Central venous pressure swing outperforms diaphragm ultrasound as a measure of inspiratory effort during pressure support ventilation in COVID-19 patients

Consecutive patients were enrolled if they had been admitted for acute respiratory failure in COVID-19, and undergoing PSV with a positive end-expiratory pressure (PEEP) > 5 cmH₂O. Exclusion criteria were: hemodynamic instability requiring vasopressors, hypoxemia requiring PEEP > 10 cmH₂O and/or FiO₂ > 60%, PS > 10 cmH₂O, Richmond Agitation and Sedation Scale score <  − 1, history of COPD. Ethical approval for this study (Rep. Int. 11649/2020) was provided by the Comitato Etico per le Sperimentazioni Cliniche of the Azienda Provinciale per i Servizi Sanitari di Trento (Chairperson dott. Angelo del Favero) on 22 June 2020; written informed consent was obtained according to Italian regulations.

Before enrolment in the study, all patients were treated according to National and local recommendations [25]. In particular, whenever a patient improved his/her oxygenation (PaO₂/FiO₂ > 200 with a PEEP ≤ 12 cmH₂O in supine position for at least 12 h) interruption of muscle relaxant use was suggested, targeting a light level of sedation and attempting assisted ventilation. An initial pressure support level of 8–12 cmH₂O was suggested, to obtain a VT between 5 and 8 ml/Kg predicted body weight; subsequent support titration was suggested to maintain respiratory rate < 35 breaths per minute. The assessment of P0.1 was advised to identify patients with a high respiratory drive. Figure 1 shows the measurement of inspiratory effort in a representative patient.

Patients were studied in the semirecumbent position. The lungs of the patients were ventilated in pressure support with the following settings: flow-triggering at 2 l/min, pressure ramp 200 ms, cycling-off at 25% of the peak inspiratory flow. Automatic-tube compensation was not used.

Esophageal pressure was measured using a balloon, graduated feeding catheter (NutriVent®, Mirandola, Modena, Italy), which was positioned under general anesthesia during mechanical ventilation. To check for the correct position of the esophageal balloon, a dynamic occlusion test was performed to ensure that Pes was changing in concert with the airway pressure (Paw) when making breathing against a closed airway [26]. The dynamic occlusion test was performed by a dedicated monitor (OptiVent®, Mirandola, Modena, Italy) which provides the level of confidence related to the quality of the occlusion maneuver and the R-square value of the relationship between Paw and Pes. The optimal inflation volume of the balloon was obtained by means of a specific calibration procedure carried out by the OptiVent® system, which inflates the balloon to different volumes while recording the respective pressure values.

CVP was measured from the distal port of a triple-lumen central venous catheter, whose distal end was located radiographically in the superior vena cava. The distal port was connected to a differential pressure transducer which was filled with a 0.9% saline solution. CVP measurements were zeroed at mid-thoracic position at the level of the fifth rib [27], and the value was taken during either end-inspiration or end-expiration, at the base of “c” wave [28]. Transmural CVP was calculated as the difference between CVP and Pes [29]. ΔCVP was recorded on a dedicated multi-parametric monitor (Carescape B850, GE Healthcare, Little Chalfont, UK).

In each step, tidal volume, respiratory rate, minute ventilation, and the tidal swing in esophageal and central venous pressures (ΔPes and ΔCVP, respectively) were measured. Mean values were computed over three consecutive breaths. End-expiratory and end-inspiratory occlusions were performed to measure P0.1 [30] and Pmusc, index [31]: the airway pressure decrease in the first 100 ms after the onset of inspiration following an end-expiratory occlusion (P0.1) was measured, reflecting the patient’s respiratory drive. The estimated pressure developed by the inspiratory muscles at the end of an inspiratory effort (Pmusc), expressed as the Pmusc, index (PMI) was calculated as follows:where Pel, rsi is the elastic recoil pressure of the respiratory system at the end of inspiration, measured as the airway plateau pressure during an end-inspiratory occlusion maneuver.

Ultrasonography was performed by the same trained operator (AS), with 5 years of experience and qualifications in respiratory ultrasound, using a General Electrics Vivid T8 Ultrasound System with a 12 MHz linear probe (GE Healthcare, Little Chalfont, UK). Images were recorded for a subsequent, computer-assisted quantitative analysis by a trained investigator (SM), unaware of the ventilatory condition.

Diaphragm thickness was assessed in the zone of apposition of the diaphragm to the rib cage. The linear probe was placed above the right 10^th rib in the mid-axillary line, as previously described [22]. The inferior border of the costophrenic sinus was identified as the transition from the artefactual representation of the lung to the visualization of the liver.

