We prospectively measured the end-tidal carbon dioxide concentration during 13 episodes of cardiac arrest in 10 critically ill patients receiving mechanical ventilation, to evaluate its usefulness as an indicator of circulatory status during cardiac arrest and resuscitation. The end-tidal carbon dioxide concentration decreased from a mean (+/- SD) of 1.4 +/- 0.9 to 0.4 +/- 0.4 percent after the onset of cardiac arrest. During precordial compression, it increased to 1.0 +/- 0.5 percent. The decreases in the end-tidal carbon dioxide concentration were associated with increases in systemic oxygen extraction (r = 0.79). Spontaneous circulation was restored on seven occasions. This was heralded by a rapid increase in the end-tidal carbon dioxide concentration, from 1.3 +/- 0.5 percent to an overshoot value of 3.7 +/- 2.1 percent, within approximately 30 seconds. The concentration then declined to a stable value of 2.4 +/- 1.8 percent four minutes after resuscitation. However, it remained 0.7 +/- 0.4 percent in six patients in whom resuscitative efforts failed to restore spontaneous circulation. These observations are consistent with experimental studies of cardiopulmonary resuscitation in pigs, in which the end-tidal carbon dioxide concentration varied directly with the cardiac output produced by precordial compression. We therefore propose that measurement of the end-tidal carbon dioxide concentration may be a practical, non-invasive method for monitoring blood flow generated by precordial compression during cardiopulmonary resuscitation and an almost immediate indicator of successful resuscitation.