Non-invasive method to detect high respiratory effort and transpulmonary driving pressures in COVID-19 patients during mechanical ventilation

Abstract

Background

High respiratory drive in mechanically ventilated patients with spontaneous breathing effort may cause excessive lung stress and strain and muscle loading. Therefore, it is important to have a reliable estimate of respiratory effort to guarantee lung and diaphragm protective mechanical ventilation. Recently, a novel non-invasive method was found to detect excessive dynamic transpulmonary driving pressure (∆P_L) and respiratory muscle pressure (P_mus) with reasonable accuracy. During the Coronavirus disease 2019 (COVID-19) pandemic, it was impossible to obtain the gold standard for respiratory effort, esophageal manometry, in every patient. Therefore, we investigated whether this novel non-invasive method could also be applied in COVID-19 patients.

Methods

∆P_L and P_mus were derived from esophageal manometry in COVID-19 patients. In addition, ∆P_L and P_mus were computed from the occlusion pressure (∆P_occ) obtained during an expiratory occlusion maneuver. Measured and computed ∆P_L and P_mus were compared and discriminative performance for excessive ∆P_L and P_mus was assessed. The relation between occlusion pressure and respiratory effort was also assessed.

Results

Thirteen patients were included. Patients had a low dynamic lung compliance [24 (20–31) mL/cmH₂O], high ∆P_L (25 ± 6 cmH₂O) and high P_mus (16 ± 7 cmH₂O). Low agreement was found between measured and computed ∆P_L and P_mus. Excessive ∆P_L > 20 cmH₂O and P_mus > 15 cmH₂O were accurately detected (area under the receiver operating curve (AUROC) 1.00 [95% confidence interval (CI), 1.00–1.00], sensitivity 100% (95% CI, 72–100%) and specificity 100% (95% CI, 16–100%) and AUROC 0.98 (95% CI, 0.90–1.00), sensitivity 100% (95% CI, 54–100%) and specificity 86% (95% CI, 42–100%), respectively). Respiratory effort calculated per minute was highly correlated with ∆P_occ (for esophageal pressure time product per minute (PTP_es/min) r² = 0.73; P = 0.0002 and work of breathing (WOB) r² = 0.85; P < 0.0001).

Conclusions

∆P_L and P_mus can be computed from an expiratory occlusion maneuver and can predict excessive ∆P_L and P_mus in patients with COVID-19 with high accuracy.

Background

Maintaining spontaneous breathing effort in mechanically ventilated patients limits respiratory muscle disuse and atrophy [1,2,3,4]. Too high respiratory effort may lead to excessive lung stress and strain causing lung injury on one hand. On the other hand, it may lead to excessive muscle loading causing muscle injury (mainly diaphragm injury) leading to muscle dysfunction [5]. High respiratory drive and effort frequently exist in critically ill patients, mainly due to insufficient ventilator assistance and sedation, but evidence also suggests biological predisposition (e.g., pulmonary and systemic inflammation, lung mechanical heterogeneity) plays a role as well. Therefore, it is important to have a reliable estimate of respiratory effort to enable lung and diaphragm protective mechanical ventilation [6,7,8].

The gold standard to obtain respiratory effort is esophageal manometry. This technique is minimally invasive, requires appropriate equipment and expertise, and can be time consuming. Other monitoring techniques or parameters only reflect respiratory drive (P_0.1 and electrical activity of the diaphragm) or muscle loading (diaphragm ultrasound) and provide only limited information about lung stress and strain (plateau pressure and driving pressure) [7]. Recently, Bertoni et al. [9] demonstrated that dynamic transpulmonary driving pressure (∆P_L) and respiratory muscle pressure (P_mus) can be estimated from the maximal decline in airway pressure (P_aw) from positive end-expiratory pressure (PEEP) during an expiratory occlusion maneuver (∆P_occ). Direct estimates of ∆P_L and P_mus were unreliable, excessive ∆P_L and P_mus, however, could be predicted with reasonable accuracy.

