Purpose This study was designed to optimize the latest generation venovenous (vv)-extracorporeal membrane oxygenation (ECMO)-circuit configuration and settings based on the evaluation of blood oxygenation and CO 2 removal determinants in patients with severe acute respiratory distress syndrome (ARDS) on ultraprotective mechanical ventilation. Methods Blood gases and hemodynamic parameters were evaluated after changing one of three ECMO settings, namely, circuit blood flow, FiO 2ECMO (fraction of inspired oxygen in circuit), or sweep gas flow ventilating the membrane, while leaving the other two parameters at their maximum setting. Results Ten mechanically ventilated ARDS patients (mean age 44 ± 16 years; 6 males; mean hemoglobin 8.0 ± 1.8 g/dL) on ECMO for a mean of 9.0 ± 3.8 days) receiving femoro–jugular vv-ECMO were evaluated. vv-ECMO blood flow and FiO 2ECMO determined arterial oxygenation. Decreasing the ECMO flow from its baseline maximum value (5.8 ± 0.8 L/min) to 40 % less (2.4 ± 0.3 L/min) significantly decreased mean PaO 2 (arterial oxygen tension; 88 ± 24 to 45 ± 9 mm Hg; p  < 0.001) and SaO 2 (oxygen saturation; 97 ± 2 to 82 ± 10 %; p  < 0.001). When the ECMO flow/cardiac output was >60 %, SaO 2 was always >90 %. Alternatively, the rate of sweep gas flow through the membrane lung determined blood decarboxylation, while PaCO 2 (arterial carbon dioxide tension) was unaffected when the ECMO blood flow and FiO 2ECMO were reduced to <2.5 L/min and 40 %, respectively. In three additional patients evaluated before and after red blood cell transfusion, O 2 delivery increased after transfusion, allowing lower ECMO flows to reach adequate SaO 2 . Conclusions For severe ARDS patients receiving femoro–jugular vv-ECMO, blood flow was the main determinant of arterial oxygenation, while CO 2 elimination depended on sweep gas flow through the oxygenator. An ECMO flow/cardiac output >60 % was constantly associated with adequate blood oxygenation and oxygen transport and delivery.