Clinical review: Update on neurally adjusted ventilatory assist--report of a round-table conference.

Conventional mechanical ventilators rely on pneumatic pressure and flow sensors and controllers to detect breaths. New modes of mechanical ventilation have been developed to better match the assistance delivered by the ventilator to the patient's needs. Among these modes, neurally adjusted ventilatory assist (NAVA) delivers a pressure that is directly proportional to the integral of the electrical activity of the diaphragm recorded continuously through an esophageal probe. In clinical settings, NAVA has been chiefly compared with pressure-support ventilation, one of the most popular modes used during the weaning phase, which delivers a constant pressure from breath to breath. Comparisons with proportional-assist ventilation, which has numerous similarities, are lacking. Because of the constant level of assistance, pressure-support ventilation reduces the natural variability of the breathing pattern and can be associated with asynchrony and/or overinflation. The ability of NAVA to circumvent these limitations has been addressed in clinical studies and is discussed in this report. Although the underlying concept is fascinating, several important questions regarding the clinical applications of NAVA remain unanswered. Among these questions, determining the optimal NAVA settings according to the patient's ventilatory needs and/or acceptable level of work of breathing is a key issue. In this report, based on an investigator-initiated round table, we review the most recent literature on this topic and discuss the theoretical advantages and disadvantages of NAVA compared with other modes, as well as the risks and limitations of NAVA.

which can cause limitation when dynamic hyperinfl ation is present and when the inspiratory trigger is delayed due to intrinsic end-expiratory pressure.
Th e other support mode is NAVA, which will be discussed in this article. Th ere are several similarities between PAV and NAVA, but this fi rst round-table meeting focused on NAVA. A vast literature also exists concerning PAV, but this topic would require a whole chapter and will not be discussed in this current paper; hopefully PAV will be the topic of a diff erent round table.
Th e present article is based on an investigator-initiated round-table meeting. Th e article aims to review the available knowledge on the physiological rationale and feasibility of the recently introduced NAVA MV modality. Th roughout the article, we place emphasis on the most recent fi ndings concerning adjustment of the NAVA settings; on the one hand considering specifi c issues associated with assisted modes of MV, and on the other considering the expecta tions placed upon NAVA.
NAVA is an assist mode of MV that delivers a pressure proportional to the integral of the electrical activity of the diaphragm (EAdi) [18], and therefore proportional to the neural output of the patient's central respiratory command. Th e level of pressure delivered is thus determined by the patient's respiratory-center neural output. With NAVA, the ventilator is triggered and cycled-off based on the EAdi value, which directly refl ects the activity of the neural respiratory command. Th e inspiratory airway pressure applied by the ventilator is determined by the following equation: where Paw is the instantaneous airway pressure (cmH 2 O), EAdi is the instantaneous integral of the diaphragmatic electrical activity signal (μV), and the NAVA level (cmH 2 O/μV or per arbitrary unit) is a proportionality constant set by the clinician.
In February 2011, several European and Canadian investigators with clinical results about NAVA available in publication or in abstract format organized a roundtable discussion at the Geneva University Hospital to describe and discuss recent advances regarding NAVA. A representative of the company that commercializes the NAVA machine (Maquet Critical Care SA, Sölna, Sweden) was invited to attend the meeting in order to answer only technical questions. Maquet Critical Care SA agreed to sign a disclosure form before the meeting specifying that neither the minutes of the meeting nor the content of the report could be modifi ed and/or used for commercial purposes. Th e main purpose of this meeting was for all of the investigators and participants to expose their standpoint and questions about NAVA, and to share the main results of their studies. Maquet Critical Care SA agreed to provide fi nancial support for organiz ing the meeting, as detailed at the end of the manuscript, but was not responsible for choosing participants and did not take any part in the writing of this report. We here describe the content of the round-table discussion, focusing on a selection of the most recent studies [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] (Table 1).