Coronavirus disease 2019 (COVID-19) is a new type of lung disease [10,11,12] originating from Wuhan, China, in December 2019. Because of the sheer number of mechanically ventilated patients with severe lung disease, it was impossible to measure esophageal pressure to assess respiratory mechanics in every patient. Therefore, we estimated ∆P_L and P_mus according to Bertoni et al. [9] in every COVID-19 patient with spontaneous breathing effort as part of standard patient care. If computed ∆P_L and/or P_mus were excessive (i.e., higher than P_mus 13–15 cmH₂O and ∆P_L 16–17 cmH₂O), or if patients received prolonged mechanical ventilation with no progress (i.e., ≥14 days) or if patients remained hypercapnic (PaCO₂ ≥ 60 mmHg), respiratory mechanics was assessed by esophageal manometry for clinical purposes.

The aim of this paper is to describe respiratory mechanics in mechanically ventilated COVID-19 patients with spontaneous breathing effort, to compute ∆P_L and P_mus from ∆P_occ and assess the discriminative performance for excessive ∆P_L and P_mus, and to assess the relation between ∆P_occ and respiratory effort.

Methods

Study population

Dynamic transpulmonary driving pressure and respiratory muscle pressure were assessed in COVID-19 patients admitted to the Intensive Care Unit of the Radboud University Medical Center according to Bertoni et al. [9] as follows:

• (1) computed ∆P_L = (peak P_aw − PEEP)—2/3 × ∆P_occ.

• (2) computed P_mus = − 3/4 × ∆P_occ.

If patients had high respiratory effort and/or high dynamic transpulmonary driving pressure (i.e., computed P_mus 13–15 cmH₂O and ∆P_L 16–17 cmH₂O or higher), prolonged mechanical ventilation without clinical progress (i.e.,  ≥ 14 days) or remained hypercapnic (PaCO₂ ≥ 60 mmHg), esophageal manometry was obtained as part of our standard clinical protocol. Patients or their legal representatives were informed about the measurements.

Study protocol

This was an observational study. All patients were ventilated with a Servo-i/u ventilator (Getinge, Sölna, Sweden). Ventilator settings were set by the treating intensivist. Patients received a nasogastric catheter with esophageal balloon [Cooper (Cooper Surgical, Trumbull, USA) or Neurovent (NeuroVent Research Inc, Toronto, Canada)] to obtain esophageal pressure (P_es). Catheter position was validated using the dynamic occlusion test [13]. A total of 3–4 manual expiratory occlusions (lasting ~ 1–2 s) were performed during a 10–15 min recording per patient. After the recordings, ventilator settings or sedation strategies were adjusted, if deemed necessary, in accordance with the treating intensivist. Being an observational study, the effect of different ventilator settings or sedatives was not investigated.

Data acquisition

Ventilator flow and airway pressure (P_aw) were obtained (sample frequency 100 Hz) by connecting a RS-232 cable via the serial port of the Servo-i/u to a dedicated measurement set-up using Servotracker software (Servotracker release 4.2, Getinge, Sölna, Sweden). The esophageal balloon (i.e., P_es) and a T-piece connected to the expiration port of Servo-i/u (i.e., P_aw) were coupled to pressure transducers and acquired (sample frequency 100 Hz) using a dedicated measurement set-up (Biopac MP160, BIOPAC Inc., USA). Signals were synchronized offline based on P_aw tracings that were acquired using both software programs. Brief manual expiratory occlusions (lasting ~ 1–2 s) were performed to enable offline synchronization. Data were processed and analyzed offline using Matlab R2018a (Mathworks, Natick, MA, USA).

