New setting of neurally adjusted ventilatory assist for noninvasive ventilation by facial mask: a physiologic study

Background Noninvasive ventilation (NIV) is generally delivered using pneumatically-triggered and cycled-off pressure support (PSP) through a mask. Neurally adjusted ventilatory assist (NAVA) is the only ventilatory mode that uses a non-pneumatic signal, i.e., diaphragm electrical activity (EAdi), to trigger and drive ventilator assistance. A specific setting to generate neurally controlled pressure support (PSN) was recently proposed for delivering NIV by helmet. We compared PSN with PSP and NAVA during NIV using a facial mask, with respect to patient comfort, gas exchange, and patient-ventilator interaction and synchrony. Methods Three 30-minute trials of NIV were randomly delivered to 14 patients immediately after extubation to prevent post-extubation respiratory failure: (1) PSP, with an inspiratory support ≥8 cmH2O; (2) NAVA, adjusting the NAVA level to achieve a comparable peak EAdi (EAdipeak) as during PSP; and (3) PSN, setting the NAVA level at 15 cmH2O/μV with an upper airway pressure (Paw) limit to obtain the same overall Paw applied during PSP. We assessed patient comfort, peak inspiratory flow (PIF), time to reach PIF (PIFtime), EAdipeak, arterial blood gases, pressure-time product of the first 300 ms (PTP300-index) and 500 ms (PTP500-index) after initiation of patient effort, inspiratory trigger delay (DelayTR-insp), and rate of asynchrony, determined as asynchrony index (AI%). The categorical variables were compared using the McNemar test, and continuous variables by the Friedman test followed by the Wilcoxon test with Bonferroni correction for multiple comparisons (p < 0.017). Results PSN significantly improved patient comfort, compared to both PSP (p = 0.001) and NAVA (p = 0.002), without differences between the two latter (p = 0.08). PIF (p = 0.109), EAdipeak (p = 0.931) and gas exchange were similar between modes. Compared to PSP and NAVA, PSN reduced PIFtime (p < 0.001), and increased PTP300-index (p = 0.004) and PTP500-index (p = 0.001). NAVA and PSN significantly reduced DelayTR-insp, as opposed to PSP (p < 0.001). During both NAVA and PSN, AI% was <10% in all patients, while AI% was ≥10% in 7 patients (50%) with PSP (p = 0.023 compared with both NAVA and PSN). Conclusions Compared to both PSP and NAVA, PSN improved comfort and patient-ventilator interaction during NIV by facial mask. PSN also improved synchrony, as opposed to PSP only. Trial registration ClinicalTrials.gov, NCT03041402. Registered (retrospectively) on 2 February 2017.


Background
Noninvasive ventilation (NIV) is increasingly used for treating acute respiratory failure (ARF) [1,2] and is commonly applied using a facial mask [3] and pneumatically triggered and cycled-off pressure support (PS P ) [4]. Although better tolerated than invasive mechanical ventilation, NIV is characterized by drawbacks such as poor patient-ventilator interaction and discomfort [5], which are major determinants of NIV failure.
In particular, the pneumatic signals, i.e., flow, volume and airway pressure (Paw), are leak-sensitive [6] and frequently cause patient-ventilator asynchrony [7]. The only mode not utilizing pneumatic signals to trigger and drive the ventilator is neurally adjusted ventilator assist (NAVA). In fact, with NAVA the ventilator assistance is under the control of the diaphragm electrical activity (EAdi) [8]. In contrast to PS P , NAVA has been repeatedly shown to improve patient-ventilator interaction and reduce asynchronies, both during invasive ventilation [9,10] and NIV [4,[11][12][13][14][15]. However, NAVA is characterized by a lower rate of pressurization than PS P [4].
Recently, a specific NAVA setting has been proposed to generate EAdi-controlled pressure support (PS N ) in patients receiving either invasive ventilation [16] or NIV by helmet [4]. PS N consists of increasing the usercontrolled gain factor (NAVA level) at the maximum level, while limiting peak airway pressure (Paw peak ) by adjusting the upper pressure limit [4,16].
During NIV delivered by helmet, compared to both PS P and NAVA, PS N results in better pressurization and triggering performance, which improves patient comfort while reducing EAdi, without affecting the respiratory rate and gas exchange [4]. Due to the different characteristics of helmets and masks, it is unclear whether these advantages could be extended to NIV delivered by mask. This physiological study aims at comparing PS N with PS P and NAVA, with respect to the patient's comfort (primary endpoint), breathing pattern, respiratory drive, gas exchange, pressurization and triggering performance and patientventilator synchrony (additional endpoints).

