Neurally Adjusted Ventilatory Assist (NAVA) or Pressure Support Ventilation (PSV) during spontaneous breathing trials in critically ill patients: a crossover trial

We conducted a crossover trial in the respiratory ICU of a university hospital in São Paulo, Brazil, from June 2011 to September 2013. The institution’s ethics committee, CAPpesq (University of Sao Paulo Medical School Ethics’ committee) approved the study (ID 0336/10) and a family member of each participant provided informed consent before inclusion. The study was registered at ClinicalTrials.gov (NCT01337271).

We included a convenience sample of 20 patients since data on the performance of NAVA during SBTs was not available in the literature. We prospectively enrolled consecutive patients admitted to the ICU during the study period who had received mechanical ventilation for more than 48 h and who the ICU team considered to be ready to undergo an SBT. Patients were included only if this was the first SBT attempt. Exclusion criteria were age less than 18 years, pregnancy, tracheotomy, participation in other clinical trials, and contraindications to the placement of the esophageal catheter (nasal pathologies, facial trauma or burns, or esophageal varicose or gastro esophageal bleeding in the past 30 days).

Patients were ventilated with the Servoi Ventilator (Maquet, Sweden), using heated humidification. We left the ventilator settings at baseline to the discretion of the ICU team.

As previously described [23, 24], we measured EAdi using a dedicated NAVA catheter (Maquet, Sweden), which has a multiple array of electrodes placed at its distal end to capture diaphragmatic electrical activity. A specific function of the ventilator was used to guide the correct positioning of the catheter.

Before initiating the SBTs, we set PSV to 5 cmH₂O and positive end-expiratory pressure (PEEP) to 5 cmH₂O for 5 min and we titrated the NAVA level to generate a peak inspiratory pressure of 10 cmH₂O [14].

All patients underwent two SBTs, one in PSV, and one in NAVA, separated by a washout period of 1 h, during which they received ventilation with baseline settings. We randomized the order of the SBTs using a computer generated randomization list, concealed by sequentially numbered, opaque, and sealed envelopes.

SBTs in PSV used a pressure support of 5 cmH₂0 and PEEP of 5 cmH₂O. We did not change the baseline flow-trigger sensitivity, cycling-off criteria, or FIO₂. SBTs in NAVA used the NAVA level titrated as described above, a PEEP of 5 cmH₂O, and baseline FIO₂ and flow-trigger sensitivity. NAVA triggering sensitivity and cycling off are fixed at 0.5 μV and 70% of EAdi peak, respectively. NAVA could be triggered either by EAdi or flow triggering, whichever happened first.

The SBTs lasted 30 min each. We suctioned the endotracheal tube before the beginning of each test and made no changes in ventilator setting during the SBTs. The ICU team was responsible for evaluating the patient during the SBT and deciding whether or not the patient had passed each test. We blinded the ICU team to the order of the SBTs to prevent bias on the SBT evaluation. We partially covered the ventilator screen during the SBTs to hide the ventilator waveforms in order to prevent unbinding. Respiratory rate, tidal volume, and minute volume were visible on the right side of the screen. We instructed clinicians caring for the patient to interrupt the test according to the standard of care criteria for SBT failure: respiratory rate greater than 35 rpm or less than 6 rpm; hypoxemia; changes in mental status; new onset of cardiac arrhythmia, tachycardia, or bradycardia; diaphoresis; or signs of respiratory distress, but investigators did not interfere with clinicians decision to interrupt the test in any circumstance . If the test was not interrupted for one of these objective criteria, investigators asked ICU clinicians if the patient had failed or passed the SBT at the end of the SBT. If the patient failed the first SBT, the ICU team was instructed to allow the second SBT after the washout period unless they considered it to be unsafe for the patient. The investigators did not participate in the decision to interrupt an SBT or to extubate the patient at the end of the second SBT. At the end of the protocol, we disclosed the order of the SBTs to the ICU team and instructed them to use the result of the SBT in PSV to decide whether or not to extubate the patient. We followed patients until ICU discharge and defined extubation failure as the need for reintubation in the first 48 h after extubation.

We performed blood gas analysis at baseline, at the end of the washout period, and at the end of each SBT. Dedicated software (NAVA Tracker V.4.0; Maquet, Sweden) acquired the airway pressure, flow, and EAdi from the ventilator at a sampling rate of 100 Hz for periods of 3 min at each of the following times: at baseline, during the last 10 min of the washout period, and every 10 min during each SBT. We processed and analyzed the data using MatLab (Mathworks, MA, USA), which automatically detected the initiation and termination of inspiratory efforts and ventilator cycles, and calculated peak airway pressure (Paw), neural inspiratory time (TIneural), ventilator inspiratory time (TIvent), respiratory rate (RR), tidal volume (V_T) in mL/kg of ideal body weight, and ΔEAdi, which is defined as EAdi peak minus EAdi minimal. Neuro ventilatory efficiency was calculated as tidal volume (in mL) divided by ΔEAdi, in μV [25]. We excluded cycles with artifacts and, for every subject, we averaged all the cycles in each three-minute recording to generate mean values for the above variables.

Detection of major asynchrony events employed both visual inspection and processing of the recordings of Paw, flow, and EAdi waveforms. We define these events below.

As previously described [26, 27], we calculated the asynchrony index (AI) as the number of cycles with asynchrony divided by the number of monitored neural cycles, expressed as a percentage.