Signal analysis

The occlusion pressure (∆P_occ) was defined as the maximal deflection in P_aw from positive end-expiratory pressure (PEEP) during an expiratory occlusion maneuver (Fig. 1). The decrease in P_es during the first 100 ms of this maneuver was computed as P_0.1. Transpulmonary pressure (P_L) was determined by subtracting P_es from P_aw. Dynamic transpulmonary driving pressure (∆P_L) was computed from onset to peak during inspiration. Dynamic lung compliance (C_dyn) was calculated as tidal volume divided by the increase in P_L between points of zero flow. Chest wall elastance (E_cw) was estimated based on predicted vital capacity [9, 14], from this chest wall elastic recoil pressure (P_cw) was computed as the product of tidal volume and E_cw. The pressure generated by the respiratory muscles (P_mus) was calculated as P_cw minus P_es. The integral of the product of P_mus and tidal volume represents work of breathing (WOB), calculated per liter and per minute. The integral of P_mus over time is defined as esophageal pressure–time product (PTP_es), calculated per breath and per minute [14, 15].

Fig. 1

Flow and pressure tracings showing an expiratory occlusion maneuver. From top to bottom: flow (in mL/s), airway pressure (P_aw), esophageal pressure (P_es), transpulmonary pressure (P_L) (P_aw – P_es), chest wall elastic recoil pressure (P_cw) (tidal volume × estimated chest wall elastance) and respiratory muscle pressure (P_mus) (P_cw – P_es) (pressures in cmH₂O). During an expiratory occlusion maneuver the patient inhales against a closed valve, resulting in a decrease in airway pressure. The maximal deflection in P_aw from positive end-expiratory pressure is defined as occlusion pressure (∆P_occ). From this ∆P_L and P_mus were computed and compared with true dynamic lung stress (increase in P_L from onset to peak during inspiration) and true respiratory effort (peak P_mus during inspiration)

Data obtained during expiratory occlusion maneuvers were averaged. Data were analyzed on a breath-by-breath basis and averaged over at least a 4-min period free of artifacts or esophageal contractions. Only recordings where ∆P_es/∆P_occ was between 0.8 and 1.2 were included in the analysis.

Statistical analysis

Normality was tested and data are presented accordingly as mean ± standard deviation (SD) or as median [interquartile range (IQR)]. Measured and computed ∆P_L and P_mus were compared using Bland–Altman analysis. Receiver operating characteristic curve analysis was performed and sensitivity and specificity were computed to assess the accuracy of computed ∆P_L and P_mus to detect excessive ∆P_L > 20 cmH₂O and P_mus > 10 and > 15 cmH₂O. Linear regression analysis was performed to assess the relationship between ∆P_occ and respiratory effort. For all tests a two-tailed P < 0.05 was considered statistically significant. Statistical analyses were performed with Prism 5 (Graphad software, San Diego, USA).

Results

Patient characteristics

Esophageal manometry was obtained in 15 COVID-19 patients between April and July 2020. Two patients were excluded from analysis due to incorrect ∆P_es/∆P_occ. Patient characteristics at time of measurement are shown in Table 1. In general, patients were 61 ± 9 years old, had high PaCO₂ (63 ± 17 mmHg) and received prolonged mechanical ventilation (41 ± 32 days). Respiratory failure was the main problem.

Table 1 Patient characteristics

Respiratory parameters are shown in Table 2. Only in patient 7 it was not possible to analyze a 4-min period due to continuous esophageal contractions. Patients had a low C_dyn [24 (20–31) mL/cmH₂O], high ∆P_L (25 ± 6 cmH₂O) and high P_mus (16 ± 7 cmH₂O).

Table 2 Respiratory parameters

Computed ∆P _L and P _mus

Bland–Altman analysis showed low bias, but wide limits of agreement between measured and computed ∆P_L [− 1.1 ± 5.9 cmH₂O (bias ± 95% limits of agreement)] (Fig. 2a). Bias between measured and computed P_mus was higher and limits of agreement were equally wide (2.3 ± 6.0 cmH₂O) (Fig. 2b). This means there is poor agreement between measured and computed ∆P_L and P_mus.

