Previous animal and clinical studies suggest that bystander-initiated cardiac-only resuscitation may be superior to conventional cardiopulmonary resuscitation (CPR) for out-of-hospital cardiac arrests. Our hypothesis was that both cardiac-only bystander resuscitation and conventional bystander CPR would improve outcomes from out-of-hospital cardiac arrests of < or = 15 minutes' duration, whereas the addition of rescue breathing would improve outcomes for cardiac arrests lasting > 15 minutes. We carried out a prospective, population-based, observational study involving consecutive patients with emergency responder resuscitation attempts from May 1, 1998, through April 30, 2003. The primary outcome measure was 1-year survival with favorable neurological outcome. Multivariable logistic regression analysis was performed to evaluate the relationship between type of CPR and outcomes. Among the 4902 witnessed cardiac arrests, 783 received conventional CPR, and 544 received cardiac-only resuscitation. Excluding very-long-duration cardiac arrests (> 15 minutes), the cardiac-only resuscitation yielded a higher rate of 1-year survival with favorable neurological outcome than no bystander CPR (4.3% versus 2.5%; odds ratio, 1.72; 95% CI, 1.01 to 2.95), and conventional CPR showed similar effectiveness (4.1%; odds ratio, 1.57; 95% CI, 0.95 to 2.60). For the very-long-duration arrests, neurologically favorable 1-year survival was greater in the conventional CPR group, but there were few survivors regardless of the type of bystander CPR (0.3% [2 of 624], 0% [0 of 92], and 2.2% [3 of 139] in the no bystander CPR, cardiac-only CPR, and conventional CPR groups, respectively; P<0.05). Bystander-initiated cardiac-only resuscitation and conventional CPR are similarly effective for most adult out-of-hospital cardiac arrests. For very prolonged cardiac arrests, the addition of rescue breathing may be of some help.