Three subsequent measures were averaged. The diaphragm thickening ratio (TR) was calculated as:

Patients who were judged ready for a spontaneous breathing trial [32], underwent a trial of three levels of PSV, lasting 30 min each. The first level was set at 10 cmH₂O (PS10). Pressure support was then reduced to 5 (PS5) and 0 cmH₂O (PS0), in this order. PEEP and FiO₂ were unchanged, as was the sedation level. During the last 5 min of each step, the pattern of breathing and indices of respiratory effort were assessed, diaphragm ultrasound was performed, and hemodynamic parameters were recorded. To avoid the confounding effect of the possible development of fatigue at low levels of assistance, which might have then influenced the following steps, we decided to perform a decremental pressure-support test and we did not randomize the sequence of pressure support levels. The protocol was allowed to be stopped, if patients developed signs of respiratory distress (respiratory rate > 35 breaths/min, SpO₂ < 90%, heart rate > 140 beats/min, systolic blood pressure > 180 mmHg, diaphoresis or anxiety).

Since, to our knowledge, no previous publications have addressed a similar topic, a formal sample size calculation was not performed, and we enrolled a convenience sample of consecutive patients with a similar size to other physiological investigations. Data were analyzed using Stata 13.0 (StataCorp, College Station, Texas, USA) for Windows. Normality was assessed by the Shapiro-Francia test. Descriptive results are reported as mean (standard deviation) if normally distributed, or median [25–75th percentiles] otherwise. The analysis on the variables recorded over the three steps (PS10, PS5, PS0) was performed by analysis of variance for repeated measurements, with step as a within-subject factor, and the statistical significance of the within-subject factors was corrected with the Greenhouse–Geisser method. Non-parametric variables were analyzed using Friedman test. Pairwise, post-hoc multiple comparisons were carried out according to Tukey method. Regression was conducted by a linear, fixed-effects model for repeated measures to deal with the longitudinal structure of our data set (patients with repeated measurements over time). The association between variables was expressed as the coefficient of determination (R²). The diagnostic performance of the CVP swing and diaphragm TR for detecting either a low or a high inspiratory effort (arbitrarily defined as an esophageal pressure swing < 5 and  >  8 cmH₂O, respectively) [10, 11] was assessed by calculating the area under the Receiver Operating Characteristic (ROC) curve, sensitivity, specificity, positive and negative predictive value. To calculate positive and negative predictive values, the prevalence of the condition in the population was assumed to be equal to the prevalence in the sample. Two-tailed P-values < 0.05 were considered for statistical significance.

Fourteen consecutive patients were enrolled. Patient characteristics at baseline and the duration of the disease are reported in Table 1. Patients were studied after an average total ICU stay of 26 [22; 29] days. At the time of enrolment, patients had been assisted in pressure support ventilation for 6 [3; 9] days. Mechanical ventilation settings and gas exchange on the study day were as follows: PEEP 6 [5; 7] cmH₂O, FiO₂ 0.40 ± 0.05; PaO₂ 103.4 ± 24.7 mmHg, pH 7.45 ± 0.06, PaCO₂ 41.6 ± 6.8 mmHg, Base excess 4.8 ± 3.9. The level of pressure support, as clinically set by the intensivists caring for the patients, was 10 [8; 10]. The average score on the Richmond agitation-sedation scale was 0 [0; 0]. Diaphragm ultrasound examinations could be performed in all patients, and no patients developed signs of distress during the study.

Table 2 reports the haemodynamic parameters and indices of respiratory drive and effort during the three steps. No significant changes in the haemodynamic parameters was detected with the reduction in the support level. Reduction of pressure support was associated with a significant increase of respiratory rate, as well as a significant decrease in tidal volume, so that minute ventilation was not modified.

The progressive reduction of support was not associated with any change in end-expiratory Pes, while end-inspiratory Pes was significantly reduced. As a result, we found a significant increase in ΔPes (5 [3; 8] vs. 8 [4; 13] vs. 12 [6; 16] cmH₂O, p < 0.0001) (Fig. 2). Similarly, P0.1 and Pmusc,index significantly increased with lowering pressure support.

End-expiratory CVP was not statistically different during the different steps of the study, while the reduction of pressure support was associated with a significant decrease in the end-inspiratory CVP. As a result, the decrease in pressure support was associated with a significant increase in the ΔCVP (4 [3; 7] vs. 8 [5; 9] vs. 10 [7; 11] cmH₂O, p < 0.0001) (Fig. 2). ΔCVP was significantly associated with ΔPes (R² = 0.810, p < 0.001) (Fig. 3).

