEEP and VT according to our EIT-based protocol did not lead to overinflation of non-dependent lung areas.
Furthermore, we found significant improvements in oxygenation and lower values of SDRVD consistent with a reduction in alveolar cycling. The average value of absolute end-expiratory transpulmonary pressure was negative before optimizing ventilation according to the EIT-based approach but became positive after 4 hours of its use. We found no changes in cardiac output and vasopressor dosing, indicating that there was no relevant hemodynamic compromise despite the higher PEEP levels applied after EIT-based optimization.
Our protocol for prospective optimization of PEEP and VT with EIT was largely based on bedside assessment of changes in regional Crs. An increase in regional Crs following a sustained-inflation recruitment maneuver was interpreted as indicator for alveolar recruitment and a higher PEEP level was selected to keep the recruited lung volume. With this approach, we identified alveolar recruitment in 15 out of 20 patients after the initial recruitment maneuver, indicating a high potential for recruitment in a large proportion of patients studied.
If regional Crs decreased during a brief reduction in VT, this was interpreted as an indicator for regional alveolar cycling and PEEP was increased by a further 3 mbar to counteract this phenomenon. On the opposite, an increase in regional Crs during a brief reduction in VT was interpreted as overdistension with the previously applied VT. In this case, the therapeutic consequence was to decrease VT, provided this did not lead to severe acidosis. In a surprisingly large proportion of patients, we still identified regional overdistension with a VT of 6 ml/kg PBW. This EIT finding led to further reductions of VT to levels below 6 ml/kg PBW in 10 out of 20 patients.
Our Crs-based approach differs from the approach that was employed for prospective optimization of PEEP with EIT in a study published by Eronia and Coworkers . In this study, the time-course of end-expiratory lung impedance was analyzed for determining changes in EELV associated with PEEP. A slow decrease in end-expiratory lung impedance following a recruitment maneuver was interpreted as derecruitment, and PEEP was increased to counteract this phenomenon until a stable level of end-expiratory impedance was achieved. As end-expiratory lung impedance appears to be a reasonably accurate measure for changes in EELV , this approach allows straightforward bedside assessment of recruitment and derecruitment. However, it is highly susceptible to artifacts: For instance, the pulsation therapy with inflatable mattresses can cause substantial artifacts in end-expiratory impedance ; the same applies to changes in torso and arm position  and even intravenous fluid therapy, which is a rather common intervention in ICU patients [26, 27]. These interferences render EIT-based analyses of EELV difficult to interpret and error-prone. Moreover, while observing end-expiratory lung impedance may facilitate bedside assessment of recruitment and derecruitment, it provides no information on regional overdistension, which can be easily identified by analyzing regional changes in Crs [11, 12].
Another Crs-based approach that has been applied for EIT-based optimization of PEEP in patients with ARDS  and with Covid-19 induced acute respiratory failure [29, 30] relies on analyzing pixel-wise changes in Crs during a decremental PEEP-trial [12, 31]. The main disadvantage of this approach is that it requires a decremental PEEP trial that must be started at relatively high PEEP levels that may be associated with overdistension while applied. The PEEP trial must then be carried on until very low PEEP levels (that may lead to alveolar collapse and atelectasis formation) are reached. Therefore, it cannot be repeated on a regular basis to adapt ventilator settings to the changing conditions of a patient’s lung. In contrast, our approach is primarily based on brief changes in VT for diagnosing regional overdistension and alveolar cycling, and can thus be repeated whenever a new assessment of these phenomena is clinically required.
Our study has several limitations. First, it was not a randomized study. Therefore, we can make no assumptions on whether individualized optimization of mechanical ventilation with our EIT-based protocol has an influence on actual clinical outcomes. Instead, we tried to carefully monitor and describe the physiologic effects of individualized adjustment of ventilator settings using the EIT-based protocol by analyzing changes in transpulmonary pressure and EELV.
Also, we did not directly measure recruitment and alveolar cycling with a reference method like CT. Our measurements of EELV using the modified nitrogen dilution technique  indicate that recruitment-adjusted FRC that was derived by subtracting the assumed PEEP volume from EELV, was significantly increased. The ventilation delay index SDRVD, that was not used for optimization of ventilation but served as a secondary outcome parameter, indicated a possible reduction in alveolar cycling with our EIT-based optimization. Nevertheless, independent reference methods like CT might be necessary to confirm a reduction in alveolar cycling following optimization of mechanical ventilation with the EIT-based protocol.
The majority of patients included in this study presented with moderate ARDS. Following a meta-analysis by Briel and coworkers , recent recommendations suggest using a higher PEEP / FiO2 strategy for patients with moderate to severe ARDS . Nevertheless, comparatively low levels of PEEP in patients with moderate to severe ARDS are still common clinical practice in many centers around the world .
Valid assessment of Crs, which is a prerequisite for our EIT-based approach for individualization of PEEP and VT, typically requires a paralyzed patient, even though it is possible to assess Crs and ΔPaw in many patients on assisted ventilation using an inspiratory-hold maneuver [35, 36]. The patients included in our study exhibited no spontaneous breathing activity. It is uncertain whether a Crs-based approach using inspiratory hold maneuvers during assisted modes of ventilation will yield similar results in patients with spontaneous breathing activity.
In conclusion, we presented a protocol for prospective adaption of PEEP and VT taking into account EIT-derived information on recruitability, overdistension and alveolar cycling. Mechanical ventilation adjusted according to the EIT protocol resulted in global values of lung stress and strain within the physiological limits and was associated with improvements in oxygenation and a reduction in regional ventilation delay inhomogeneity.