Methods
The present physiologic, crossover, randomized study was conducted from March to September 2013 in the Intensive Care Units (ICUs) of the University Hospital "Maggiore della Carità" (Novara, Italy) and the ZhongDa Hospital, Southeast University (Nanjing, China). The study was approved by the local Ethics Committees "A.O.U Maggiore della Carità" in Novara, Italy (protocol n°64/12) and the Research Ethics Board of Zhongda Hospital, Southeast University, Nanjing, China (2013ZDSYLL097.0). Written informed consent was obtained from the patients for publication of their individual details and accompanying images in this manuscript. The consent forms are held by the authors and are available for review by the Editor-in-Chief. At the time the study was conducted, trial registration was not mandatory for this type of investigation; however, the trial was retrospectively registered at ClinicalTrials.gov (NCT03041402). We followed the Consolidated Standards of Reporting Trials (CONSORT) recommendations for reporting of randomized trials [17].

Patients
We considered any patient eligible who was ≥18 years of age and admitted to the ICU, and who was orally intubated and undergoing invasive mechanical ventilation for at least 48 hours. The inclusion criteria were: (1) consciousness, as indicated by a Glasgow Coma Scale (GCS) of 11 (i.e. spontaneous eye opening, response to command and no verbal response because of the endotracheal tube in place); (2) no infusion of midazolam or propofol in the previous 24 hours or 4 hours, respectively; and (3) readiness for extubation with indication, prior to extubation, to receive NIV to prevent postextubation respiratory failure. The patients were considered to be eligible for the spontaneous breathing trial if they met the following criteria [18]: (1) GCS ≥8; (2) presence of clearly audible cough during suctioning; (3) tracheal suctioning ≤2/hour; (4) normal sodium blood values; (5) core temperature <38.5°C during the previous 8 hours; (6) arterial oxygen tension (PaO 2 ) to fraction of inspired oxygen (FiO 2 ) ratio (PaO 2 /FIO 2 ) ≥200 with positive end-expiratory pressure (PEEP) ≤5 cmH 2 O; (7) FiO 2 ≤ 0.4; (8) heart rate ≤125 beats/min; and (9) systolic blood pressure >90 mmHg without epinephrine or norepinephrine infusion and with dopamine infusion ≤5 mcg/kg/min. The patients considered to be at risk of extubation failure exhibited at least one of the following: (1) more than one consecutive failure of the weaning trial [19]; (2) arterial partial pressure of carbon dioxide (PaCO 2 ) >45 mmHg at the end of the 30-min spontaneous breathing trial [20]; (3) chronic respiratory disorders [19]; and (4) chronic heart failure [19].

