We found that irrespective of the assistance, arterial blood gases were not significantly different between PSV and NAVA. The differences in breathing pattern, ventilator assistance, and respiratory drive and timing between PSV and NAVA were small at the two lower assist levels, but significant at the highest assist. In the range of settings studied, compared to PSV, NAVA limited the risk of over-assistance, avoided patient–ventilator asynchrony, and improved overall patient–ventilator interaction.
When introducing a new mode of ventilation it is necessary to compare it with the established treatment, which in our case was PSV. We set the reference PSV (PSV100) in order to achieve a V
t between 6 and 8 ml/kg. During PSV, however, the patient might produce a small, brief effort, just sufficient to trigger the ventilator and then relax, therefore being passively ventilated for the large majority of the inspiratory phase and not retaining control of his/her breathing pattern [25]. To overcome this drawback we used the non-invasive technique described by Foti et al. [18]. Moreover, when comparing modes of ventilation, equivalent levels of assistance should be used making the choice of the matching variable crucial. This choice, however, is not simple and is anyway susceptible to criticisms [26]. We opted for using the specific function (NAVA Preview) available on the ventilator to estimate the NAVA level necessary to deliver the assistance equivalent to a given preset PSV. PSV100 and NAVA100 did not show significant differences in Paw,peak (Table 2) and mean inspiratory Paw (18.2 ± 4.3 and 16.1 ± 5.1, for PSV100 and NAVA100, respectively, P = 0.23); moreover, V
t/kg, breathing frequency, EAdi and PTPaw also showed no significant difference between PSV100 and NAVA100.
We included patients with different underlying diseases who had been receiving invasive mechanical ventilation for a variable time. We are aware that evaluating a more selected population of patients might result in more homogeneous results. In a previous study aimed to evaluate whether and to what extent the underlying disease affected diaphragmatic effort and patient–ventilator synchrony, Nava et al. [5] found that the application of different levels of PSV resulted in comparable diaphragmatic efforts in patients with acute respiratory failure due to different pathologies. Also, patient–ventilator asynchrony occurred with all underlying diseases, although it was more pronounced in patients with chronic obstructive pulmonary disease [5]. To date, no study reporting the use of NAVA in ICU patients, rather than in animals or healthy subjects, has been published yet.
Patients receiving mechanical ventilation require sedation and analgesia for anxiety and pain experienced during the time they are intubated. Accordingly, all patients included in our study received a continuous infusion of propofol, either alone or in association with remifentanil, to maintain a Ramsay score of 3 [21]. By depressing the neural drive, which controls the ventilator during NAVA, sedatives might more markedly affect NAVA than PSV. We found that PSV and NAVA had similar effects on gas exchange, irrespective of the assist level, suggesting that NAVA was not associated with an increased risk of hypoventilation secondary to sedative infusion. A total of 3 patients received propofol alone and 11 received both propofol and remifentanil. No patient received fentanyl. Propofol infusion rate ranged between 1 and 4 mg/(kg h); remifentanil infusion rate ranged between 0.05 and 0.2 μg/(kg min). Mean infusion rates were 2.2 ± 1.3 mg/(kg h) and 0.12 ± 0.05 μg/(kg min) for propofol and remifentanil, respectively. Noteworthy, in no patient propofol infusion exceeded 4 mg/(kg h) and only in three patients remifentanil infusion exceeded 0.125 μcg/(kg min). Indeed, the risk of central drive depression at these dosages is very low. We cannot exclude, however, that deeper levels of sedation might determine different results. Moreover, our protocol does not address whether or not varying the rate of sedative infusion would affect the two modes.
It has been repeatedly reported that increasing the level of PSV augments V
t, decreases breathing frequency, neural drive, and inspiratory effort, and may worsen patient–ventilator synchrony [7, 10, 11]. We found that with both modes increasing the assist level produced significant increments in PTPaw and V
t, and reduction in EAdi. While, at the two lower assist levels PSV and NAVA showed no significant differences in either the amount of assistance provided (i.e. PTPaw), breathing pattern, or neural drive (i.e. EAdi,Peak), PSV150 significantly increased V
t and reduced breathing frequency, compared to NAVA150. Also, although Peak Paw values were not different between PSV150 and NAVA150, with the former, as opposed to the latter, the ventilator assistance was greater and the neural drive smaller. These results indicate that, overall, compared to PSV, NAVA has the potential to limit the risk of over-assistance.
Both neural- and flow-based respiratory timing was unaffected by the NAVA level. T
i,neu was also unaffected by the support level in PSV. Increasing PSV, however, significantly lengthened T
i,flow. With PSV, a prolonged mechanical insufflation exceeding T
i,neu frequently occurs, especially at higher levels of support [1, 8, 27]. This is recognized to be the underlying mechanism of patient–ventilator asynchrony [28]. When the onset of the neural expiration is impeded, the response of the respiratory centers is to prolong the expiratory time [6, 29]. Accordingly, we found that increasing PSV lengthened both T
e,neu and T
e,flow. As a result, we observed significant differences in T
i,flow, T
e,neu, T
e,flow, and T
i/T
tot,neu between PSV150 and NAVA150. Because T
i/T
tot,flow, was unchanged between the two modes regardless of the level of assistance, NAVA did not increase, when compared to PSV, the risk of dynamic hyperinflation.
Patient–ventilator asynchrony is a maladaptation of the ventilator to patient’s neural respiratory timing [28]. As expected, with NAVA we could not observe any asynchrony, regardless of the level of assistance. On the contrary, more than one-third of the PSV trials were characterized by ineffective efforts, as evidenced by an AI > 10%, with a rate that increased at the highest level of assistance. As recent work showed that patients with an AI exceeding 10% are subject to longer duration of mechanical ventilation and increased recourse to tracheostomy [3], these findings may be clinically relevant. Clearly, our results are valid in the range of setting studied and one could argue that further increasing the NAVA level might produce an additional decrease in EAdi and cause ineffective efforts to occur. Also, because of the mixed patient population, we chose not to vary the expiratory trigger threshold and leave the ventilator default value; we cannot exclude a lower rate of AI had occurred varying this PSV setting on an individual basis. Moreover, we do not know whether the use of an EAdi-based trigger during PSV would contain the occurrence of ineffective efforts and how the overall physiologic response to PSV would be consequently affected.
In a comparison between PSV and proportional assist ventilation (PAV), Wrigge et al. [23] found that V
t variability was higher during PAV and rose when increasing the assist level. The authors considered these results as a consequence of the improved ability to vary V
t by modulating the inspiratory effort. In our study, when compared to PSV, NAVA showed a higher V
t variability regardless of the assist level. EAdi variability in NAVA was lower than in PSV, a difference that achieved statistical significance at the highest assist level. Furthermore, in contrast to the aforementioned study, when increasing the assistance the variability in breathing frequency progressively increased with PSV and decreased with NAVA. This was statistically significant when comparing PSV150 to NAVA150. Overall, as opposed to PSV, increasing NAVA augmented V
t variability and reduced RRflow variability, mimicking the behavior of healthy individuals [22].