Agreement of treatment effects in decentralised trials versus traditional trials: meta-epidemiological study

AbstractObjectiveTo compare treatment effects from trials designed as decentralised trials or as traditional non-decentralised trials.DesignMeta-epidemiological study.Data sourcePubMed database.Eligibility criteriaTrials identified as decentralised and included in meta-analyses alongside corresponding non-decentralised trials, restricted to drug interventions across all populations and outcome types.Results51 decentralised trials and 86 non-decentralised trials regarding 11 clinical questions were compared. Decentralised trials were larger than non-decentralised trials (median sample size 1175 v 236 participants) and more recent than non-decentralised trials (median year of publication 2012 v 2007). The level of significance agreed between decentralised and non-decentralised trials in nine of 11 clinical questions (82%). No systematic difference between effects from decentralised and non-decentralised trials was found (summary ratio of odds ratio 1.01; 95% confidence interval 0.93 to 1.09; heterogeneity I2=8.6%) with a median absolute deviation of 1.3-fold (absolute ratio of odds ratio 1.30; interquartile range 1.13-1.51).ConclusionDecentralised trials provide similar answers to clinical questions as traditional, non-decentralised randomised trials. They do not systematically overestimate or underestimate effect estimates, but modest absolute deviations occur that may lead to different conclusions. Beyond direct benefits to participants, decentralisation opens new avenues for randomised trials, including the facilitation of larger trials.