Article type
Year
Abstract
Background: Risk of bias is a critical consideration when interpreting and applying the results of randomized controlled trials (RCTs).
Objective: To describe the risk of bias in RCTs in pediatric critical care.
Methods: As part of a scoping review we searched MEDLINE, EMBASE, LILACS and CENTRAL (from inception to January 4, 2013). We included RCTs and quasi-randomized trials published in English that administered any intervention to children in a critical care unit. We excluded trials enrolling exclusively newborns, cross-over trials, and those only published as abstracts. Pairs of reviewers independently screened studies for eligibility, abstracted data, and used the Cochrane Risk of Bias Tool to describe the risk of bias for the included trials.
Results: We included 221 trials. Figure 1 shows the assessments for the individual domains of the Cochrane Risk of Bias tool. 97 trials (44%) were assessed as high risk of bias for at least one domain. 10 trials (5%) were assessed as low risk of bias for all domains. All trials at low risk of bias were published since 2006 and the proportion at low risk of bias increased over time (p for trend < 0.001). All trials at low overall risk of bias were published since 2006. Trials at high risk of bias less frequently reported a commercial source of funding (10% vs. 23%; p = 0.02) and less frequently studied medications (52% vs. 70%; p = 0.01) than trials at low or unclear risk of bias. The median (IQR) number of children randomized was not different between trials at high 48 (28–97) and low or unclear risk of bias [46 (36–92); p = 0.89].
Conclusions: Many trials in pediatric critical care are at high risk of bias, most commonly because of lack of blinding. The proportion of trials at low risk of bias is increasing over time.
Objective: To describe the risk of bias in RCTs in pediatric critical care.
Methods: As part of a scoping review we searched MEDLINE, EMBASE, LILACS and CENTRAL (from inception to January 4, 2013). We included RCTs and quasi-randomized trials published in English that administered any intervention to children in a critical care unit. We excluded trials enrolling exclusively newborns, cross-over trials, and those only published as abstracts. Pairs of reviewers independently screened studies for eligibility, abstracted data, and used the Cochrane Risk of Bias Tool to describe the risk of bias for the included trials.
Results: We included 221 trials. Figure 1 shows the assessments for the individual domains of the Cochrane Risk of Bias tool. 97 trials (44%) were assessed as high risk of bias for at least one domain. 10 trials (5%) were assessed as low risk of bias for all domains. All trials at low risk of bias were published since 2006 and the proportion at low risk of bias increased over time (p for trend < 0.001). All trials at low overall risk of bias were published since 2006. Trials at high risk of bias less frequently reported a commercial source of funding (10% vs. 23%; p = 0.02) and less frequently studied medications (52% vs. 70%; p = 0.01) than trials at low or unclear risk of bias. The median (IQR) number of children randomized was not different between trials at high 48 (28–97) and low or unclear risk of bias [46 (36–92); p = 0.89].
Conclusions: Many trials in pediatric critical care are at high risk of bias, most commonly because of lack of blinding. The proportion of trials at low risk of bias is increasing over time.
Images