Main problems with conventional ventilation modalities in the ICU
Assisted modes generally aim at synchronizing the ventilator insuffl ation to the patient's eff ort, both to optimize comfort and to minimize the work of breathing. Th e price to pay for this strategy is a risk of patientventilator asynchrony, which can be defi ned as a mismatch between the patient's neural output and the ventilator's inspiratory and expiratory times [34][35][36][37]. Th ille and colleagues reported that one-quarter of patients had high rates of asynchrony during assisted ventilation [34]. Frequent asynchrony is associated with a longer duration of MV [34,38].
Compelling evidence accumulated over the last decade also supports the use of tidal volume (V T ) values that are lower than those traditionally used. Lower V T values than traditionally used have several main advantages: they diminish the risk of ventilator-induced lung injury [39][40][41]; they preserve spontaneous breathing by avoiding respiratory alkalosis, thus preventing diaphragmatic disuse atrophy associated with MV [42][43][44][45][46][47][48]; they diminish several types of patient-ventilator asynchrony [49]; and they may improve the effi ciency of gas exchange [50]. Assisted modes of ventilation that maintain at least part of the patient's spontaneous breathing activity contribute to preventing these pulmonary and muscular complications.
Th e new challenge in developing ventilation strategies thus consists of minimizing the risk of lung injury, avoiding disuse atrophy of the diaphragm, and improving the match between the patient's needs and the assistance delivered by the ventilator [6]. New ventilation modes have been designed to meet this challenge [5], and NAVA is a pressure-assisted mode in which the pressure delivered by the ventilator is proportional to the electrical activity of the diaphragm recorded continuously through an esophageal probe [18]. NAVA theoretically delivers pressure proportional to the neural output of the patient's central respiratory command. During NAVA, however, reliable positioning of the catheter is mandatory in order to obtain a representative EAdi signal from the diaphragm. Barwing and colleagues have evaluated whether a formula based on the measurement from nose to ear lobe to xiphoid process of the sternum (the NEX distance) modifi ed for the EAdi catheter (NEXmod) is adequate for predicting the accurate position of the esophageal probe [51]. Th ey observed in 18 of 25 patients (72%) that at NEXmod the EAdi signal was suitable for running NAVA. Th e NAVA mode was possible at the optimal position in four patients -the optimal position being defi ned by checking three criteria: stable EAdi signals, electrical activity highlighted in central leads of the catheter positioning tool, and an absence of the pwave in the distal lead. Th e authors thus concluded that positioning the EAdi catheter using NEXmod gives a good approximation in most of the patients. Moreover, the body position, positive end-expiratory pressure (PEEP) and intra-abdominal pressure are factors known to infl uence the position of the diaphragm. Barwing and colleagues therefore enrolled 20 patients in order to evaluate the eff ects of these factors on catheter position [52]. Th ey evaluated six diff erent situations regarding the PEEP, body position and intra-abdominal pressure. Th eir results demonstrated that these factors may modify the EAdi catheter optimal position, although not compromising a stable signal due to the wide electrode array. One can therefore conclude that the optimal catheter position should be adjusted after major changes in ventilator settings, clinical condition or patient positioning.

Management of patient-ventilator synchrony
Th e time lag between the neural inspiratory input and the occurrence of a ventilator breath aff ects all steps of the respiratory cycle (initiation, insuffl ation, and cyclingoff for expiration) [53]. Among the diff erent forms of asynchrony, ineff ective triggering (also known as wasted eff ort) is the most common during invasive MV. During noninvasive ventilation (NIV), leaks at the patientventilator interface impair the function of the pneumatic trigger and cycling system [54], thus promoting specifi c asynchronies (autotriggering and prolonged insuffl ation) [55].
Ineff ective eff orts are explained both by patients' characteristics and by ventilator settings. Th e presence of intrinsic PEEP increases the patient eff ort required to trigger the ventilator, thereby increasing the likelihood that the patient's inspiratory eff ort will fail to trigger a ventilator breath [36,53,56]. A weak inspiratory eff ort, which may occur during situations of low respiratory drive such as excessive ventilation, is also a risk factor and is common in patients receiving high assist levels [22] or sedation [38]. An excessive level of pressure support is also associated with prolonged insuffl ation, thus promoting hyperinfl ation and intrinsic PEEP. Reduction of ineff ective eff orts is often possible through a careful optimization of ventilator settings, at least in short-term studies. Reducing V T during PSV can improve most factors contributing to ineff ective eff orts [49]. Th ille and colleagues showed that wasted eff orts could be decreased without increasing the patient's work of breath ing, with the main goal of decreasing the pressuresupport level to obtain V T values of about 6 ml/kg predicted body weight [49]. Because high pressuresupport levels are associated with prolonged insuffl ation beyond the end of the patient's neural inspiratory time, another useful means of decreasing wasted eff orts consists of adjusting the inspiratory time by increasing the fl ow threshold of the cycling criterion [49,57].

