Article type
Year
Abstract
Background: Recruitment into randomised trials in emergency settings may be difficult because of the inherent time restrictions encountered when enrolling, randomising and treating participants. Although quasi-randomisation might mitigate some of these problems, systematic reviewers are likely to judge such trials as having a high risk of bias, due to the possibility of selection bias. However, selection bias may not actually occur since the very nature of emergency settings might preclude opportunities for bias.
Objectives: To compare quasi-randomisation with random sequence generation when used in emergency setting trials, on the prevalence of possible selection bias, and on trial recruitment outcomes.
Methods: Systematic reviews of participants with an acute injury or illness, requiring intervention as quickly as was clinically practicable were identified using the ‘emergency medicine’ topic in the Cochrane Library. Eligible reviews incorporated at least one trial using truly random sequence generation and at least one trial using quasi-randomisation. Evidence of possible selection bias was identified by assessment of group baseline characteristics. Clinical advice and published studies were used to identify important prognostic indicators, and important magnitudes of group difference.
Results: The seven eligible reviews included 27 eligible trials: 11 used quasi-randomisation and 16 used random sequence generation. Important group imbalance was identified in two of the 11 quasi-randomised trials (18%) and four of the 16 trials using true sequence generation (25%); of the latter, three trials described appropriate methods to conceal treatment allocation, though all three also had small sample sizes. Clinical heterogeneity and poor reporting limited the assessment of trial recruitment outcomes.
Conclusions: In emergency care settings quasi-randomised trials may be no more likely to result in biased recruitment of participants than trials using true randomisation. The likelihood of chance imbalances affecting results is arguably more important to consider, for both systematic reviewers and trial investigators.
Objectives: To compare quasi-randomisation with random sequence generation when used in emergency setting trials, on the prevalence of possible selection bias, and on trial recruitment outcomes.
Methods: Systematic reviews of participants with an acute injury or illness, requiring intervention as quickly as was clinically practicable were identified using the ‘emergency medicine’ topic in the Cochrane Library. Eligible reviews incorporated at least one trial using truly random sequence generation and at least one trial using quasi-randomisation. Evidence of possible selection bias was identified by assessment of group baseline characteristics. Clinical advice and published studies were used to identify important prognostic indicators, and important magnitudes of group difference.
Results: The seven eligible reviews included 27 eligible trials: 11 used quasi-randomisation and 16 used random sequence generation. Important group imbalance was identified in two of the 11 quasi-randomised trials (18%) and four of the 16 trials using true sequence generation (25%); of the latter, three trials described appropriate methods to conceal treatment allocation, though all three also had small sample sizes. Clinical heterogeneity and poor reporting limited the assessment of trial recruitment outcomes.
Conclusions: In emergency care settings quasi-randomised trials may be no more likely to result in biased recruitment of participants than trials using true randomisation. The likelihood of chance imbalances affecting results is arguably more important to consider, for both systematic reviewers and trial investigators.