We did not observe lower tidal volumes or higher respiratory rates in NAVA compared to PSV. This may be because these changes were described in studies in which PSV levels were high and over-assistance was common [13‐19, 34], while in our study, patients were receiving minimal support in both PSV and NAVA. We observed greater Paw in NAVA than in PSV (Table 2) despite having titrated NAVA before randomization to generate equivalent Paw in NAVA and PSV. But because Paw in NAVA is proportional to patient effort, greater Paw during the SBT in NAVA occurred because ΔEAdi was greater during the SBTs (both NAVA and PSV) than during the titration period. However, this difference is unlikely to have resulted in more assistance during the SBTs in NAVA compared to SBTs in PSV, because ΔEAdi, RR and V_T were similar in NAVA and SBT; moreover, if ventilatory support during the SBT in NAVA had been considerably greater than during the SBT in PSV, we would expect lower rates of SBT failure in NAVA, which was not the case.

These results suggest that patients being ventilated with NAVA during the weaning period may be screened for extubation readiness with NAVA using the same objective criteria for SBT failure used for SBTs in PSV or a T-tube.

Asynchrony events were common among this population and were, for the most part, overlooked by the clinicians caring for the patients [35]. The AI was greater than 10%, which is considered clinically significant, in 12 (71%) patients in PSV and 11 (64%) patients in NAVA. Compared to PSV, NAVA decreased total AI, which agrees with most previous studies comparing NAVA and PSV [14‐16, 19, 36]. What is novel in our results is that the AI was lower in NAVA even in the absence of over-assistance with PSV. In previous studies, NAVA decreased the total asynchrony index mostly by decreasing ineffective efforts [13‐16, 36]. In our study, ineffective efforts were uncommon because over-assistance, its most important risk factor, is not expected during SBTs. Only five patients showed ineffective efforts during PSV, which were abolished with NAVA.

Triggering delay was the most prevalent type of asynchrony and NAVA significantly reduced its occurrence compared to PSV. Few previous authors have included triggering delay into the computation of total AI [37] because its detection requires direct monitoring of patient inspiratory effort. Since triggering delay may cause considerable discomfort [38] and is easily measured with the NAVA catheter [36, 39, 40], we added it to our computation of total AI. Therefore, comparisons of our results to previous studies that did not include this type of asynchrony should take this difference into account.

NAVA had no significant impact in cycling asynchrony in comparison with PSV. Some previous studies showed a reduction in cycling delay with NAVA [13‐16] but we observed an overall low incidence of cycling delay, probably because inspiratory assistance during the SBTs was minimal. Double triggering was more common in NAVA compared to PSV, which agrees with previous results [13, 19]. Therefore, clinicians caring for patients ventilated with NAVA should pay special attention to this type of asynchrony and take measures to minimize it. Interestingly, double triggering in NAVA did not result in breath staking and high tidal volumes, which is a concern related to double triggering in assisted-controlled modes [41], because the second breath usually resulted in zero flow, as shown in Fig. 3. No flow was present on the second breath because, during inspiration, NAVA delivers inspiratory flow to generate a target airway pressure equal to EAdi times the NAVA level. When double triggering occurs, if the tidal volume delivered in the first breath has not been completely exhaled, airway pressure at the beginning of the second breath may be greater than the target airway pressure calculated by NAVA, and thus the ventilator does not deliver more flow.

The SBT success rate in NAVA was high and comparable to PSV. All patients passed the SBT in PSV, which was unexpected, since most reports point to approximately 20% failure in the first SBT [2, 42, 43]. Possible explanations for such a high success rate could be that our study population was not too sick or that the ICU team’s decision to submit the patients to an SBT was delayed. However, our population had high SAPS 3 scores at admission, almost half had a history of chronic lung disease and, despite passing the SBTs in PSV, 25% of them required reintubation. We applied low values of inspiratory support and PEEP during the SBTs [1, 2, 8, 44] and ΔEAdi values during the trials were similar to values reported during mechanical ventilation weaning [21, 45]. However, we can’t rule out that this level of inspiratory support prevented clinicians from identifying patients not ready to breathe without assistance [46]. We believe that a reasonable explanation for the high success rate of SBTs in our population is that clinicians tend to perform badly when evaluating patients in SBTs and may have missed clinical signs of intolerance [47, 48].

Three patients failed the SBT in NAVA but, because our protocol mandated that clinicians used the result of the SBT in PSV, they were extubated. One of these patients needed to be reintubated. Given the preliminary nature of our trial, we were underpowered to detect differences in sensitivity and specificity to predict extubation failure between NAVA and PSV, and therefore we consider it premature to speculate if NAVA may improve the sensitivity and positive predictive value and reduce the specificity and negative predictive value of the SBT. The reintubation rate was relatively high, and could be a reflection of the severity of illness at ICU admission, measured by the SAPS3, and of the ICU profile, a unit that is specialized in care for patients with respiratory failure and need for mechanical ventilation in a large university hospital. Moreover, given the small number of participants, the confidence interval around the reintubation rate is large, and although Boles et all report an average 13% reintubation rate in 2007 [2], a large observational, multicentric study reported 29% of extubation failure [49].

Our study has several limitations. We recruited a small, convenience sample, because we were using NAVA during SBTs for the first time. This prevented the study from having enough power to detect differences in the success rate of SBTs in NAVA and PSV, and from comparing the predictive values for the two types of SBTs. The patients came from a single respiratory ICU that admits high-complexity patients, which might explain the high incidence of patient ventilator asynchrony on the day of extubation. The success of the SBT is a subjective outcome, dependent on observer experience, and may have been influenced by knowledge that participants were in a clinical trial. Finally, we used the EAdi, not electromyography, to identify patients’ inspiratory efforts; therefore, inspiratory efforts initiated by accessory muscles were not detected.