This study retrospectively analyzed the pattern of sodium bicarbonate (SB) use during cardiopulmonary resuscitation (CPR) in the Brain Resuscitation Clinical Trial III (BRCT III). BRCT III was a prospective clinical trial, which compared high-dose to standard-dose epinephrine during CPR. SB use was left optional in the study protocol. Records of 2915 patients were reviewed. Percentage, timing and dosage of SB administration were correlated with demographic and cardiac arrest variables and with times from collapse to Basic Life Support, to Advanced Cardiac Life Support (ACLS) and to the major interventions performed during CPR. SB was administered in 54.5% of the resuscitations. The rate of SB use decreased with increasing patient age-primarily reflecting shorter CPR attempts. Mean time intervals from arrest, from start of ACLS and from first epinephrine to administration of the first SB were 29+/-16, 19+/-13, and 10.8+/-11.1 min, respectively. No correlation was found between the rate of SB use and the pre-ACLS hypoxia times. On the other hand, a direct linear correlation was found between the rate of SB use and the duration of ACLS. We conclude that when SB was used, the time from initiation of ACLS to administration of its first dose was long and severe metabolic acidosis probably already existed at this point. Therefore, if SB is used, earlier administration may be considered. Contrary to physiological rationale, clinical decisions regarding SB use did not seem to take into consideration the duration of pre-ACLS hypoxia times. We suggest that guidelines for SB use during CPR should emphasize the importance of pre-ACLS hypoxia time in contributing to metabolic acidosis and should be more specific in defining the duration of "protracted CPR or long resuscitative efforts", the most frequent indication for SB administration.