Study protocol
After the patient's enrolment in the study, the nasalgastric feeding tube in place was replaced by the EAdi catheter (Maquet Critical Care, Solna, Sweden) [9]. The correct positioning was ascertained as previously described [9]. The study was performed using a standard Servo-I ventilator (Maquet Critical Care, Solna, Sweden) equipped with NAVA module and NIV software for air leaks. The facial mask was individually selected for each patient based on their anthropometric characteristics to minimize air leaks and optimize patient tolerance; the facial mask was selected from among three different models: FreeMotion RT041 Non Vented Full Face Mask (Fisher and Paykel, Auckland, New Zealand); Ultra Mirage FFM-NV (ResMed, San Diego, CA, USA); and PerforMax Face Mask (Philips Respironics, Murrysville, PA, USA).
Immediately after extubation, we performed a 15-min PS P trial, setting the inspiratory pressure support ≥8 cmH 2 O to obtain a tidal volume of 6-8 mL · kg -1 of ideal body weight, with the fastest rate of pressurization and I/E cycling at 35% of peak inspiratory flow (PIF). All patients subsequently underwent three 30-min trials in random order: (1) PS P , with the settings obtained in the aforementioned trial; (2) NAVA, adjusting the NAVA level in order to achieve a comparable peak EAdi (EAdi peak ) as during the PS P trial, with a safety Paw upper limit of 30 cmH 2 O [4,15]; and (3) PS N , setting the NAVA level at its maximum (i.e., 15 cmH 2 O/μV), and an upper Paw limit to obtain the same overall Paw applied during the PS P trial [4,16,21]. During both NAVA and PS N , the trigger sensitivity was set at 0.5 μV while the default cycling-off was 70% EAdi peak , as fixed by the manufacturer [21]. PEEP was set by the attending physicians in a range between 5 cmH 2 O and 10 cmH 2 O, and it remained unmodified throughout the entire study period. The FiO 2 was regulated to obtain peripheral oxygen saturation (SpO 2 ) between 94% and 96%, before starting the protocol, and it remained unmodified throughout the study period.
The three modes of ventilation were applied according to a computer-generated random sequence using sealed, opaque, numbered envelopes. The envelopes were kept in the head nurse's office in both institutions. The envelope was opened by the nurse in charge of the patient, and the prescribed sequence of modes was communicated to the investigators.

Data acquisition and analysis
Airflow, Paw and EAdi were acquired from the ventilator using an RS232 interface at a sampling rate of 100 Hz and were recorded on a computer using dedicated software (ServoTracker V. 4.0, Maquet Critical Care, Solna, Sweden). The last minute of each trial was manually analysed off-line using customized software based on Microsoft Excel, as previously described [9].
Comfort was assessed through an 11-point numeric rating scale (NRS), as previously reported [4,[22][23][24]. Before protocol initiation, all patients received a detailed explanation of the NRS. The patients were asked to evaluate their comfort level, indicating a number between 0 (worst possible comfort) and 10 (best possible comfort) using an ICU-adapted large-printed scale including numbers and descriptors [23]. The scores obtained were recorded without additional indications or comments [24].
Breathing pattern was assessed by determining (1) mechanical inspiratory time (TI mec ), breath duration (TTOT mec ) and rate of ventilator cycling (RR mec ) from the flow tracing, and (2) the patient's own (neural) inspiratory time (TI neu ), breath duration (TTOT neu ) and respiratory rate (RR neu ) from the EAdi tracing. The mechanical (TI/TTOT mec ) and neural (TI/TTOT neu ) inspiratory duty cycles were also calculated [15,25]. Air leaks were computed over one minute as the difference between inspiratory and expiratory tidal volumes times RR mec and were expressed as percentage of the exhaled volume over one minute [15,25]. Moreover, we measured Paw peak , peak inspiratory flow (PIF) and the time to reach PIF from the onset of the patient's effort (PIF time ). EAdi peak was also determined as an index of respiratory drive [26]. Gas exchange was assessed at the end of each trial by sampling arterial blood from a catheter already inserted for clinical purposes.
To evaluate the pressurization performance, we computed the pressure-time product (PTP) of the first 200 ms from the onset of the ventilator pressurization (PTP 200 ), and the PTP of the first 300 ms and 500 ms from the onset of the neural effort, expressed as the percentage of the area of ideal pressurization (PTP 300-index and PTP 500-index , respectively) [4,24,27,28]. The ideal PTP was computed considering a perfectly squared rectangle on the Paw-time tracing, with the height of the actual Paw above PEEP and the width of the time window considered (i.e., 0.3 second and 0.5 second from the onset of the inspiratory effort, assessed from the EAdi tracing, for PTP 300-index and PTP 500-index , respectively) [4,24,27,28]. The triggering performance was evaluated by determining the pressure drop (ΔP trigger ) and PTP of Paw (PTP t ) during the triggering phase [4,24,27,28].
To assess patient-ventilator synchrony, we computed the inspiratory trigger delay (Delay TR-insp ), as the time lag between the onsets of neural inspiration and ventilator support, and the expiratory trigger delay (Delay TRexp ), as the time lag between the fall towards baseline of EAdi and the end of ventilator support. The time during which respiratory effort and ventilator assistance were synchronous, indexed to the TI neu (Time synch /TI neu ), was also computed [4,24,27]. The asynchrony index (AI%) was calculated as the total number of asynchronies (i.e., ineffective efforts, auto-triggers and double-triggers) divided by the sum of triggered and non-triggered breaths [7]. An AI% ≥10% was considered to indicate a clinically relevant rate of asynchronies [7].

