This study was conducted to observe effects of two methods of setting positive end-expiratory pressure (PEEP) based on the pressure-volume (PV) curve. After lung injury was induced by oleic acid in six mongrel adult dogs, the inflation PV curve was traced and the lower inflection point (LIP) was measured. The 'PEEP(INF)' was defined as LIP+2 cmH(2)O. After recruitment maneuver to move the lung physiology to the deflation limb of PV curve, decremental PEEP was applied. The lowest level of PEEP that did not result in a significant drop in PaO(2) was defined as the 'PEEP(DEF)'. Arterial blood gases, lung mechanics, hemodynamics, and lung volumes (measured on computed tomography during end-expiratory pause) were measured at PEEP of 0 cmH(2)O, PEEP(INF) and PEEP(DEF) sequentially. The median PEEP(INF) was 13.4 cm H(2)O (interquartile range, 12.5-14.3) and median PEEP(DEF) was 12.0 cm H(2)O (10.0-16.5) (p=0.813). PEEP(DEF) was associated with significantly higher PaO(2) and lung volumes, and significantly lower shunt fraction and cardiac index when compared to PEEP(INF) (p<0.05). Setting the PEEP based on the deflation limb of the PV curve was useful in improving oxygenation and lung volumes in a canine lung injury model.