Patients receiving endotracheal mechanical ventilation for more than 24 h for acute respiratory failure of respiratory cause (de novo hypoxaemic respiratory failure, acute cardiogenic pulmonary oedema or acute-on-chronic respiratory failure) were eligible when they met the following criteria: ability to sustain PSV for at least 30 min with a total level of inspiratory pressure below 30 cmH2O, estimated remaining duration of mechanical ventilation greater than 48 h, level of sedation at most 4 on the Ramsay scale in the absence of medical decision to increase the level of sedation, fraction of inspired oxygen (FiO2) at most 50 % with positive end-expiratory pressure (PEEP) no greater than 8 cmH2O and absence of administration of high-dose vasopressor therapy defined as norepinephrine above 0.3 µg kg−1 min−1 or dopamine above 10 µg kg−1 min−1. Exclusion criteria were age younger than 18 years, known pregnancy, participation in another trial within the 30 days preceding completion of the eligibility criteria, contraindication to placement of the NAVA oesophageal tube (i.e. any contraindication to placement of a gastric tube or repositioning of a tube already in place, recent gastrointestinal suture, rupture of oesophageal varices with gastrointestinal bleeding during the 4 days prior to inclusion) and decision to withhold life-sustaining treatment.
Patients were enrolled from April 2012 to June 2013 at 11 intensive care units in France, comprising six medical ICU, four medical surgical ICUs and one surgical ICU. The study protocol was approved for all centres by the Comité de Protection des Personnes Ile de France 6 (independent ethics committee), according to French law. Written informed consent was obtained from the patients or their surrogates before study inclusion. The trial was monitored by an independent data and safety monitoring board. Designated investigators at each centre enrolled patients. Blinding of treating clinicians to group assignment was not feasible, but all analyses were conducted blindly.
As soon as patients were included, the standard nasogastric feeding tube was removed and replaced by an EAdi catheter. Patients were then randomly assigned to blocks stratified by centre to receive either PSV or NAVA. Allocation concealment was achieved using sequentially numbered, opaque, sealed envelopes.
The pressure support level in the PSV group and the NAVA level in the NAVA group were set to obtain a tidal volume of 6–8 ml/kg of ideal body weight. In both groups, the physician in charge set partial pressure of arterial oxygen (FiO2) and positive end-expiratory pressure (PEEP) according to local guidelines. NAVA or PSV were continued for 14 days unless the patients met criteria for switching to controlled mechanical ventilation or for weaning and subsequent extubation. After 14 days, patients of the NAVA group were switched to PSV.
Criteria for switching to assist-control mechanical ventilation in the two groups were (1) respiratory distress, hypoxaemia or hypercapnic acidosis despite optimization of ventilator settings; (2) severe hypotension, shock or arrhythmias; (3) increased need for sedation for agitation or for patient–ventilatory asynchrony; (4) an investigation requiring increased sedation; (5) procedures that necessitated a level of sedation requiring controlled mechanical ventilation (gastrointestinal endoscopy, transoesophageal echocardiography, surgery). As soon as the criteria for switching to assist-control mechanical ventilation and the inclusion criteria were restored, partial ventilatory support with either NAVA or PSV was re-instituted according to the randomization group.
Weaning from the ventilator was performed according to the same protocol in both groups. Although weaning could be started before the second day following randomization if deemed appropriate by the attending physician, the protocol required daily screening of prerequisite criteria for weaning at least from the second day following randomization: improvement of the acute respiratory failure causative factor, SpO2 greater than 92 % with FiO2 less than 50 % and PEEP less than 5 cmH2O, no infusion of vasopressor agents or sedatives and adequate responses to simple commands. If prerequisite criteria for weaning were present, a spontaneous breathing trial was performed. This test consisted of breathing spontaneously for 30 min to 1 h disconnected from the ventilator on a T piece, or in pressure-support ventilation with an inspiratory pressure of 7 cmH2O and zero end-expiratory pressure. The test was interrupted if any of the following signs of poor tolerance were observed: respiratory rate greater than 35/min, SpO2 less than 90 %, heart rate below 140/min, arterial systolic blood pressure above 180 or below 90 mmHg. If no signs of poor tolerance were observed, the patient was extubated. Post-extubation NIV could be instituted according to the decision of the physician in charge.
