OBJECTIVES: Emergency medical research performed under federal regulation 21 section sign CFR 50.24 provides a means to protect human subjects and investigate novel time-sensitive treatments. Although prospective individual consent is not required for studies conducted under this regulation, consent from a legally authorized representative (LAR) or the patient at the earliest feasible opportunity is required to obtain short- and long-term outcome data. The objective of this study was to determine which demographic, cardiac arrest, and patient outcome characteristics predicted the likelihood of obtaining informed consent following enrollment under exception from informed consent in a multicenter cardiac arrest study. METHODS: This investigation was an analysis of data collected during a multisite, randomized, controlled, out-of-hospital cardiac arrest clinical trial performed under 21 section sign CFR 50.24. Research personnel attempted to obtain informed consent from LARs and subjects for medical records review of primary outcome data, as well as consent for neurologic outcome assessments up to 1 year post-cardiac arrest. Hospital discharge and neurologic status were obtained from public records and/or medical records up until the time consent was formally denied, in accordance with federal regulations and guidance. Local institutional review boards also allowed medical records review for cases where consent was neither obtained nor declined despite multiple consent attempts. Patient demographic, cardiac arrest, and clinical outcome characteristics were analyzed in univariate multinomial regression models, with consent status (obtained, denied, neither obtained nor denied) as the dependent variable. A multivariate multinomial logistic regression was then performed. An exploratory secondary analysis following the same process was performed after assigning patients who neither consented nor declined to the declined consent group. RESULTS: Among a total study population of 1,655 cardiac arrest subjects, 457 were transported and had consent attempted (27.6%). The survival status and neurologic function at the time of hospital discharge were known in 440 of 457 (96%) subjects. In the multivariate analysis, initial rhythm of ventricular fibrillation/ventricular tachycardia (VF/VT) and survival with good neurologic outcome were strong predictors of obtaining consent (odds ratio [OR] = 3.15, 95% confidence interval [CI] = 1.73 to 5.75; OR = 7.64, 95% CI = 2.28 to 25.63, respectively). The exploratory secondary analysis also showed initial rhythm of VF/VT and survival with good neurologic outcome as strong predictors of obtaining consent (OR = 1.86, 95% CI = 1.17 to 2.95; OR = 4.52, 95% CI = 2.21 to 9.26, respectively). CONCLUSIONS: Initial arrest rhythm and survival with good neurologic outcome were highly predictive of obtaining consent in this cardiac arrest trial. This phenomenon could result in underrepresentation of outcome data in the study arm with the worse outcome and represents a significant potential confounder in studies performed under 21 section sign CFR 50.24. Future revisions to the exception from informed consent regulations should allow access to critical survival data recorded as part of standard documentation, regardless of patient consent status.