We have used the intravascular oxygenator (IVOX) in eight patients (14-76 years) with acute respiratory failure. At the time of referral for IVOX all patients were on inotropic support, five had pneumothoraces and two were on haemofiltration. Ventilatory data before IVOX were: ventilated for 2-14 days (median 6), PEEP 5-12 cm H2O (12), PIP 34-95 cm H2O (49), FiO2 0.7-1 (1), PaO2 5.4-26 KPa (8.5) and PaCO2 3.7-23 KPa (8). The intravascular oxygenator was used for 6 h-12 days (median 4), the oxygen transfer achieved was 68-140 ml/min (85) and the carbon dioxide removed was 40-106 ml/min (59). It was possible to decrease the ventilator settings with a significant increase in the PaO2 by 48 h (P < 0.05) but the fall in PaCO2 did not reach significance. As the IVOX was turned from "off" to "on" the mean PaO2 increase was 0.35 KPa +/- 0.14 SEM (P = 0.04) and the mean PaCO2 decrease 0.7 KPa +/- 0.2 SEM (P = 0.02) without any significant change in cardiac output. Two patients survived and six died from multisystem failure with three patients being hypoxic at the time of death. Postmortem examinations on five patients did not reveal any IVOX-related complications. We conclude that IVOX is safe but it is not as efficient in gas exchange as extracorporeal membrane oxygenation at present. Further improvement in the gas exchange efficiency of this prototype could render IVOX a very useful device.