Neurally adjusted ventilatory assist and asynchrony
NAVA involves the transesophageal recording of diaphrag matic electrical activity using specifi cally designed technology to minimize measurement errors. Th e EAdi signal reliably monitors and controls the ventilatory assist [58]. During NAVA, the EAdi triggers the assist when the patient initiates an inspiratory eff ort -even during expiration with intrinsic PEEP -and a decrease in EAdi terminates the assist. NAVA does not therefore depend on measurements of airway pressure or fl ow and keeps the assist synchronous with the inspiratory eff orts (independent of the presence of leaks or intrinsic PEEP) [19,21,22,25,29,59]. NAVA thus has two important features: the delivered pressure is, in theory, synchronous with the diaphragmatic activity, and the V T is completely controlled by the output of the patient's respiratory control center [18]. A frequent form of minor patient-ventilator asynchrony is a long inspiratory trigger delay (time lag between the onset of neural inspiration, then the detection of a breath initiated by the patient and, fi nally, the onset of ventilator pressurization). Several factors may increase the inspiratory trigger delay during PSV, including the presence of intrinsic PEEP and suboptimal ventilator performance [60]. Th e cycling-off delay is the time diff erence between the end of the neural inspiratory ramp and the end of ventilator pressurization. Piquilloud and colleagues compared these delays and their consequences between NAVA and PSV in a group of 22 patients intubated for acute respiratory failure. Th e inspiratory trigger delay, the excess inspiratory time, and the frequency of patient-ventilator asynchrony were compared between the two modes [26]. Compared with PSV, NAVA substantially improved patient-ventilator synchrony by reducing the inspiratory trigger delay and the total number of asynchrony events, and by improving expiratory cycling-off .
Increasing the level of ventilatory assist with standard modes may expose the patient to potentially dangerous levels of volume and pressure, and to uncoupling between the patient's neural output and ventilator assistance. In contrast to PSV, there is good evidence that NAVA off ers protection against excessive Paw and V T values because there is a downregulation of EAdi in response to increasing assistance levels: the net result is a decrease in the amount of assistance provided [20,21,[61][62][63]. Th e absence of a V T increase with increasing NAVA levels suggests that the Hering-Breuer refl ex is operative [64], stopping the output from the respiratory control center at the same V T level, irrespective of the NAVA level. Unloading of the respiratory muscles is always partial, as some level of spontaneous activity is maintained, and patientventilator synchrony is improved.
Several studies have evaluated the impact of increasing PSV levels versus NAVA levels using similar methods of setting the ventilator [20][21][22]25]. Inspiratory pressure support was titrated in order to obtain 6 to 8 ml/kg predicted body weight during active inspiration. During PSV, the ventilator function 'NAVA Preview' estimates the NAVA level that would achieve the same peak inspiratory pressure. All studies performed in the ICU consistently showed that NAVA, in contrast to PSV, averted the risk of overassistance when the assist level was increased gradually. NAVA also improved patientventilator synchrony, in contrast to PSV, regardless of the underlying diagnosis. Very high levels of NAVA, however, might result in unstable periodic breathing patterns with delivery of high tidal volume followed by periods of apnea and signs of discomfort [65]. To separate the eff ects of neural triggering and those of proportional assis tance, Terzi and colleagues studied a selected popula tion of patients recovering from acute respiratory distress syndrome, using NAVA with two inspiratory triggers: the EAdi signal and the inspiratory fl ow threshold used previously for PSV [25] (Figure 1a,b). Not only proportional assistance but also neural triggering improved patient-ventilator synchrony in these patients during the weaning process.
All of the available studies of NAVA in ICU patients have limitations regarding the clinical applicability of the results. Except for two studies [19,25], the patient population was heterogeneous in terms of the cause of respiratory failure. Th e evaluation time was relatively short in eight studies, but not for two studies [24,27].

