Neurally adjusted ventilatory assist (NAVA) versus pressure support ventilation: patient-ventilator interaction during invasive ventilation delivered by tracheostomy

Difficult weaning from ventilator support was defined by the consensus conference published in the European Respiratory Journal in 2007 [1]. Nearly half of the patients who fail at the first single-breathing trial still require mechanical ventilation at day 7 [2, 3]. Prolonged weaning significantly increases the incidence of ventilator-associated pneumonia (VAP) [4] and mortality [5].

Prolonged weaning could be caused by respiratory or heart failure, intensive care unit (ICU)-acquired weakness [1, 2], or be related to poor patient-ventilator interaction (called patient-ventilator asynchrony). According to research, tracheostomy could reduce the duration of mechanical ventilation [6, 7] and both ICU [6, 8] and hospital mortality [9].

Patient-ventilator asynchronies occur in approximately 25% of mechanically ventilated patients, thus increasing the duration of mechanical ventilation and the length of ICU or hospital stays [10, 11].

Asynchronies can be divided into macro- and micro-asynchronies. Macro-asynchronies are generally comprised of ineffective triggering, some double triggering and, more rarely, auto-triggering [10, 12]. Micro-asynchronies comprise inspiratory trigger delay, premature, and late cycling. They may reflect a delay between a diaphragm signal and a respirator response [13]. An inspiratory trigger delay is too big a delay between the diaphragmatic signal and respiratory insufflation. Late cycling, initially called “prolonged cycle” [10], is too long a respiratory insufflation while the patient is already in expiration time (diaphragmatic relaxation) [14, 15]. These prolonged inspirations favour other asynchronies such as ineffective triggering and are potentially responsible for dynamic hyperinflation that increases both respiratory work and the duration of the ventilator support [16]. We examined a third micro-asynchrony called premature cycling [13], which is defined by a ventilator insufflation stop while the diaphragmatic signal is still in inspiration.

Pressure support ventilation (PSV) is a normal weaning mode. New assisted proportional ventilation modes such as neurally adjusted ventilator assist (NAVA) could help optimise weaning. NAVA mode is based on the detection of a diaphragm electromyographic signal (Eadi) [17] which triggers the ventilator and provides proportional assistance.

The benefits on patient-ventilator synchrony have been well established on intubated patients [13, 18‐22].

To our knowledge, no study has compared the occurrence of various asynchronies in NAVA versus PSV on ventilated tracheotomised patients [23].

The objective of our study was to show a reduction of asynchronies in NAVA compared with PSV in mechanically ventilated patients with tracheostomies. The secondary objectives were to compare the different types of asynchronies between NAVA and PSV and other respiratory parameters.

We conducted a prospective, sequential, non-randomized, single-centre study in the ICU of the University Hospital of Rangueil in Toulouse, France. The study was approved by our institution’s Ethics Research Committee (No. 50–0614). The inclusions started in June 2014 for a period of 2 years and 9 months. All ventilated tracheotomised patients were eligible for inclusion. The exclusion criteria were: patients younger than 18 years old, pregnant women, classical contraindications to Eadi catheter placement (recent oesophageal or gastric surgery, presence of oesophageal varices), a progressive infectious process defined by the association with clinical examination or imaging of an infected area with systemic inflammatory response syndrome (SIRS), haemodynamic failure (mean arterial pressure less than 65 mmHg or catecholaminergic treatment), decisions to withhold life-sustaining treatment, and the presence of a guardianship.

Difficult weaning from ventilator support requires optimal ventilation settings. Here, we demonstrated in a prospective study that NAVA reduced total patient-ventilator asynchronies mainly through a decrease of micro-asynchronies in tracheotomised patients in comparison with the PSV mode.

Generally, tracheostomy is performed in the ICU for difficult weaning and prolonged ventilation or to protect the upper airways in brain-injured patients. In our population, a majority of the patients were tracheotomised for difficult weaning.

Frutos-Vivar et al. [8] showed in a large multicentric observational study that tracheostomy is mainly performed in the context of respiratory diseases. A minority of 22% of the patients included were tracheotomised due to a coma.

In comparison to this previous study, there were several comas (less than 10%), except for post-anoxic coma, because of the coexistence of a neurosurgical ICU in the same University Hospital.

Another specific feature of our study was the important role of patients with cardiac and respiratory diseases, these representing a particular population for whom weaning would be difficult.

We did not precisely quantify the ICU-acquired weakness but, as recently established, weaning failure is not associated with a poor Medical Research Council (MRC) score [25].

Unfortunately, not all the tracheotomised patients were included (almost 50%), with the majority of them presenting contraindications to the nasogastric tube.

We also lost some potentially eligible patients because of the small number of Servo-i® devices in our unit. When a patient benefited from NAVA using another interface (orotracheal tube, non-invasive ventilation (NIV)), it would have been unethical to exchange the machine for the gain of a potentially included patient. When several patients were tracheotomised and several Servo-i® ventilators were available, we favoured patients with the most difficult weaning. In addition, we had to deal with catheter shortages for a few weeks.

