Abstract 353: The Uncertainty Principle in Cardiac Arrest and Critical Care Trials

Abstract

Introduction: A challenge with clinical trials is the capacity to identify a population that may potentially benefit from the intervention that is not “too well” (will recover) or “too sick” (will not survive) regardless of intervention. While inclusion criteria attempt to obtain this, populations enrolled in trials include pts in which bedside clinicians would be able to categorize in terms of likelihood of survival independent of study intervention. We further hypothesize that patients categorized as having an uncertain outcome would be most likely to benefit from the intervention. Methods: We present subgroup analyses of patients from the Thiamine as a Metabolic Resuscitator in Cardiac Arrest (THACA) trial in patients with out of hospital cardiac arrest and the Ascorbic Acid, Corticosteroids and Thiamine in Sepsis (ACTS) trial in patients with septic shock. The treating physician predicted at the time of enrollment whether the patient is “very likely”, “very unlikely”, or ”uncertain” to be alive at 30 days. Results: Among 31 pts in the THACA trial with physician predictions, 12/31 (39%) were predicted very unlikely to be alive on day 30, 3/31 (10%) very likely, and 16/31 (52%) uncertain. Among 200 pts in the ACTS trial, 13/200 (7%) of patients were predicted as very unlikely, 72/201 (36%) very likely and 115/201 (57%) uncertain. Predicted mortality was associated with observed mortality in both trials(p = 0.04, p<0.001). When evaluating the primary outcome in the population categorized as uncertain, there was a non-significant shift in favor of the intervention in THACA (p = 0.69); and a difference favoring the intervention in ACTS (p = 0.03) as opposed to the overall cohort which did not reach statistical significance (p = 0.12). Conclusion: In two RCTs in patients with cardiac arrest and septic shock, physicians reliably predicted mortality. Restricting analyses to pts with an uncertain outcome, the primary outcome in one trial changed from trending toward favoring the intervention to a statistically significant difference. The second trial shifted in favor of the outcome but did not reach statistical significance. These findings suggest that physician prediction may be a potential method to target patients with the greatest chance to benefit from clinical trials.