Matching alveolar ventilation to metabolic demand: role for the neural controller -variability
Interestingly, and for reasons that are not yet fully understood, NAVA compared with PSV seemed to improve the partial pressure of oxygen in arterial blood in some studies independent of changes in the partial pressure of carbon dioxide in arterial blood (PaCO 2 ) [25,27]. One hypo thesis is that the continuous spon taneous inspira tory activity during NAVA improves the matching between ventilation and perfusion. Earlier studies had established that partial ventilatory support allowing some degree of spontaneous breathing activity using modes of ventilation other than NAVA improved the ventilation/perfusion relationship compared with fully controlled MV [66]. In addition, NAVA allows a more natural breathing pattern characterized by greater variability, which may also contribute to improve gas exchange [67] (see below).
According to the principle of homeostasis, the closed loop that regulates PaCO 2 comprises: sensors (or detectors), which are chemoreceptors; a controller (or compara tor), which is the central respiratory command; and  eff ectors, which are the respiratory muscles. Each compo nent controls the next component in the loop, and the eff ectors change their activity (that is, adapt) to keep the PaCO 2 value relatively constant. In other words, EAdi and therefore the breathing pattern must adapt to a variety of conditions to maintain PaCO 2 within the normal range. Another regulatory mechanism is optimiza tion of the work of breathing. For example, the rate and/or the depth of breathing can be adjusted to minimize the energy expenditure at a given respiratory eff ort and/or to minimize the stretch on the lungs.
Any strategy based on automated feedback control of ventilatory support should ideally require neural information on the lung volume, rate of lung volume change, and transpulmonary pressure, which are provided by mechano receptors in the lungs and chest wall. Finally, the varia bility and complexity of the breathing pattern are infl u enced by several factors, including the load-capacity relationship of the respiratory system [68][69][70], vagal aff erent traffi c to the brain [71], and the activity of the central pattern generators [72].
Ventilatory activity is nonlinear in nature and exhibits chaos-like mathematical complexity [72,73]. Variability is a mathematically complex notion, often expressed using the coeffi cient of variation, which is the ratio of the standard deviation over the mean. However, the complexity of fl ow and EAdi variability can also be described using noise titration, the largest Lyapunov exponent, Kolmogorov-Sinai entropy, and three-dimensional phase portraits [74,75]. Schmidt and colleagues used these methods to compare respiratory variability and complexity during PSV and NAVA [23]. Compared with PSV, NAVA increased breathing pattern variability and fl ow complexity without changing EAdi complexity. Accordingly, when the NAVA level was increased from zero to a high level in healthy individuals, they adapted their inspiratory activity to the NAVA level in order to control V T and to regulate PaCO 2 over a broad range of NAVA settings [63]. In contrast, during high-level PSV, V T became almost entirely determined by the ventilator and hypocapnia developed as previously shown in healthy subjects [76,77]. Th ese diff erences between NAVA and PSV establish that with NAVA, even at a high level of assis tance, V T is not imposed by the ventilator but remains under the control of the patient's central respiratory command. NAVA therefore decreases the risk of overassistance. Th e extent to which the preserved variability associated with NAVA is benefi cial remains to be established. Whether variability restoration could be used to adapt NAVA settings also warrants further studies, as well as the development of specifi c tools for assessing variability at the bedside.
Patients with respiratory failure probably adjust their breathing activity to achieve the best compromise between the muscular eff ort needed to breathe and the sensory cost of tolerating elevated PaCO 2 levels. NAVA acts as an additional external cost-free muscle controlled by the central respiratory command. NAVA therefore does not seem to alter the closed loop that controls the PaCO 2 and respiratory pattern optimization. Accordingly, when introducing NAVA in patients with respiratory failure, progressively increasing the NAVA level allows the PaCO 2 (that is, V T ) to improve to the optimal value. Further NAVA level increases then lead to respiratory eff ort adjustments aimed at maintaining this optimal PaCO 2 value, but do not change V T [20].
Moreover, Karagiannidis and colleagues intended recently to evaluate the physiological eff ect of extra corporeal membrane oxygenation on the pattern of breathing in patients with severe lung failure treated with NAVA [78]. Th ey demonstrated that a downregulation of extra corporeal exchange gas transfer caused an immediate upregulation of ventilation. Eucapnia under NAVA was preserved because the patients adjusted their minute ventilation to their needs. Th ese interesting data highlighted once again that the ventilatory adaptation to maintain normocapnia remains under NAVA.

