Driving pressure during proportional assist ventilation: an observational study

The driving pressure of respiratory system (ΔP) during passive mechanical ventilation is defined as the difference between static end-inspiratory plateau pressure (Pplat) and static positive end-expiratory pressure (PEEP) and equals the ratio of tidal volume (V_T) to respiratory system compliance (Crs). Therefore, ΔP reflects the extent of lung stretch at end inspiration better than V_T alone (when V_T is set), because it takes into account patient’s respiratory system compliance. Despite the fact that ΔP represents a global measurement of lung stretch and thus cannot capture lung inhomogeneity, recent studies have shown that ΔP is a main determinant of ventilator-induced lung injury (VILI), and it is associated with mortality in ARDS patients, particularly at ΔP values above 14 cmH₂O [1–6]. In addition, even in patients with uninjured lungs, an association between high ΔP and increased morbidity has been postulated [4, 7, 8].

Although ΔP as a risk factor for VILI has been exclusively studied in patients under controlled mechanical ventilation, the potentially harmful effects of high ΔP are probably present in any mode of ventilation. Recently, the concept of self-inflicted lung injury has been introduced, referring to patients in assisted ventilation [9–11]. As the beneficial effects of spontaneous breathing during mechanical ventilation are well established, it becomes increasingly important to identify patients at risk of self-inflicted lung injury during assisted ventilation [9–11]. To identify patients at risk and prevent self-inflicted lung injury, monitoring of ΔP during assisted ventilation might be helpful. Experimental data indicate that during assisted ventilation vigorous spontaneous efforts increase transpulmonary driving pressure and worsen lung injury [12, 13]. However, limited information is available on the presence of high ΔP in patients ventilated in assisted modes, mainly because measuring ΔP requires valid estimation of Crs, a complicated task with conventional assisted modes of ventilation such as volume assist and pressure support [14].

In a recent study [15], we have reported data on ΔP obtained using proportional assist ventilation with load-adjustable gain factors (PAV+), a mode validated to measure end-inspiratory quasi-static Pplat, and compute Crs [16–19]. In this study, using single measurements of ΔP obtained when patients were switched from controlled ventilation to PAV+, we found that ΔP was mostly below 15 cmH₂O, while V_T was usually higher than that set during controlled ventilation [15]. Nevertheless, because in spontaneously breathing patients there is considerable variability in breathing patterns, prolonged and continuous measurements of ΔP would be required to fully capture the spectrum of ΔP during assisted ventilation.

In the current study, we described ΔP over time, aiming to explore whether and when high ΔP occurs in everyday clinical practice in patients placed in assisted ventilation, using continuous measurements of ΔP obtained in PAV+ mode. We hypothesized that sustained high ΔP (≥ 15 cmH₂O), and hence increased risk of injury, would be present during periods of relative hyperventilation, when tidal volume and minute ventilation would be high, and/or during periods when Crs would be relatively low, and sought to identify potential safe thresholds for V_T and/or Crs.

This observational study reports, for the first time to our knowledge, continuous and prolonged measurements of driving pressure in everyday clinical practice in critically ill patients during proportional assist ventilation. The main findings of our study are: (1) For most of the analyzed time (95%), driving pressure and tidal volume were below 15 cmH₂O and 11 mL/kg, respectively. (2) The incidence of prolonged high driving pressure (≥ 15 cmH₂O) was 8%, and this was not associated with either very high tidal volume (mean 7.5 mL/kg, max. 9.5 mL/kg) or minute ventilation (mean 10 L/min, max. 13 L/min). (3) Independent of tidal volume, episodes of sustained high driving pressure were very unlikely to occur when respiratory system compliance was above 30 mL/cmH₂O.

Certain methodological issues of the study should be discussed first. To begin with, the measurement of ΔP relies on the measurement of compliance used by PAV+ software. Studies have shown that respiratory system mechanics, as measured with PAV+, are similar to those measured during passive mechanical ventilation using standard techniques [16, 17, 20]. Particularly, provided that the level of assist is greater than 20%, Paw measured at 0.3 s from the onset of end-inspiratory occlusion in PAV+ provides a reliable estimate of passive elastic recoil pressure at the corresponding V_T, independent of respiratory drive, making the calculation of Crs and ΔP during active breathing possible and accurate [16–20]. Secondly, driving pressure is the pressure dissipated against the elastic recoil of total respiratory system (ΔP = ΔP_{chest wall} plus ΔP_lung), while it is well known that the injurious effects of high ΔP are related to high transpulmonary driving pressure (ΔP_lung = end-inspiratory minus end-expiratory transpulmonary pressure) [13, 21–23]. Although in our study driving transpulmonary pressures were not measured, the ΔP must always be higher than ΔP_lung. As it has been shown that during passive mechanical ventilation a ΔP ≥ 15 cmH₂O can detect lung overstress with an acceptable accuracy [24], it follows that, during PAV+, a ΔP below 15 cmH₂O should be associated with low lung stress. Finally, the study entry criteria (estimated need for mechanical ventilation for at least 1 day after inclusion, and exclusion of patients requiring low levels of assist) resulted in a population of severely ill patients (APACHE-II score on admission 25). Most of the patients had ICU-acquired infections, and mild or moderate ARDS. Although patients were not formally identified as having difficult weaning, the prolonged duration of mechanical ventilation in this study group (median 18 days) should be acknowledged, emphasizing that the observed incidence of high ΔP is derived from a subset of critically ill patients with high severity scores and need for prolonged mechanical ventilation. Presumably, high driving pressure would be even rarer in patients with uncomplicated course and simple weaning.