The expiratory thickness of the diaphragm was unchanged in the different steps, whereas the inspiratory thickness, and hence the thickening ratio, significantly increased with lowering levels of support (9.2 ± 6.1 vs. 17.6 ± 7.2 vs. 28.0 ± 10.0%, p < 0.0001) (Fig. 2). Diaphragm thickening ratio significantly correlated with ΔPes (R² = 0.399, p < 0.001) (Fig. 3).

In a multivariable, linear, fixed-effects regression, only ΔCVP was significantly associated with ΔPes (coefficient: 1.45 ± 0.16, p < 0.001), while diaphragm TR was not (coefficient: − 0.07 ± 0.05, p = 0.184).

The areas under the ROC curves for the detection of a low inspiratory effort were 0.783 [0.613; 0.954] for the central venous pressure swing and 0.736 [0.554; 0.918] for diaphragm TR, p = 0.6218 (Supplementary Figure S2). The best cutoffs were 5 cmH₂O for ΔCVP and 8% for diaphragm TR. As for the detection of a high inspiratory effort, the areas under the ROC curves were 0.743 [0.582; 0.903] for diaphragm TR and 0.866 [0.766; 0.986] for ΔCVP, p = 0.0484. The best cutoffs were 9 cmH₂O for ΔCVP and 20% for diaphragm TR. Supplementary Table S1 shows the diagnostic performance of the best cutoffs for central venous pressure swing and diaphragm thickening ratio for detecting either a low or a high inspiratory effort.

Several recommendations have been published to guide the ventilator management of patients admitted to the ICU for COVID-19-related acute respiratory failure [2‐5]. However, the weaning process has not been fully covered, and suggestions have been made to follow the general criteria for weaning in any type of respiratory failure.

In the Lombardy region, at the peak of the emergency, more than twice the number of preexisting ICU beds were occupied by patients with severe COVID-19 [33]. Contributing to this resource scarcity is the prolonged ventilator dependence and the difficult weaning associated to prolonged prone positioning, deep sedation and use of muscle relaxants and corticosteroids [1]. Given the lack of ICU beds, many clinicians attempted weaning from mechanical ventilation, well aware of the difficult balance between the detrimental and beneficial effects of spontaneous breathing effort [12], i.e. the potential protection against diaphragm contractile dysfunction [34, 35] and the possibility of self-inflicted lung injury from an excessive respiratory drive [36].

However, neither a breathing pattern with utilization of accessory muscles, nor a rapid or shallow breathing, or the inspection of ventilator vaweforms allow any quantitative assessment of breathing effort [13, 37] and an appropriate bedside monitoring of inspiratory effort is strongly required.

When the inspiratory muscles contract, the size of the ribcage increases, reducing pleural pressure, which can be accurately estimated at the bedside by the assessment of ΔPes [38, 39]. We recently demonstrated how, in patients undergoing assisted mechanical ventilation after acute hypoxemic respiratory failure, this parameter is strongly correlated with gold-standard indices of respiratory effort over a relatively wide range of loading conditions [40].

Diaphragm ultrasonography has been recently suggested to monitor respiratory workload [22, 23], as the inspiratory thickening ratio has shown fair correlation to respiratory pressure generation (i.e. the ΔPes) [40‐43]. The non-invasive nature, low-costs, steep learning curve and straightforward calculations are its main advantages.

As expected, we found that TR increased with decreasing levels of support. However, the correlation despite being statistically significant, was far from ideal. First of all, TR is insensitive to duration and frequency of contractions. Moreover, TR may differ from inspiratory effort in two directions: a low TR in the presence of a high ΔPes, and the opposite. The first case (high inspiratory effort, low diaphragm thickening) can arise when diaphragm dysfunction is present [40]. Less intuitive is the second condition, i.e. the coexistence of a high diaphragm thickening with a low inspiratory effort. The relationship between diaphragm thickening and pressure generation may in fact depend on the pattern of thoracoabdominal motion, so that more thickening is expected for any given ΔPes when the inspiration is predominantly accommodated by descent of the diaphragm rather than expansion of the ribcage. Although the thoracoabdominal pattern was not recorded in the present investigation, we can speculate that this issue might be a factor contributing to our results.

Since the superior vena cava is a highly compliant, intrathoracic vein, indwelling central venous catheters have been used to estimate the respiratory effort without resorting to additional catheters. That CVP and Pes fluctuate during ventilation in a similar way is not a new finding [44]. During positive-pressure breathing in passive conditions, tidal changes of Pes and CVP were almost identical in size, and linearly correlated [19].