Statistical analysis
To detect an increase in comfort of 2.5 [4], with α risk of 0.05 and β risk of 0.20, a sample of 12 patients was deemed necessary. Because this calculation was based on a pairwise comparison and we actually compared three conditions, we applied the Bonferroni correction, which reduced the α risk from 0.05 to 0.017, increasing the sample size up to 14 patients. We used non-parametric tests because of the relatively small number of patients. The data are reported as median values (25-75% interquartile), unless otherwise specified. All continuous variables were compared between modes using the Friedman test and then by the Wilcoxon test; the Bonferroni correction was applied for multiple comparisons (p < 0.017). We compared the categorical data using the McNemar test. The Spearman rank correlation test was used to ascertain the correlation between each individual comfort score and the corresponding PTP 300-index , PTP 500-index , PTP t , Delay TR-insp , PIF and PIF time . For these comparisons, we considered two-sided p values <0.05 significant. All statistical analyses were performed using the Sigmaplot v. 12.0 (Systat Software Inc., San Jose, CA, USA). No interim analysis has been planned or conducted.

Results
We enrolled 14 consecutive patients. The patients' study flow is shown in Fig. 1. All patients completed the study protocol without any complication and were included in the data analysis. No patient required either sedative or analgaesic drugs during the study period. No patients met any criteria for post-extubation respiratory failure requiring re-intubation. The patients' demographic and anthropometric characteristics are shown in Table 1.

Breathing pattern, respiratory drive and gas exchange
As reported in Table 2, the breathing pattern was not different between modes. Only TI/TTOT mec was significantly lower during PS P , as opposed to both NAVA (p = 0.007) and PS N (p = 0.010). Paw peak (p = 0.607), air leaks (p = 0.395) and respiratory drive, as indicated by the EAdi peak (p = 0.931), were also not different between modes. PIF did not differ between the three modes of ventilation (p = 0.109), while PIF time was significantly Fig. 1 Enrolment of the study participants. The flow of patient assessment and inclusion in the protocol is shown. A total of 54 patients were considered eligible for the study, having met all inclusion criteria: 40 patients were excluded from the study because they met one or more of the exclusion criteria. Therefore, 14 patients were included in the study. No protocol discontinuations were recorded. EAdi diaphragm electrical activity reduced by PS N , as opposed to both PS P and NAVA (p < 0.001 for both comparison), with no differences between PS P and NAVA (p = 0.217). Figure 3 shows, from top to bottom, Paw, flow and EAdi tracings of one representative patient undergoing PS P (left), NAVA (middle) and PS N (right). The arrow indicates an ineffective inspiratory effort during PS P . The median group values are presented in Table 2.
Gas exchanges were no different between trials ( Table 2).
Pressurization and triggering performance Figure 4 depicts Paw profiles of individual breaths during PS P (solid line), NAVA (dotted line) and PS N (dashed line) from another patient. The arrow indicates the beginning of the patient's own (neural) effort. PS P and PS N have similar Paw profiles, characterized by a fast rate of pressurization; however, during PS N the beginning of pressurization is notably anticipated and closer to the onset of the patient's effort. NAVA is characterized by a slower rate of pressurization. Consistent with these findings, PS N improved both PTP 300-index and PTP 500-index , as opposed to both PS P and NAVA (Table 2), whereas PTP 200 was lower during NAVA, as compared to both PS P and PS N (p < 0.001 for both comparisons), with no significant difference between PS P and PS N (p = 0.761). Shown also in Table 2, NAVA and PS N significantly reduced Delay TR-insp , PTPt and ΔP trigger , in contrast to PS P (p < 0.001 for all comparisons). Delay TR-exp was no different between modes (p = 0.395).

