Neurally adjusted ventilatory assist versus pressure support ventilation: a randomized controlled feasibility trial performed in patients at risk of prolonged mechanical ventilation

Neurally adjusted ventilatory assist (NAVA) uses electromyographic signals from the diaphragm (Edi) detected via a specialized naso-gastric feeding catheter (Getinge, Solna, Sweden), as a measure of neural respiratory drive and as a means of controlling the delivery of inspiratory support by a mechanical ventilator [1]. Over the last 12 years, studies have suggested several important physiological benefits associated with NAVA, including improved breath synchronization and effective assist that is proportional to neural respiratory drive [2, 3]. Studies have not yet, however, demonstrated improved patient outcomes as was originally suggested [1], with only two, relatively small trials finding non-statistically significant effects on secondary outcomes [3, 4].

Although there is a demand for adequately powered trials of NAVA [5], uncertainties remain in relation to its use over prolonged durations. Estimates of intervention adherence and protocol compliance provide critical information for the calculation of statistical power [6]. Randomized controlled clinical trials are expensive [7], and poor adherence risks a type II error, with an intervention deemed ineffective when it was actually not delivered with sufficient fidelity [8, 9]. Although good NAVA mode adherence was observed up to 48 h from randomization in a recent efficacy RCT [3], conflicting results were reported in a subsequent physiological RCT where NAVA failed in 7/20 (35%) participants across the same time period. To date, no trial has investigated adherence to a NAVA trial protocol beyond 48 h or determined the reasons for adherence or lack thereof.

The primary aim of this study was, therefore, to assess the feasibility of a trial protocol comparing NAVA to PSV, via an assessment of mode adherence. Our primary feasibility objectives were to assess mode adherence (proportion of time adherent to assigned mode) and protocol compliance (binary—≥ 65% mode adherence) during the entire study period. In contrast to previous trials, recruitment was commenced soon after intubation (ability to trigger the ventilator was not a requirement for inclusion), and the intervention was continued to the end of MV support or D28. In addition, reasons for mode cross-over and poor adherence were recorded. Some study results have been previously reported in the form of abstracts [10, 11].

An open-label, parallel-group, randomized feasibility trial was undertaken in four ICUs (surgical, general medical, neuro/trauma, and liver) comprising 75 beds at a university-affiliated hospital in London, UK. The trial was prospectively registered (NCT01826890) and was approved by the London Westminster ethics committee (13/LO/0012).

Over a 45-month enrolment period, 774 invasively ventilated patients with COPD, HF, or ARDS were identified, 112 eligible patients were approached, and 78 participants (39 NAVA and 39 PSV) were recruited, a consent rate of 72% (3 patients consented outside the recruitment window and were subsequently excluded) (Fig. 1). Baseline characteristics of the two groups were similar (Table 1) with good balance across groups for COPD (stratification factor), HF, and ARDS diagnoses. Median durations of MV in any mode and in PSV before randomization were similar (Table 1).

This four ICU, single-centre, parallel group, RCT has demonstrated protocol feasibility by achieving acceptable adherence to the assigned mode and an acceptable proportion of compliant patients. The study has also demonstrated protocol acceptability by low rates of physician recruitment refusal and high participant consent rates and protocol safety via the absence of adverse events. Acknowledging the possibility of chance findings in this relatively small study, the exploratory clinical outcomes suggest increased VFDs, reduced time to breathing without assistance and reduced time to ICU discharge in the NAVA group, and improved sedation management while in the NAVA mode. These feasibility data improve our understanding of NAVA compared to PSV beyond 48 h of application, improving the chances of success in any subsequent trial. The clinical results are consistent with those reported by Demoule et al. [3] and suggest that a fully powered trial is justified.

Despite satisfactory mode adherence and protocol compliance across groups, these proportions were lower in the NAVA group, and substantial (albeit for a short duration) cross-over from NAVA was observed. Of the reported reasons for cross-over, human factors such as trial awareness, clinician preference, and lack of NAVA experience were most common, implying low confidence in the application of NAVA and need for further training. These issues may have been expected in a large group of clinical staff (over 300) despite attempts to embed NAVA technology into unit practice. This finding also highlights the complexity of NAVA and the likely variation in clinical application caused by ICU contexts and human interactions.

