Article Text
Abstract
Background Allocation concealment is essential to avoid selection bias in randomised trials. The ACUTE pilot trial compared the effectiveness of pre-hospital CPAP vs standard oxygen therapy for acute respiratory failure (ARF). The randomisation schedule was implemented, and treatment allocation concealed, with a novel method using identical boxes.
Methods Investigation of allocation concealment in the ACUTE trial proceeded in 4 stages. Firstly, the characteristics of recruited ACUTE trial patients were compared across arms. Secondly, the findings of a weekly trial box audit log were examined. Thirdly, allocation concealment was explored in paramedic focus groups and survey. Finally, a convenience sample of West Midlands and Yorkshire Ambulance Service paramedics were presented with a random pair of ACUTE boxes from each trial arm, and asked to identify any differences. If a difference was noted, they were asked to indicate which box contained CPAP and why.
Results The ACUTE study enrolled 77 participants (42 CPAP, 35 control). Baseline characteristics were similar, although patients in the CPAP arm appeared to have slightly more severe ARF. In week 10, audit of trial boxes revealed that intervention arm boxes were ‘rattling’, secondary to deflation of CPAP masks packaged under tension. All boxes were consequently re-packaged and resupplied. No comments in the paramedic survey, but a focus group participant stated that one box had appeared different when shaken. 278 paramedics participated in the observational study. 115 participants (41.4%) felt they were able to tell which box contained CPAP, predominantly due to weight difference. Of these, 81 paramedics (70%, p<0.05) were correct.
Conclusions Although the majority of paramedics were unable to determine box contents, a significant minority correctly determined the CPAP arm. Taken together with the development of a ‘rattle’ during the trial, and slight baseline imbalance in characteristics of trial participants, loss of allocation concealment cannot be excluded.