PURPOSE: Non-invasive ventilation is largely used to treat acute and chronic respiratory failure. This ventilation encounters a non-negligible rate of failure related to the used interface/mask, but the reasons for this failure remain unclear. In order to shed light on this issue and to better understand the effects of the geometrical design of interfaces, we aimed to quantify flow, pressure and gas composition in terms of CO(2) and O(2) at the passage through different types of interface (oronasal mask, integral mask and helmet). In particular, we postulated that due to specific gas flow passing throughout the interface, the effective dead space added by the interface is not always related to the whole gas volume included in the interface. METHODS: Numerical simulations, using computational fluid dynamics, were used to describe pressure, flow and gas composition during ventilation with the different interfaces. RESULTS: Between the different interfaces the effective dead spaces differed only modestly (110-370 ml), whereas their internal volumes were markedly different (110-10,000 ml). Effective dead space was limited to half the tidal volume for the most voluminous interface, whereas it was close to the interface gas volume for the less voluminous interfaces. Pressure variations induced by the flow ventilation throughout the interface were negligible. CONCLUSIONS: Effective dead space is not related to the internal gas volume included in the interface, suggesting that this internal volume should not be considered as a limiting factor for their efficacy during non-invasive ventilation. Patient's comfort and synchrony have also to be taken into account.