The purpose of this study was to determine whether serum potassium concentration (SK) can predict resuscitation outcome in a canine model of severe hypothermic cardiac arrest. Fifteen adult mongrel anesthetized dogs were immersed to the neck in a 4 degrees C water bath and ventilated with room air, with ventilation halved at 45 min and stopped at 90 min. After cardiac arrest, 14 of the dogs were kept in the water bath for periods of 2-7 h, and another was held in arrest for 13 h. Following 10 min of closed chest cardiopulmonary resuscitation (CPR) (simulating a short transport time to a hospital), animals were placed on cardiopulmonary bypass and rapidly rewarmed. With appearance of ventricular fibrillation, animals were defibrillated up to three times. Standard advanced cardiac life support was initiated at a core temperature (Tc) of 30 degrees C. Eight of the 15 dogs had return of spontaneous circulation (ROSC), at Tc ranging from 30.4 to 36.5 degrees C. The eight dogs with ROSC did not differ from the seven without ROSC in time to arrest (128 +/- 48 versus 128 +/- 23 min) (mean +/- SD) or Tc at arrest (18.1 +/- 2.2 versus 17.9 +/- 3.1 degrees C), but had higher Tc at the end of the arrest period (9.7 +/- 3.0 versus 5.2 +/- 2.0 degrees C), reflecting a shorter arrest period in the dogs with ROSC (225 +/- 95 versus 420 +/- 193 min). SK (mEq liter(-1)) did not differ between dogs with and without ROSC at baseline (3.5 +/- 0.4 versus 3.7 +/- 0.4) or at arrest (3.4 +/- 0.7 versus 4.3 +/- 2.2), but there was a trend toward higher SK at the end of arrest in the group without ROSC (4.6 +/- 1.5 versus 9.4 +/- 6.3; range 3.2-7.8 versus 3.5-21.4; p = .053). SK was similar after 10 min of CPR in the groups with and without ROSC (6.6 +/- 2.9 versus 9.0 +/- 2.4; range 2.5-11.1 versus 4.5-11.0; p = .107). SK after 10 min of CPR was higher in some animals with ROSC (9.6 and 11.1) than in others which did not have ROSC (4.5 and 7.9). We conclude that very high SK following prolonged hypothermic cardiac arrest may be suggestive of an inability to resuscitate. However, SK after both prolonged hypothermic cardiac arrest and a brief period of CPR is not a good predictor of resuscitation using cardiopulmonary bypass rewarming in an animal model.