While the influence of human factors described above may be modified with increased/improved training and changes to study documents, of more concern for trial feasibility is the difficulty in acquiring and maintaining a satisfactory Edi signal, which occurred in 10 out of 36 (27.8%) NAVA participants. Edi signal problems have been reported previously. Di Mussi et al. reported NAVA mode failure in seven out of 20 patients (35%) due to loss of ‘Edi synchrony’ or low Edi activity, despite having obtained a reliable Edi signal at baseline [23]. Demoule et al. reported high levels of mode adherence across the first 48 h in NAVA (median 44.1 h [IQR 33 to 47.8]), but do not report on cross-over or its reasons. Recruitment after successful commencement of PSV in the study by Demoule et al. may be relevant. However, the reasons for the disparities between these trials and for the Edi difficulties described remain unclear. Varying levels of clinical expertise across a large staff group may be expected and relevant, but the results may also suggest limitations in NAVA technology (including anatomical abnormalities such as hiatal hernia) and/or a technical complexity in its clinical application over and above those associated with PSV. Notably, both trials were conducted in academic centres; therefore, these issues may be further magnified in non-academic centres with less experience of MV and NAVA.

In contrast to previous studies, the exploratory analysis of ventilation and clinical outcomes found potential benefit in the NAVA arm with increased VFDs to D28, decreased time to breathing without assistance and to alive ICU discharge, and reduced hospital mortality. Acknowledging the low power of these analyses, certain characteristics of our study that are different to previous studies may be relevant to the interpretation of these results. We specifically selected patients with risk factors for prolonged MV [12] and conditions where NAVA has potential physiological benefits [24‐26]. In addition, Edi monitoring was used in the NAVA arm only, meaning the trial assessed the combined effects of monitoring and NAVA. Neural respiratory drive (Edi) monitoring is a potential advantage of NAVA [27] as it may encourage adjustment and improvement of ventilator settings, decrease sedative use, and expedite clinical assessment of weaning readiness. Clinical factors such as renal replacement therapy for example, which differed between the groups, or the lack of an extubation protocol, may also be of relevance to the results. Accepting the need for cautious interpretation, outcomes favouring NAVA are consistent with most published physiological data on NAVA and with trends toward superior clinical outcomes in two previous NAVA RCTs [3, 4]. As such, our findings strengthen the support of a future definitive, adequately powered RCT.

In addition to benefits in ventilation and clinical outcomes and despite no difference in sedation dose and RASS during in all MV modes, RASS scores were closer to zero in the NAVA mode compared to the PSV mode (p = 0.020), potentially indicating improved sedation management. Decreases in sedation may be driven by the need to optimise the Edi signal and ensure optimal levels of respiratory drive are present prior to the initiation of NAVA [28, 29]. Improved synchrony and patient comfort when in the NAVA mode may further reduce sedative requirements. Reduced sedation [30] and improved COMFORT scores [31‐33] have been suggested in paediatric trials, but there are limited data available from adult studies. Demoule et al. [3] showed improved dyspnoea at day 1, but no overall difference in adaptation to intensive care environment (ATICE) comfort scores [34] in patients ventilated with NAVA versus PSV. Coisel et al. [35] found no difference in RASS in a small randomized cross-over study in 14 patients ventilated post-operatively. However, while sedation load and RASS outcomes may provide insight into the potential mechanisms underlying beneficial effects of NAVA, the individual clinical relevance and relative clinical value remain unclear.

This study is the first to our knowledge to report NAVA mode adherence rates beyond 48 h and reasons for cross-over, methodological evidence which is key to the success of future clinical trials. Together with the clinical benefit suggested in secondary outcomes, this study suggests that it is both feasible and justified to conduct a definitive randomized controlled trial to establish the effectiveness of NAVA in patients with a high likelihood of prolonged MV.