Fig. 2

Bland–Altman plots with regression lines in which measured and computed dynamic transpulmonary driving pressure (∆P_L) (a) and respiratory muscle pressure (P_mus) (b) are compared. Computed ∆P_L overestimates measured ∆P_L (a) (− 1.1 ± 5.9 cmH₂O (bias ± 95% limits of agreement), while computed P_mus underestimates measured P_mus (b) (2.3 ± 6.0 cmH₂O). Limits of agreement are large for both parameters. There was no significant trend in differences (for ∆P_L r² = 0.27; P = 0.06 and for P_mus r² = 0.18; P = 0.15)

Receiver operating characteristic curve analysis was performed to assess the discriminative performance to predict excessive dynamic lung stress and respiratory effort (Fig. 3; Table 3). Excessive ∆P_L > 20 cmH₂O was accurately predicted by computed ∆P_L > 19 cmH₂O [with area under the receiver operating curve (AUROC) 1.00 (95% confidence interval (CI), 1.00–1.00), sensitivity 100% (95% CI, 72–100%) and specificity 100% (95% CI, 16–100%)]. Discriminative performance for P_mus > 10 cmH₂O was only moderate, but was acceptable for P_mus > 15 cmH₂O with computed P_mus > 13 cmH₂O [with AUROC 0.98 (95% CI, 0.90–1.00), sensitivity 100% (95% CI, 54–100%) and specificity 86% (95% CI, 42–100%)] (Fig. 3).

Fig. 3

Receiver operating characteristic curves (ROC) showing the discriminative performance of computed transpulmonary driving pressure (∆P_L) to detect measured excessive (> 20 cmH₂O) ∆P_L [area under the ROC (AUROC) 1.00 (95% confidence interval (CI), 1.00–1.00)] (a) and computed respiratory muscle pressure (P_mus) to detect measured P_mus > 10 cmH₂O [AUROC 0.94 (95% CI, 0.81–1.00)] (b) and 15 cmH₂O (AUROC 0.98 (95% CI, 0.90–1.00)] (c). Threshold values are shown as points on the curves

Table 3 Discriminative performance

∆P _occ and respiratory effort

∆P_occ was correlated with respiratory effort (Fig. 4). Only moderate correlations were found between ∆P_occ and PTP_{es breath} (r² = 0.51; P = 0.0060) and WOB (calculated per liter) (r² = 0.68; P = 0.0005). Respiratory effort calculated per minute showed much better correlations with ∆P_occ (for PTP_es/min r² = 0.73; P = 0.0002 and WOB r² = 0.85; P < 0.0001).

Fig. 4

Relation between occlusion pressure (∆P_occ) and respiratory effort. ∆P_occ is moderately correlated with work of breathing (WOB) calculated per liter (a) and pressure time product of esophageal pressure per breath (PTP_{es breath}) (c). Higher correlations were found between ∆P_occ and respiratory effort when respiratory effort was calculated per minute (b, d)

Discussion

We demonstrate that the mechanically ventilated COVID-19 patients with spontaneous breathing effort included in this study received prolonged mechanical ventilation, had a low dynamic lung compliance, high dynamic transpulmonary driving pressures and high respiratory effort. Dynamic transpulmonary driving pressure and respiratory muscle pressure were estimated from the maximal decline in airway pressure from PEEP during an expiratory occlusion maneuver. Computed ∆P_L and P_mus are unreliable for direct estimates of ∆P_L and P_mus derived from esophageal manometry, as analysis showed poor agreement between computed and measured values. However, they can predict excessive ∆P_L (> 20 cmH₂O) and P_mus (> 15 cmH₂O) with high sensitivity and specificity. The occlusion pressure is highly correlated with respiratory effort per minute.