How can the optimal NAVA level be determined?
Determining the optimal NAVA level remains challenging, and several methods have been suggested. Contrary to PSV and as already described, NAVA generates V T levels that can remain constant independent of the assist level once the patient's ventilation needs appear to be satisfi ed [20]. Consequently, NAVA settings cannot be adjusted based solely on V T (and/or the corresponding PaCO 2 target).
Brander and colleagues tried to fi nd the best NAVA level using breathing pattern analysis during a titration procedure [20]. Titration consisted of starting at a minimal assist level of around 3 cmH 2 O and then increasing the NAVA level every 3 minutes in steps of 1 cmH 2 O per arbitrary unit (the amount of microvolts recorded from the EAdi signal). Th e response in terms of V T and Paw was biphasic. During the fi rst phase, V T and Paw increased while the esophageal pressure-time product (that is, inspiratory muscle eff ort) and EAdi decreased. Further increases in the NAVA level (second phase) did not signifi cantly change Paw or V T but continued to decrease the esophageal pressure-time product and EAdi. Th e fi rst phase may thus indicate an insuffi cient NAVA level to supplement the patient's weak breathing eff ort, while the beginning of the second phase may correspond to the minimal assist level that satisfi es the patient's respiratory demand. Th e optimal (or adequate) NAVA level may thus be indicated by the infl ection point of the airway pressure trend graph during a stepwise increase in the NAVA level ( Figure 2). In this study the patients were ventilated with these settings for 3 hours without experiencing adverse hemodynamic or respira tory events [20]. Interestingly, the optimal NAVA level occurred at about 75% of the highest EAdi obtained with the minimal NAVA level and PEEP [20].
As suggested, titration of the NAVA level may be performed by systematically increasing the NAVA level to determine the optimal setting with regard to unloading patient's respiratory muscles [20,61,79]. During a recent observational study, transferring patients to NAVA was uneventful and the NAVA level contributed to adjustments of the preset NAVA level [80]. Interpretation of several interacting physiological parameters might be diffi cult in cases in which there is no marked decrease in EAdi during NAVA titration [80]. An automated approach enabled faster identifi cation of the best NAVA level with a good accuracy [81].
Instead of stepwise titration, Rozé and colleagues tried to fi nd the best NAVA level using an EAdi target of 60% of the highest EAdi value recorded during spontaneous breathing [24]. Th is measurement was reassessed daily using a spontaneous breathing trial with a pressuresupport level of 7 cmH 2 O and no PEEP. Th is method proved feasible and well tolerated until extubation ( Figure 3). Th e 60% of the highest EAdi value threshold was based on a muscular rehabilitation protocol developed using data on diaphragmatic electromyogram activation during exercise [82]. Whether this approach is also optimal during assisted ventilation needs further evaluation. It is worth noting that EAdi measured during the daily spontaneous breathing trial increased steadily over time in all patients until successful extubation [24]. Th is improvement probably originated in multiple factors, including discontinuation of sedative agents and gradual restoration of the functional electrophysiologic activity of the diaphragm. Monitoring diaphragmatic activity may be of clinical interest and could be achieved using the NAVA electrode.
Using EAdi analysis to titrate NAVA is an interesting approach that could potentially be easier to use than the breathing pattern analysis method of Brander and colleagues (V T change during titration) [20]. Th e EAdi target of 60% of the highest EAdi value with 7 cmH 2 O of PSV proposed by Rozé and colleagues should be used cautiously [24], as Brander and colleagues found that the EAdi at the optimal NAVA level was equal to 75% of the highest EAdi value recorded with minimal NAVA (inspiratory Paw above PEEP = 3 cmH 2 O) [20]. Further studies are clearly needed to better determine the optimal EAdi target.

