Temporal compliance trends in a cluster randomization with crossover trial of out-of-hospital cardiac arrest

Abstract

Background  Low compliance to randomized nondrug interventions can affect treatment estimates of clinical trials. Cluster-randomized crossover may be appropriate for increasing compliance in the out-of-hospital cardiac arrest setting.

Purpose  The purpose was to determine whether the elapsed time from start of a nonblinded treatment period to episode enrollment date in a cluster-randomized crossover trial is associated with compliance to either a period of brief cardiopulmonary resuscitation (CPR) with electrocardiogram (ECG) rhythm analysis or a period of longer CPR with a delayed ECG rhythm analysis in patients with out-of-hospital cardiac arrest.

Methods  The Resuscitation Outcomes Consortium PRIMED Analyze Late (AL) versus Analyze Early (AE) trial was a cluster-randomized crossover trial at 10 North American regional sites. Clusters were created based on local service preference with treatment periods varying from 3 to 12 months depending on the expected enrollment rate of each randomizing unit. Episodes on the AL arm had a target of 180 s from CPR start to shock assessment and were deemed compliant if total time was between 150 and 210 s. Episodes on the AE arm had a target of <30 s from CPR start to shock assessment and were deemed compliant if total time was <60 s. We used logistic regression to examine the association between compliance (yes/no) and the elapsed number of days from the start of the treatment period to the episode in the framework of generalized estimating equations, controlling for randomized treatment (Late, reference = Early) and treatment period length (reference = 3, 4–5, 6, 7–11, and 12 months).

Results  We had 8769 episodes in our analysis population. Overall compliance to the randomized arm was 63.5%. After adjusting for treatment arm and treatment period length, the odds of compliance for episodes occurring >300 days from treatment period start were 33% lower (odds ratio (OR): 0.67; 95% confidence interval (CI): 0.52, 0.86) than for those <60 days from treatment period start. There was no significant difference in compliance between episodes before and immediately after a cluster crossed over to the opposite arm (OR: 0.81; 95% CI: 0.57, 1.16).

Limitations  A major challenge was the lack of synchronicity between training cycles and agency crossover dates.

Conclusion  We found a significant decrease in compliance to the AL versus AE cardiac arrest intervention as the elapsed time from start of treatment period increased. We did not find a difference in compliance immediately before and after a crossover. While these results suggest that future cluster with crossover trials in the out-of-hospital setting be designed with short treatment periods and frequent crossovers, provider logistical concerns must also be considered.