Dynamic lung stress and respiratory effort

Maintaining spontaneous breathing effort during mechanical ventilation has become increasingly important in recent years, due to accumulating evidence for over-assistance myotrauma not only during controlled mechanical ventilation, but also during high levels of pressure support ventilation [1,2,3,4,5]. Too high respiratory effort, however, can also cause lung and/or diaphragm injury. This might not be that obvious when relying only on plateau and driving pressures on the ventilator screen. The pressure generated by the respiratory muscles (i.e., P_mus) might in fact be quite high and thus the pleural pressure (i.e., P_es) quite negative, despite high levels of pressure support. Indirect evidence suggests that high P_mus may cause load-induced muscle injury and dysfunction [5, 6]. Negative pleural pressures in an already injured lung increase transpulmonary pressures and thus lung stress and strain and worsen vascular leakage [i.e., patient-self inflicted lung injury (P-SILI)] [16]. In our study, patients had a relatively high P_mus and PaCO₂. Apparently, they were not able to increase P_mus to achieve a normal PaCO₂. Patients had a high respiratory frequency, but this was insufficient in most patients to meet ventilatory demands as they had high dead space ventilation reflecting severe gas exchange disorders (Additional file 1: Table S1) [17]. ∆P_occ was only moderately correlated with PTP_{es breath} and WOB (J/L), but highly correlated when respiratory effort was multiplied with respiratory frequency [i.e., PTP_es/min and WOB (J/min)]. Telias et al. [18] observed something similar for P_0.1, which correlated better with respiratory effort per minute as compared to respiratory effort per breath. Together, the data from our study and the study of Telias et al. [18] suggest that in response to high respiratory drive critically ill patients increase respiratory frequency rather than tidal volume, probably due to a combination of respiratory muscle weakness and decreased lung compliance, limiting the ability to increase effort per breath [7, 19, 20].

Clinical implications

Bertoni et al. [9] provided a novel non-invasive method to compute ∆P_L and P_mus from ∆P_occ in mechanically ventilated patients with spontaneous breathing effort. We demonstrated that this novel method can also be applied in COVID-19 patients. In accordance with Bertoni et al. [9], computed ∆P_L and P_mus cannot directly replace ∆P_L and P_mus derived from esophageal manometry. In the external validation cohort they found reasonable discriminative performance for ∆P_L > 15 cmH₂O and P_mus > 10 cmH₂O. In this study, we were able to show that computed values can also be used to predict excessive ∆P_L (> 20 cmH₂O) and P_mus (> 15 cmH₂O). This is very useful when it is not feasible to perform esophageal manometry for various reasons.

COVID-19 patients have severely injured lungs and are prone to high respiratory effort, necessitating close monitoring to enable lung and diaphragm protective ventilation [6, 8]. If computed ∆P_L and/or P_mus are/is excessively high, one can decide to measure esophageal pressures. If that is not feasible, ventilator settings should be changed followed by appropriate sedation to keep computed ∆P_L and P_mus within the clinically acceptable range based on most recent studies and reviews [6, 8, 9]. Excessive sedation, however, can lead to insufficient respiratory effort (i.e., diminished ∆P_occ) and increased patient ventilator asynchronies [8].

Limitations

This study has some limitations. First, the relatively small sample size. However, many physiological studies with critically ill patients are limited in sample size. For example, the external validation cohort in the study by Bertoni et al. [9] only included 12 patients. Second, there is a selection bias. Only patients with computed high respiratory effort and/or high dynamic transpulmonary driving pressure, prolonged mechanical ventilation and/or who were hypercapnic, were included in the study. Therefore, we found relatively high measured ∆P_L and P_mus. Third, limitations in measured ∆P_L and P_mus. ∆P_L is the dynamic transpulmonary driving pressure, therefore it may overestimate lung stress due to the resistance component. Some studies suggest to perform an end-inspiratory occlusion maneuver in the presence of spontaneous breathing activity to obtain semi-static pressure measurements [21, 22]. For P_mus calculations the chest wall elastance was estimated based on predicted vital capacity. Bertoni et al. [9] demonstrated that predicted values approximated measured values of chest wall elastance.

Conclusions

In mechanically ventilated COVID-19 patients with spontaneous breathing effort ∆P_L and P_mus can be computed from an expiratory occlusion maneuver. Computed ∆P_L and P_mus cannot replace ∆P_L and P_mus derived from esophageal manometry, but they can predict excessive ∆P_L and P_mus with high accuracy. The occlusion pressure is highly correlated with respiratory effort per minute.