Data collection and outcome measures
The primary outcome was the proportion of patients who remained continuously in partial ventilatory support (either NAVA or PSV) during the first 48 h following inclusion or until extubation without any return to assist-control ventilation, which was defined as “success” (of partial ventilator support). In contrast, “failure” was defined as at least one episode of return to assist-control mechanical ventilation. To accurately assess the primary outcome, we precisely determined the time spent in each ventilator mode each day: data were collected each day from the ventilator with a ventilation record card, and were then further processed and stored.
Secondary outcomes were the prevalence of patient–ventilator asynchrony, the incidence of ventilator-acquired pneumonia according to predefined criteria , ventilator-free days, duration of mechanical ventilation from either intubation or inclusion to successful extubation, defined as extubation not followed by another intubation within 48 h, ICU and hospital length of stay, ICU and 28-day mortality. Application and duration of post-extubation noninvasive mechanical ventilation was also collected. There were no predefined criteria for the institution of post-extubation noninvasive mechanical ventilation, which was based on the decision of the physician in charge of the patient.
Data were collected at the time of randomization to characterize the severity of the underlying medical conditions, the severity of the acute illness, ventilatory settings, arterial blood gases and the cause of acute respiratory failure. Patients were monitored daily for dyspnoea as quantified by a visual analogue scale, comfort as evaluated by the ATICE score , ventilatory conditions and arterial blood gases.
In order to quantify patient–ventilator asynchrony, airway pressure, flow and EAdi were acquired at 100 Hz for 20 min from the ventilator via an RS232 interface connected to a computer using commercially available software (Servo-i® RCR, version 3.6.2, Maquet Critical Care) 12, 24, 36 and 48 h after inclusion. The five main types of asynchronies were quantified off-line by the two same investigators (A.D. and C.R.D.) according to previously published definitions  and are described in Table E2 (see the Online Data Supplement): (1) ineffective efforts, (2) auto-triggering, (3) premature cycling (4) late cycling and (5) double triggering. The number of episodes of each type of asynchrony was reported as the total number of each event per minute. A global asynchrony index (AI) was computed as previously published , defined as the number of asynchrony events divided by the total respiratory rate computed as the sum of the number of ventilator cycles (triggered or not) and of wasted efforts: asynchrony index (expressed in percentage) = number of asynchrony events/total respiratory rate (ventilator cycles + wasted efforts) × 100.
Sample size calculation was based on a previous report  as follows: assuming 78 % of patients remaining in partial ventilator support in the PSV group during the first 48 h, 58 patients per group would provide 80 % power at a two-sided α level of 0.05 to detect a 17 % absolute increase in the probability of continuously remaining on partial ventilator support in the NAVA group without any return to assist-control ventilation. With an estimated 10 % failure of ventilator data collection, the final calculated sample size was 128 patients.
All analyses were conducted on an intention-to-treat basis at a 5 % level of statistical significance. Continuous data are reported as median (interquartile range [IQR]) and categorical data as number of events (percentages). Differences between groups were assessed with the t test or the Mann–Whitney test for continuous variables and the χ
2 test for categorical variables. Adjustments were performed with the Cochran–Mantel–Haenszel test. Probability of remaining continuously in an assisted mode, time to successful extubation defined as extubation not followed by reintubation within 48 h, and time to unassisted breathing (including noninvasive ventilation) curves were constructed (Kaplan–Meier method) and differences between groups were compared using the log rank test. Patients’ characteristics and baseline variables between failure and success patients were compared. All reported p values are two-sided. Statistical analysis was conducted using SAS version 9.2 statistical software (SAS Institute Inc., Cary NC, USA).