Standard external cardiopulmonary resuscitation (CPR) steps A-B-C produce a low blood flow that may or may not preserve brain viability during prolonged cardiac arrest. A dog model was used with ventricular fibrillation (VF) of 20 minutes, reperfusion with brief cardiopulmonary bypass, controlled ventilation to 20 hours, and intensive care to 96 hours. A retrospective comparison was made of the results of one series, now called "group I" (n = 10)--which received CPR basic life support interposed from VF 10 to 15 minutes, and CPR advanced life support with epinephrine (without defibrillation) from VF 15 to 20 minutes--to the results of another series, now "control group II" (n = 10)--which received VF no flow (no CPR) for 20 minutes. All 20 dogs within protocol were resuscitated. All 10 of group I and 7 of 10 of group II survived to 96 hours. Pupillary light reflex returned after the start of cardiopulmonary bypass at 7.7 +/- 3.7 minutes in CPR group I, versus 16.3 +/- 7.4 minutes in control group II (P = .032). At 96 hours postarrest, final overall performance categories (1, normal; 5, brain death) were better in group I. Six of 10 dogs achieved normality (overall performance category 1) in group I, as compared with none of 10 in group II (P = .004). Final neurologic deficit score (0%, best; 100% worst) was lower (better) in group I (15% +/- 20%) than in group II (51% +/- 6%; P less than .001).(ABSTRACT TRUNCATED AT 250 WORDS)