Our results exhibited some macro-asynchronies in PSV mode, compared with the 10% reported in the literature [10, 26]. This benefit can be attributed to the combined effects of tracheostomy and a careful personalized ventilator setting.

Thille et al. showed that optimized PSV settings help avoid patient-ventilator asynchronies and are related to a reduced weaning period, even in tracheotomised patients [10]. The same team also demonstrated that reducing pressure support or inspiratory duration eliminated ineffective triggering in a majority of the patients with a high percentage of ineffective efforts [16]. Consequently, we showed no difference in AI between the PSV and NAVA in macro-asynchronies (1.6% in PSV versus 1.64% in NAVA) despite the proportional assistance applied by NAVA which is known to reduce asynchronies [13, 17, 18, 20, 21, 27, 28].

As previously published by our team [21] and by Piquilloud et al. [13], there was more double triggering in NAVA compared with PSV. The existence of a biphasic aspect of the Eadi signal could explain a successive respiratory cycle during diaphragmatic relaxation (type 1 double triggering) [13], but the occurrence of double triggering was very rare in both modes (0.3% in PSV and 0.4% in NAVA).

Contrary to research on NIV [27] and intubated mechanically ventilated (IMV) patients [10, 13, 18, 29], we did not identify a decrease in the rate of ineffective triggering in NAVA versus PSV. But, as previously noticed, theses asynchronies were rare in our work.

To our knowledge, one study [23] compared patient-ventilator interactions in NAVA versus PSV on tracheotomised patients during the weaning process. The authors assessed the physiological response to varying levels of NAVA and PSV in 13 tracheotomised patients. Patient-ventilator asynchrony was not fully investigated. They only explored ineffective triggering rates, which were not significantly different between the two modes, independently from the level of assistance. This could be explained by the specific interface (tracheostomy) which allows for a decrease in dead space and airway resistance in both ventilator modes. Consequently, it could limit dynamic lung hyperinflation and auto-PEEP.

Our study shows a significant decrease in the total asynchrony index in NAVA versus PSV (2.1% vs 14%, p < 0.0001). Thille et al. showed that patients whose asynchrony index was greater than 10% had a longer duration of mechanical ventilation [10] and higher mortality [28]. But the difference on AI only focuses on micro-asynchronies. The micro-asynchrony index was statistically significant lower in NAVA than in PSV (0.35% versus 9.8%).

Despite particular attention being paid to ventilator settings to limit dynamic lung hyperinflation in PSV (personalization of inspiratory trigger, low pressure support level), there was less late cycling in NAVA. In NAVA, the level of assistance is proportional to the patient request and the expiratory trigger is neural, preset to 80% of the Eadi peak. These specificities allow for an optimization of the respiratory parameters from one cycle to the next, with a continuous variability of the pressure support level (cmH₂O/μV) and the expiratory timing. This result strengthens previous studies that underline the importance of a personal and adapted cycling-off to avoid dynamic hyperinflation, especially in COPD patients [14, 29].

There was little premature cycling in NAVA. This makes sense because, in NAVA, expiratory triggers are systematically neural.

We defined an asynchrony as a delay of 150 ms between the origin of the Eadi signal and the start of the ventilator insufflation, which represents the conscious perception threshold that could be source of discomfort. Consequently, we added a 50-ms safety margin. In other studies, inspiratory delays were measured and the durations were compared based on the ventilator mode [13, 18, 20, 27, 30].

The median respiratory rate (26 versus 28 breaths/min) and the median tidal volume were lower in NAVA than in PSV (5.8 ml/kg versus 6.2 ml/kg) (p < 0.001). Furthermore, the Eadi peak was similar in both modes. It has been shown that over-assistance is reflected by a decrease in the Eadi peak [19, 23, 30]. This eliminates the possibility of a disparity in the level of assistance between the two modes. PEEP was lower in NAVA (5 versus 6 cmH₂O). Better expiratory cycling may limit overdistension and allow for a decrease in PEEP.

Our study has some limitations. This was a single-centre observational study, and was non-randomized. This study demonstrated a reduction in asynchronies in NAVA versus PSV but these results could not lead to an assumption of a benefit concerning weaning duration, hospital length of stay, or survival, mainly due to the small population studied. Prospective clinical studies comparing the weaning from ventilator support in tracheotomised patients with NAVA versus PSV are necessary.

This is the first study to compare patient-ventilator asynchronies in NAVA and PSV in tracheotomised patients presenting with difficult weaning criteria. There were fewer asynchronies in NAVA with reduced auto-triggering, premature cycling-off, late cycling-off, and inspiratory trigger delay compared with PSV. An asynchrony index greater than 10%, which is related to an increased duration of ventilation, was lower in NAVA. Further studies are required to determine the clinical impact of NAVA on major outcomes in tracheotomised patients.