In our previous study [15], 108 patients were switched from controlled mechanical ventilation to PAV+ and a median of eight measurements of ΔP per patient within 48 h of assisted mechanical ventilation was analyzed. These measurements showed that critically ill patients control their ΔP below 15 cmH₂O by sizing V_T to individual respiratory system compliance. This is achieved by appropriate feedback systems (reflex: Hering–Breuer and chemical: ventilatory response to CO₂). Indeed, it has been shown that, with proportional modes of support, these feedback mechanisms allow to maintain a safe range of tidal volume even at high assist [25, 26], since a decrease in patient effort through activation of chemical feedback and/or Hering reflex results in a proportional decrease in ventilator pressure. The current observational study, using continuous and prolonged measurements of ΔP, demonstrated that V_T and ΔP varied significantly over time. For brief periods of time (2.5 min), ΔP values ≥ 15 cmH₂O occurred in many patients, but prolonged periods of high ΔP were observed in only 8% of patients. Although in these patients the contribution of chest wall to ΔP is not known, their clinical characteristics indicate that high ΔP is likely associated with high transpulmonary driving pressure. One patient had cryptogenic organizing pneumonia (COP) and another four had primary ARDS or decompensated congestive heart failure, conditions that decrease lung compliance and thus increase the contribution of transpulmonary pressure to ΔP values. Patients who died in the ICU overall had more time with ΔP above 15 cmH₂O, yet, due to the small number of patients with prolonged high ΔP, no threshold of high ΔP duration associated with adverse outcome could be identified, and no causality could be established.

This study has some important clinical implications, which, however, should be evaluated in larger, randomized trials. Firstly, the incidence of high driving pressure, albeit small (8%), is not negligible, considering that these patients fulfilled criteria to be placed and maintained on assisted ventilation. Second, we showed that the presence of high ΔP was not associated with high tidal volume or high minute ventilation. The observed tidal volumes were in the range of 5–11 mL/kg, and not greater than 9.5 mL/kg during high ΔP periods. These results indicate that, in patients meeting criteria for assisted ventilation, the control of breathing mechanisms, chemical and reflex feedback mechanisms [27–29], often allows V_T to be higher than the recommended ‘safe’ range of 6–8 mL/kg. More importantly, high ΔP was strictly associated with low compliance; a threshold of 30 mL/cmH₂O was identified, above which high ΔP is very rare. Additionally, ΔP was always high when compliance was below 20 mL/cmH₂O and in half of the cases when compliance was below 25 mL/cmH₂O. Therefore, provided that with conventional modes of support (assist volume control or pressure support) assist is not excessive, high ΔP is very unlikely to occur when respiratory system compliance is above 30 mL/cmH₂O, even if V_T is higher than 8 mL/kg. On the other hand, while proportional modes such as PAV+ or neurally adjusted ventilatory assist (NAVA), are expected to provide a more protective ventilation [30], by allowing the operation of chemical and reflex feedback mechanisms [27–29, 31], this study indicates that when compliance is below 30 mL/cmH₂O the protective mechanisms of control of breathing system may be overridden. Additionally, experimental and clinical data indicate that vigorous inspiratory efforts may promote lung injury, especially in the presence of severe underlying lung injury [9, 12, 13, 32]. Taken together, these findings suggest that when patients with lung injury and compliance below 30 mL/cmH₂O are ventilated in assisted modes, they are at risk of developing high driving pressure, and physicians should consider monitoring driving or transpulmonary pressures.

This study has certain limitations that should be considered. The study included a group of patients with high disease severity scores, and prolonged mechanical ventilation, from a single center. Patients were studied whenever the primary physician placed them on PAV+, and not specifically when first placed in assisted mode. Moreover, chest wall mechanics were not evaluated, and thus, in some patients, high ΔP may not correspond to high transpulmonary pressure, due to low chest wall compliance. The driving pressure could also be overestimated in the presence of PEEPi (as PEEPi was not included in the calculation of compliance). In the population studied, the median PEEPi was low (0.3 cmH₂O), and results were qualitatively the same when PEEPi was included in calculations. Most patients included in the study, as well as most patients admitted in the ICU, were overweight or obese, and patients with prolonged high driving pressure had even higher BMI. Finally, this study does not establish a causative relationship between high ΔP and mortality, but indicates that, given the small incidence of prolonged high ΔP identified, a very large study would be required to investigate this. Yet, this study identifies for the first time a safety threshold for respiratory system compliance during assisted ventilation at 30 mL/cmH₂O, below which high driving pressures are more likely to occur. However, the clinical significance of this finding should be prospectively investigated.

Continuous measurements of driving pressure in critically ill patients fulfilling criteria for assisted ventilation showed that sustained high driving pressure occurs, even in such patients, in 8% of cases. High driving pressure is not associated with high tidal volume or high minute ventilation, but with low compliance. A threshold of compliance greater than 30 mL/cmH₂O was identified to exclude the presence of high driving pressure with 90% sensitivity and specificity.