Noninvasive ventilation, sleep and NAVA
NIV is a specifi c clinical situation during which the occurrence of leaks may greatly aff ect patient-ventilator interactions, thereby complicating the determination of optimal ventilator settings. In a study by Vignaux and colleagues, more than 40% of patients experienced various types of asynchrony during conventional NIV and the asynchrony rate correlated with the level of leakage [83]. With NAVA, assistance is delivered based on neural triggering, which is not aff ected by leakage. NAVA may thus, in theory, diminish asynchrony events, thereby improving the tolerance of NIV. New software for NIV has been developed using NAVA technology. With this specifi cally designed algorithm, NIV assistance is triggered and cycled-off by the neural diaphragmatic activity, which would be expected to improve patientventilator synchrony during NIV. Th is hypothesis has not yet been fully investigated.
A study of NIV-PSV with a helmet interface in healthy volunteers compared asynchrony with a neural trigger and a conventional pneumatic trigger [59]. Increasing PSV levels and respiratory rates applied with neural trigger ing and cycling-off produced signifi cantly less impair ment of synchrony, trigger eff ort, and breathing  comfort, compared with conventional pneumatic triggering and cycling-off . Cammarotta and colleagues recently compared NAVA and NIV-PSV delivered through a helmet interface in postextubation hypoxemic patients [32]. Ten patients underwent three 20-minute trials of helmet NIV in PSV, NAVA, and PSV again. Th e authors demonstrated that there was less asynchrony during NAVA than during PSV and no diff erence in gas exchange, although there were more leaks during NAVA. Moreover it is important to underline that the PSV mode chosen was specifi cally dedicated to NIV, whereas the NAVA mode dedicated to NIV that is now currently available did not exist at the time of this study.
Recent data obtained in low-birth-weight infants indicate that NAVA can maintain synchrony -both in terms of timing and proportionality -even after extubation in patients with an excessively leaky interface under NIV (all infants in this study were ventilated using a single nasal prong) [29].
Another consideration for NIV that deserves attention in the near future is the impact on swallowing, phonation, and sleep quality, most notably when NIV is used for several days. Improvements in swallowing performance have been reported in neuromuscular patients receiving MV compared with spontaneous breathing [84,85]. Th e close relationship between the muscles involved in swallowing and those contributing to inspiration was evidenced by Orlikowski and colleagues using an original method of tongue-strength measurement. Th e signifi cant tongue weakness observed in 16 weak patients with Guillain-Barré syndrome correlated with the alterations in respiratory parameters [86]. Additional physiological studies are required to document the potential benefi ts of NAVA on swallowing-breathing interactions during NIV.
Sleep quality during NIV has been shown to be a predictor of success or failure [87]. Sleep quality can also be improved compared with standard NIV settings by careful physiological titration of the ventilator settings [88]. Patient-ventilator asynchrony can cause sleep disrup tion. Bosma and colleagues demonstrated that PAV, a mode of partial ventilatory support in which the ventilator applies pressure in proportion to the inspiratory load, was more eff ective than PSV in matching the ventilatory requirements to the level of ventilator assistance, thereby resulting in fewer patient-ventilator asynchronies and better quality of sleep [11]. Delisle and colleagues recently obtained sleep recordings during a crossover study comparing NAVA and PSV in 14 mechanically ventilated patients [89]. Each condition was studied for 4 hours, and recordings were obtained over 19 consecutive hours in all. Patient-ventilator asynchrony varied signifi cantly across sleep stages, and no asynchrony occurred with NAVA. Overassistance occurred only with PSV, which probably explained the improvements in physiological indices of sleep quality observed with NAVA.

Neurally adjusted ventilatory assist in children and infants
MV in children and in low-birth-weight infants is more diffi cult to apply than in adults and has several speciicities. First infants take a very small tidal volume, have a rapid respiratory rate, have a limited chest wall musculature, and have variable and fl uctuating lung compliance. Second, most neonatal units use uncuff ed tracheal tubes for fears of pressure necrosis and air leak is always present, making reliable measurements and triggering problematic. Th ird, ventilators that are effi cient in adults are not systematically effi cient in children and infants, mainly because the inspiratory triggers are not suffi ciently sensitive for early detection of infants'/children's inspiratory eff ort [90].
Whether or not the respiratory drive of the preterm infant is suitable to control MV is unknown. Beck and colleagues fi rst evaluated patient-ventilator interaction with NAVA in seven very-low-birth-weight infants [29]. As suggested by previous animal studies [91], they demonstrated that NAVA could be implemented for a short-term period, both invasively and noninvasively, in infants with body weight as low as 640 g up to 3 years old. During invasive ventilation with NAVA, EAdi and ventilator pressure were correlated and patient-ventilator synchrony was improved compared with the other mode. Moreover, this synchrony persisted after extubation while ventilating the patient with an excessively leaky interface. After this fi rst physiological demonstration, Bengtsson and Edberg demonstrated the clinical feasibility and safety with use of NAVA in pediatric patients [30]. Similarly, Breatnach and colleagues compared NAVA (with a neural trigger) and PSV (with a pneumatic trigger) in 16 ventilated infants [31]. Th is prospective crossover comparison demonstrated that ventilation with NAVA improved patient-ventilator synchrony.
Furthermore, Alander and colleagues recently compared NAVA with pressure-controlled ventilation for newborns and with pressure-regulated controlled ventilation for children older than 3 months (with conventional trigger modes: pressure and fl ow trigger) [92]. In this prospective cross-over study, 18 patients requiring MV were randomized for 10 minutes with the diff erent modes. During NAVA, the peak airway pressure was lower, the respira tory rate was 10 breaths/minute higher than in the pressure group, and patient-ventilator synchronization was improved. However, there were no diff erences in tidal volume and in oxygen saturation.
To evaluate the eff ects of the neural trigger on trigger delay, ventilator response time, or work of breathing, Clement and colleagues conducted a study in 23 pediatric patients aged 0 to 24 months with a diagnosis of bronchio litis presenting respiratory failure requiring MV [33]. Th e authors compared the neural trigger and the pneumatic trigger using similar NAVA assistance, and observed that the trigger delay, the ventilator response time, and the work of breathing were reduced by the neural trigger.
Finally, all of these studies seem to demonstrate the feasibility of and a potential advantage for NAVA in children compared with the other assisted ventilatory modes. Because patient-ventilator synchrony is improved with NAVA, the children may require lower doses of sedation with this mode of MV [93], which could reduce the time of MV.