Patient-ventilator synchrony
Compared to PS P , both NAVA (p = 0.005) and PS N (p = 0.002) improved Time synch /TI neu , with no differences between the two (p = 0.08) ( Table 2). The median values of AI% are reported in Table 2. As expected, during both NAVA and PS N , the AI% was <10% in all patients, whereas it was ≥10% in 7 patients (50%) with PS P (p = 0.023, compared to both NAVA and PS N ).

Discussion
This physiologic study shows that in patients receiving NIV by facial mask, compared to both PS P and NAVA, PS N improves pressurization and triggering performance, resulting in better comfort, while not affecting respiratory drive, Arterial Blood Gases ABGs and respiratory rate. Both PS N and NAVA equally improve patient-ventilator synchrony, in contrast to PS P .
To the best of our knowledge, this investigation is the first to evaluate PS N for delivery of NIV using a mask. In a study evaluating intubated patients with COPD and intrinsic PEEP, compared to PS P , PS N improved patientventilator interaction and synchrony, and counterbalanced the extra load due to intrinsic PEEP without the need for externally applied PEEP [16]. In healthy volunteers, comfort was reduced when increasing the level of support [29], whereas it was improved by EAdi triggering, as opposed to pneumatic triggering, during NIV delivered by helmet [30]. In a recent study comparing PS N with PS P and NAVA during NIV delivered by helmet in an analogous patient population, PS N improved comfort, pressurization and triggering performance, and reduced EAdi, without affecting gas exchange [4].
Consistent with the results of these investigations, in the present study PS N outperforms PS P with respect to  [4,16,30], and comfort [4,30]. In accordance with Cammarota et al. [4], who compared the same three modes delivering NIV by helmet, PS N improved pressurization PTP 300-index and PTP 500-index , and comfort with respect to both PS P and NAVA, while in contrast to that study, PS N neither increased PTP 200 , compared to PS P , nor reduced EAdi, compared to both PS P and NAVA. These discrepancies are likely due to the different physical properties of mask and helmet, the latter being characterized by more problematic triggering and pressurization performance [31]. Nonetheless, we found improvements in triggering and pressurization performance to ameliorate comfort, which is a major determinant of NIV outcome. Indeed, NIV can be complicated by discomfort, which is associated with increased rate of failure and worsened patient outcome [32]. PIF was not different between modes, while PIF time was shortened by PS N , as opposed to both PS P and NAVA. In intubated patients with acute on chronic respiratory failure undergoing PS P , Bonmarchand et al. evaluated the effects of varying Paw rates of pressurization; they found that the fastest rate generated the highest PIF and was associated with greater  PS P pneumatically triggered and cycled-off pressure support, NAVA neurally adjusted ventilatory assist, PS N neurally controlled pressure support, RR mec ventilator respiratory rate, RR neu patient's respiratory rate, TI mec inspiratory time of the ventilator, TI neu inspiratory time of the patient, TI/TOT mec ventilator inspiratory duty cycle, TI/TOT neu patient's inspiratory duty cycle, Paw peak peak airway pressure, PIF peak inspiratory flow, PIF time time to reach the PIF, EAdi electrical activity of the diaphragm, EAdi peak peak value of EAdi, PaCO 2 arterial partial pressure of carbon dioxide, PaO 2 /FiO 2 ratio between arterial partial pressure and inspired fraction of oxygen, PTP pressure time product, PTP 300-index PTP of the first 300 ms since the effort of the patient indexed to the ideal PTP, PTP 500-index PTP of the first 500 ms since the effort of the patient indexed to the ideal PTP, PTP 200 PTP of the first 200 ms since the beginning of pressurization, PTPt PTP of the trigger, ΔP trigger drop of pressure during triggering phase, Delay TR-insp inspiratory trigger delay, Dealy TR-exp expiratory trigger delay, Time synch /TI neu synchronous time between respiratory effort and ventilator assistance, indexed to the TI neu , AI% asynchrony index. *p < 0.017 PS P vs. NAVA, # p < 0.017 PS P vs. PS N , § p < 0.017 NAVA vs. PS N reduction in the work of breathing [33]. Similar results were obtained during invasive PS N in restrictive patients [34] and in patients recovering from hypoxemic ARF [35].
To explain the differences between these studies and our investigation, it is important to note the different computational approach to the pressurization indexes [27]. PTP 200 reflects the sole rate of pressurization rate, i.e., the slope of Paw after triggering, which affects the PIF. Both PTP 300-index and PTP 500-index instead consider not only the pressurization rate but also the triggering performance, which influences PIF time , without affecting PIF. We found PTP 200 no different between PS N and PS P , while triggering performance was significantly improved by PS N , as indicated by the lower values of PTPt and Delay TR-insp . Notably, while patient comfort is improved when flow delivery by the ventilator meets the patient's demand [36], excessively high PIF may worsen the patient's comfort during both invasive ventilation [37] and NIV [36].

