Bias and sample size in intensive care unit trials: Protocol for a meta-epidemiological study.

Anthon, Carl Thomas; Granholm, Anders; Perner, Anders; Laake, Jon Henrik; Møller, Morten Hylander

BACKGROUND Systematic errors (bias) and random errors result in inflated and imprecise intervention effect estimates in randomised clinical trials (RCT) and meta-analyses. We aim to assess time trends in the Cochrane risk of bias domains and sample size in RCTs of intensive care unit (ICU) interventions. METHODS AND DESIGN We will conduct a meta-epidemiologic study of RCTs included in Cochrane systematic reviews assessing any intervention used in adult patients admitted to the ICU. We will search the Cochrane Database of Systematic Reviews and include reviews published from March 2011 corresponding to the latest update in the Cochrane risk of bias tool. We will extract data on risk of bias judgments in the seven Cochrane bias domains and trial sample sizes and evaluate time trends using run charts. The primary outcome will be time trends in the annual proportion of trials with overall low risk of bias (low risk of bias in all bias domains). The secondary outcomes include time trends in the annual median trial sample size, and the annual proportion of trials with low-, unclear- and high risk of bias in each of the seven Cochrane bias domains. DISCUSSION The outlined meta-epidemiologic study will assess time trends in risk of bias and sample sizes in RCTs assessing ICU interventions. This will inform researchers, healthcare personnel and policymakers on the general reliability of findings from RCTs of ICU interventions over time, and inform future RCT design and reporting.