The mechanism by which mechanical ventilation (MV) with positive end-expiratory pressure (PEEP) improves hypoxemia in patients with acute respiratory failure (ARF) is unclear, and may be attributed in part to a decrease in cardiac output inducing by itself a reduction of the shunt. Using the multiple inert gas elimination technique we evaluated the effects of PEEP on ventilation-perfusion (VA/Q) distribution in 8 patients while cardiac output was maintained at control value by means of a dopamine infusion. In each patient, evaluation was performed during MV without PEEP (control) then with PEEP (17 +/- 2 cm H2O) and dopamine. After application of PEEP, PaO2, PvO2, and oxygen transport (TO2) increased significantly, whereas venous admixture decreased from 37.5 +/- 5 to 17 +/- 2% (p less than 0.01). Comparison of VA/Q distribution during PEEP and zero end-expiratory pressure documented a redistribution of pulmonary blood flow; the shunt decreased markedly from 30 +/- 4 to 13 +/- 2% (p less than 0.001), whereas the fraction of cardiac output distributed to "normal" VA/Q ratio units (0.1 to 10) increased from 62 to 78.5% (p less than 0.001). Dead space increased slightly with PEEP, from 44 to 49% (p less than 0.01) of total ventilation. The pattern of ventilation distribution was essentially unaltered; specifically, no additional high VA/Q mode was observed during PEEP. It is concluded that cardiac output maintenance with dopamine infusion during PEEP does not suppress the beneficial effects of PEEP on gas exchange, but induces a redistribution of pulmonary blood toward the main VA/Q ratio.