Future research
Clinical studies obtained in critically ill patients confi rm many of the expected short-term physiological benefi ts associated with NAVA, as discussed above.
Particularly, NAVA seems to markedly improve the problems of nonsynchronization between the patient and the ventilator and the problems of risk of overventilationincluding the risk of ineff ective or missed inspiratory eff orts due to intrinsic PEEP observed in chronic obstructive pulmonary disease patients, or to a rapid breathing frequency with a very small tidal volume observed in pediatric patients.
In addition, NAVA minimizes the risk of overinfl ation because the duration and level of pressurization remain under respiratory-center control, and minimizes the risk of diaphragmatic inactivity because the presence of pressure assistance requires the presence of this inspiratory activity.
A preserved respiratory muscle function is pivotal for weaning from MV [44]. By using NAVA, which outperforms the previous modes of MV for adequately assisting the patient's inspiratory eff ort without inducing patient/ventilator dyssynchrony, a reduction in the duration of MV could be expected. Studies are needed to evaluate the best time to begin the weaning process with NAVA.
Th e NAVA setting is an important question not yet fully resolved. If clinicians are accustomed to set a PSV level, this is not the case for NAVA. Furthermore, because the breathing pattern is less modifi ed by the NAVA setting than during PSV, it is much less informative for NAVA adjustment. As described above, the literature suggests that the adjustment should consider the electromyographic activity of the diaphragm, but this method is not simple. As recently proposed, a direct evalu ation of patient comfort and sense of dyspnea for the NAVA setting should be evaluated [23].
Finally, the next research step will be to evaluate NAVA over longer periods, in order to know whether this mode can replace the modes usually used during MV and the weaning period, like PSV. Appreciating the safety, the feasibility and the constraints of this technology will be useful. It is therefore necessary to test, during the total weaning period, the eff ectiveness of the esophageal probe and to know whether it is regularly necessary to adjust the probe position. One of the most diffi cult questions to address, however, is in which situations it is not desirable to let the respiratory centers drive the ventilation. Situations of severe metabolic acidosis, of high respiratory drive and of high catecholamine levels may induce situations of extreme hyperventilation, which may be dangerous for the lungs. When sedation and/or paralysis become necessary is therefore an important question to address before widespread use of this mode [94].

Conclusion
NAVA, which is based on an original physiological concept, adds new knowledge on patient-ventilator interactions during spontaneous breathing, thus helping to unravel the complex mechanisms involved in breathing control during MV. Th ere is compelling evidence that NAVA, as well as the PAV+ software, improves patientventilator interactions and increases respiratory variability in comparison with PSV. Th is advantage holds potential for many applications. Th e short-term and long-term experi ence with NAVA, however, remains scant. Further clinical studies are needed to assess the feasibility and safety of NAVA. A key challenge is how to determine the best NAVA settings according to the patient's ventilatory needs and the acceptable level of work of breathing.

Abbreviations
EAdi, electrical activity of the diaphragm; MV, mechanical ventilation; NAVA, neurally adjusted ventilatory assist; NEXmod, nose to ear lobe to xiphoid process of the sternum distance modifi ed for the EAdi catheter; NIV, noninvasive ventilation; PaCO 2 , partial pressure of carbon dioxide in arterial blood; PAV, proportional-assist ventilation; Paw, airway pressure; PEEP, positive end-expiratory pressure; PSV, pressure-support ventilation; V T , tidal volume.

Author contributions
NT, J-CMR and LB initiated and wrote the manuscript; all authors contributed to the revision of this. All authors read and approved the fi nal manuscript.