Cluster Versus Individual Randomization in Prehospital, Emergency and Critical Care Trials: A Meta-Epidemiological Study

Objectives: Cluster-randomized clinical trials (cRCTs) are pooled with individually randomized clinical trials (iRCTs) in meta-analyses. However, cRCTs are associated with methodological characteristics potentially inducing systematic differences compared with iRCTs. In acute settings, trial inclusion is time-sensitive, exacerbating inclusion-based biases. We conducted a meta-epidemiological study to assess deviations in effect estimates between cRCTs and iRCTs in these settings. Data Sources: A systematic literature search was conducted in PubMed and Embase. Data Selection: We screened and identified all meta-analyses in prehospital, emergency, or critical care within the last decade that included at least one cRCT and one iRCT. Data Extraction: Data were extracted in duplicate from the meta-analyses. If overlapping trials were identified, results were pooled. The primary outcome was mortality; the secondary outcome comprised length-of-stay- or days-alive-type outcomes. Data Synthesis: Using absolute ln(odds ratio [OR]) for mortality and absolute Hedges’ g for the secondary outcome, mean differences (MDs) in effect estimates between cRCTs and iRCTs were analyzed using mixed regression models, presented with 95% CIs. We identified 94 eligible meta-analyses; 44.7% reported on randomization type (cRCT or iRCT). We included 65 meta-analyses in quantitative analyses, encompassing 212 randomized clinical trials (39 cRCTs and 173 iRCTs) and 315,276 patients. No statistically significant difference in absolute ln(OR) for mortality between cRCTs and iRCTs was found (MD, –0.110; 95% CI, –0.093 to 0.313). For length-of-stay- or days-alive-type outcomes, a significant difference in absolute Hedges’ g was found (MD, –0.202; 95% CI, –0.392 to –0.012). Conclusions: A systematic difference between cRCTs and iRCTs in length-of-stay- or days-alive-type outcomes, but not mortality, was identified, with cRCTs reporting smaller effect size magnitudes than iRCTs. With under half of identified meta-analyses reporting on randomization modality, these results highlight the need for improved transparency and statistical consideration of randomization type in future meta-analyses in prehospital, emergency, and critical care.