Objective To compare the rates of death or bronchopulmonary dysplasia (BPD) in infants who received nasal intermittent positive pressure ventilation (NIPPV) delivered by a conventional mechanical ventilator (CMV) or a bilevel device.

Design A preplanned non-randomised comparison of infants randomised to the NIPPV arm of the NIPPV trial.

Setting Thirty-six tertiary neonatal units in three continents.

Patients Infants <1000 g and <30 weeks gestational age at birth.

Interventions Infants received treatment with CMV NIPPV or bilevel NIPPV, as a primary mode of respiratory support or following first extubation.

Results 241 received mainly bilevel NIPPV and 215 mainly CMV NIPPV. No difference was found in death or BPD at 36 weeks corrected age (adjusted OR 0.88 (95% CI 0.57 to 1.35)). More deaths occurred in infants receiving bilevel NIPPV (9.4%) than in CMV NIPPV (2.3%) (adjusted OR 5.01: 95% CI 1.74 to 14.4). There was a corresponding but not statistically significant decrease in BPD in the bilevel NIPPV group (30% vs 37%) (adjusted OR 0.64 (95% CI 0.41 to 1.02)). No difference was observed in extubation failure or age at last extubation. A post hoc test of interaction between device type and synchronisation was not statistically significant.

Conclusions We did not observe a statistically significant difference in the composite outcome of death or BPD between infants who received mostly bilevel NIPPV compared with mostly CMV NIPPV. Differences in component outcomes of morbidity and BPD may be due to the competing nature of these outcomes or differences in baseline characteristics of infants.

Trial registration number NCT00433212.