Quite surprisingly, however, the use of ΔCVP to estimate inspiratory effort has seldom been reported, and is generally not used in the everyday clinical practice. Flemale et al. measured CVP and Pes with identical fluid-filled systems in healthy volunteers during inspiratory efforts against a closed airway in seated, supine, right lateral and left lateral positions, and concluded that CVP reflects Pes, with a ΔCVP/ΔPes ratio close to the unity [18]. Similar results were recently found in 10 patients undergoing progressive reduction of inspiratory pressure support, in whom the ratio of ΔCVP to ΔPes was on average 1.1, with a mean difference of 1 cmH₂O [20]. We found that ΔCVP was an adequate estimate of inspiratory muscle force generation across a wide range of loading conditions.

Potential factors which may affect the concordance between Pes and CVP include the fact that CVP is measured with a fluid-filled system, while Pes is measured with an air-filled system. While fluid-filled systems have a theoretical better frequency response, only small differences were found between the two systems [45]. Moreover, a poor transmission of the pleural pressure changes into the superior vena cava is also possible, and a reduction of the compliance of the vena cava has been suggested as a mechanism [18]. Since the filling of the superior vena cava may potentially influence its mechanical characteristics [21], a secondary outcome of the present investigation was to investigate the effect of CVP (both as the absolute and the transmural value) on the correspondence between ΔCVP and ΔPes. We found that the association between the two variables was significant independently of the value of CVP.

Indeed, it is known that weaning-induced pulmonary oedema is characterized by a progressive increase in transvascular pressure, which is then the cause of the clinical picture [46‐48]. In our population, the three-step, sequential reduction of the pressure support levels was not associated with any statistically significant increase in the transmural CVP, likely because none of the patients had severe heart disease [48]; moreover, even at the lowest level of support (PS0), a CPAP was always guaranteed [49].

We investigated the diagnostic characteristics of central venous pressure swing and diaphragm thickening ratio in detecting a low or a high inspiratory effort, as defined by widely accepted threshold values of esophageal pressure swing [10, 11]. Concerning a low level of inspiratory effort, both indices had a similar diagnostic performance, as assessed by the area under the ROC curve. As far as the detection of a high inspiratory effort is concerned, ΔCVP had a significantly higher overall diagnostic performance as compared to diaphragm TR. Central venous pressure swing proved a specific rather than a sensitive test, both for low and high inspiratory effort: this means that it can effectively be used to rule in the specific condition with virtually no false negatives and an overall acceptable discriminative ability (i.e. a CVP swing < 5 cmH₂O rules in a low inspiratory effort, whereas a CVP swing > 8 cmH₂O rules in a high inspiratory effort).

Monitoring of respiratory drive and effort is increasingly recognized as necessary during assisted mechanical ventilation, as both excessively low and high levels can have detrimental consequences in terms of lung injury and diaphragm myotrauma [11]. However, despite the availability of monitoring techniques (mainly used for research purposes), inspiratory effort is seldom measured directly [50]. In the current investigation we showed how the inspiratory swing of CVP, a near-ubiquitous monitoring in the ICU, can be used to track inspiratory effort with an acceptable accuracy. Since monitoring of inspiratory effort was suggested for patients at higher risk of complications from injurious breathing [10], such as those with more severe lung injury or systemic inflammation, and patients with COVID-19-associated acute respiratory failure have a form of injury that was found similar to patients with ARDS unrelated to COVID-19 [51], we speculate that using ΔCVP to assess inspiratory effort may be a useful tool in other patients with acute hypoxemic respiratory failure.

Our study has some limitations: first, we studied a relatively small population, which is, however, comparable with that of other physiological studies [43, 52]. Moreover, patients were observed over a limited time-frame: we are unaware if and how much the results would change over time and their clinical relevance. The level of pressure support at the time of enrolment in the study was set by the clinician according to local recommendations. Indeed, it was shown how the clinical setting of pressure support often results in higher levels than needed by the patients to preserve respiratory muscle function while at the same time preventing passive lung inflation, and several physiological model based decision support systems may help in the selection of appropriate ventilator settings [53].

In the current investigation, esophageal pressure was used as an index of global inspiratory effort. Indeed, we did not measure transdiaphragmatic pressure, which is the gold-standard measurement of the pressure-generating ability of the diaphragm. Despite the exclusion of patients with a story of COPD, intrinsic PEEP was not measured, and we cannot exclude the effect of hyperinflation on our findings. While the population included in the present study suffered from COVID-19-related acute respiratory failure and the virtual absence of any relevant comorbidity in the patients warns against direct extrapolation of our results to other groups of patients, the strong physiological basis of our findings make us speculate in favour of a possible utility of these measurements in other categories of patients. Further studies are however warranted to verify this point.