Neurally adjusted ventilatory assist (NAVA), a mode of mechanical ventilation controlled by diaphragmatic electrical activity (EAdi), may improve patient-ventilator interaction. We examined patient-ventilator interaction by comparing EAdi to ventilator pressure during conventional ventilation (CV) and NAVA delivered invasively and non-invasively. Seven intubated infants [birth weight 936 g (range, 676-1266 g); gestational age 26 wk (range, 25-29)] were studied before and after extubation, initially during CV and then NAVA. NAVA-intubated and NAVA-extubated demonstrated similar delays between onset of EAdi and onset of ventilator pressure of 74 +/- 17 and 72 +/- 23 ms (p = 0.698), respectively. During CV, the mean trigger delays were not different from NAVA, however 13 +/- 8.5% of ventilator breaths were triggered on average 59 +/- 27 ms before onset of EAdi. There was no difference in off-cycling delays between NAVA-intubated and extubated (32 +/- 34 versus 28 +/- 11 ms). CV cycled-off before NAVA (120 +/- 66 ms prior, p < 0.001). During NAVA, EAdi and ventilator pressure were correlated [mean determination coefficient (NAVA-intubated 0.8 +/- 0.06 and NAVA-extubated 0.73 +/- 0.22)]. Pressure delivery during conventional ventilation was not correlated to EAdi. Neural expiratory time was longer (p = 0.044), and respiratory rate was lower (p = 0.004) during NAVA. We conclude that in low birth weight infants, NAVA can improve patient-ventilator interaction, even in the presence of large leaks.