Neurally adjusted ventilatory assist (NAVA) is a mode of ventilation wherein the delivered assistance is proportional to diaphragm electrical activity (EAdi) throughout inspiration. We assessed the physiologic response to varying levels of NAVA and pressure support ventilation (PSV) in 13 tracheostomised patients with prolonged weaning. Each patient randomly underwent 8 trials, at four levels of assistance either in PSV and NAVA. i - high (no dyspnoea and/or distress); iv - low (associated with dyspnoea and/or distress; ii and iii - at ∼75% and ∼25% of the difference between high and low support respectively. We measured tidal volume (VT), peak EAdi, (EAdipeak) and airway pressure, ineffective efforts and breathing pattern variability. With both NAVA and PSV, decreasing assistance resulted in parallel significant increase in EAdipeak associated with a concomitant reduction in VT and minute ventilation in PSV, but not in NAVA. VT variability significantly increased when reducing ventilatory assistance in PSV only, while remained unchanged varying the NAVA level. The ineffective triggering index was not significantly different between the two modes. In patients with prolonged weaning, with the specific settings adopted, compared to PSV, NAVA reduced the risk of over-assistance and overall improved patient-ventilator interaction, while not significantly affecting patient-ventilator synchrony.