Conclusions
Compared to both PS P and NAVA, in patients receiving NIV by facial mask, PS N improves triggering performance and patient-ventilator synchrony, thereby ameliorating the patient's comfort. It remains to be determined whether these physiologic benefits may also occur in other categories of patients and translate into improved clinical outcomes.

Funding
Maquet Critical Care (Solna, Sweden) provided the NAVA module and catheters used for the study. A portion of the results from this study was presented in abstract form at the International Symposium on Intensive Care and Emergency Medicine in Brussels (2014).

Availability of data and materials
The full protocol and raw data are available at longhini.federico@gmail.com.
Authors' contributions FL was responsible for conception and design of the study, acquisition, analysis and interpretation of the data and for drafting and revising the article for final approval of the version to be published. CP was responsible for the conception and design of the study, acquisition, analysis and interpretation of data and for drafting and revising the article for final approval of the version to be published. JX and GC were responsible for the acquisition of data and for revising the article for final approval of the version to be published. AB and EG were responsible for analysis and interpretation of the data and for drafting and revising the article for final approval of the version to be published. YY participated in the design of the stud, acquisition and analysis of the data and in revising the article for final approval of the version to be published. PN and HQ were responsible for the conception and design of the study, analysis and interpretation of data and for drafting and revising the article for important intellectual content and final approval of the version to be published. All authors have read and approved the final manuscript.

Ethics approval and consent to participate
The study was approved by the local Ethics Committees "A.O.U Maggiore della Carità" in Novara, Italy (protocol n°64/12) and the Research Ethics Board of Zhongda Hospital, Southeast University, Nanjing, China (2013ZDSYLL097.0). Written informed consent was obtained from each participant before inclusion in the study, according to the local regulations and principles outlined in the Helsinki declaration. At the time the study was conducted, trial registration was not mandatory for this type of investigation.

Consent for publication
All patients gave consent for data publication according to national regulations.